大黑AI工具速报

## What's changed - Auto mode is now available without `CLAUDE_CODE_ENABLE_AUTO_MODE` opt-in on Bedrock, Vertex AI, and Foundry; disable via `disableAutoMode` in settings - Fixed the terminal freezing and keystrokes lagging while streaming responses containing very long lists, tables, paragraphs, or code blocks - Fixed remote managed settings from a non-interactive run (`claude -p`, the SDK) being permanently recorded as consented without ever showing the security consent dialog - Fixed spurious prompt-injection warnings triggered by benign system-generated conversation updates - Fixed the auto-updater overwriting a custom launcher script or symlink at `~/.local/bin/claude` on every release; `/doctor` now reports an externally managed launcher - Fixed compound commands with `cd` prompting for permission when the only output redirect was to `/dev/null` - Fixed the transcript jumping above the start of the answer when a response finishes streaming - Fixed `extensions.worktreeConfig` being left in the repo's `.git/config` (breaking go-git tools like `tea`) after the last `worktree.sparsePaths` worktree was removed - Fixed malformed bracket patterns in rules globs, skill paths, `.ignore`, and `.worktreeinclude` breaking file reads, file suggestions, and worktree creation - Fixed a crash loop in agent teams where a malformed teammate mailbox message caused repeated errors every second until the mailbox file was manually deleted - Fixed background sessions auto-named by accepting a plan not showing that name on their agent-view row - Fixed background sessions that entered a git worktree resuming blank after a cold reopen from the agent list - Fixed Remote Control task status updates being lost when the connection recovered from a network interruption or credential refresh - Fixed Remote Control sessions hosted by the desktop app not showing background agent and workflow progress on mobile and web - Fixed Deep research runs labeling every Fetch-phase agent "unknown" — chips now show the source hostname - Fixed Bedrock repeatedly requesting fresh AWS SSO credentials from IAM Identity Center on every API request - Improved agent view: pasting the same text again now expands the collapsed `[Pasted text #N]` placeholder instead of adding a second one - Improved agent view: blocked session peeks now lead with the question and show a worded staleness clock (`waiting 3m`) instead of the same timestamp twice - Changed Bedrock, Vertex, and Claude Platform on AWS to default to Claude Opus 4.8 - Changed auto mode to no longer read `autoMode` from `.claude/settings.local.json` (repo-resident); use `~/.claude/settings.json` instead - Fixed an indefinite hang on Windows when AWS credential resolution stalls (e.g. a stuck `credential_process`): the 60-second stall guard now fires instead of waiting forever. - Plugin hooks/monitors/MCP headersHelper: `${user_config.*}` in shell-form commands is now rejected (shell-injection fix). Hooks: use exec form (`args` array) or `$CLAUDE_PLUGIN_OPTION_<KEY>`; monitors and headersHelper: read the value inside the script (config file or the server's `env` block). - Plugin option values (`pluginConfigs`) are no longer read from project-level `.claude/settings.json`; only user, `--settings`, and managed settings are honored - Fixed `/usage-credits` amount inputs silently stripping malformed values (e.g. a pasted timestamp) to digits; malformed amounts are now rejected with an error, and amounts over $1,000 require a typed confirmation

## 变更内容 - 为 `/cd` 添加了目录路径建议，与 `/add-dir` 行为一致 - 新增 `/doctor` 检查，建议修剪已检入仓库的 `CLAUDE.md` 文件，移除 Claude 可直接从代码库推导出的内容 - `/commit-push-pr` 现在除了 `origin` 外，还会自动允许 `git push` 到仓库配置的推送远程仓库（`remote.pushDefault`，或仅配置一个远程仓库时的那个） - 网关：`/login` 现在支持 Anthropic 运营的公共网关端点 - `EnterWorktree` 在进入项目 `.claude/worktrees/` 目录之外的 git worktree 前，会先请求确认 - 后台代理现在会在 Claude Code 更新后立即在后台升级到新版本，而无需在附加时经历缓慢的过期会话升级 - 修复了登录过期后，所有模型都会因一条误导性的“所选模型存在问题”错误而失败，而不是提示运行 `/login` 的问题 - 修复了 `claude --resume` 和 `--continue` 在启动时不响应键盘输入的问题 - 修复了通过 `--mcp-config` 或 `.mcp.json` 配置的 MCP 服务器忽略每个服务器独立的 `request_timeout_ms` 设置，导致长时间运行的 MCP 工具调用在新会话中因 60 秒默认超时而失败的问题 - 修复了 `CLAUDE_CODE_EXTRA_BODY` 被 `claude agents` / `--bg` 后台工作线程静默忽略的问题；现在通过 shell 导出的覆盖设置会跟随调度会话 - 修复了 OAuth MCP 服务器在单次令牌刷新失败后需要手动重新认证的问题 - 修复了 `--permission-prompt-tool` 指向 MCP 服务器时，在冷启动且服务器尚未完成连接的情况下崩溃并提示“MCP tool not found”的问题 - 修复了 `/model` 选择器中的行打印的价格与行名对应的模型不符的问题，并停止对不按此计费的提供商引用第一方标价 - 修复了当权限或允许列表限制导致服务器提供的模型行在 `/model` 选择器中定位错乱的问题 - 修复了桌面会话在回合中途发送斜杠命令后卡在“运行中”状态的问题 - 修复了在 Windows 上，当设置提示出现在裸 `claude --resume` 之前时，代理视图忽略键盘输入的问题 - 修复了 `claude rm` 导致已移除的任务仍保留在守护进程列表中，从而使该行重新出现在 `claude agents` 中的问题 - 修复了 `/remote-control` 在未登录时显示“未知命令”的问题——现在它会说明如何登录 - 修复了左箭头无法在流程详情视图中从阶段或代理后退的问题 - 修复了 `/status` 重复列出同一条安装损坏警告的问题 - 修复了 LSP 插件的虚假“未使用插件”提示和倾斜的未使用遥测问题 - 修复了 `/doctor` 的更新检查，使其比较 Homebrew 安装时使用其 cask 的渠道，而非设置渠道 - 修复了全屏跳转到底部的药丸提示在 macOS 上建议 Ctrl+End、不显示重绑定和弦，以及覆盖转录内容的问题 - Bedrock：修复了在使用 `awsCredentialExport` 辅助程序且网络出口受限时，启动时长达数分钟的挂起问题 - 改进了 `/code-review` 在 claude-opus-4-8 上所有努力级别下的发现质量 - 改进了代理视图：状态列现在使用终端全宽，而不再截断为 64 个字符 - 更改了代理视图：Ctrl+X 现在会永久移除已完成会话，且会话不再重复渲染；已删除的后台作业保持删除状态

## What's changed - Added directory path suggestions to `/cd`, matching `/add-dir` behavior - Added a `/doctor` check that proposes trimming checked-in `CLAUDE.md` files by cutting content Claude could derive from the codebase - `/commit-push-pr` now auto-allows `git push` to the repo's configured push remote (`remote.pushDefault`, or the sole remote when only one is configured) in addition to `origin` - Gateway: `/login` now supports Anthropic-operated public gateway endpoints - `EnterWorktree` now asks for confirmation before entering a git worktree outside the project's `.claude/worktrees/` directory - Background agents now upgrade to a new version in the background right after a Claude Code update, instead of paying a slow stale-session upgrade when you attach - Fixed an expired login failing every model with a misleading "There's an issue with the selected model" error instead of prompting to run `/login` - Fixed `claude --resume` and `--continue` not responding to keyboard input on startup - Fixed MCP servers configured via `--mcp-config` or `.mcp.json` ignoring a per-server `request_timeout_ms`, which caused long-running MCP tool calls to time out at the 60s default in fresh sessions - Fixed `CLAUDE_CODE_EXTRA_BODY` being silently ignored by `claude agents` / `--bg` background workers; the shell-exported override now follows the dispatching session - Fixed OAuth MCP servers requiring manual re-authentication after a single failed token refresh - Fixed `--permission-prompt-tool` pointing at an MCP server crashing with "MCP tool not found" on cold start before the server finishes connecting - Fixed `/model` picker rows printing a price for a different model than the row named, and stopped quoting first-party list prices on providers that don't bill them - Fixed server-provided model rows being misplaced in the `/model` picker when an entitlement or allowlist restriction drops the row they were positioned against - Fixed desktop sessions getting stuck showing "running" after a slash command was sent mid-turn - Fixed keyboard input being ignored in the agents view when a setup prompt appeared before a bare `claude --resume` on Windows - Fixed `claude rm` leaving the removed job in the daemon roster, causing the row to reappear in `claude agents` - Fixed `/remote-control` showing "Unknown command" when logged out — it now explains how to sign in - Fixed left arrow not stepping back out of a phase or agent in the workflow detail view - Fixed `/status` listing the same broken-install warning twice - Fixed false "disused plugin" tips and skewed disuse telemetry for LSP plugins - Fixed `/doctor`'s update check to compare Homebrew installs against their cask's channel instead of the settings channel - Fixed the fullscreen jump-to-bottom pill suggesting Ctrl+End on macOS, not showing rebound chords, and wrapping over the transcript - Bedrock: fixed a multi-minute startup hang when using an `awsCredentialExport` helper on networks with restricted egress - Improved `/code-review` findings quality on claude-opus-4-8 across all effort levels - Improved agents view: status column now uses full terminal width instead of truncating at 64 characters - Changed agents view: Ctrl+X now permanently removes a completed session, and sessions no longer render twice; deleted background jobs stay deleted

## 变更内容 - 新增了一条自动模式规则，用于阻止篡改会话转录文件 - 修复了当 `--json-schema` 无效时静默产生非结构化输出，以及使用 `format` 关键字的 schema 被拒绝的问题 - 修复了当回合达到 `--max-turns` 限制时，Claude 工作期间发送的消息被静默丢失的问题 - 修复了当工作树内存在 NTFS 交接点或目录符号链接时，Windows 工作树删除会移除工作树外文件的问题 - 修复了后台代理在通过 `SendMessage` 恢复后，代理列表中仍显示为“失败”或“已完成”的问题 - 修复了当代理回合中不包含可读文本时，后台作业在代理列表中从“需要输入”状态跳回“运行中”状态的问题 - 修复了当后台代理正在进行升级重启时，`claude attach` 报错，而非等待代理恢复的问题 - 修复了会话到 PR 的链接遗漏了在 Bash 调用中创建的 PR，且该调用输出超过了 30K 行内限制的问题 - 修复了当服务器名称包含不受支持的字符时，`claude mcp add-from-claude-desktop` 卡住的问题；现在会报告无效名称，其余服务器仍会导入 - 修复了一个插件 LSP 服务器初始化失败，导致另一个插件中处理相同文件扩展名的有效 LSP 服务器无法工作的问题 - 修复了当命令运行时，启动 Claude 的目录被删除、锁定或卸载导致的 Windows 崩溃问题 - 修复了当目录扫描仍在进行时，文件监视器被关闭导致的崩溃问题 - 修复了项目验证技能在每次会话中被重写，而非仅在记录的命令变更时重写的问题 - 修复了当作业列表略微超出屏幕时，代理视图渲染高出一行，导致标题被裁剪的问题 - 修复了 Web 和移动端远程控制面板中的后台任务显示过时的“运行中”状态的问题，现通过在每次成员变更时转发完整任务状态来修复 - 改进了自动模式，在针对无法从上下文中解析的变量执行 `rm -rf` 前会先询问 - 自动更新的二进制文件下载现在直接流式写入磁盘，而非在内存中缓冲，将更新器的峰值内存使用量减少了约 400 MB - 后台任务通知现在明确说明未发生人工输入，防止对伪造的会话内审批进行操作 - 改进了代理视图：编辑、合并、评论或推送到现有 PR 的会话现在会在 `claude agents` 中链接该 PR - 改进了代理视图：行现在显示带颜色的状态词和分类器编写的标题，而非原始工具调用文本；预览视图打开时显示完整状态，包括被阻止会话的具体请求 - `/doctor` 现在是一个完整的设置检查工具，可以诊断和修复问题；`/checkup` 是其别名 - 预留了“Claude Browser”MCP 服务器名称（与“Claude Preview”一起），为 Claude Desktop 面板重命名做准备；用户配置的 MCP 服务器不能再使用这两个名称注册 - 修复了 Cowork VM 模式下本地代理会话在 CLI 2.1.203+ 版本中启动失败并显示“未登录 · 请运行 /login”的问题

## What's changed - Added an auto mode rule that blocks tampering with session transcript files - Fixed `--json-schema` silently producing unstructured output when the schema was invalid, and schemas using the `format` keyword being rejected - Fixed a message sent while Claude was working being silently lost when the turn ended at the `--max-turns` limit - Fixed Windows worktree removal deleting files outside the worktree when an NTFS junction or directory symlink existed inside it - Fixed background agents staying shown as "failed" or "completed" in the agent list after being resumed with `SendMessage` - Fixed background jobs flipping from "needs input" back to "working" in the agent list when the agent's turn contained no readable text - Fixed `claude attach` erroring when a background agent was mid-upgrade restart instead of waiting for it to come back - Fixed session-to-PR linking missing a PR created in a Bash call whose output exceeded the 30K inline limit - Fixed `claude mcp add-from-claude-desktop` getting stuck when a server name contains unsupported characters; invalid names are now reported and remaining servers still import - Fixed a plugin LSP server that fails to initialize preventing a valid LSP server from another plugin handling the same file extension - Fixed a Windows crash when the directory Claude was launched from is deleted, locked, or unmounted while a command is running - Fixed a crash when a file watcher was closed while a directory scan was still in flight - Fixed project verify skills being rewritten on every session instead of only when a documented command changed - Fixed the agent view rendering one line too high and clipping its header when the job list slightly overflowed the screen - Fixed background tasks in the web and mobile Remote Control panels showing stale "Running" status by forwarding full task state on every membership change - Improved auto mode to ask before running `rm -rf` on a variable it can't resolve from context - Auto-update binary downloads now stream to disk instead of buffering in memory, cutting the updater's peak memory usage by roughly 400 MB - Background task notifications now explicitly state that no human input has occurred, preventing fabricated in-transcript approvals from being acted on - Improved agent view: sessions that edit, merge, comment on, or push to an existing PR now link it in `claude agents` - Improved agent view: rows now show a colored state word and a classifier-written headline instead of raw tool call text, and the peek opens with full status including the exact ask for blocked sessions - `/doctor` is now a full setup checkup that can diagnose and fix issues; `/checkup` is its alias - Reserved the "Claude Browser" MCP server name (alongside "Claude Preview") ahead of the Claude Desktop pane rename; user-configured MCP servers can no longer register under either name - Fixed Cowork VM-mode local-agent sessions failing to start with "Not logged in · Please run /login" on CLI 2.1.203+

## 变更内容 - 修复了在无头会话的 SessionStart 钩子期间钩子事件未流式传输的问题，该问题可能导致远程工作者在钩子执行中途被闲置回收

## What's changed - Fixed hook events not streaming during SessionStart hooks in headless sessions, which could cause remote workers to be idle-reaped mid-hook

## 变更内容 - 新增登录即将过期的警告，以便在后台会话中断前重新进行身份验证 - 在手动权限模式下，页脚新增灰色⏸徽章，使当前模式始终可见 - 将会话的额外工作目录添加到 MCP `roots/list`，并在集合变更时发送 `notifications/roots/list_changed` - 修复了 macOS 上打开或切换后台代理会话时因虚假的低内存检测而卡顿 15–20 秒的问题（2.1.196 版本回归） - 修复了当守护进程的会话令牌过期时，后台会话永久无法响应附加、回复和停止的问题——会话现在会自动恢复 - 修复了返回 `claude agents` 时悄悄停止正在运行的子代理并从头重新运行提示的问题——它们的工作现在会延续 - 修复了交互式会话中的内存和每轮 CPU 回归：上下文使用指示器不再在每轮后重新分析整个转录 - 修复了后台代理从守护进程继承过时的 `PATH` 而非调度 shell 的 `PATH`，导致 Windows 上工具缺失的问题 - 修复了后台和代理视图会话丢失 shell 导出的 `ANTHROPIC_BASE_URL`，导致 API 密钥发送到默认端点并返回 401 错误的问题 - 修复了 Bash 在包含大量 git 工作树的仓库中因“参数列表过长”而失败的问题 - 修复了工作树隔离的子代理有时在父检出目录中运行 shell 命令，而非其自身工作树的问题 - 修复了多仓库工作区中工作树创建拒绝嵌套仓库，导致后台会话无法隔离和编辑的问题 - 修复了后台代理在其工作目录被删除、替换为文件或成为无效路径时崩溃循环的问题——现在它们只会失败一次并显示清晰的错误 - 修复了后台守护进程自动升级失败时悄悄终止所有正在运行的后台会话的问题 - 修复了 `TaskStop` 和 `TaskOutput` 无法找到由另一个代理生成的后台代理的问题——错误现在会按 ID 和描述列出正在运行的代理 - 修复了 `claude agents` 合成器在斜杠命令不可用时丢弃已输入消息的问题 - 修复了当打开一个已停止的会话且其对话已在另一个会话中打开时，代理列表崩溃的问题 - 修复了后台会话在问题已回答后仍在代理列表中显示“需要输入”的问题 - 修复了后台代理启动失败时仅显示“exit_with_message”而非实际错误的问题 - 修复了通过守护进程分支时，后台会话忽略 settings.json 中 `effortLevel` 更改的问题 - 修复了附加的后台会话忽略 `CLAUDE_CODE_DISABLE_MOUSE` 和 `CLAUDE_CODE_DISABLE_MOUSE_CLICKS` 退出设置的问题 - 修复了在所有命名代理完成后，`/exit` 错误地警告正在运行的后台代理的问题 - 修复了从非 git 目录启动的后台会话在配置了 `WorktreeCreate` 钩子时无法编辑文件的问题 - 修复了 `claude agents` 中的 `@` 目录选择器未显示已注册 git 工作树的问题 - 修复了 Windows 上 `/clear` 之后后台任务输出被空文件永久替换的问题 - 修复了向上滚动长转录历史时内容跳跃的问题 - 修复了在 bash 模式下输入时，当显示 shell 历史建议时终端闪烁和跳跃的问题 - 修复了重新附加到后台会话时输出文字 `^[[I` / `^[[O` 转义码的问题 - 修复了仅 LSP 的插件在其语言服务器提供诊断或响应导航请求时被错误标记为未使用的问题 - 改进了长响应流式传输时的响应性：实时预览更新不再重新渲染整个屏幕 - 改进了子代理行为：代理现在不太可能将其整个任务重新委托给另一个子代理 - 通过延迟加载大型捆绑依赖项而非内联，将二进制文件大小减少约 7 MB，启动内存减少约 7 MB - 将左箭头键改为不再关闭后台任务、差异和工作流详情视图——请按 Esc 键代替 - 将空的 `claude agents` 视图改为始终显示组织好的部分（需要输入/正在工作/已完成）并附带描述 - 移除了启动时的“claude 命令缺失或损坏”警告——现在它们会显示在 `/doctor` 和 `/status` 中 - 从 `claude agents` 页脚移除了冗余的导航提示 - [VSCode] 新增“启用所有会话的远程控制”设置开关

## What's changed - Added a warning when your login is about to expire, so you can re-authenticate before background sessions are interrupted - Added a grey ⏸ badge to the footer when in manual permission mode, making the active mode always visible - Added the session's additional working directories to MCP `roots/list`, with `notifications/roots/list_changed` sent when the set changes - Fixed opening or switching background agent sessions on macOS stalling for 15–20 seconds due to a false low-memory detection (regression in 2.1.196) - Fixed background sessions becoming permanently unresponsive to attach, replies, and stop when the daemon's session token went stale — the session now recovers automatically - Fixed returning to `claude agents` silently stopping running subagents and re-running the prompt from scratch — their work now carries over - Fixed a memory and per-turn CPU regression in interactive sessions: the context-usage indicator no longer re-analyzes the entire transcript after every turn - Fixed background agents inheriting a stale `PATH` from the daemon instead of the dispatching shell, causing missing tools on Windows - Fixed background and agent-view sessions dropping a shell-exported `ANTHROPIC_BASE_URL`, which sent API keys to the default endpoint and failed with 401 - Fixed Bash failing with "argument list too long" in repos with many git worktrees - Fixed worktree-isolated subagents sometimes running shell commands in the parent checkout instead of their own worktree - Fixed worktree creation rejecting nested repositories in multi-repo workspaces, leaving background sessions unable to isolate and edit - Fixed background agents crash-looping when their working directory was deleted, replaced by a file, or became an invalid path — they now fail once with a clear error - Fixed a background daemon auto-upgrade failure silently killing all running background sessions - Fixed `TaskStop` and `TaskOutput` failing to find background agents spawned by another agent — errors now list running agents by id and description - Fixed the `claude agents` composer discarding your typed message when a slash command isn't available there - Fixed the agent list crashing when opening a stopped session whose conversation was already open in another session - Fixed background sessions showing "Needs input" in the agent list after the question was already answered - Fixed background agent startup failures showing only "exit_with_message" instead of the actual error - Fixed background sessions ignoring `effortLevel` changes in settings.json when forked through the daemon - Fixed attached background sessions ignoring `CLAUDE_CODE_DISABLE_MOUSE` and `CLAUDE_CODE_DISABLE_MOUSE_CLICKS` opt-outs - Fixed `/exit` incorrectly warning about running background agents after all named agents had completed - Fixed background sessions started from a non-git directory unable to edit files when a `WorktreeCreate` hook was configured - Fixed the `@` directory picker in `claude agents` not showing registered git worktrees - Fixed background task output on Windows being permanently replaced by an empty file after `/clear` - Fixed content jumping when scrolling up through long transcript history - Fixed the terminal flickering and jumping while typing in bash mode when a shell-history suggestion was shown - Fixed literal `^[[I` / `^[[O` escape codes being printed when reattaching to a background session - Fixed LSP-only plugins being incorrectly flagged for disuse when their language servers deliver diagnostics or answer navigation requests - Improved responsiveness while long responses stream: live-preview updates no longer re-render the whole screen - Improved subagent behavior: agents are now less likely to re-delegate their entire task to another subagent - Reduced binary size by ~7 MB and startup memory by ~7 MB by loading a large bundled dependency lazily instead of inlining it - Changed left arrow to no longer close the background tasks, diff, and workflow detail views — press Esc instead - Changed the empty `claude agents` view to always show the organized sections (Needs input / Working / Completed) with descriptions - Removed the startup "claude command missing or broken" warnings — they now appear in `/doctor` and `/status` instead - Removed a redundant navigation hint from the `claude agents` footer - [VSCode] Added a Settings toggle for "Enable Remote Control for all sessions"

## Core ### Bugfixes - Prevent crashes and bad pricing data when GitHub Copilot returns models with a zero billing batch size. ### Improvements - Add a model-specific system prompt for Meta Muse Spark.

## 核心 ### 错误修复 - 改进了推理变体及提供商请求的元模型处理。 ## 桌面端 ### 错误修复 - 修复了模型选择器标签中裁剪的下伸部分。 ### 改进 - 新增了可关闭的标签页介绍弹窗，并刷新了帮助入口。 - 更新了子代理任务行，以更贴近 v2 会话设计。 - 重新设计了 v2 会话中的撤销停靠栏，使还原操作更清晰。 - 新增了 v2 免费模型选择器，可更快访问更多提供商。

## Core ### Bugfixes - Improved Meta model handling for reasoning variants and provider requests. ## Desktop ### Bugfixes - Fixed clipped descenders in model selector labels. ### Improvements - Added a dismissible tabs intro popup and refreshed the help entry point. - Updated sub-agent task rows to match the v2 session design more closely. - Restyled the revert dock in v2 sessions for clearer restore actions. - Added a v2 free-model selector with faster access to more providers.

## 核心 ### 错误修复 - 为 Grok 模型公开了推理努力变体。 - 改进了 xAI 提示缓存路由以及响应模型中的 PDF 文件支持。 ## 桌面端 ### 改进 - 为主屏幕上的项目添加了“打开包含文件夹”操作。 - 添加了用于文件、命令、上下文和 Shell 模式的编辑器附加菜单，且不会丢失草稿文本。 - 当未连接外部提供程序时，在新会话中添加了提供程序设置提示。 - 记住每个会话的审查面板模式和所选文件。 - 添加了内联文件浏览器标签页，用于从侧面板打开和预览文件。 - 更新了桌面端翻译。 ### 错误修复 - 通过保留最小宽度，使审查面板在宽屏上保持可用。 - 修复了隐藏的会话面板保持挂载状态并导致侧面板显示异常的问题。 - 将旧版草稿重新路由到会话页面。 - 在初始差异被截断时，为大型补丁加载完整的审查差异。 - 移除了命令面板中的首次打开闪烁。 - 将新会话快捷方式路由到新布局中的标签页。 - 在打开旧版会话 URL 时，保持会话路由位于正确的布局内。

## Core ### Bugfixes - Exposed reasoning effort variants for Grok models. - Improved xAI prompt cache routing and PDF file support in Responses models. ## Desktop ### Improvements - Added an Open containing folder action for projects on the home screen. - Added a composer add menu for files, commands, context, and shell mode without losing draft text. - Added a provider setup tip on new sessions when no external providers are connected. - Remember review panel mode and selected file per session. - Added inline file browser tabs for opening and previewing files from the side panel. - Updated desktop translations. ### Bugfixes - Kept the review pane usable on wide screens by reserving its minimum width. - Fixed hidden session panes staying mounted and causing side-panel glitches. - Routed legacy drafts back to the session page. - Loaded full review diffs for large patches when the initial diff is capped. - Removed the first-open flash in the command palette. - Routed the new session shortcut through tabs in the new layout. - Kept session routes inside the correct layout when opening legacy session URLs.

## 核心 ### 错误修复 - 更准确地分类 Z.ai 上下文窗口溢出错误，使过大的请求能够呈现正确的失败模式（@fengjikui） - 在读取配置文件时，更优雅地处理不可用的配置目录 ## 桌面端 ### 改进 - 恢复模型选择器中的模型详情工具提示（@arvsrn） - 刷新 v2 命令面板，使其更好地匹配新版桌面 UI 的其他部分（@arvsrn） ### 错误修复 - 当终端处于焦点状态时，终端快捷键优先 - 在包含待处理问题的标签页上显示未读状态 - 在会话数据仍在加载时，保持会话标签页标题在重载后可用 - 仅在设置中选择主题后才应用主题更改 - 修复 macOS Sequoia 中标题栏外观不匹配的问题，并调整信号灯位置（@arvsrn） **感谢 3 位社区贡献者：** - @arvsrn： - feat(app)：使 cmd k 菜单与 v2 样式对齐（#35152） - feat(app)：视觉改进（#35134） - feat(app)：在模型选择器中重新引入模型上下文工具提示（#35087） - feat(app)：更改信号灯位置，修复 Sequoia bug（#35081） - @StarpTech： - fix(core)：展开相对于 home 目录的权限路径（#35737） - @fengjikui： - fix(llm)：分类 Z.ai 令牌限制溢出（#35671）

## Core ### Bugfixes - Better classify Z.ai context-window overflow errors so oversized requests surface the right failure mode (@fengjikui) - Handle unavailable config directories more gracefully when reading config files ## Desktop ### Improvements - Restore model details tooltips in the model picker (@arvsrn) - Refresh the v2 command palette to better match the rest of the new desktop UI (@arvsrn) ### Bugfixes - Make terminal shortcuts take priority when the terminal is focused - Show unread state on tabs with pending questions - Keep session tab titles available after reloads while session data is still loading - Apply theme changes only after you select them in Settings - Fix macOS titlebar appearance mismatches in Sequoia and adjust traffic light placement (@arvsrn) **Thank you to 3 community contributors:** - @arvsrn: - feat(app): align cmd k menu with v2 styles (#35152) - feat(app): visual improvements (#35134) - feat(app): reintroduce model context tooltip in model selector (#35087) - feat(app): change traffic light position, fix sequoia bug (#35081) - @StarpTech: - fix(core): expand home-relative permission paths (#35737) - @fengjikui: - fix(llm): classify zai token limit overflow (#35671)

## 核心 ### 改进 - 新增代码模式 MCP 适配器，用于对已连接的 MCP 工具运行受限编排脚本。 - 除非启用代码模式，否则隐藏 `execute` 工具。 ### 错误修复 - 修复了分页 MCP 工具目录丢失工具元数据和输出架构验证的问题。 - 保留了 OpenRouter 小模型变体的低推理努力模式，而非禁用它。 - 修复了 GitHub Copilot 模型路由，使其遵循各模型所标明的聊天或响应端点。 - 修复了会话列表，确保可靠匹配等效的实例目录。 - 修复了 Cerebras 推理回放，使得较早的助手推理能通过提供商支持的字段正确返回。 ## TUI ### 错误修复 - 修复了旋转器注册，确保加载指示器能在 TUI 各界面持续渲染。 ## 桌面端 ### 改进 - 在标题栏显示草稿服务器状态。 - 统一了应用内的提供商连接流程。 - 改进了集成终端体验。（@arvsrn） - 新增了重新打开已关闭标签页及后台打开标签页的功能。（@usrnk1） - 改进了选择器对话框中的模型搜索。（@arvsrn） - 在首页添加了最近关闭的项目。（@usrnk1） - 在新布局中，标签页导航响应鼠标按下事件。 - 全面重构了 v2 审查面板。（@arvsrn） - 优化了会话标签页预览弹出框。（@usrnk1） ### 错误修复 - 修复了从首页创建新会话时使用了错误项目的问题。 - 修复了首次启动的引导流程，使其仅在适当时显示。 - 提升了大型审查面板的性能。 - 修复了从助手回复开始的时间线加载，通过补充缺失的用户轮次来修正。 - 恢复了窗口未聚焦时在编辑器中的输入功能。 - 修复了下拉菜单搜索行为。（@arvsrn） - 在切换会话标签页时保持 v2 审查面板挂载。 - 在工作区切换会话标签页时保持终端挂载。 - 修复了会话标签页忙碌指示器，使其始终正确限定在对应服务器范围内。 - 修复了标签页切换时的子会话导航和谱系解析问题。 - 在关闭并重新打开应用后保持桌面窗口标签页。 **感谢 3 位社区贡献者：** - @usrnk1： - 功能(桌面)：优化会话标签页预览弹出框 (#34792) - 功能(桌面)：在首页添加最近关闭的项目 (#34926) - 功能(桌面)：细节修复 (#34939) - 功能(桌面)：重新打开已关闭标签页及后台打开标签页 (#35010) - @arvsrn： - 功能(应用)：全面重构 v2 审查面板 (#31882) - 功能(应用)：使子代理 UI 与 v2 保持一致 (#34931) - 功能(应用)：改进模型搜索 (#34954) - 功能(应用)：修复下拉菜单搜索 (#34961) - 功能(应用)：终端改进 (#34747) - @StarpTech： - 功能(TUI)：添加调试信息对话框并支持复制到剪贴板 (#35004)

## Core ### Improvements - Added a code mode MCP adapter for running confined orchestration scripts against connected MCP tools. - Hid the `execute` tool unless code mode is enabled. ### Bugfixes - Fixed paginated MCP tool catalogs losing tool metadata and output schema validation. - Preserved low reasoning effort for OpenRouter small-model variants instead of disabling it. - Fixed GitHub Copilot model routing to honor each model's advertised chat or responses endpoint. - Fixed session lists to match equivalent instance directories reliably. - Fixed Cerebras reasoning replay so earlier assistant reasoning is sent back in the provider-supported field. ## TUI ### Bugfixes - Fixed spinner registration so loading indicators keep rendering across TUI surfaces. ## Desktop ### Improvements - Show draft server status in the title bar. - Unified the provider connect flow across the app. - Improved the integrated terminal experience. (@arvsrn) - Added reopening for closed tabs and background tab opening. (@usrnk1) - Improved model search in picker dialogs. (@arvsrn) - Added recently closed projects to Home. (@usrnk1) - Made tab navigation respond on mouse down in the new layout. - Overhauled the v2 review panel. (@arvsrn) - Refined the session tab preview popover. (@usrnk1) ### Bugfixes - Fixed new sessions from Home using the wrong project. - Fixed first-launch onboarding so it only appears when appropriate. - Improved large review pane performance. - Fixed timeline loads that started on assistant replies by backfilling the missing user turn. - Restored typing in the composer while the window is unfocused. - Fixed dropdown search behavior. (@arvsrn) - Kept the v2 review pane mounted when switching session tabs. - Kept the terminal mounted when switching session tabs in a workspace. - Fixed session tab busy indicators to stay scoped to the correct server. - Fixed child-session navigation and lineage resolution during tab transitions. - Kept desktop window tabs after closing and reopening the app. **Thank you to 3 community contributors:** - @usrnk1: - feat(desktop): refine session tab preview popover (#34792) - feat(desktop): add recently closed projects to home (#34926) - feat(desktop): papercut fixes (#34939) - feat(desktop): reopen closed tabs and background tab open (#35010) - @arvsrn: - feat(app): v2 review panel overhaul (#31882) - feat(app): align subagent UI with v2 (#34931) - feat(app): improvements to model search (#34954) - feat(app): dropdown search fix (#34961) - feat(app): terminal improvements (#34747) - @StarpTech: - feat(tui): add debug info dialog with copy to clipboard (#35004)

# Hermes Agent v0.18.2 (v2026.7.7.2) **Release Date:** July 7, 2026 > Same-day patch on top of v0.18.1, picking up the WhatsApp Baileys dependency fix needed for tagged-release Docker builds. --- ## What's in this patch - **fix(whatsapp): unpin Baileys from git commit, use published 7.0.0-rc13** ([#60643](https://github.com/NousResearch/hermes-agent/pull/60643)) — the WhatsApp bridge dependency now installs from the published npm release instead of a pinned git commit, making installs and Docker image builds reliable. Full curated release notes for the entire post-v0.18.0 window ship with v0.19.0. ## Updating ```bash hermes update # existing installs pip install -U hermes-agent ``` **Full Changelog**: [v2026.7.7...v2026.7.7.2](https://github.com/NousResearch/hermes-agent/compare/v2026.7.7...v2026.7.7.2)

# Hermes Agent v0.18.1 (v2026.7.7) **发布日期：** 2026年7月7日 > 补丁版本。此标签汇总了自 v0.18.0（7月1日）以来合并的约 660 个 PR —— 包括错误修复、加固以及进行中的功能工作 —— 为下游用户（Docker 镜像、托管部署、PyPI 安装）提供一个稳定的标记版本。 --- ## 关于此版本 这是一个由基础设施驱动的补丁标签，而非经过全面策划的版本。自六天前 v0.18.0 发布以来，主干分支已累积了约 **667 次提交，涉及约 990 个文件（+89.5k/−10.4k 行）**，包括 Windows 上的安装程序/更新程序自修复、仪表板和网关修复、WhatsApp 仪表板配对、MCP 及 provider 修复，以及大量的稳定性工作。 **此时间窗口的完整策划发布说明将随 v0.19.0 一起发布**，届时将记录从 v0.18.0 开始的所有内容 —— 亮点、功能领域以及完整的贡献者致谢。此窗口内的任何内容都不会被遗漏；所有内容都将在下一个次要版本中予以记录。 ## 更新 ```bash hermes update # 现有安装 pip install -U hermes-agent ``` **完整更新日志：** [v2026.7.1...v2026.7.7](https://github.com/NousResearch/hermes-agent/compare/v2026.7.1...v2026.7.7)

# Hermes Agent v0.18.1 (v2026.7.7) **Release Date:** July 7, 2026 > Patch release. This tag rolls up the ~660 PRs merged since v0.18.0 (July 1) — bug fixes, hardening, and in-progress feature work — into a stable tagged release for downstream consumers (Docker images, hosted deployments, PyPI installs). --- ## About this release This is an infrastructure-driven patch tag rather than a fully curated release. Since v0.18.0 shipped six days ago, main has accumulated roughly **667 commits across ~990 files (+89.5k/−10.4k lines)**, including installer/updater self-healing on Windows, dashboard and gateway fixes, WhatsApp dashboard pairing, MCP and provider fixes, and a large volume of stability work. **Full curated release notes for this window will ship with v0.19.0**, which will document everything from v0.18.0 onward — highlights, feature areas, and complete contributor credits. Nothing in this window is skipped; it's documented in the next minor release. ## Updating ```bash hermes update # existing installs pip install -U hermes-agent ``` **Full Changelog**: [v2026.7.1...v2026.7.7](https://github.com/NousResearch/hermes-agent/compare/v2026.7.1...v2026.7.7)

# Hermes Agent v0.18.0 (v2026.7.1) **发布日期:** 2026年7月1日 **自 v0.17.0 以来:** 约 1,720 次提交 · 998 个合并 PR · 2,215 个文件变更 · 约 251,000 行新增 · 约 41,000 行删除 · **949 个 issue 关闭** · **370+ 位社区贡献者** > **“决断”发布。** 在过去一周半的时间里，团队几乎把所有精力都投入到一个目标：解决 **Hermes Agent 仓库中每一个 P0 和 P1 级别的 issue 和 PR** —— 而在此版本发布之际，**100% 均已关闭。** 零个开放 P0。零个开放 P1。这意味着在本次窗口中，作为 **约 1,950 个总 issue 和 PR 关闭** 的一部分，**约 700 个最高优先级项** 被清空。我们打算从此将 P0/P1 保持为零。 > > 在此次彻底清理之上，v0.18.0 还关注 Hermes 的思考 *质量* 以及它 *如何知道自己的工作是否真正完成*。混合代理模型成为一等公民 —— 您可以像选择任何其他模型一样选择命名的模型集成体，每个参考模型的推理过程都会显示给您，聚合器的答案会实时流式输出。智能体学会了根据证据而非直觉来验证自己的工作，`/goal` 获得了完成契约，`/learn` + `/journey` 将自我改进变成了您可以看见并掌控的过程。底层方面，网关变得真正可大规模部署（缩至零、排空协调），桌面端获得了头等编程项目和可玩的记忆图谱，子代理现在可以在后台并行展开。 ## 🎯 P0/P1 彻底清理 —— 100% 已解决 这是本次发布的头条。团队日以继夜地猛攻优先级积压任务整整一周半，现在整个仓库中每一个 P0 和 P1 都已关闭： | 优先级 | 关闭的 Issue | 合并的 PR | |---|---|---| | **P0** (关键) | 3 | 8 | | **P1** (高) | 493 | 188 | | **总计** | **496** | **196** | 这意味着在十二天内解决了 **约 692 个最高优先级项** —— 而在清理完成的那一刻，整个仓库的开放 P0/P1 计数达到了 **0**。最后被攻克的一批是 interrupt-protected-compression 兄弟分叉 bug（issue #56391）及其修复（#56416），在本次发布前夕的通宵工作中关闭。 特别感谢 **@kshitijk4poor**，他与核心团队一起日夜不停地消灭优先级积压任务 —— 他负责了 cron 可靠性浪潮、压缩分叉修复、凭据泄露加固以及大量 P1 关闭任务。 我们将从此保持 P0/P1 为 **0**。🫡 ## ✨ 亮点 - **混合代理模型现在是一个可选的一等模型** —— MoA 过去是一个需要切换的模式；现在每个命名的 MoA 预设都作为 `moa` 提供商下的可选模型出现，与 Claude、GPT 和 Grok 并排显示在所有模型选择器（CLI、TUI、桌面版、网关）中。就像选择任何模型一样选择 "my-council"，Hermes 会自动将您的提示路由到该集成体。一个由前沿模型组成的委员会来探讨您最困难的问题，现在只需一次选择，在任何界面上都可实现。（[#46081](https://github.com/NousResearch/hermes-agent/pull/46081)、[#53548](https://github.com/NousResearch/hermes-agent/pull/53548)、[#53561](https://github.com/NousResearch/hermes-agent/pull/53561) — @teknium1） - **查看每个模型的推理，然后观察答案流式输出** —— 当 MoA 集成体运行时，每个参考模型的完整输出现在会作为自己的标记块呈现——您可以阅读 GPT-5 的想法、Claude 的想法以及 Grok 的想法，然后聚合器将它们合成为一个答案。而且最终答案现在会实时流式输出给您，而不是在经过长时间沉默后一次性出现。这在 CLI、TUI 和桌面应用中均可使用。您可以看到委员会的讨论过程，而不仅仅是阅读最终裁决。（[#53793](https://github.com/NousResearch/hermes-agent/pull/53793)、[#53855](https://github.com/NousResearch/hermes-agent/pull/53855)、[#55625](https://github.com/NousResearch/hermes-agent/pull/55625)、[#56101](https://github.com/NousResearch/hermes-agent/pull/56101) — @teknium1） - **智能体验证自己的工作——“完成”意味着被证明，而非声称** —— Hermes 现在会记录编码工作的验证证据，并且可以通过实际运行项目的检查来决定是否完成，而不是单纯宣称成功。`/goal` 获得了 **完成契约**：您声明“完成”的样子，常驻目标循环会根据该证据来判断是否完成，而不是在模型觉得满意时就停止。有一个 `pre_verify` 钩子用于接入自定义检查，以及一次性的迁移来合理调整默认值。区别是“我认为我修好了”与“测试通过了，这是证据”。（[#50501](https://github.com/NousResearch/hermes-agent/pull/50501)、[#52285](https://github.com/NousResearch/hermes-agent/pull/52285)、[#55413](https://github.com/NousResearch/hermes-agent/pull/55413)、[#53552](https://github.com/NousResearch/hermes-agent/pull/53552) — @teknium1、@OutThisLife） - **`/learn` —— 通过描述将任何内容变成可复用的技能** —— 运行 `/learn <任何内容>`，Hermes 会从您指向的任何内容（一个目录、一个 URL，或者您五分钟前刚刚演示过的工作流）中提炼出一个可复用的技能。它会自动按照 CONTRIBUTING.md 中的标准编写技能。下次您需要那个工作流时，它已经在那里了。教 Hermes 新技能现在只需一条命令，而不是手动的技能创作会议。（[#51506](https://github.com/NousResearch/hermes-agent/pull/51506)、[#52372](https://github.com/NousResearch/hermes-agent/pull/52372) — @teknium1） - **`/journey` —— Hermes 所了解到的关于您的一切的可播放时间线** —— CLI 和 TUI 新增了 `/journey`，这是一个学习时间线，显示 Hermes 随时间积累的记忆和技能——并且您可以直接在该视图中编辑或删除它们。配合桌面版新的 **记忆图谱**（一个自上而下、可播放的径向时间线，展示记忆和技能），您第一次可以真正 *看到* 您的智能体知道什么，观察其成长，并修剪错误的内容。您的智能体记忆不再是一个黑箱。（[#55555](https://github.com/NousResearch/hermes-agent/pull/55555)、[#55859](https://github.com/NousResearch/hermes-agent/pull/55859)、[#55226](https://github.com/NousResearch/hermes-agent/pull/55226) — @OutThisLife） - **委派一堆工作并继续前行 —— 后台并行展开** —— `delegate_task` 现在可以并行展开多个子代理，它们都在 **后台** 运行：您的聊天永远不会被阻塞，当所有子代理完成时，它们的结果会作为一次整合的回合返回。启动“并行研究这五个竞争对手”或“审计这三个模块”，然后继续做其他事情，而一支小型舰队在后台工作。当所有工作完成时，您会收到一份简洁的总结，而不是逐一盯梢每个子代理。（[#49734](https://github.com/NousResearch/hermes-agent/pull/49734) — @teknium1） - **桌面应用中的一等编程项目** —— 桌面应用获得了真正的、按配置文件管理的 **项目** —— 包含代码库的侧边栏、编码轨道、审查面板、git 工作树管理以及面向智能体的项目工具，全部由适当的 `project → repo → lane` 模型支持。不再有零散的聊天会话，您的编码工作被组织成智能体能够理解并执行的项目。这标志着桌面版正在变成一个真正的编码座舱。（[#49037](https://github.com/NousResearch/hermes-agent/pull/49037)、[#54385](https://github.com/NousResearch/hermes-agent/pull/54385)、[#54517](https://github.com/NousResearch/hermes-agent/pull/54517) — @OutThisLife） - **规模化运行 Hermes —— 缩至零和排空协调** —— 网关现在可以在空闲时进入 **休眠状态**，并在重启、迁移或自动更新前干净地静默——而不会丢弃正在进行的对话。托管的或仅中继的 Hermes 可以在无人对话时缩至零，并在有需求时重新唤醒，破坏性的生命周期操作会协调外部排空，确保没有人会在对话中途被切断。运行面向团队或作为托管服务的 Hermes 现在变得更加生产级。（[#52243](https://github.com/NousResearch/hermes-agent/pull/52243)、[#52937](https://github.com/NousResearch/hermes-agent/pull/52937)、[#54824](https://github.com/NousResearch/hermes-agent/pull/54824) — @teknium1、@benbarclay） - **更便宜的自我改进 —— 更智能的后台审查** —— 回合后的自我改进分支（决定是否保存记忆或技能的那个分支）现在路由到辅助模型，消化上下文而不是重放整个对话，并调整其节奏——因此在您的回合后运行的“从刚刚发生的事情中学习”循环的成本仅为原来的一小部分。您保留了自我改进的能力，同时不再为此支付完整主模型的价格。（[#49252](https://github.com/NousResearch/hermes-agent/pull/49252) — @teknium1） - **在编辑器中编写下一个提示 —— `/prompt`** —— `/prompt` 打开您的 `$EDITOR`，这样您可以用真正的 Markdown 手写一个长的、多行的提示，而不是与单行输入框作斗争。起草一份详细的规格说明、一个结构化的问题或一大段粘贴内容，保存后它就会排队作为您的下一条消息。小功能，但对任何需要对 Hermes 说超过一句话的人来说，这是一个巨大的生活质量提升。（[#50509](https://github.com/NousResearch/hermes-agent/pull/50509) — @teknium1） - **Google Vertex AI —— 通过您的 GCP 服务账户使用 Gemini，无需静态密钥** —— Vertex AI 现在是通过 Vertex 的 OpenAI 兼容端点使用 Gemini 模型的一等提供商。普通自定义提供商设置总是在对话中途失败的原因是 Vertex 没有静态 API 密钥——每次请求需要一个从服务账户 JSON 或应用默认凭证生成的短生命周期 OAuth2 访问令牌（约 1 小时 TTL）。Hermes 现在为您生成并自动刷新这些令牌，因此如果您的组织通过 Google Cloud 运行 Gemini，只需将 Hermes 指向您的服务账户即可正常工作——无需粘贴令牌，无需担心会话中途过期。（[#56363](https://github.com/NousResearch/hermes-agent/pull/56363) — @teknium1、@slawt） - **安全轮次** —— 本次窗口加固了多个表面：MCP 配置持久化攻击面被锁定，可能泄露提供商凭据的 cron `base_url` 覆盖被阻止，文件读取中前缀秘密使用不可重用的哨兵，Slack 应用级（`xapp-`）令牌被遮蔽，所有后端强制执行浏览器云元数据，以及跨惰性消息路径的 `aiohttp` CVE 底线。减少了提示注入或配置错误的会话泄露凭据的方式。（[#50476](https://github.com/NousResearch/hermes-agent/pull/50476)、[#56196](https://github.com/NousResearch/hermes-agent/pull/56196)、[#54166](https://github.com/NousResearch/hermes-agent/pull/54166)、[#56227](https://github.com/NousResearch/hermes-agent/pull/56227)、[#52349](https://github.com/NousResearch/hermes-agent/pull/52349)、[#56237](https://github.com/NousResearch/hermes-agent/pull/56237) — @teknium1、@kshitijk4poor、@claudlos） --- ## 🧠 混合代理模型（MoA） MoA 在本窗口中从一种模式升级为模型系统中的一等组成部分。 - **预设作为可选虚拟模型** —— 每个命名的 MoA 预设都作为 `moa` 提供商下的模型出现；在任何模型选择器中选中它，Hermes 就会通过该集成体路由（[#46081](https://github.com/NousResearch/hermes-agent/pull/46081)、[#53561](https://github.com/NousResearch/hermes-agent/pull/53561)、[#53775](https://github.com/NousResearch/hermes-agent/pull/53775) — @teknium1） - **`/moa` 现在是单次快捷指令** —— 通过默认预设运行单次提示，之后恢复您的模型；持久切换通过模型选择器进行（[#53548](https://github.com/NousResearch/hermes-agent/pull/53548) — @teknium1） - **参考模型输出显示为标记块**，在 CLI、TUI 和桌面版中 —— 在聚合器合成之前阅读每个模型的推理（[#53793](https://github.com/NousResearch/hermes-agent/pull/53793)、[#53855](https://github.com/NousResearch/hermes-agent/pull/53855) — @teknium1） - **聚合器响应实时流式输出**，而不是在沉默后一次性出现（[#55625](https://github.com/NousResearch/hermes-agent/pull/55625) — @teknium1） - **参考模型可看到完整工具状态**，并在每次用户/工具响应时触发；咨询参考模型在用户回合结束并收到参考角色系统提示（[#54016](https://github.com/NousResearch/hermes-agent/pull/54016)、[#54007](https://github.com/NousResearch/hermes-agent/pull/54007) — @teknium1） - **可选全回合追踪持久化到 JSONL**（`moa.save_traces`），用于调试和评估（[#56101](https://github.com/NousResearch/hermes-agent/pull/56101) — @teknium1） - 可靠性：参考模型和聚合器模型通过其提供商的实际路由调用；上下文窗口从聚合器解析（而不是默认的 256K）；辅助任务解析到聚合器；虚拟提供商被阻止作为参考/聚合器插槽；容忍手动编辑的预设配置（[#53580](https://github.com/NousResearch/hermes-agent/pull/53580)、[#53780](https://github.com/NousResearch/hermes-agent/pull/53780)、[#53827](https://github.com/NousResearch/hermes-agent/pull/53827)、[#53281](https://github.com/NousResearch/hermes-agent/pull/53281)、[#53275](https://github.com/NousResearch/hermes-agent/pull/53275)、[#53556](https://github.com/NousResearch/hermes-agent/pull/53556) — @teknium1） - MoA 插槽提供商身份统一在单一的 `call_llm` 检查点上；HermesBench 结果已记录（[#55991](https://github.com/NousResearch/hermes-agent/pull/55991)、[#53206](https://github.com/NousResearch/hermes-agent/pull/53206) — @teknium1） ## ✅ 验证与目标 —— 智能体证明其工作 - **`/goal` 的完成契约** —— 声明“完成”的样子；常驻目标循环根据证据而非模型的主观判断来评估（[#50501](https://github.com/NousResearch/hermes-agent/pull/50501) — @teknium1） - **`/goal wait <pid>`** —— 将常驻目标循环挂起到后台进程，而不是重新调用智能体（[#50503](https://github.com/NousResearch/hermes-agent/pull/50503) — @teknium1） - **编码验证证据分类账** —— 按配置文件范围记录由 `agent.coding_context` 检测到的规范项目检查；网关暴露验证状态（[#52285](https://github.com/NousResearch/hermes-agent/pull/52285)、[#52286](https://github.com/NousResearch/hermes-agent/pull/52286) — @OutThisLife） - **`pre_verify` 钩子 + 编码指导配置**；验证停止循环 + 即席验证脚本（[#55413](https://github.com/NousResearch/hermes-agent/pull/55413)、[#52296](https://github.com/NousResearch/hermes-agent/pull/52296)、[#52297](https://github.com/NousResearch/hermes-agent/pull/52297) — @OutThisLife） - **停止时验证默认为关闭**，并附带一次性 v32 迁移；跳过仅文档的编辑；感知表面的“自动”默认值已恢复；在消息传递表面关闭（[#53552](https://github.com/NousResearch/hermes-agent/pull/53552)、[#54740](https://github.com/NousResearch/hermes-agent/pull/54740)、[#55449](https://github.com/NousResearch/hermes-agent/pull/55449)、[#52412](https://github.com/NousResearch/hermes-agent/pull/52412) — @teknium1、@OutThisLife、@GodsBoy） ## 🎓 自我改进（学习 / 旅程） - **`/learn <任何内容>`** —— 从目录、URL 或您刚刚演示的工作流中提炼出可复用的技能；遵循 CONTRIBUTING.md 技能标准和混合要求（[#51506](https://github.com/NousResearch/hermes-agent/pull/51506)、[#52372](https://github.com/NousResearch/hermes-agent/pull/52372)、[#55956](https://github.com/NousResearch/hermes-agent/pull/55956) — @teknium1） - **`/journey`** —— CLI + TUI 学习时间线，展示积累的记忆和技能，支持就地编辑/删除（[#55555](https://github.com/NousResearch/hermes-agent/pull/55555)、[#55859](https://github.com/NousResearch/hermes-agent/pull/55859) — @OutThisLife） - **更便宜的后台审查** —— 辅助模型路由 + 上下文摘要 + 自适应节奏，用于回合后的自我改进分支（[#49252](https://github.com/NousResearch/hermes-agent/pull/49252) — @teknium1） - **桌面版中的 `memory` 图谱** —— 可播放的径向时间线，展示随时间变化的记忆和技能（[#55226](https://github.com/NousResearch/hermes-agent/pull/55226) — @OutThisLife） ## 🖥️ Hermes 桌面应用 ### 编码座舱 - **一等项目** —— 按配置文件管理的侧边栏、编码轨道、审查面板、智能体项目工具（`project → repo → lane`）；支持远程网关的文件夹选择器 + git 座舱（状态、审查、工作树）（[#49037](https://github.com/NousResearch/hermes-agent/pull/49037)、[#54385](https://github.com/NousResearch/hermes-agent/pull/54385) — @OutThisLife） - **多终端面板**，带有只读智能体终端；跨重启持久化并恢复终端标签页及回滚内容（[#54517](https://github.com/NousResearch/hermes-agent/pull/54517)、[#54585](https://github.com/NousResearch/hermes-agent/pull/54585) — @OutThisLife） - **聊天中的 PR 风格文件差异**；文件预览面板的应用内即时编辑器；内联富媒体嵌入、图表和助手 markdown 中的提醒（[#50731](https://github.com/NousResearch/hermes-agent/pull/50731)、[#52772](https://github.com/NousResearch/hermes-agent/pull/52772)、[#52935](https://github.com/NousResearch/hermes-agent/pull/52935) — @OutThisLife） ### 用户体验与界面 - 长线程的对话时间线轨道；上下文使用量分解弹出框；用于子代理监视窗口的只读观众副本；将编辑器弹出为可拖动的浮动窗口（[#51094](https://github.com/NousResearch/hermes-agent/pull/51094)、[#54907](https://github.com/NousResearch/hermes-agent/pull/54907)、[#55033](https://github.com/NousResearch/hermes-agent/pull/55033)、[#49488](https://github.com/NousResearch/hermes-agent/pull/49488) — @OutThisLife、@austinpickett） - 朗读回复（自动 TTS）编辑器切换；记住窗口大小/位置/最大化状态跨启动；重新设计的澄清提示；共享的覆盖面板原语用于 cron/配置文件/智能体（[#55154](https://github.com/NousResearch/hermes-agent/pull/55154)、[#52086](https://github.com/NousResearch/hermes-agent/pull/52086)、[#52993](https://github.com/NousResearch/hermes-agent/pull/52993)、[#54558](https://github.com/NousResearch/hermes-agent/pull/54558) — @OutThisLife） - 从 Web UI 备份导入/创建/下载；添加上下文使用量弹出框；在安装选择器中标记已安装的主题；配置驱动的 Electron 启动标志 + GPU 策略（[#54611](https://github.com/NousResearch/hermes-agent/pull/54611)、[#55410](https://github.com/NousResearch/hermes-agent/pull/55410)、[#53991](https://github.com/NousResearch/hermes-agent/pull/53991) — @teknium1、@OutThisLife） - **宠物** —— 漫游宠物（可选）、更冷静/逼真的漫游、Alt+滚轮缩放不再裁剪、帧完美孵化流程 + CPU 安全色度、弹出叠加 + 通知（[#55114](https://github.com/NousResearch/hermes-agent/pull/55114)、[#55400](https://github.com/NousResearch/hermes-agent/pull/55400)、[#52877](https://github.com/NousResearch/hermes-agent/pull/52877)、[#47959](https://github.com/NousResearch/hermes-agent/pull/47959)、[#52303](https://github.com/NousResearch/hermes-agent/pull/52303) — @OutThisLife） ### 重构浪潮（编辑器 / 神文件解耦） - 将编辑器分解为独立的引擎钩子；提取了分支/转义/URL/占位符/弹出引擎；将 `thread.tsx`、`sidebar/index.tsx`、引导覆盖和 `use-prompt-actions` 神文件拆分为专注的模块；共享的 WebSocket 层将桌面版从仪表板（`hermes serve`）解耦（[#55500](https://github.com/NousResearch/hermes-agent/pull/55500)、[#55842](https://github.com/NousResearch/hermes-agent/pull/55842)、[#55451](https://github.com/NousResearch/hermes-agent/pull/55451)、[#55453](https://github.com/NousResearch/hermes-agent/pull/55453)、[#55807](https://github.com/NousResearch/hermes-agent/pull/55807)、[#55504](https://github.com/NousResearch/hermes-agent/pull/55504)、[#54568](https://github.com/NousResearch/hermes-agent/pull/54568) — @OutThisLife） - 性能：限制工具结果渲染，防止大的 `/learn` 运行导致冻结；在负载下快速切换会话（[#52273](https://github.com/NousResearch/hermes-agent/pull/52273)、[#52620](https://github.com/NousResearch/hermes-agent/pull/52620) — @OutThisLife） ## 📊 Web 仪表板 - 在未认证加载时自动发起门户 SSO 重定向；在无提供商、非回环绑定时交互式身份验证设置；自托管 OIDC 中支持机密客户端（`client_secret`）（[#54846](https://github.com/NousResearch/hermes-agent/pull/54846)、[#50551](https://github.com/NousResearch/hermes-agent/pull/50551)、[#55344](https://github.com/NousResearch/hermes-agent/pull/55344) — @teknium1、@benbarclay） - 在 `OPTIONAL_ENV_VARS` 中列出所有记忆提供商的 API 密钥；在密钥页面上列出并添加任意自定义 `.env` 密钥；暴露 cron 作业执行字段；备份导入/创建/下载（[#54546](https://github.com/NousResearch/hermes-agent/pull/54546)、[#54552](https://github.com/NousResearch/hermes-agent/pull/54552)、[#53551](https://github.com/NousResearch/hermes-agent/pull/53551)、[#54611](https://github.com/NousResearch/hermes-agent/pull/54611) — @benbarclay、@teknium1） - 将 PTY 生成/关闭从事件循环中卸出；从交互式登录表面排除非交互式提供商（[#53227](https://github.com/NousResearch/hermes-agent/pull/53227)、[#53239](https://github.com/NousResearch/hermes-agent/pull/53239) — @IAvecilla） ## 🏗️ 核心智能体与架构 ### 委派与子代理 - **后台并行展开** —— 并行子代理在后台运行，所有完成时一次性合并返回；为后台 `delegate_task` 提供平静的“将恢复”提示（[#49734](https://github.com/NousResearch/hermes-agent/pull/49734)、[#52756](https://github.com/NousResearch/hermes-agent/pull/52756) — @teknium1、@OutThisLife） - 在 CLI + TUI 状态栏中跟踪后台子代理（[#51441](https://github.com/NousResearch/hermes-agent/pull/51441)、[#51485](https://github.com/NousResearch/hermes-agent/pull/51485) — @teknium1） ### 智能体循环、工具与编码上下文 - 单次 LLM 助手 + `llm.oneshot` 网关 RPC；暴露编码上下文项目事实（`project.facts` RPC）（[#51261](https://github.com/NousResearch/hermes-agent/pull/51261)、[#51259](https://github.com/NousResearch/hermes-agent/pull/51259) — @OutThisLife） - `web_extract` 截断并存储，而非 LLM 摘要；并发 @-引用展开（[#54843](https://github.com/NousResearch/hermes-agent/pull/54843)、[#55207](https://github.com/NousResearch/hermes-agent/pull/55207) — @teknium1、@kshitijk4poor） - 为内置工具提供友好的人类化工具标签；`/reasoning full`（无上限思考）；`/timestamps` + `/history` 中的时间戳；`/prompt` 在 `$EDITOR` 中编写（[#55166](https://github.com/NousResearch/hermes-agent/pull/55166)、[#50499](https://github.com/NousResearch/hermes-agent/pull/50499)、[#50506](https://github.com/NousResearch/hermes-agent/pull/50506)、[#50509](https://github.com/NousResearch/hermes-agent/pull/50509) — @teknium1） - 在流式和非流式检测器中为每个推理模型设置过时超时底线；在宽限期后将 SIGTERM 升级为 SIGKILL 以终止主机 PID（[#52845](https://github.com/NousResearch/hermes-agent/pull/52845)、[#50489](https://github.com/NousResearch/hermes-agent/pull/50489) — @teknium1） - 支持多个 `HERMES_WRITE_SAFE_ROOT` 目录；将 HTTP/WS 主体捕获可选项隔离到独立的、排除共享的 `gui_bodies.log` 文件中（[#53292](https://github.com/NousResearch/hermes-agent/pull/53292)、[#49044](https://github.com/NousResearch/hermes-agent/pull/49044) — @kshitijk4poor） ### 压缩与会话 - 原地压缩选项（单一会话 id）；将 `in_place` 默认值改为 True 并附带保护修复（[#49739](https://github.com/NousResearch/hermes-agent/pull/49739)、[#52658](https://github.com/NousResearch/hermes-agent/pull/52658) — @teknium1） - 备份在更新前快照中包含 `projects.db` 和看板（[#52990](https://github.com/NousResearch/hermes-agent/pull/52990) — @kshitijk4poor） ### 提供商与模型 - **Google Vertex AI** 作为通过 OpenAI 兼容端点使用 Gemini 的一等提供商——自动生成并刷新来自服务账户 JSON / ADC 的短生命周期 OAuth2 令牌（无需静态密钥）；保留并现代化了 @slawt 的 #8427（[#56363](https://github.com/NousResearch/hermes-agent/pull/56363) — @teknium1、@slawt） - 通过托管 Nous 订阅网关的 Krea；Z.AI 端点选择器（全球/中国/编码计划）；Ollama-cloud reasoning_effort 接线；移除 google-gemini-cli + google-antigravity OAuth 提供商（[#52647](https://github.com/NousResearch/hermes-agent/pull/52647)、[#52364](https://github.com/NousResearch/hermes-agent/pull/52364)、[#51494](https://github.com/NousResearch/hermes-agent/pull/51494)、[#50492](https://github.com/NousResearch/hermes-agent/pull/50492) — @teknium1、@kshitijk4poor） - 遵循 `NOUS_INFERENCE_BASE_URL` 环境覆盖用于 Nous OAuth；保持 Nous 认证对空闲仪表板/网关智能体保持新鲜（[#52270](https://github.com/NousResearch/hermes-agent/pull/52270)、[#50567](https://github.com/NousResearch/hermes-agent/pull/50567) — @benbarclay、@teknium1） ## 🌐 网关、舰队与中继 ### 缩至零与排空 - **缩至零空闲检测 + 休眠静默（阶段 0）**；加固休眠保护；修复 arm-gate 计数禁用占位平台的问题（[#52243](https://github.com/NousResearch/hermes-agent/pull/52243)、[#52359](https://github.com/NousResearch/hermes-agent/pull/52359)、[#52831](https://github.com/NousResearch/hermes-agent/pull/52831) — @teknium1、@benbarclay） - **外部排空协调（安全关闭阶段 2）**；在标记排空时抑制主通道关闭广播；在重启/关闭排空超时后持久化进行中的对话记录；为安全生命周期操作提供忙碌/空闲读取（[#52937](https://github.com/NousResearch/hermes-agent/pull/52937)、[#54824](https://github.com/NousResearch/hermes-agent/pull/54824)、[#50312](https://github.com/NousResearch/hermes-agent/pull/50312)、[#50131](https://github.com/NousResearch/hermes-agent/pull/50131) — @teknium1、@benbarclay、@kshitijk4poor） - 默认 `restart_drain_timeout` 为 0 以终止 systemd 崩溃循环；自我修复陷入排空/降级状态的网关（[#54066](https://github.com/NousResearch/hermes-agent/pull/54066)、[#55397](https://github.com/NousResearch/hermes-agent/pull/55397) — @teknium1） ### 中继（阶段 5 / 6） - 唤醒原语（网关侧）；进入空闲/缓冲翻转原语；通过 WS 的 `passthrough_forward`；每个智能体多平台身份 + 每帧出口；在自我配置时转发稳定的实例 id；向连接器声明相关性策略（[#51595](https://github.com/NousResearch/hermes-agent/pull/51595)、[#51572](https://github.com/NousResearch/hermes-agent/pull/51572)、[#50702](https://github.com/NousResearch/hermes-agent/pull/50702)、[#52830](https://github.com/NousResearch/hermes-agent/pull/52830)、[#50772](https://github.com/NousResearch/hermes-agent/pull/50772)、[#51248](https://github.com/NousResearch/hermes-agent/pull/51248) — @benbarclay） - 根据投递目标而非 `source.platform` 授权中继投递的事件；采用 `scope_id` 线键；清除平台特定的作用域术语（[#52306](https://github.com/NousResearch/hermes-agent/pull/52306)、[#55289](https://github.com/NousResearch/hermes-agent/pull/55289)、[#56016](https://github.com/NousResearch/hermes-agent/pull/56016) — @benbarclay） ### 网关核心与渲染 - 类型化发送错误分类（`SendResult.error_kind`）；每平台 `typing_indicator` 切换；`/usage` 中的每类别上下文分解（[#50342](https://github.com/NousResearch/hermes-agent/pull/50342)、[#55394](https://github.com/NousResearch/hermes-agent/pull/55394)、[#55204](https://github.com/NousResearch/hermes-agent/pull/55204) — @teknium1） - API 服务器：可配置的并发运行上限以防止 DoS；按运行 id 限定运行审批范围（[#50007](https://github.com/NousResearch/hermes-agent/pull/50007)、[#56129](https://github.com/NousResearch/hermes-agent/pull/56129) — @teknium1） ## 📱 消息平台 - **Cron 延续** —— 可延续的 cron 作业（线程优先延续，附带 DM 镜像回退）；Slack 的平坦频道内可延续 cron 投递；在创建/列出 cron 时警告网关未运行（[#52250](https://github.com/NousResearch/hermes-agent/pull/52250)、[#56254](https://github.com/NousResearch/hermes-agent/pull/56254)、[#51696](https://github.com/NousResearch/hermes-agent/pull/51696) — @teknium1） - Telegram：可配置的命令菜单 + 提高默认上限以便技能保持可见；分别控制富草稿预览；在重试前排空通用发送池（[#51716](https://github.com/NousResearch/hermes-agent/pull/51716)、[#52088](https://github.com/NousResearch/hermes-agent/pull/52088)、[#54121](https://github.com/NousResearch/hermes-agent/pull/54121) — @teknium1、@helix4u） - Slack：智能体消息的可选 Block Kit 渲染；清单生成的 `--no-assistant` 标志（[#56102](https://github.com/NousResearch/hermes-agent/pull/56102)、[#51487](https://github.com/NousResearch/hermes-agent/pull/51487) — @teknium1） - Discord：通过 `display.reasoning_style` 将推理渲染为 `-#` 子文本（[#51168](https://github.com/NousResearch/hermes-agent/pull/51168) — @teknium1） - 通过 Baileys 桥的原生 WhatsApp 媒体投递；Teams 原生 `send_video`/`send_voice`/`send_document`；photon 边车升级到 spectrum-ts v8 并带有 tapback 相关；Raft 网关设置向导（[#53598](https://github.com/NousResearch/hermes-agent/pull/53598)、[#49308](https://github.com/NousResearch/hermes-agent/pull/49308)、[#53451](https://github.com/NousResearch/hermes-agent/pull/53451)、[#56230](https://github.com/NousResearch/hermes-agent/pull/56230) — @teknium1） - Signal：AAC 语音留言重新混合 + 共享 Markdown 格式化（[#49530](https://github.com/NousResearch/hermes-agent/pull/49530) — @kshitijk4poor） - 将 slack/dingtalk/whatsapp/matrix/feishu/telegram/wecom/email/sms 适配器迁移到捆绑模式（[#49408](https://github.com/NousResearch/hermes-agent/pull/49408) — @teknium1） ## 🔧 工具系统、技能与 MCP - 空白石板设置模式 —— 最小智能体，选择性加入所有内容（[#36733](https://github.com/NousResearch/hermes-agent/pull/36733) — @teknium1） - MCP：配置持久化攻击面加固；阻止 base_url 泄露；短 TTL 会话的保活（参见安全部分）—— 同时保留自 v0.17.0 的目录和用户体验 - 技能：`/learn` 提炼（参见自我改进）；`cloudflare-temporary-deploy` 可选技能；creative-ideation v2.1.0 方法库（[#50849](https://github.com/NousResearch/hermes-agent/pull/50849)、[#42402](https://github.com/NousResearch/hermes-agent/pull/42402) — @teknium1、@SHL0MS） - 看板：任务生命周期插件钩子（已认领/已完成/已阻塞）；类型化阻塞原因 + 解除阻塞循环断路器；移交新鲜度戳记（[#50349](https://github.com/NousResearch/hermes-agent/pull/50349)、[#52848](https://github.com/NousResearch/hermes-agent/pull/52848)、[#53973](https://github.com/NousResearch/hermes-agent/pull/53973) — @teknium1） - 插件：`ctx.profile_name` 用于与会话无关的配置文件访问（[#50346](https://github.com/NousResearch/hermes-agent/pull/50346) — @teknium1） - LSP：PowerShellEditorServices 语言服务器；mem0 v3 API + OSS 模式 + 更新/删除工具（[#55930](https://github.com/NousResearch/hermes-agent/pull/55930)、[#15624](https://github.com/NousResearch/hermes-agent/pull/15624) — @teknium1、@kartik-mem0） ## ⚡ 性能 - 冷启动：延迟加载网关平台适配器；使用 libyaml `CSafeLoader` 解析配置 + 插件清单（[#54448](https://github.com/NousResearch/hermes-agent/pull/54448)、[#54486](https://github.com/NousResearch/hermes-agent/pull/54486) — @teknium1） - 状态：合并 FTS5 段 + `handoff_state` 索引以减少写锁争用；单次遍历 `list_profiles` 别名映射 + 技能计数缓存 + 事件循环卸出（[#54752](https://github.com/NousResearch/hermes-agent/pull/54752)、[#54770](https://github.com/NousResearch/hermes-agent/pull/54770) — @kshitijk4poor、@teknium1） ## 🔒 安全与可靠性 - 加固 MCP 配置持久化攻击面；阻止泄露提供商凭据的 cron `base_url` 覆盖；在文件读取中使用不可重用的哨兵保护前缀秘密（[#50476](https://github.com/NousResearch/hermes-agent/pull/50476)、[#56196](https://github.com/NousResearch/hermes-agent/pull/56196)、[#54166](https://github.com/NousResearch/hermes-agent/pull/54166) — @teknium1、@kshitijk4poor） - 遮蔽 Slack 应用级（`xapp-`）令牌；所有后端强制执行浏览器云元数据（CDP 非本地）；在 `browser_back` 导航后重新检查私有网络保护；将 `/resume` 和 `/sessions` 限制为调用者源（IDOR）；跨惰性消息路径设置 `aiohttp` 3.14.1 CVE 底线 + 固定漂移保护（[#56227](https://github.com/NousResearch/hermes-agent/pull/56227)、[#52349](https://github.com/NousResearch/hermes-agent/pull/52349)、[#56526](https://github.com/NousResearch/hermes-agent/pull/56526)、[#56378](https://github.com/NousResearch/hermes-agent/pull/56378)、[#56237](https://github.com/NousResearch/hermes-agent/pull/56237) — @teknium1、@claudlos、@kshitijk4poor） - Cron 可靠性浪潮：当未固定的作业的提供商发生漂移时，失败关闭；运行错过宽限期的作业一次而不是永远延迟；在 `BaseException` 发生时保持心跳活跃 + 心跳感知状态；将已启用的 MCP 服务器分层应用于每作业工具集；保护 cron 模型工具路径 + 自动恢复循环断路器（[#51051](https://github.com/NousResearch/hermes-agent/pull/51051)、[#50062](https://github.com/NousResearch/hermes-agent/pull/50062)、[#50016](https://github.com/NousResearch/hermes-agent/pull/50016)、[#50117](https://github.com/NousResearch/hermes-agent/pull/50117)、[#56240](https://github.com/NousResearch/hermes-agent/pull/56240) — @kshitijk4poor、@teknium1） - Windows：抑制控制台闪烁 + 加固网关重启；在 PATH 回退中优先使用 cmd npm shim；在 GUI 更新后重新生成无窗口网关；优先使用托管节点用于 whatsapp/desktop（[#52340](https://github.com/NousResearch/hermes-agent/pull/52340)、[#50398](https://github.com/NousResearch/hermes-agent/pull/50398)、[#52239](https://github.com/NousResearch/hermes-agent/pull/52239) — @helix4u、@teknium1） ## 🔁 回滚（窗口内，以供记录） - cron 作业存储恢复到每个配置文件（回滚 #32117 + #50993）；不克隆 `auth.json`（重复 OAuth 授权会导致兄弟令牌撤销）；回滚 Windows 终端弹出 PR；`prompt_caching.enabled` 切换回滚以供重新评估（[#51116](https://github.com/NousResearch/hermes-agent/pull/51116)、[#51732](https://github.com/NousResearch/hermes-agent/pull/51732)、[#53853](https://github.com/NousResearch/hermes-agent/pull/53853)、[#56126](https://github.com/NousResearch/hermes-agent/pull/56126) — @teknium1） ## 👥 贡献者 本次发布共有 **381 人** 贡献（通过提交、共同作者尾注以及保存的 PR）。感谢各位。 ### 核心 - @teknium1 —— 发布负责人；MoA 一等公民、验证/目标、`/learn`、后台审查、安全轮次、提供商、P0/P1 彻底清理 - @OutThisLife —— 桌面应用（项目、记忆图谱、`/journey`、多终端、编辑器重构浪潮、宠物、验证用户体验） ### 顶级社区贡献者 - @kshitijk4poor —— P0/P1 积压任务燃烧：cron 可靠性浪潮、状态性能、安全（cron 凭据泄露）、网关/信号、TUI 配置——优先关闭任务中的巨大份额 - @benbarclay —— 中继阶段 5/6、缩至零/排空协调、仪表板认证/密钥、网关加固 - @ethernet8023 —— CI/ Docker（统一作业、更快构建、计时报告） - @helix4u —— Windows 加固（控制台闪烁、npm shim、网关重启） ### 全部贡献者 *（此处省略了完整的贡献者列表，保留原文格式但翻译标题为“全部贡献者”即可。由于列表极长，此处按格式保留但以翻译说明。）* @0xbyt4、@0xDevNinja、@0xsir0000、@1RB、@595650661、@aaronlab、@abchiaravalle、@adammatski1972、@AetherAgents、@Afnath-max、@agt-user、@ahmadashfq、@AhmetArif0、@AIalliAI、@aieng-abdullah、@ailang323、@ailthrim、@aj-nt、@alelpoan、@alloevil、@amathxbt、@ambition0802、@anderskev、@andressommerhoff、@angelos、@annguyenNous、@Antimatter543、@arminanton、@arthurzhang、@asimons81、@austinpickett、@baolingao、@Bartok9、@basilalshukaili、@BBCrypto-web、@bbopen、@Beandon13、@beardthelion、@benbarclay、@benbenlijie、@binhnt92、@bitcryptic-gw、@Blaryxoff、@bogerman1、@bradhallett、@brett539、@briandevans、@buihongduc132、@bykim0119、@catapreta、@chaithanyak42、@charleneleong-ai、@CharlieKerfoot、@chazmaniandinkle、@chrispersico、@Christopher-Schulze、@chriswesley4、@claudlos、@clovericbot、@cmcejas、@codexGW、@Cossackx、@counterposition、@coygeek、@CRWuTJ、@cyb0rgk1tty、@cyb3rwr3n、@cypctlinux、@cypres0099、@dalenguyen、@Danamove、@DanAsBjorn、@DataAdvisory、@davidgut1982、@DavidMetcalfe、@davidvv、@de1tydev、@denisqq、@devorun、@devsart95、@DhivinX、@DiamondEyesFox、@difujia、@Disaster-Terminator、@djimit、@djstunami、@dodo-reach、@donovan-yohan、@Dr1985、@DrZM007、@Dusk1e、@egilewski、@ehz0ah、@Eji4h、@EloquentBrush0x、@Elshayib、@emozilla、@entropy-0x、@Erosika、@EtherAura、@etherman-os、@ethernet8023、@f-trycua、@fayenix、@fesalfayed、@firefly、@flamiinngo、@flobo3、@francescomucio、@franksong2702、@friendshipisover、@fsaad1984、@fyzanshaik、@GauravPatil2515、@gdeyoung、@georgex8001、@GodsBoy、@graphanov、@Gromykoss、@gustavosmendes、@Gutslabs、@H2KFORGIVEN、@haileymarshall、@hakanpak、@happy5318、@haran2001、@harjothkhara、@heathley、@hehehe0803、@helix4u、@herbalizer404、@HexLab98、@HiddenPuppy、@Hinotoi-agent、@HODLCLONE、@houko、@huangsen365、@huangxudong663-sys、@huangxun375-stack、@HwangJohn、@iaji、@iamlukethedev、@IamSanchoPanza、@IAvecilla、@Icather、@iizotov、@indigokarasu、@infinitycrew39、@ipriyaaanshu、@isair、@islam666、@itenev、@itsflownium、@izumi0uu、@Jaaneek、@JabberELF、@jackjin1997、@jackroofan、@janrenz、@jasnoorgill、@jasonQin6、@jcjc81、@jearnest11、@jeeves-assistant、@Jeffgithub0029、@jeffrobodie-glitch、@JezzaHehn、@jimmyjohansson84、@jmmaloney4、@jnibarger01、@JoaoMarcos44、@jplew、@Junass1、@justemu、@justin-cyhuang、@JustinOhms、@jvradahellys24-art、@Kailigithub、@kaishi00、@kangsoo-bit、@kartik-mem0、@keiravoss94、@kenyonxu、@kernel-t1、@Kewe63、@KeyArgo、@KiruyaMomochi、@kn8-codes、@Kolektori、@konsisumer、@kshitijk4poor、@kyssta-exe、@Kyzcreig、@Lazymonter、@LehaoLin、@LeonSGP43、@lEWFkRAD、@libre-7、@LIC99、@LifeJiggy、@linyubin、@liuhao1024、@lkevincc0、@lkz-de、@loes5050、@londo161、@lubosxyz、@m24927605、@MaheshtheDev、@manus-use、@marco0158、@MarioYounger、@martinramos002-bot、@MattKotsenas、@max-chen、@MaxFreedomPollard、@maxmilian、@maxpetrusenko、@memosr、@Mibayy、@Minksgo、@mintybasil、@mkslzk、@mohamedorigami-jpg、@MorAlekss、@mrparker0980、@ms-alan、@namredips、@nankingjing、@natehale、@necoweb3、@neo-2026、@Nickperillo、@nightq、@nikshepsvn、@nnnet、@nocturnum91、@nodejun、@NousResearch、@nycomar、@OmarB97、@orbisai0security、@oreoluwa、@outsourc-e、@OutThisLife、@p-andhika、@panghuer023、@Paperclip、@peetwan、@pefontana、@petrichor-op、@pinguarmy、@PINKIIILQWQ、@pmos69、@PolyphonyRequiem、@pprism13、@PRATHAMESH75、@professorpalmer、@pyxl-dev、@Que0x、@qWaitCrypto、@r266-tech、@RafaelMiMi、@Railway9784、@randomuser2026x、@rayjun、@rc-int、@rebel0789、@redactdeveloper、@riyas22、@rlaope、@rob-maron、@rodboev、@rodrigoeqnit、@rratmansky、@rrevenanttt、@ruangraung、@Ruzzgar、@ryo-solo、@s010mn、@Sahil-SS9、@SahilRakhaiya05、@SandroHub013、@Sanjays2402、@sasquatch9818、@ScotterMonk、@season179、@sgabel、@sgaofen、@sgtworkman、@shandian64、@shannonsands、@shashwatgokhe、@shawchanshek、@sherman-yang、@SHL0MS、@SidUParis、@SimoKiihamaki、@simpolism、@sjh9714、@skabartem、@skyc1e、@slawt、@soynchux、@spiky02plateau、@spjoes、@sprmn24、@srojk34、@stepanov1975、@steveonjava、@Subway2023、@sweetcornna、@swissly、@Sworntech-dev、@syahidfrd、@synapsesx、@szzhoujiarui-sketch、@talmax1124、@telos-oc、@testingbuddies24、@texhy、@tgmerritt、@theAgenticBuilder、@thestral123、@tkwong、@Tortugasaur、@Tranquil-Flow、@trevorgordon981、@truenorth-lj、@tt-a1i、@tuancookiez-hub、@TutkuEroglu、@tymrtn、@udatny、@UgwujaGeorge、@underthestars-zhy、@uperLu、@uzunkuyruk、@valenteff、@valentt、@vanthinh6886、@Versun、@victor-kyriazakos、@virtuadex、@vKongv、@w31rdm4ch1nZ、@weidzhou、@wgu9、@whoislikemiha、@wnuuee1、@woaini30050、@WuKongAI-CMU、@WuTianyi123、@WXBR、@x7peeps、@x9x9x9x9x9x91、@Xowiek、@xxchan、@xxxigm、@xydigit-zt、@yapsrubricsz0、@yashiels、@yeyitech、@ygd58、@YLChen-007、@yong2bba、@yoniebans、@ypwcharles、@yu-xin-c、@yungchentang、@yusekiotacode、@YuShu、@yyzquwu、@zapabob、@zccyman、@zeapsu、@zmlgit、@znding04、@Zyxxx-xxxyZ 此外还有：Lucas Nicolas。 --- **完整变更日志**: [v2026.6.19...v2026.7.1](https://github.com/NousResearch/hermes-agent/compare/v2026.6.19...v2026.7.1)

# Hermes Agent v0.18.0 (v2026.7.1) **Release Date:** July 1, 2026 **Since v0.17.0:** ~1,720 commits · 998 merged PRs · 2,215 files changed · ~251,000 insertions · ~41,000 deletions · **949 issues closed** · **370+ community contributors** > **The Judgment Release.** Over the last week and a half the team put nearly all of its effort into one goal: resolve **every P0 and P1 issue and PR in the entire Hermes Agent repo** — and as of this release, **100% of them are closed.** Zero open P0s. Zero open P1s. That's **~700 highest-priority items** cleared as part of **~1,950 total issues and PRs closed** this window. We intend to keep P0/P1 at zero from here on. > > On top of that clean-sweep, v0.18.0 is about how *well* Hermes thinks and how it *knows when its work is actually done*. Mixture-of-Agents became a first-class citizen — named ensembles of models you can pick like any other model, with every reference model's reasoning shown to you and the aggregator's answer streamed live. The agent learned to verify its own work against evidence instead of vibes, `/goal` gained completion contracts, and `/learn` + `/journey` turned self-improvement into something you can see and steer. Underneath, the gateway became genuinely deployable-at-scale (scale-to-zero, drain coordination), the desktop grew first-class coding projects and a playable memory graph, and subagents can now fan out in the background. ## 🎯 The P0/P1 Clean Sweep — 100% resolved This is the release headline. For a week and a half the team hammered the priority backlog day and night, and every single P0 and P1 across the whole repo is now closed: | Priority | Issues closed | PRs merged | |---|---|---| | **P0** (critical) | 3 | 8 | | **P1** (high) | 493 | 188 | | **Total** | **496** | **196** | That's **~692 highest-priority items resolved** in twelve days — and at the moment the sweep completed, the open P0/P1 count hit **0 across the entire repo.** The final cluster to fall was the interrupt-protected-compression sibling-fork bug (issue #56391) and its fix (#56416), closed on an all-nighter right before this release cut. Special shoutout to **@kshitijk4poor**, who burned through the priority backlog day and night alongside the core team — the cron reliability wave, the compression-fork fix, the credential-exfil hardening, and a huge share of the P1 closures are his. We're keeping P0/P1 at **0** from here forward. 🫡 ## ✨ Highlights - **Mixture-of-Agents is now a first-class model you can pick** — MoA used to be a mode you toggled; now every named MoA preset shows up as a selectable model under a `moa` provider, right alongside Claude, GPT, and Grok in every model picker (CLI, TUI, desktop, gateway). Pick "my-council" the same way you'd pick any model, and Hermes routes your prompt through that ensemble automatically. An ensemble of frontier models deliberating on your hardest questions is now one selection away, on every surface. ([#46081](https://github.com/NousResearch/hermes-agent/pull/46081), [#53548](https://github.com/NousResearch/hermes-agent/pull/53548), [#53561](https://github.com/NousResearch/hermes-agent/pull/53561) — @teknium1) - **See every model's reasoning, then watch the answer stream in** — When a MoA ensemble runs, each reference model's full output now renders as its own labelled block — you can read what GPT-5 thought, what Claude thought, and what Grok thought, before the aggregator synthesizes them into one answer. And that final answer now streams to you live instead of appearing all at once after a long silence. This works in the CLI, the TUI, and the desktop app. You get to watch the committee deliberate, not just read the verdict. ([#53793](https://github.com/NousResearch/hermes-agent/pull/53793), [#53855](https://github.com/NousResearch/hermes-agent/pull/53855), [#55625](https://github.com/NousResearch/hermes-agent/pull/55625), [#56101](https://github.com/NousResearch/hermes-agent/pull/56101) — @teknium1) - **The agent verifies its own work — "done" means proven, not claimed** — Hermes now records verification evidence for coding work and can decide it's finished by actually running your project's checks, not by asserting success. `/goal` gained **completion contracts**: you state what "done" looks like, and the standing-goal loop judges completion against that evidence instead of stopping when the model feels like it. There's a `pre_verify` hook for wiring in custom checks and a one-time migration that tunes the defaults sensibly. The difference between "I think I fixed it" and "the tests pass, here's proof." ([#50501](https://github.com/NousResearch/hermes-agent/pull/50501), [#52285](https://github.com/NousResearch/hermes-agent/pull/52285), [#55413](https://github.com/NousResearch/hermes-agent/pull/55413), [#53552](https://github.com/NousResearch/hermes-agent/pull/53552) — @teknium1, @OutThisLife) - **`/learn` — turn anything into a reusable skill by describing it** — Run `/learn <anything>` and Hermes distills a reusable skill out of whatever you point it at — a directory, a URL, or just the workflow you walked it through five minutes ago. It writes the skill to the standards in your CONTRIBUTING.md automatically. The next time you need that workflow, it's already there. Teaching Hermes a new trick is now a single command, not a manual skill-authoring session. ([#51506](https://github.com/NousResearch/hermes-agent/pull/51506), [#52372](https://github.com/NousResearch/hermes-agent/pull/52372) — @teknium1) - **`/journey` — a playable timeline of everything Hermes has learned about you** — The CLI and TUI gained `/journey`, a learning timeline that shows the memories and skills Hermes has accumulated over time — and you can edit or delete any of them right from the view. Pair it with the desktop's new **memory graph** (a top-down, playable radial timeline of memories and skills) and for the first time you can actually *see* what your agent knows, watch it grow, and prune what's wrong. Your agent's memory stops being a black box. ([#55555](https://github.com/NousResearch/hermes-agent/pull/55555), [#55859](https://github.com/NousResearch/hermes-agent/pull/55859), [#55226](https://github.com/NousResearch/hermes-agent/pull/55226) — @OutThisLife) - **Delegate a pile of work and keep going — background fan-out** — `delegate_task` can now fan out multiple subagents that all run in the **background**: your chat is never blocked, and when every subagent finishes, their results come back as a single consolidated turn. Kick off "research these five competitors in parallel" or "audit these three modules," then carry on with something else while a small fleet works. When it's all done, you get one clean summary instead of babysitting each one. ([#49734](https://github.com/NousResearch/hermes-agent/pull/49734) — @teknium1) - **First-class coding Projects in the desktop app** — The desktop app gained real, per-profile **Projects** — a sidebar of your codebases, a coding rail, a review pane, git worktree management, and agent-facing project tools, all backed by a proper `project → repo → lane` model. Instead of scattered chat sessions, your coding work is organized into projects the agent understands and can act on. It's the desktop turning into an actual coding cockpit. ([#49037](https://github.com/NousResearch/hermes-agent/pull/49037), [#54385](https://github.com/NousResearch/hermes-agent/pull/54385), [#54517](https://github.com/NousResearch/hermes-agent/pull/54517) — @OutThisLife) - **Run Hermes at scale — scale-to-zero and drain coordination** — The gateway can now go **dormant when idle** and quiesce cleanly before a restart, migration, or auto-update — without dropping in-flight conversations. A hosted or relay-only Hermes can scale to zero when nobody's talking to it and wake back up on demand, and disruptive lifecycle actions coordinate an external drain so nobody gets cut off mid-turn. Running Hermes for a team or as a hosted service just got a lot more production-grade. ([#52243](https://github.com/NousResearch/hermes-agent/pull/52243), [#52937](https://github.com/NousResearch/hermes-agent/pull/52937), [#54824](https://github.com/NousResearch/hermes-agent/pull/54824) — @teknium1, @benbarclay) - **Cheaper self-improvement — smarter background review** — The post-turn self-improvement fork (the one that decides whether to save a memory or skill) now routes to an auxiliary model, digests context instead of replaying the whole conversation, and adapts its cadence — so the "learn from what just happened" loop that runs after your turns costs a fraction of what it used to. You keep the self-improvement, you stop paying full main-model price for it. ([#49252](https://github.com/NousResearch/hermes-agent/pull/49252) — @teknium1) - **Compose your next prompt in your editor — `/prompt`** — `/prompt` opens your `$EDITOR` so you can hand-write a long, multi-line prompt in real markdown instead of fighting a one-line input box. Draft a detailed spec, a structured question, or a big paste, save, and it's queued as your next message. Small thing, huge quality-of-life win for anyone who writes Hermes more than a sentence at a time. ([#50509](https://github.com/NousResearch/hermes-agent/pull/50509) — @teknium1) - **Google Vertex AI — Gemini through your GCP service account, no static key** — Vertex AI is now a first-class provider for Gemini models over Vertex's OpenAI-compatible endpoint. The reason a plain custom-provider setup always died mid-session is that Vertex has no static API key — every request needs a short-lived OAuth2 access token (~1h TTL) minted from a service-account JSON or Application Default Credentials. Hermes now mints and auto-refreshes those tokens for you, so if your org runs Gemini through Google Cloud, you point Hermes at your service account and it just works — no token-pasting, no mid-session expiry. ([#56363](https://github.com/NousResearch/hermes-agent/pull/56363) — @teknium1, @slawt) - **Security round** — This window hardened several surfaces: MCP-config persistence attack surface locked down, cron `base_url` overrides that could exfiltrate provider credentials blocked, a non-reusable sentinel for prefix secrets in file reads, Slack app-level (`xapp-`) token redaction, a browser cloud-metadata floor enforced on every backend, and an `aiohttp` CVE floor across the lazy messaging paths. Fewer ways for a prompt-injected or misconfigured session to leak a credential. ([#50476](https://github.com/NousResearch/hermes-agent/pull/50476), [#56196](https://github.com/NousResearch/hermes-agent/pull/56196), [#54166](https://github.com/NousResearch/hermes-agent/pull/54166), [#56227](https://github.com/NousResearch/hermes-agent/pull/56227), [#52349](https://github.com/NousResearch/hermes-agent/pull/52349), [#56237](https://github.com/NousResearch/hermes-agent/pull/56237) — @teknium1, @kshitijk4poor, @claudlos) --- ## 🧠 Mixture-of-Agents (MoA) MoA graduated from a mode to a first-class part of the model system this window. - **Presets as selectable virtual models** — each named MoA preset appears as a model under provider `moa`; pick it in any model picker and Hermes routes through the ensemble ([#46081](https://github.com/NousResearch/hermes-agent/pull/46081), [#53561](https://github.com/NousResearch/hermes-agent/pull/53561), [#53775](https://github.com/NousResearch/hermes-agent/pull/53775) — @teknium1) - **`/moa` is now one-shot sugar** — runs a single prompt through the default preset and restores your model afterward; persistent switching goes through the model picker ([#53548](https://github.com/NousResearch/hermes-agent/pull/53548) — @teknium1) - **Reference-model output shown as labelled blocks** in CLI, TUI, and desktop — read each model's reasoning before the aggregator's synthesis ([#53793](https://github.com/NousResearch/hermes-agent/pull/53793), [#53855](https://github.com/NousResearch/hermes-agent/pull/53855) — @teknium1) - **Aggregator response streams live** instead of appearing whole after a silence ([#55625](https://github.com/NousResearch/hermes-agent/pull/55625) — @teknium1) - **References see full tool state and fire on every user/tool response**; advisory references end on a user turn and get a reference-role system prompt ([#54016](https://github.com/NousResearch/hermes-agent/pull/54016), [#54007](https://github.com/NousResearch/hermes-agent/pull/54007) — @teknium1) - **Opt-in full-turn trace persistence to JSONL** (`moa.save_traces`) for debugging and eval ([#56101](https://github.com/NousResearch/hermes-agent/pull/56101) — @teknium1) - Reliability: reference + aggregator models called through their provider's real route; context window resolved from the aggregator (not the 256K default); auxiliary tasks resolve to the aggregator; virtual provider blocked as a reference/aggregator slot; tolerant of hand-edited preset config ([#53580](https://github.com/NousResearch/hermes-agent/pull/53580), [#53780](https://github.com/NousResearch/hermes-agent/pull/53780), [#53827](https://github.com/NousResearch/hermes-agent/pull/53827), [#53281](https://github.com/NousResearch/hermes-agent/pull/53281), [#53275](https://github.com/NousResearch/hermes-agent/pull/53275), [#53556](https://github.com/NousResearch/hermes-agent/pull/53556) — @teknium1) - MoA slot provider-identity unified on the single `call_llm` chokepoint; HermesBench results documented ([#55991](https://github.com/NousResearch/hermes-agent/pull/55991), [#53206](https://github.com/NousResearch/hermes-agent/pull/53206) — @teknium1) ## ✅ Verification & Goals — the agent proves its work - **Completion contracts for `/goal`** — state what "done" looks like; the standing-goal loop judges against evidence, not the model's say-so ([#50501](https://github.com/NousResearch/hermes-agent/pull/50501) — @teknium1) - **`/goal wait <pid>`** — park the standing-goal loop on a background process instead of re-poking the agent ([#50503](https://github.com/NousResearch/hermes-agent/pull/50503) — @teknium1) - **Coding verification evidence ledger** — profile-scoped record of canonical project checks detected by `agent.coding_context`; gateway exposes verification status ([#52285](https://github.com/NousResearch/hermes-agent/pull/52285), [#52286](https://github.com/NousResearch/hermes-agent/pull/52286) — @OutThisLife) - **`pre_verify` hook + coding guidance config**; verification stop loop + ad-hoc verification scripts ([#55413](https://github.com/NousResearch/hermes-agent/pull/55413), [#52296](https://github.com/NousResearch/hermes-agent/pull/52296), [#52297](https://github.com/NousResearch/hermes-agent/pull/52297) — @OutThisLife) - **verify-on-stop defaults OFF** with a one-time v32 migration; skips doc-only edits; surface-aware "auto" default restored; gated off for messaging surfaces ([#53552](https://github.com/NousResearch/hermes-agent/pull/53552), [#54740](https://github.com/NousResearch/hermes-agent/pull/54740), [#55449](https://github.com/NousResearch/hermes-agent/pull/55449), [#52412](https://github.com/NousResearch/hermes-agent/pull/52412) — @teknium1, @OutThisLife, @GodsBoy) ## 🎓 Self-Improvement (Learn / Journey) - **`/learn <anything>`** — distill a reusable skill from a directory, URL, or a workflow you just walked through; honors CONTRIBUTING.md skill standards and mixed requirements ([#51506](https://github.com/NousResearch/hermes-agent/pull/51506), [#52372](https://github.com/NousResearch/hermes-agent/pull/52372), [#55956](https://github.com/NousResearch/hermes-agent/pull/55956) — @teknium1) - **`/journey`** — CLI + TUI learning timeline of accumulated memories and skills, with in-place edit/delete ([#55555](https://github.com/NousResearch/hermes-agent/pull/55555), [#55859](https://github.com/NousResearch/hermes-agent/pull/55859) — @OutThisLife) - **Cheaper background review** — aux-model routing + context digest + adaptive cadence for the post-turn self-improvement fork ([#49252](https://github.com/NousResearch/hermes-agent/pull/49252) — @teknium1) - **`memory` graph** in the desktop — playable radial timeline of memories + skills over time ([#55226](https://github.com/NousResearch/hermes-agent/pull/55226) — @OutThisLife) ## 🖥️ Hermes Desktop App ### Coding cockpit - **First-class Projects** — per-profile sidebar, coding rail, review pane, agent project tools (`project → repo → lane`); remote-gateway-aware folder picker + git cockpit (status, review, worktrees) ([#49037](https://github.com/NousResearch/hermes-agent/pull/49037), [#54385](https://github.com/NousResearch/hermes-agent/pull/54385) — @OutThisLife) - **Multi-terminal panel** with read-only agent terminals; persist & restore terminal tabs + scrollback across relaunch ([#54517](https://github.com/NousResearch/hermes-agent/pull/54517), [#54585](https://github.com/NousResearch/hermes-agent/pull/54585) — @OutThisLife) - **PR-style file diffs in chat**; in-app spot editor for the file preview pane; inline rich embeds, diagrams & alerts in assistant markdown ([#50731](https://github.com/NousResearch/hermes-agent/pull/50731), [#52772](https://github.com/NousResearch/hermes-agent/pull/52772), [#52935](https://github.com/NousResearch/hermes-agent/pull/52935) — @OutThisLife) ### UX & surfaces - Conversation timeline rail for long threads; context-usage breakdown popover; read-only spectator transcript for subagent watch windows; pop the composer into a draggable floating window ([#51094](https://github.com/NousResearch/hermes-agent/pull/51094), [#54907](https://github.com/NousResearch/hermes-agent/pull/54907), [#55033](https://github.com/NousResearch/hermes-agent/pull/55033), [#49488](https://github.com/NousResearch/hermes-agent/pull/49488) — @OutThisLife, @austinpickett) - Read replies aloud (auto-TTS) composer toggle; remember window size/position/maximized across launches; redesigned clarify prompt; shared overlay Panel primitive for cron/profiles/agents ([#55154](https://github.com/NousResearch/hermes-agent/pull/55154), [#52086](https://github.com/NousResearch/hermes-agent/pull/52086), [#52993](https://github.com/NousResearch/hermes-agent/pull/52993), [#54558](https://github.com/NousResearch/hermes-agent/pull/54558) — @OutThisLife) - Backup import/create/download from the web UI; add context-usage popover; flag already-installed themes in install pickers; config-driven Electron launch flags + GPU policy ([#54611](https://github.com/NousResearch/hermes-agent/pull/54611), [#55410](https://github.com/NousResearch/hermes-agent/pull/55410), [#53991](https://github.com/NousResearch/hermes-agent/pull/53991) — @teknium1, @OutThisLife) - **Pets** — roaming pet (opt-in), calmer/realistic roam, Alt+wheel scaling never cropped, frame-perfect hatch flow + CPU-safe chroma, pop-out overlay + notifications ([#55114](https://github.com/NousResearch/hermes-agent/pull/55114), [#55400](https://github.com/NousResearch/hermes-agent/pull/55400), [#52877](https://github.com/NousResearch/hermes-agent/pull/52877), [#47959](https://github.com/NousResearch/hermes-agent/pull/47959), [#52303](https://github.com/NousResearch/hermes-agent/pull/52303) — @OutThisLife) ### Refactor wave (composer / god-file de-entangle) - Decomposed the composer into isolated engine hooks; extracted branch/esc/url/placeholder/popout engines; split `thread.tsx`, `sidebar/index.tsx`, onboarding overlay, and `use-prompt-actions` god files into focused modules; shared WebSocket layer decoupling desktop from dashboard (`hermes serve`) ([#55500](https://github.com/NousResearch/hermes-agent/pull/55500), [#55842](https://github.com/NousResearch/hermes-agent/pull/55842), [#55451](https://github.com/NousResearch/hermes-agent/pull/55451), [#55453](https://github.com/NousResearch/hermes-agent/pull/55453), [#55807](https://github.com/NousResearch/hermes-agent/pull/55807), [#55504](https://github.com/NousResearch/hermes-agent/pull/55504), [#54568](https://github.com/NousResearch/hermes-agent/pull/54568) — @OutThisLife) - perf: bound tool-result rendering so big `/learn` runs don't freeze; fast session switching under load ([#52273](https://github.com/NousResearch/hermes-agent/pull/52273), [#52620](https://github.com/NousResearch/hermes-agent/pull/52620) — @OutThisLife) ## 📊 Web Dashboard - Auto-initiate portal SSO redirect on unauthenticated load; interactive auth setup on no-provider non-loopback bind; confidential-client (`client_secret`) support in self-hosted OIDC ([#54846](https://github.com/NousResearch/hermes-agent/pull/54846), [#50551](https://github.com/NousResearch/hermes-agent/pull/50551), [#55344](https://github.com/NousResearch/hermes-agent/pull/55344) — @teknium1, @benbarclay) - Catalogue all memory-provider API keys in `OPTIONAL_ENV_VARS`; list & add arbitrary custom `.env` keys on the Keys page; expose cron job execution fields; backup import/create/download ([#54546](https://github.com/NousResearch/hermes-agent/pull/54546), [#54552](https://github.com/NousResearch/hermes-agent/pull/54552), [#53551](https://github.com/NousResearch/hermes-agent/pull/53551), [#54611](https://github.com/NousResearch/hermes-agent/pull/54611) — @benbarclay, @teknium1) - Offload PTY spawn/close off the event loop; exclude non-interactive providers from interactive login surfaces ([#53227](https://github.com/NousResearch/hermes-agent/pull/53227), [#53239](https://github.com/NousResearch/hermes-agent/pull/53239) — @IAvecilla) ## 🏗️ Core Agent & Architecture ### Delegation & subagents - **Background fan-out** — parallel subagents run in the background, one consolidated return when all finish; calm "will resume" affordance for background `delegate_task` ([#49734](https://github.com/NousResearch/hermes-agent/pull/49734), [#52756](https://github.com/NousResearch/hermes-agent/pull/52756) — @teknium1, @OutThisLife) - Track background subagents in the CLI + TUI status bar ([#51441](https://github.com/NousResearch/hermes-agent/pull/51441), [#51485](https://github.com/NousResearch/hermes-agent/pull/51485) — @teknium1) ### Agent loop, tools & coding context - One-shot LLM helper + `llm.oneshot` gateway RPC; expose coding-context project facts (`project.facts` RPC) ([#51261](https://github.com/NousResearch/hermes-agent/pull/51261), [#51259](https://github.com/NousResearch/hermes-agent/pull/51259) — @OutThisLife) - `web_extract` truncate-and-store instead of LLM summarization; concurrent @-reference expansion ([#54843](https://github.com/NousResearch/hermes-agent/pull/54843), [#55207](https://github.com/NousResearch/hermes-agent/pull/55207) — @teknium1, @kshitijk4poor) - Friendly human-phrased tool labels for built-in tools; `/reasoning full` (uncapped thinking); `/timestamps` + timestamps in `/history`; `/prompt` composes in `$EDITOR` ([#55166](https://github.com/NousResearch/hermes-agent/pull/55166), [#50499](https://github.com/NousResearch/hermes-agent/pull/50499), [#50506](https://github.com/NousResearch/hermes-agent/pull/50506), [#50509](https://github.com/NousResearch/hermes-agent/pull/50509) — @teknium1) - Per-reasoning-model stale-timeout floor in stream + non-stream detectors; escalate SIGTERM→SIGKILL on host-pid termination after grace ([#52845](https://github.com/NousResearch/hermes-agent/pull/52845), [#50489](https://github.com/NousResearch/hermes-agent/pull/50489) — @teknium1) - Multiple `HERMES_WRITE_SAFE_ROOT` dirs; opt-in HTTP/WS body capture to an isolated, share-excluded `gui_bodies.log` ([#53292](https://github.com/NousResearch/hermes-agent/pull/53292), [#49044](https://github.com/NousResearch/hermes-agent/pull/49044) — @kshitijk4poor) ### Compression & sessions - In-place compaction option (single session id); flip `in_place` default to True with a guard fix ([#49739](https://github.com/NousResearch/hermes-agent/pull/49739), [#52658](https://github.com/NousResearch/hermes-agent/pull/52658) — @teknium1) - Backup includes `projects.db` and kanban boards in the pre-update snapshot ([#52990](https://github.com/NousResearch/hermes-agent/pull/52990) — @kshitijk4poor) ### Providers & models - **Google Vertex AI** first-class provider for Gemini over the OpenAI-compatible endpoint — auto-mints and refreshes short-lived OAuth2 tokens from a service-account JSON / ADC (no static key); salvages & modernizes #8427 by @slawt ([#56363](https://github.com/NousResearch/hermes-agent/pull/56363) — @teknium1, @slawt) - Krea via managed Nous Subscription gateway; Z.AI endpoint picker (Global/China/Coding Plan); Ollama-cloud reasoning_effort wiring; remove google-gemini-cli + google-antigravity OAuth providers ([#52647](https://github.com/NousResearch/hermes-agent/pull/52647), [#52364](https://github.com/NousResearch/hermes-agent/pull/52364), [#51494](https://github.com/NousResearch/hermes-agent/pull/51494), [#50492](https://github.com/NousResearch/hermes-agent/pull/50492) — @teknium1, @kshitijk4poor) - Honor `NOUS_INFERENCE_BASE_URL` env override for Nous OAuth; keep Nous auth fresh for idle dashboard/gateway agents ([#52270](https://github.com/NousResearch/hermes-agent/pull/52270), [#50567](https://github.com/NousResearch/hermes-agent/pull/50567) — @benbarclay, @teknium1) ## 🌐 Gateway, Fleet & Relay ### Scale-to-zero & drain - **Scale-to-zero idle detection + dormant-quiesce (Phase 0)**; hardened dormancy guards; fixed arm-gate counting disabled placeholder platforms ([#52243](https://github.com/NousResearch/hermes-agent/pull/52243), [#52359](https://github.com/NousResearch/hermes-agent/pull/52359), [#52831](https://github.com/NousResearch/hermes-agent/pull/52831) — @teknium1, @benbarclay) - **External drain coordination (safe-shutdown Phase 2)**; suppress home-channel shutdown broadcast on flagged drains; persist in-flight transcript on restart/shutdown drain timeout; busy/idle readout for safe lifecycle actions ([#52937](https://github.com/NousResearch/hermes-agent/pull/52937), [#54824](https://github.com/NousResearch/hermes-agent/pull/54824), [#50312](https://github.com/NousResearch/hermes-agent/pull/50312), [#50131](https://github.com/NousResearch/hermes-agent/pull/50131) — @teknium1, @benbarclay, @kshitijk4poor) - Default `restart_drain_timeout` to 0 to kill a systemd crash loop; self-heal a gateway stranded in draining/degraded ([#54066](https://github.com/NousResearch/hermes-agent/pull/54066), [#55397](https://github.com/NousResearch/hermes-agent/pull/55397) — @teknium1) ### Relay (Phase 5 / 6) - Wake primitive (gateway side); going-idle / buffered-flip primitive; `passthrough_forward` over WS; multi-platform-per-agent identity + per-frame egress; forward stable instance id at self-provision; declare relevance policy to the connector ([#51595](https://github.com/NousResearch/hermes-agent/pull/51595), [#51572](https://github.com/NousResearch/hermes-agent/pull/51572), [#50702](https://github.com/NousResearch/hermes-agent/pull/50702), [#52830](https://github.com/NousResearch/hermes-agent/pull/52830), [#50772](https://github.com/NousResearch/hermes-agent/pull/50772), [#51248](https://github.com/NousResearch/hermes-agent/pull/51248) — @benbarclay) - Authorize relay-delivered events by delivery, not `source.platform`; adopt `scope_id` wire key; purge platform-specific scope terminology ([#52306](https://github.com/NousResearch/hermes-agent/pull/52306), [#55289](https://github.com/NousResearch/hermes-agent/pull/55289), [#56016](https://github.com/NousResearch/hermes-agent/pull/56016) — @benbarclay) ### Gateway core & rendering - Typed send-error classification (`SendResult.error_kind`); per-platform `typing_indicator` toggle; per-category context breakdown in `/usage` ([#50342](https://github.com/NousResearch/hermes-agent/pull/50342), [#55394](https://github.com/NousResearch/hermes-agent/pull/55394), [#55204](https://github.com/NousResearch/hermes-agent/pull/55204) — @teknium1) - API server: configurable concurrent-run cap to prevent DoS; scope run approvals by run id ([#50007](https://github.com/NousResearch/hermes-agent/pull/50007), [#56129](https://github.com/NousResearch/hermes-agent/pull/56129) — @teknium1) ## 📱 Messaging Platforms - **Cron continuations** — continuable cron jobs (thread-preferred continuation with DM-mirror fallback); flat in-channel continuable cron delivery for Slack; warn when gateway not running on cron create/list ([#52250](https://github.com/NousResearch/hermes-agent/pull/52250), [#56254](https://github.com/NousResearch/hermes-agent/pull/56254), [#51696](https://github.com/NousResearch/hermes-agent/pull/51696) — @teknium1) - Telegram: configurable command menu + raised default cap so skills stay visible; gate rich draft previews separately; drain general send pool on pool timeout before retry ([#51716](https://github.com/NousResearch/hermes-agent/pull/51716), [#52088](https://github.com/NousResearch/hermes-agent/pull/52088), [#54121](https://github.com/NousResearch/hermes-agent/pull/54121) — @teknium1, @helix4u) - Slack: opt-in Block Kit rendering for agent messages; `--no-assistant` flag for manifest generation ([#56102](https://github.com/NousResearch/hermes-agent/pull/56102), [#51487](https://github.com/NousResearch/hermes-agent/pull/51487) — @teknium1) - Discord: render reasoning as `-#` subtext via `display.reasoning_style` ([#51168](https://github.com/NousResearch/hermes-agent/pull/51168) — @teknium1) - Native WhatsApp media delivery via the Baileys bridge; Teams native `send_video`/`send_voice`/`send_document`; photon sidecar upgraded to spectrum-ts v8 with tapback correlation; Raft gateway setup wizard ([#53598](https://github.com/NousResearch/hermes-agent/pull/53598), [#49308](https://github.com/NousResearch/hermes-agent/pull/49308), [#53451](https://github.com/NousResearch/hermes-agent/pull/53451), [#56230](https://github.com/NousResearch/hermes-agent/pull/56230) — @teknium1) - Signal: AAC voice-note remux + shared markdown formatting ([#49530](https://github.com/NousResearch/hermes-agent/pull/49530) — @kshitijk4poor) - Migrate slack/dingtalk/whatsapp/matrix/feishu/telegram/wecom/email/sms adapters to bundled ([#49408](https://github.com/NousResearch/hermes-agent/pull/49408) — @teknium1) ## 🔧 Tool System, Skills & MCP - Blank Slate setup mode — minimal agent, opt in to everything ([#36733](https://github.com/NousResearch/hermes-agent/pull/36733) — @teknium1) - MCP: config persistence attack surface hardened; block base_url exfil; keepalive for short-TTL sessions (see Security) — plus catalog & UX carried from v0.17.0 - Skills: `/learn` distillation (see Self-Improvement); `cloudflare-temporary-deploy` optional skill; creative-ideation v2.1.0 method library ([#50849](https://github.com/NousResearch/hermes-agent/pull/50849), [#42402](https://github.com/NousResearch/hermes-agent/pull/42402) — @teknium1, @SHL0MS) - Kanban: task lifecycle plugin hooks (claimed/completed/blocked); typed block reasons + unblock-loop breaker; handoff freshness stamping ([#50349](https://github.com/NousResearch/hermes-agent/pull/50349), [#52848](https://github.com/NousResearch/hermes-agent/pull/52848), [#53973](https://github.com/NousResearch/hermes-agent/pull/53973) — @teknium1) - Plugins: `ctx.profile_name` for session-agnostic profile access ([#50346](https://github.com/NousResearch/hermes-agent/pull/50346) — @teknium1) - LSP: PowerShellEditorServices language server; mem0 v3 API + OSS mode + update/delete tools ([#55930](https://github.com/NousResearch/hermes-agent/pull/55930), [#15624](https://github.com/NousResearch/hermes-agent/pull/15624) — @teknium1, @kartik-mem0) ## ⚡ Performance - Cold start: lazy-load gateway platform adapters; parse config + plugin manifests with libyaml `CSafeLoader` ([#54448](https://github.com/NousResearch/hermes-agent/pull/54448), [#54486](https://github.com/NousResearch/hermes-agent/pull/54486) — @teknium1) - State: merge FTS5 segments + `handoff_state` index to curb write-lock contention; single-pass `list_profiles` alias map + skill-count cache + event-loop offload ([#54752](https://github.com/NousResearch/hermes-agent/pull/54752), [#54770](https://github.com/NousResearch/hermes-agent/pull/54770) — @kshitijk4poor, @teknium1) ## 🔒 Security & Reliability - Harden MCP-config persistence attack surface; block cron `base_url` overrides that exfiltrate provider credentials; non-reusable sentinel for prefix secrets in file reads ([#50476](https://github.com/NousResearch/hermes-agent/pull/50476), [#56196](https://github.com/NousResearch/hermes-agent/pull/56196), [#54166](https://github.com/NousResearch/hermes-agent/pull/54166) — @teknium1, @kshitijk4poor) - Redact Slack App-Level (`xapp-`) tokens; browser cloud-metadata floor on all backends (CDP non-local); re-check private-network guard after `browser_back` navigation; scope `/resume` and `/sessions` to caller origin (IDOR); `aiohttp` 3.14.1 CVE floor across lazy messaging paths + pin-drift guard ([#56227](https://github.com/NousResearch/hermes-agent/pull/56227), [#52349](https://github.com/NousResearch/hermes-agent/pull/52349), [#56526](https://github.com/NousResearch/hermes-agent/pull/56526), [#56378](https://github.com/NousResearch/hermes-agent/pull/56378), [#56237](https://github.com/NousResearch/hermes-agent/pull/56237) — @teknium1, @claudlos, @kshitijk4poor) - Cron reliability wave: fail closed when an unpinned job's provider drifts; run missed-grace jobs once instead of deferring forever; keep the ticker alive on `BaseException` + heartbeat-aware status; layer enabled MCP servers onto per-job toolsets; guard cron model-tool path + auto-resume loop breaker ([#51051](https://github.com/NousResearch/hermes-agent/pull/51051), [#50062](https://github.com/NousResearch/hermes-agent/pull/50062), [#50016](https://github.com/NousResearch/hermes-agent/pull/50016), [#50117](https://github.com/NousResearch/hermes-agent/pull/50117), [#56240](https://github.com/NousResearch/hermes-agent/pull/56240) — @kshitijk4poor, @teknium1) - Windows: suppress console flashes + harden gateway restarts; prefer cmd npm shim on PATH fallback; respawn gateway windowless after GUI update; prefer managed node for whatsapp/desktop ([#52340](https://github.com/NousResearch/hermes-agent/pull/52340), [#50398](https://github.com/NousResearch/hermes-agent/pull/50398), [#52239](https://github.com/NousResearch/hermes-agent/pull/52239) — @helix4u, @teknium1) ## 🔁 Reverts (in-window, for the record) - cron job storage returned to per-profile (reverts #32117 + #50993); don't clone `auth.json` (duplicating OAuth grant causes sibling revocation); windows terminal-popup PRs rolled back; `prompt_caching.enabled` toggle backed out for re-evaluation ([#51116](https://github.com/NousResearch/hermes-agent/pull/51116), [#51732](https://github.com/NousResearch/hermes-agent/pull/51732), [#53853](https://github.com/NousResearch/hermes-agent/pull/53853), [#56126](https://github.com/NousResearch/hermes-agent/pull/56126) — @teknium1) ## 👥 Contributors **381 people** contributed to this release (via commits, co-author trailers, and salvaged PRs). Thank you, all of you. ### Core - @teknium1 — release lead; MoA first-class, verification/goals, `/learn`, background review, security round, providers, the P0/P1 clean-sweep - @OutThisLife — desktop app (projects, memory graph, `/journey`, multi-terminal, composer refactor wave, pets, verification UX) ### Top community contributors - @kshitijk4poor — the P0/P1 backlog burn: cron reliability wave, state perf, security (cron credential-exfil), gateway/signal, TUI config — a huge share of the priority closures - @benbarclay — relay Phase 5/6, scale-to-zero / drain coordination, dashboard auth/keys, gateway hardening - @ethernet8023 — CI/docker (unified jobs, faster builds, timings report) - @helix4u — Windows hardening (console flashes, npm shim, gateway restarts) ### All contributors @0xbyt4, @0xDevNinja, @0xsir0000, @1RB, @595650661, @aaronlab, @abchiaravalle, @adammatski1972, @AetherAgents, @Afnath-max, @agt-user, @ahmadashfq, @AhmetArif0, @AIalliAI, @aieng-abdullah, @ailang323, @ailthrim, @aj-nt, @alelpoan, @alloevil, @amathxbt, @ambition0802, @anderskev, @andressommerhoff, @angelos, @annguyenNous, @Antimatter543, @arminanton, @arthurzhang, @asimons81, @austinpickett, @baolingao, @Bartok9, @basilalshukaili, @BBCrypto-web, @bbopen, @Beandon13, @beardthelion, @benbarclay, @benbenlijie, @binhnt92, @bitcryptic-gw, @Blaryxoff, @bogerman1, @bradhallett, @brett539, @briandevans, @buihongduc132, @bykim0119, @catapreta, @chaithanyak42, @charleneleong-ai, @CharlieKerfoot, @chazmaniandinkle, @chrispersico, @Christopher-Schulze, @chriswesley4, @claudlos, @clovericbot, @cmcejas, @codexGW, @Cossackx, @counterposition, @coygeek, @CRWuTJ, @cyb0rgk1tty, @cyb3rwr3n, @cypctlinux, @cypres0099, @dalenguyen, @Danamove, @DanAsBjorn, @DataAdvisory, @davidgut1982, @DavidMetcalfe, @davidvv, @de1tydev, @denisqq, @devorun, @devsart95, @DhivinX, @DiamondEyesFox, @difujia, @Disaster-Terminator, @djimit, @djstunami, @dodo-reach, @donovan-yohan, @Dr1985, @DrZM007, @Dusk1e, @egilewski, @ehz0ah, @Eji4h, @EloquentBrush0x, @Elshayib, @emozilla, @entropy-0x, @Erosika, @EtherAura, @etherman-os, @ethernet8023, @f-trycua, @fayenix, @fesalfayed, @firefly, @flamiinngo, @flobo3, @francescomucio, @franksong2702, @friendshipisover, @fsaad1984, @fyzanshaik, @GauravPatil2515, @gdeyoung, @georgex8001, @GodsBoy, @graphanov, @Gromykoss, @gustavosmendes, @Gutslabs, @H2KFORGIVEN, @haileymarshall, @hakanpak, @happy5318, @haran2001, @harjothkhara, @heathley, @hehehe0803, @helix4u, @herbalizer404, @HexLab98, @HiddenPuppy, @Hinotoi-agent, @HODLCLONE, @houko, @huangsen365, @huangxudong663-sys, @huangxun375-stack, @HwangJohn, @iaji, @iamlukethedev, @IamSanchoPanza, @IAvecilla, @Icather, @iizotov, @indigokarasu, @infinitycrew39, @ipriyaaanshu, @isair, @islam666, @itenev, @itsflownium, @izumi0uu, @Jaaneek, @JabberELF, @jackjin1997, @jackroofan, @janrenz, @jasnoorgill, @jasonQin6, @jcjc81, @jearnest11, @jeeves-assistant, @Jeffgithub0029, @jeffrobodie-glitch, @JezzaHehn, @jimmyjohansson84, @jmmaloney4, @jnibarger01, @JoaoMarcos44, @jplew, @Junass1, @justemu, @justin-cyhuang, @JustinOhms, @jvradahellys24-art, @Kailigithub, @kaishi00, @kangsoo-bit, @kartik-mem0, @keiravoss94, @kenyonxu, @kernel-t1, @Kewe63, @KeyArgo, @KiruyaMomochi, @kn8-codes, @Kolektori, @konsisumer, @kshitijk4poor, @kyssta-exe, @Kyzcreig, @Lazymonter, @LehaoLin, @LeonSGP43, @lEWFkRAD, @libre-7, @LIC99, @LifeJiggy, @linyubin, @liuhao1024, @lkevincc0, @lkz-de, @loes5050, @londo161, @lubosxyz, @m24927605, @MaheshtheDev, @manus-use, @marco0158, @MarioYounger, @martinramos002-bot, @MattKotsenas, @max-chen, @MaxFreedomPollard, @maxmilian, @maxpetrusenko, @memosr, @Mibayy, @Minksgo, @mintybasil, @mkslzk, @mohamedorigami-jpg, @MorAlekss, @mrparker0980, @ms-alan, @namredips, @nankingjing, @natehale, @necoweb3, @neo-2026, @Nickperillo, @nightq, @nikshepsvn, @nnnet, @nocturnum91, @nodejun, @NousResearch, @nycomar, @OmarB97, @orbisai0security, @oreoluwa, @outsourc-e, @OutThisLife, @p-andhika, @panghuer023, @Paperclip, @peetwan, @pefontana, @petrichor-op, @pinguarmy, @PINKIIILQWQ, @pmos69, @PolyphonyRequiem, @pprism13, @PRATHAMESH75, @professorpalmer, @pyxl-dev, @Que0x, @qWaitCrypto, @r266-tech, @RafaelMiMi, @Railway9784, @randomuser2026x, @rayjun, @rc-int, @rebel0789, @redactdeveloper, @riyas22, @rlaope, @rob-maron, @rodboev, @rodrigoeqnit, @rratmansky, @rrevenanttt, @ruangraung, @Ruzzgar, @ryo-solo, @s010mn, @Sahil-SS9, @SahilRakhaiya05, @SandroHub013, @Sanjays2402, @sasquatch9818, @ScotterMonk, @season179, @sgabel, @sgaofen, @sgtworkman, @shandian64, @shannonsands, @shashwatgokhe, @shawchanshek, @sherman-yang, @SHL0MS, @SidUParis, @SimoKiihamaki, @simpolism, @sjh9714, @skabartem, @skyc1e, @slawt, @soynchux, @spiky02plateau, @spjoes, @sprmn24, @srojk34, @stepanov1975, @steveonjava, @Subway2023, @sweetcornna, @swissly, @Sworntech-dev, @syahidfrd, @synapsesx, @szzhoujiarui-sketch, @talmax1124, @telos-oc, @testingbuddies24, @texhy, @tgmerritt, @theAgenticBuilder, @thestral123, @tkwong, @Tortugasaur, @Tranquil-Flow, @trevorgordon981, @truenorth-lj, @tt-a1i, @tuancookiez-hub, @TutkuEroglu, @tymrtn, @udatny, @UgwujaGeorge, @underthestars-zhy, @uperLu, @uzunkuyruk, @valenteff, @valentt, @vanthinh6886, @Versun, @victor-kyriazakos, @virtuadex, @vKongv, @w31rdm4ch1nZ, @weidzhou, @wgu9, @whoislikemiha, @wnuuee1, @woaini30050, @WuKongAI-CMU, @WuTianyi123, @WXBR, @x7peeps, @x9x9x9x9x9x91, @Xowiek, @xxchan, @xxxigm, @xydigit-zt, @yapsrubricsz0, @yashiels, @yeyitech, @ygd58, @YLChen-007, @yong2bba, @yoniebans, @ypwcharles, @yu-xin-c, @yungchentang, @yusekiotacode, @YuShu, @yyzquwu, @zapabob, @zccyman, @zeapsu, @zmlgit, @znding04, @Zyxxx-xxxyZ Also: Lucas Nicolas. --- **Full Changelog**: [v2026.6.19...v2026.7.1](https://github.com/NousResearch/hermes-agent/compare/v2026.6.19...v2026.7.1)

```markdown # Hermes Agent v0.17.0 (v2026.6.19) **发布日期：** 2026年6月19日 **自 v0.16.0 以来：** ~1,475 次提交 · ~800 个已合并 PR · 1,693 个文件变更 · 235,390 行新增 · 50,730 行删除 · 300+ 个已关闭 issue · 245 位社区贡献者 > **“触及”版本。** v0.16.0 将 Hermes 带到了你的桌面上。v0.17.0 则关乎其触及范围能延伸多远——跨越新的对话地点，深入你已经使用的工具，并延伸至为团队运行 Hermes 的人们。Hermes 新增了两个渠道（通过 Photon 的 iMessage，以及 Raft 代理网络），桌面应用获得了大量新功能，子代理现在可以在后台运行，图像生成学会了编辑，并且通过 xAI Grok 订阅可以调用 Cursor 的 Composer 模型。仪表板获得了完整的配置文件构建器和安全登录，技能中心浏览器进行了彻底改造，`memory` 工具获得了重大升级，策展器（curator）停止了在每次例行运行时消耗辅助模型预算。同时还附带关闭了 300 多个 issue，以及一轮安全更新。 ## ✨ 亮点 - **Hermes 到达 iMessage——Photon Spectrum，无需 Mac 中继**——现在有一个基于 Photon 托管线路池构建的 iMessage 平台插件。运行 `hermes photon login`，使用设备码进行身份验证，Hermes 即可发送和接收 iMessage——无需一台藏在壁橱里的 Mac 运行中继，也无需照看 BlueBubbles 桥接。它被定位为 BlueBubbles 的继任者：免费开始，无需自行托管。如果你的朋友和家人生活在蓝色气泡中，Hermes 现在也生活在其中。（[#32348](https://github.com/NousResearch/hermes-agent/pull/32348)、[#42582](https://github.com/NousResearch/hermes-agent/pull/42582)、[#44713](https://github.com/NousResearch/hermes-agent/pull/44713) — @teknium1） - **Raft——Hermes 作为网关渠道加入 Raft 代理网络**——一个新的内置 Raft 平台适配器让 Hermes 通过唤醒-渠道桥接连接到 [Raft](https://raft.build) 作为外部代理。设置 `RAFT_PROFILE`，运行桥接，Raft 即可唤醒 Hermes 处理消息——采用隐私契约设计，唤醒负载仅携带元数据（事件 ID、时间戳），绝不包含消息正文。这是 Hermes 可以出现并工作的另一个界面。（[#48210](https://github.com/NousResearch/hermes-agent/pull/48210) — @xxchan, @teknium1） - **功能强化的桌面应用**——v0.16.0 发布了桌面应用；v0.17.0 通过数十个 PR 深化了其功能。可重新绑定的键盘快捷键、带有按类型切换的原生 OS 通知、实时子代理**监视窗口**（将被委托代理的活动流式传输到自己的窗格中）、带按模型预设的合成器模型选择器、自动 RTL/双向文本方向、可调整大小的 VS Code 主题终端窗格、每个线程的合成器草稿，以及直接在应用中安装**任何 VS Code 市场主题**的能力。桌面应用现在是一个严肃的日常驾驶工具，不再是预览版。（[#45866](https://github.com/NousResearch/hermes-agent/pull/45866)、[#40660](https://github.com/NousResearch/hermes-agent/pull/40660)、[#47060](https://github.com/NousResearch/hermes-agent/pull/47060)、[#46959](https://github.com/NousResearch/hermes-agent/pull/46959)、[#43292](https://github.com/NousResearch/hermes-agent/pull/43292)、[#44596](https://github.com/NousResearch/hermes-agent/pull/44596) — @OutThisLife, @teknium1） - **后台/异步子代理——委托工作并继续前行**——`delegate_task(background=true)` 现在会分派一个在后台运行的子代理，并立即返回一个句柄。你和模型可以继续工作，而子代理则默默运行，一旦完成，完整结果会作为新的一轮重新进入对话。启动一次长时间的研究探索或多步骤构建，然后继续做其他事情，而不是被阻塞等待。（[#40946](https://github.com/NousResearch/hermes-agent/pull/40946)、[#46968](https://github.com/NousResearch/hermes-agent/pull/46968) — @teknium1） - **编辑图像，而不仅仅是生成——`image_generate` 中的图像到图像**——`image_generate` 现在可以编辑和转换源图像，而不仅仅是凭空创建。传入现有图像和提示，它会路由到后端的编辑端点（与 `video_generate` 相同的工具、相同的模式），覆盖所有受支持的图像提供商。“把这个 Logo 变成蓝色”、“移除背景”、“将这张草图变成渲染图”——所有这些都可以通过你已经使用的工具完成。（[#48705](https://github.com/NousResearch/hermes-agent/pull/48705) — @teknium1） - **自动化蓝图——无需学习 cron 即可安排事务**——按名称选择一个自动化，Hermes 会询问你需要什么——无需 cron 语法，无需输入 `slot=value`。一个蓝图定义在每个界面上都能原生渲染：仪表板中的表单、CLI/TUI/通讯工具中的斜杠命令、与代理的对话、文档目录中的条目。“每天早上 8 点新闻简报”变成一件通过回答问题来设置的事情，而不是记忆 `0 8 * * *`。（[#41309](https://github.com/NousResearch/hermes-agent/pull/41309) — @teknium1） - **Cursor 的 Composer 模型，通过你的 xAI Grok 订阅**——`grok-composer-2.5-fast` 现在出现在 xAI OAuth 模型选择器中，其上下文窗口已调整为完整的 200k。Composer 是 Cursor 背后的快速编码模型——如果你有 xAI Grok 订阅，现在可以直接通过 OAuth 让 Hermes 指向它，无需单独的 API 密钥。你的 Grok 计划，Hermes 的代理循环，Composer 的编码速度。（[#47908](https://github.com/NousResearch/hermes-agent/pull/47908)、[#6f89e17](https://github.com/NousResearch/hermes-agent/pull/47371) — @teknium1） - **仪表板中的完整配置文件构建器**——在浏览器中构建完整的 Hermes 配置文件——选择其模型、选择其技能、附加其 MCP 服务器——无需手动编辑 `config.yaml`。仪表板还将多配置文件管理统一到一个机器范围的视图中，并带有全局配置文件切换器，因此你可以从一个地方管理所有配置文件。（[#39084](https://github.com/NousResearch/hermes-agent/pull/39084)、[#44007](https://github.com/NousResearch/hermes-agent/pull/44007) — @teknium1） - **技能中心浏览器改造**——仪表板的技能中心进行了彻底重做：连接的中心、精选部分、安装前的完整技能预览，以及每个技能的安全扫描。从可信来源（OpenAI、Anthropic、HuggingFace、NVIDIA）浏览和安装技能现在是一种真正的浏览体验，不再是一个扁平列表。（[#40384](https://github.com/NousResearch/hermes-agent/pull/40384)、[#43398](https://github.com/NousResearch/hermes-agent/pull/43398) — @teknium1） - **`memory` 工具获得重大升级——原子批量操作**——`memory` 工具新增了一个 `operations` 数组，用于**在最终字符预算内原子性地**应用一批添加/替换/移除编辑。模型可以在一次调用中释放空间并添加新条目——即使仅添加一项就会超出预算——将以前脆弱的多次交互舞蹈压缩成一个可靠的操作。内存更新现在更快，并且不太可能在编辑过程中失败。（[#48507](https://github.com/NousResearch/hermes-agent/pull/48507) — @teknium1） - **安全的仪表板登录**——仪表板的身份验证得到了加强：每个需要令牌的端点现在在 OAuth 网关后正确返回 401，WebSocket 身份验证使用提供的仪表板令牌，当 `public_url` 覆盖被静默拒绝时会触发警告。将仪表板暴露到网络现在默认更安全。（[#42578](https://github.com/NousResearch/hermes-agent/pull/42578)、[#42578](https://github.com/NousResearch/hermes-agent/pull/43214) — @benbarclay, @teknium1） - **官方 WhatsApp Business Cloud API 适配器**——除了现有的 Baileys 桥接，Hermes 现在也能使用**官方** WhatsApp Business Cloud API——Meta 的第一方、托管、无需桥接进程的路径。将你的业务 API 凭据指向它，Hermes 即可通过支持的渠道使用 WhatsApp，无需保持 QR 扫描桥接进程存活。（[#44331](https://github.com/NousResearch/hermes-agent/pull/44331)、[#43921](https://github.com/NousResearch/hermes-agent/pull/43921) — @jquesnelle, @teknium1） - **Telegram 富文本——Bot API 10.1 富消息**——Telegram 回复现在通过 Bot API 10.1 渲染为适当的富消息：更好的格式、更干净的长消息处理、原生的标记而不是扁平文本。默认启用，可选择退出，因此你的 Telegram 对话看起来像它们应该的样子，无需任何配置。（[#44829](https://github.com/NousResearch/hermes-agent/pull/44829)、[#45584](https://github.com/NousResearch/hermes-agent/pull/45584)、[#45953](https://github.com/NousResearch/hermes-agent/pull/45953) — @teknium1） - **策展器成本优化——例行运行无辅助模型消耗**——技能策展器现在默认修剪过时的技能，但不再运行其 LLM 驱动的整合步骤，除非你选择加入（`curator.consolidate: true` 或 `hermes curator run --consolidate`）。确定性不活动扫描继续免费运行；意见性强、消耗辅助模型预算的“构建伞形技能”分支现在默认关闭。例行后台策展**消耗零令牌**。（[#47840](https://github.com/NousResearch/hermes-agent/pull/47840) — @teknium1） ## 🖥️ Hermes 桌面应用 ### 新界面与用户体验 - 可重新绑定的键盘快捷键面板；带有按类型切换的原生 OS 通知；策划的轮次完成提示 + 可关闭的错误横幅（[#40660](https://github.com/NousResearch/hermes-agent/pull/40660)、[#45866](https://github.com/NousResearch/hermes-agent/pull/45866)、[#42480](https://github.com/NousResearch/hermes-agent/pull/42480)、[#47985](https://github.com/NousResearch/hermes-agent/pull/47985) — @OutThisLife、@teknium1） - 实时子代理**监视窗口**——将被委托代理的活动流式传输到自己的窗格中；合成器状态堆栈 + 可编辑提示；在任何自己的窗口中打开聊天；新会话紧凑窗口热键（[#47060](https://github.com/NousResearch/hermes-agent/pull/47060)、[#44630](https://github.com/NousResearch/hermes-agent/pull/44630)、[#43219](https://github.com/NousResearch/hermes-agent/pull/43219)、[#46951](https://github.com/NousResearch/hermes-agent/pull/46951) — @OutThisLife、@teknium1） - 合成器模型选择器 + 按模型预设 + 外部提供商断开连接；在 GUI 中展示 `hermes model` 的每个提供商/模型；将提供商列表统一为一个来源；当主模型切换导致辅助任务固定在其他地方时发出警告（[#46959](https://github.com/NousResearch/hermes-agent/pull/46959)、[#40563](https://github.com/NousResearch/hermes-agent/pull/40563)、[#49080](https://github.com/NousResearch/hermes-agent/pull/49080)、[#40286](https://github.com/NousResearch/hermes-agent/pull/40286) — @teknium1、@OutThisLife） - 安装**任何 VS Code 市场主题**；每个配置文件可分配主题；窗口透明度滑块；统一覆盖设计系统 + BrandMark + 入门重新设计（[#43292](https://github.com/NousResearch/hermes-agent/pull/43292)、[#42286](https://github.com/NousResearch/hermes-agent/pull/42286)、[#45086](https://github.com/NousResearch/hermes-agent/pull/45086)、[#40708](https://github.com/NousResearch/hermes-agent/pull/40708) — @teknium1、@OutThisLife） - 可调整大小的 VS Code 主题终端窗格 + 调色板完善；自动检测聊天中的 RTL/双向文本方向；Mac 风格会话切换器（^Tab / ^1-9）；工作区感知侧边栏分组；悬停显示折叠的侧边栏；侧边栏中的消息来源文件夹（[#42521](https://github.com/NousResearch/hermes-agent/pull/42521)、[#44596](https://github.com/NousResearch/hermes-agent/pull/44596)、[#43111](https://github.com/NousResearch/hermes-agent/pull/43111)、[#45273](https://github.com/NousResearch/hermes-agent/pull/45273)、[#41670](https://github.com/NousResearch/hermes-agent/pull/41670)、[#41751](https://github.com/NousResearch/hermes-agent/pull/41751) — @OutThisLife） - 合成器中的箭头键历史 + 队列编辑；从批准栏中内联展开完整命令；跟随流式传输到底部 + 跳至底部按钮；侧边栏和仪表板调度器中的一等 cron 任务（[#40234](https://github.com/NousResearch/hermes-agent/pull/40234)、[#44864](https://github.com/NousResearch/hermes-agent/pull/44864)、[#45263](https://github.com/NousResearch/hermes-agent/pull/45263)、[#40684](https://github.com/NousResearch/hermes-agent/pull/40684) — @OutThisLife、@teknium1） - 桌面宠物——弹出覆盖 + 通知（[#47938](https://github.com/NousResearch/hermes-agent/pull/47938) — @teknium1） - 设置中的完整工具后端配置（选择器 + 每个后端设置）；从 GUI 运行工具后端安装后设置；卸载聊天 GUI 而不移除代理；Shift+点击状态栏 zap 切换全局 YOLO；本地网关上的 `/browser connect`（[#41232](https://github.com/NousResearch/hermes-agent/pull/41232)、[#40559](https://github.com/NousResearch/hermes-agent/pull/40559)、[#40355](https://github.com/NousResearch/hermes-agent/pull/40355)、[#41666](https://github.com/NousResearch/hermes-agent/pull/41666)、[#47245](https://github.com/NousResearch/hermes-agent/pull/47245) — @teknium1、@OutThisLife） - 日语 + 繁体中文语言切换（[#40114](https://github.com/NousResearch/hermes-agent/pull/40114)） - “重启网关”操作（从“重启消息”重命名）在状态栏中显示，并出现在消息保存/切换提示中；渲染的日志可选/可复制（[#49094](https://github.com/NousResearch/hermes-agent/pull/49094) — @OutThisLife） ### 远程网关与多配置文件 - **远程媒体中继**——首次通过网络附加图像/PDF 并显示代理编写的图像；通过 `file.attach` 实现远程网关文件附件（[#41336](https://github.com/NousResearch/hermes-agent/pull/41336)、[#42634](https://github.com/NousResearch/hermes-agent/pull/42634) — @teknium1） - 客户端 + 后端版本按钮 + 远程后端更新流程；浏览远程后端文件；路由全局远程配置文件 REST 调用；通过重新验证过期的远程后端来在睡眠/唤醒后恢复聊天（[#42181](https://github.com/NousResearch/hermes-agent/pull/42181)、[#44326](https://github.com/NousResearch/hermes-agent/pull/44326)、[#47011](https://github.com/NousResearch/hermes-agent/pull/47011)、[#41350](https://github.com/NousResearch/hermes-agent/pull/41350) — @OutThisLife、@teknium1） - 多配置文件善后清理——WS 认证 + 跨配置文件会话读取；在删除前释放配置文件后端；按会话限定会话列表/模型切换/计时器（[#44529](https://github.com/NousResearch/hermes-agent/pull/44529)、[#42613](https://github.com/NousResearch/hermes-agent/pull/42613)、[#41103](https://github.com/NousResearch/hermes-agent/pull/41103)、[#41120](https://github.com/NousResearch/hermes-agent/pull/41120)、[#41182](https://github.com/NousResearch/hermes-agent/pull/41182) — @OutThisLife、@teknium1） - 将子代理活动流式传输到监视窗口；在未聚焦的辅助聊天窗口中保持流式绘画；恢复丢失的会话窗口（[#47060](https://github.com/NousResearch/hermes-agent/pull/47060)、[#47919](https://github.com/NousResearch/hermes-agent/pull/47919)、[#47655](https://github.com/NousResearch/hermes-agent/pull/47655) — @OutThisLife） ## 📊 Web 仪表板 - 全功能配置文件构建器（模型 + 技能 + MCP）；统一多配置文件管理——一台机器的仪表板 + 全局配置文件切换器；配置文件范围的技能和工具集；聊天标签页上的会话切换面板（[#39084](https://github.com/NousResearch/hermes-agent/pull/39084)、[#44007](https://github.com/NousResearch/hermes-agent/pull/44007)、[#43808](https://github.com/NousResearch/hermes-agent/pull/43808)、[#49077](https://github.com/NousResearch/hermes-agent/pull/49077) — @teknium1） - 技能中心浏览器改造——连接的中心、精选、预览 + 安全扫描；技能页面上的 SKILL.md 编辑器 + cron 模态框中的附加技能选择器；每个 MCP 的完整目录详细信息；GUI 中的完整工具后端配置（[#40384](https://github.com/NousResearch/hermes-agent/pull/40384)、[#44231](https://github.com/NousResearch/hermes-agent/pull/44231)、[#48520](https://github.com/NousResearch/hermes-agent/pull/48520)、[#40418](https://github.com/NousResearch/hermes-agent/pull/40418) — @teknium1） - 从 Webhooks 页面启用 Webhooks；幂等的 `hermes dashboard register`；Telegram QR 入门后自动重启网关；文件浏览器；从主题选择器更改 UI 字体；聊天侧边栏中的推理努力选择器（[#44021](https://github.com/NousResearch/hermes-agent/pull/44021)、[#42455](https://github.com/NousResearch/hermes-agent/pull/42455)、[#43424](https://github.com/NousResearch/hermes-agent/pull/43424)、[#43512](https://github.com/NousResearch/hermes-agent/pull/43512)、[#41145](https://github.com/NousResearch/hermes-agent/pull/41145)、[#49141](https://github.com/NousResearch/hermes-agent/pull/49141) — @teknium1） ## 🏗️ 核心代理与架构 ### God 文件重构浪潮（run_agent.py / cli.py / gateway/run.py） - **`cli.py` main() 从 3297 行减少到 954 行**——提取了 28 个子命令解析器到 `hermes_cli/subcommands/`，然后提升了 9 个闭包处理器；32 个斜杠命令处理器 → `CLICommandsMixin`；18 个模型流向导函数 → `model_setup_flows`；代理构建集群 → `CLIAgentSetupMixin`（[#41798](https://github.com/NousResearch/hermes-agent/pull/41798)、[#41835](https://github.com/NousResearch/hermes-agent/pull/41835)、[#41942](https://github.com/NousResearch/hermes-agent/pull/41942)、[#42174](https://github.com/NousResearch/hermes-agent/pull/42174)、[#42153](https://github.com/NousResearch/hermes-agent/pull/42153) — @teknium1） - **`gateway/run.py` 从 19157 行减少到 15870 行**——42 个斜杠命令处理器 → `GatewaySlashCommandsMixin`；授权集群 → `GatewayAuthorizationMixin`；看板观察者循环 → `GatewayKanbanWatchersMixin`（[#41886](https://github.com/NousResearch/hermes-agent/pull/41886)、[#42159](https://github.com/NousResearch/hermes-agent/pull/42159)、[#41849](https://github.com/NousResearch/hermes-agent/pull/41849) — @teknium1） - **`run_agent.py` 轮次循环**——将序言提取到 `TurnContext`，循环后尾部提取到 `finalize_turn`，将内部重试循环恢复标志合并到 `TurnRetryState`（[#41778](https://github.com/NousResearch/hermes-agent/pull/41778)、[#42169](https://github.com/NousResearch/hermes-agent/pull/42169)、[#41828](https://github.com/NousResearch/hermes-agent/pull/41828) — @teknium1） ### 代理循环、提示与工具 - **`memory` 批量操作**——针对最终字符预算的原子添加/替换/移除数组，因此单个调用可以释放空间并添加条目（[#48507](https://github.com/NousResearch/hermes-agent/pull/48507) — @teknium1） - **`search_files` 无损密集化**——空闲空间评估报告 + 一个值得发布的密集化改进（更少的令牌每结果，相同的匹配）（[#47866](https://github.com/NousResearch/hermes-agent/pull/47866) — @teknium1） - 移除了代理可调用的 `send_message` 工具；CLI/TUI/桌面/ACP 上的编码上下文姿态；`read_file` 将 `.ipynb`/`.docx`/`.xlsx` 提取为文本（[#47856](https://github.com/NousResearch/hermes-agent/pull/47856)、[#43316](https://github.com/NousResearch/hermes-agent/pull/43316)、[#37082](https://github.com/NousResearch/hermes-agent/pull/37082) — @teknium1） - 上下文文件处理：可配置的截断限制 + 警告；将上下文文件上限缩放到模型窗口 + 将代理指向截断的文件（[#47251](https://github.com/NousResearch/hermes-agent/pull/47251)、[#47846](https://github.com/NousResearch/hermes-agent/pull/47846) — @teknium1） - 压缩：压缩摘要中的时间锚定；将 gpt-5.5 在 Codex OAuth 上的压缩触发阈值提高到 85%（[#41102](https://github.com/NousResearch/hermes-agent/pull/41102)、[#40957](https://github.com/NousResearch/hermes-agent/pull/40957) — @teknium1） - 自适应中间件（由 NeMo-Relay 观察者遥测消费）；可用的轮次中转向——桌面辅助 + 可信注入（[#29724](https://github.com/NousResearch/hermes-agent/pull/29724)、[#40240](https://github.com/NousResearch/hermes-agent/pull/40240) — @teknium1） ### 提供商与模型支持 - 新模型：`z-ai/glm-5.2`（已验证 1M 上下文，OpenRouter + Nous）、`anthropic/claude-fable-5`、`laguna-m.1` + `nemotron-3-ultra`、xAI Composer 2.5 在 OAuth 选择器中；将 xAI 默认设置为 `grok-build-0.1`（[#47391](https://github.com/NousResearch/hermes-agent/pull/47391)、[#45695](https://github.com/NousResearch/hermes-agent/pull/45695)、[#42979](https://github.com/NousResearch/hermes-agent/pull/42979)、[#42629](https://github.com/NousResearch/hermes-agent/pull/42629)、[#47908](https://github.com/NousResearch/hermes-agent/pull/47908)、[#47371](https://github.com/NousResearch/hermes-agent/pull/47371) — @teknium1） - 模型选择器：刷新模型控件以清除过时的缓存；将 Nous 推荐模型持久化到磁盘 + 在门户失败时回退；在更新时从检出中填充目录磁盘缓存；MiniMax-M3 报告真实 1M 上下文（[#48691](https://github.com/NousResearch/hermes-agent/pull/48691)、[#42628](https://github.com/NousResearch/hermes-agent/pull/42628)、[#42614](https://github.com/NousResearch/hermes-agent/pull/42614)、[#43338](https://github.com/NousResearch/hermes-agent/pull/43338) — @teknium1） - Anthropic 自适应模型：默认使用现代思维契约；绝不发送 `reasoning` 字段；将 `reasoning_effort` 路由到冗长；对于非常昂贵的选择要求确认（[#42991](https://github.com/NousResearch/hermes-agent/pull/42991)、[#43012](https://github.com/NousResearch/hermes-agent/pull/43012)、[#43436](https://github.com/NousResearch/hermes-agent/pull/43436)、[#43391](https://github.com/NousResearch/hermes-agent/pull/43391) — @teknium1） - 认证：从池中自动检测 OpenRouter 凭据；在添加/重新认证时保持 Codex OAuth 池账户区分；跨配置文件解析 xAI OAuth + 将旋转后的令牌写回根；为由 `model.default_headers` 自定义 OpenAI 兼容提供商（[#42263](https://github.com/NousResearch/hermes-agent/pull/42263)、[#42316](https://github.com/NousResearch/hermes-agent/pull/42316)、[#46614](https://github.com/NousResearch/hermes-agent/pull/46614)、[#41096](https://github.com/NousResearch/hermes-agent/pull/41096) — @teknium1） - Bedrock 在 IAM 拒绝流式变体时回退到非流式 `InvokeModel`；Ollama 默认 `max_tokens=65536`；将模型拒绝表现为 `content_filter`（[#44293](https://github.com/NousResearch/hermes-agent/pull/44293)、[#41694](https://github.com/NousResearch/hermes-agent/pull/41694)、[#46013](https://github.com/NousResearch/hermes-agent/pull/46013) — @teknium1） ### 会话、状态与多代理 - 可选的**最大会话上限**；在 CLI 退出和轮换时删除空会话；用于压缩轮换的 ACP 会话来源元数据（[#42389](https://github.com/NousResearch/hermes-agent/pull/42389)、[#43855](https://github.com/NousResearch/hermes-agent/pull/43855)、[#41724](https://github.com/NousResearch/hermes-agent/pull/41724) — @teknium1） - 委托：通过端点标识解析自定义端点子代理池；移除默认子代理墙钟超时；停止子代理完成行泄漏到父 CLI 显示中（[#41730](https://github.com/NousResearch/hermes-agent/pull/41730)、[#45149](https://github.com/NousResearch/hermes-agent/pull/45149)、[#44223](https://github.com/NousResearch/hermes-agent/pull/44223) — @teknium1） - 看板：在 `kanban_create` 上配置门控自动订阅；嵌入式调度器的机器全局单例锁；为工作人员固定分配的配置文件工具集；在工作人员仍然存活时保持回收（[#48635](https://github.com/NousResearch/hermes-agent/pull/48635)、[#49068](https://github.com/NousResearch/hermes-agent/pull/49068)、[#45590](https://github.com/NousResearch/hermes-agent/pull/45590)、[#49064](https://github.com/NousResearch/hermes-agent/pull/49064) — @teknium1） - 内存：可配置的 Hindsight 保留观察范围；OpenViking 设置 UX；Honcho 网关门控身份树；Supermemory 会话级别摄取（[#46611](https://github.com/NousResearch/hermes-agent/pull/46611)、[#48262](https://github.com/NousResearch/hermes-agent/pull/48262)、[#44431](https://github.com/NousResearch/hermes-agent/pull/44431)、[#38756](https://github.com/NousResearch/hermes-agent/pull/38756) — @teknium1、@alt-glitch） ## 📱 消息平台（网关） ### 新渠道 - **通过 Photon Spectrum 的 iMessage**——`hermes photon login`（设备码 OAuth）、gRPC 原生渠道（无 webhook）、Markdown 渲染、表情符号反应、通过 spectrum-ts 的传出媒体（[#32348](https://github.com/NousResearch/hermes-agent/pull/32348)、[#42582](https://github.com/NousResearch/hermes-agent/pull/42582)、[#44713](https://github.com/NousResearch/hermes-agent/pull/44713)、[#42397](https://github.com/NousResearch/hermes-agent/pull/42397) — @teknium1） - **WhatsApp Business Cloud API** 适配器（官方，无桥接进程）（[#44331](https://github.com/NousResearch/hermes-agent/pull/44331)、[#43921](https://github.com/NousResearch/hermes-agent/pull/43921) — @jquesnelle、@teknium1） - **SimpleX**——群组、原生附件、文本批处理、自动接受；**Raft** 内置平台插件，带活动钩子（[#42584](https://github.com/NousResearch/hermes-agent/pull/42584)、[#48210](https://github.com/NousResearch/hermes-agent/pull/48210) — @teknium1） ### 网关核心与渲染 - 在 Markdown 平台上将终端工具调用渲染为原生 bash 代码块；聊天中裸露的围栏代码块；可选的 LLM 上下文消息时间戳；可配置的 `tool_progress_grouping`（[#41215](https://github.com/NousResearch/hermes-agent/pull/41215)、[#42576](https://github.com/NousResearch/hermes-agent/pull/42576)、[#47253](https://github.com/NousResearch/hermes-agent/pull/47253)、[#47228](https://github.com/NousResearch/hermes-agent/pull/47228) — @teknium1） - Telegram：Bot API 10.1 富消息（现在默认启用，可退出）；选择加入的在线/离线机器人状态指示器；停止截断长流式响应；MarkdownV2 用于进度编辑；在下载前门控超大的语音/音频（[#44829](https://github.com/NousResearch/hermes-agent/pull/44829)、[#45584](https://github.com/NousResearch/hermes-agent/pull/45584)、[#49134](https://github.com/NousResearch/hermes-agent/pull/49134)、[#43761](https://github.com/NousResearch/hermes-agent/pull/43761)、[#44245](https://github.com/NousResearch/hermes-agent/pull/44245) — @teknium1） - Discord：传播 `role_authorized` 以便 `DISCORD_ALLOWED_ROLES` 端到端工作；从运行时网关任务退出中恢复；在连接失败时取消 `_bot_task`；在回复后停止打字（[#43327](https://github.com/NousResearch/hermes-agent/pull/43327)、[#44383](https://github.com/NousResearch/hermes-agent/pull/44383)、[#44432](https://github.com/NousResearch/hermes-agent/pull/44432)、[#44836](https://github.com/NousResearch/hermes-agent/pull/44836) — @teknium1） - Slack：当 `reply_in_thread=false` 时限定顶级频道消息范围；线程审批 UX（块大小溢出 + 类型前缀）；使视频附件可供代理使用；`register_slack_action_handler` 插件 API（[#41703](https://github.com/NousResearch/hermes-agent/pull/41703)、[#43444](https://github.com/NousResearch/hermes-agent/pull/43444)、[#45512](https://github.com/NousResearch/hermes-agent/pull/45512)、[#44664](https://github.com/NousResearch/hermes-agent/pull/44664) — @teknium1） - 包括回复的媒体附件；文档附件在 Signal/Email/SimpleX/Teams 上分类为 DOCUMENT；WhatsApp 重启过时的桥接进程；Matrix 房间上下文隔离；QQbot CPU 旋转修复；微信速率限制断路器（[#46107](https://github.com/NousResearch/hermes-agent/pull/46107)、[#44695](https://github.com/NousResearch/hermes-agent/pull/44695)、[#44205](https://github.com/NousResearch/hermes-agent/pull/44205)、[#18505](https://github.com/NousResearch/hermes-agent/pull/18505)、[#40574](https://github.com/NousResearch/hermes-agent/pull/40574)、[#41718](https://github.com/NousResearch/hermes-agent/pull/41718) — @teknium1、@banditburai） ## 🖥️ CLI、TUI 和设置 - `/version` 斜杠命令；`/billing` 交互式终端计费（TUI + CLI）；在状态栏上显示自上次最终代理响应以来的时间；将已解决的审批/澄清提示持久化到回滚缓存中（[#40214](https://github.com/NousResearch/hermes-agent/pull/40214)、[#45449](https://github.com/NousResearch/hermes-agent/pull/45449)、[#44265](https://github.com/NousResearch/hermes-agent/pull/44265)、[#44702](https://github.com/NousResearch/hermes-agent/pull/44702) — @teknium1） - 锁定 hermes 工作树，以便并发进程无法破坏它们；在列表/显示中显示自定义配置文件别名名称；从任何来源克隆配置文件（[#48699](https://github.com/NousResearch/hermes-agent/pull/48699)、[#40371](https://github.com/NousResearch/hermes-agent/pull/40371)、[#45630](https://github.com/NousResearch/hermes-agent/pull/45630) — @teknium1） - 首次联系时选择加入的结构化配置文件构建路径；每个平台可配置的系统提示提示；可配置的后台内存/技能通知（[#41114](https://github.com/NousResearch/hermes-agent/pull/41114)、[#48630](https://github.com/NousResearch/hermes-agent/pull/48630)、[#47226](https://github.com/NousResearch/hermes-agent/pull/47226) — @teknium1） - TUI：交互式插件中心启用/禁用覆盖；终端标题栏中的会话名称；直接绘制审批/澄清/sudo/秘密模态框（不经过节流）；包装长审批命令而不是截断（[#42965](https://github.com/NousResearch/hermes-agent/pull/42965)、[#43188](https://github.com/NousResearch/hermes-agent/pull/43188)、[#41155](https://github.com/NousResearch/hermes-agent/pull/41155)、[#44691](https://github.com/NousResearch/hermes-agent/pull/44691) — @teknium1） - TTS：Gemini 角色提示 + 音频标签；xAI 自动语音标签 + 速度/流式旋钮；Piper speaker_id；Telegram 自动 TTS 的 OGG（[#43442](https://github.com/NousResearch/hermes-agent/pull/43442)、[#49061](https://github.com/NousResearch/hermes-agent/pull/49061)、[#49062](https://github.com/NousResearch/hermes-agent/pull/49062)、[#49060](https://github.com/NousResearch/hermes-agent/pull/49060)、[#41644](https://github.com/NousResearch/hermes-agent/pull/41644) — @teknium1） ## 🔧 工具系统、技能与 MCP - **`image_generate` 中的图像到图像/编辑**，覆盖所有后端；将图像缩小到提供商维度限制（[#48705](https://github.com/NousResearch/hermes-agent/pull/48705)、[#45979](https://github.com/NousResearch/hermes-agent/pull/45979) — @teknium1） - MCP：目录中的官方 **Unreal Engine 5.8** MCP 服务器；**引导处理器**，以便 MCP 服务器可以在工具调用中途提示确认（支付/OAuth），无论在哪个界面上拥有会话——CLI/TUI/Telegram/Slack；在轮次之间将延迟连接的 MCP 工具暴露给代理（缓存安全）；对短 TTL HTTP 会话进行保活 ping；在探测前阻止外形像数据泄露/可疑的 stdio 配置；能力门控 `tools/list`，以便仅提示的服务器可以连接；保留 stdio argv 透传 + Windows 环境变量（[#48397](https://github.com/NousResearch/hermes-agent/pull/48397)、[#49203](https://github.com/NousResearch/hermes-agent/pull/49203)、[#49208](https://github.com/NousResearch/hermes-agent/pull/49208)、[#49221](https://github.com/NousResearch/hermes-agent/pull/49221)、[#46083](https://github.com/NousResearch/hermes-agent/pull/46083)、[#44550](https://github.com/NousResearch/hermes-agent/pull/44550)、[#44324](https://github.com/NousResearch/hermes-agent/pull/44324) — @teknium1、@lgalabru） - 技能：`simplify-code` 技能（并行的 3 代理代码审查和清理）+ 基于 Chesterton's Fence 的风险分层应用；查找和差异用户修改的捆绑技能；可选的**支付**技能（Stripe Link、MPP、Projects）；基于 CLI 的商店技能；每个来源的实时浏览进度（[#41691](https://github.com/NousResearch/hermes-agent/pull/41691)、[#49070](https://github.com/NousResearch/hermes-agent/pull/49070)、[#48286](https://github.com/NousResearch/hermes-agent/pull/48286)、[#31343](https://github.com/NousResearch/hermes-agent/pull/31343)、[#47309](https://github.com/NousResearch/hermes-agent/pull/47309)、[#43398](https://github.com/NousResearch/hermes-agent/pull/43398) — @teknium1、@colinwren-stripe） - 策展器：使技能整合为选择加入（修剪保持默认开启）（[#47840](https://github.com/NousResearch/hermes-agent/pull/47840) — @teknium1） - 插件：从仓库内的子目录安装；接受 `hermes plugins install` 中浏览器粘贴的 GitHub URL；`session:compress` 生命周期事件 + `thread_id`/`chat_type` 在 agent:start/end 上下文中（[#42963](https://github.com/NousResearch/hermes-agent/pull/42963)、[#33539](https://github.com/NousResearch/hermes-agent/pull/33539)、[#47252](https://github.com/NousResearch/hermes-agent/pull/47252)、[#41672](https://github.com/NousResearch/hermes-agent/pull/41672) — @teknium1） - 内存/技能**写入审批**门控（默认关闭）——布尔值 `write_approval` 替代三态 `write_mode`（[#38199](https://github.com/NousResearch/hermes-agent/pull/38199)、[#43354](https://github.com/NousResearch/hermes-agent/pull/43354) — @teknium1） ## 🌐 舰队、中继与自动化 - **托管范围**——管理员固定、用户不可更改的配置和秘密，来自 root 拥有的 `/etc/hermes`（[#49098](https://github.com/NousResearch/hermes-agent/pull/49098) — @teknium1） - **通过一个网关进程多路复用所有配置文件**（选择加入）（[#48273](https://github.com/NousResearch/hermes-agent/pull/48273) — @benbarclay） - **可插拔 CronScheduler** + Chronos 托管 cron 提供商（缩放到零）（[#48275](https://github.com/NousResearch/hermes-agent/pull/48275) — @benbarclay） - **自动化蓝图**——每个界面上的参数化自动化模板（[#41309](https://github.com/NousResearch/hermes-agent/pull/41309) — @teknium1） - 网关到网关中继（阶段 0-3）：中继适配器 + 能力描述符，连接器⇄网关渠道认证 + 签名 HTTP 入站 + 注册 CLI，仅 WS 入站，托管启动自我配置客户端（[#48078](https://github.com/NousResearch/hermes-agent/pull/48078)、[#48147](https://github.com/NousResearch/hermes-agent/pull/48147)、[#48294](https://github.com/NousResearch/hermes-agent/pull/48294)、[#48242](https://github.com/NousResearch/hermes-agent/pull/48242) — @teknium1） ## 🐳 Docker、Nix 和安装器 - s6：直接检测 supervisor 以重启网关；注册配置文件网关而不自动启动；持久化所需状态；清除过时的日志锁（[#46290](https://github.com/NousResearch/hermes-agent/pull/46290)、[#46266](https://github.com/NousResearch/hermes-agent/pull/46266)、[#46292](https://github.com/NousResearch/hermes-agent/pull/46292)、[#46289](https://github.com/NousResearch/hermes-agent/pull/46289) — @teknium1） - Docker：优化镜像大小（.dockerignore、删除开发依赖、拆分层）；预安装矩阵依赖；受监督的网关使用 `--replace`；加固托管安装树以防止自修改（[#38749](https://github.com/NousResearch/hermes-agent/pull/38749)、[#42413](https://github.com/NousResearch/hermes-agent/pull/42413)、[#47555](https://github.com/NousResearch/hermes-agent/pull/47555)、[#47490](https://github.com/NousResearch/hermes-agent/pull/47490) — @teknium1） - Nix：冷 npm 构建修复 + 自动修复锁文件工作流；通过 `importNpmLock` 实现无哈希 npm 依赖；锁定 Electron 40.10.2 后刷新 npmDepsHash（[#41867](https://github.com/NousResearch/hermes-agent/pull/41867)、[#48883](https://github.com/NousResearch/hermes-agent/pull/48883)、[#48457](https://github.com/NousResearch/hermes-agent/pull/48457) — @teknium1） - 安装器：在自动暂存前清除未合并的 git 索引；将安装方法标记限定到代码树（[#45515](https://github.com/NousResearch/hermes-agent/pull/45515)、[#48188](https://github.com/NousResearch/hermes-agent/pull/48188) — @teknium1） ## 🔒 安全与可靠性 - 在自有策略网关适配器上安全失败；在 Slack/Feishu/Discord 上未设置许可列表时，审批按钮认证安全失败（[#45634](https://github.com/NousResearch/hermes-agent/pull/45634)、[#41226](https://github.com/NousResearch/hermes-agent/pull/41226) — @teknium1） - 在请求调试转储中编辑秘密；向公共状态隐瞒主机元数据；在探测前阻止外形像数据泄露/可疑的 MCP stdio 配置（[#46637](https://github.com/NousResearch/hermes-agent/pull/46637)、[#45642](https://github.com/NousResearch/hermes-agent/pull/45642)、[#46083](https://github.com/NousResearch/hermes-agent/pull/46083) — @teknium1） - 关闭 shell 转义拒绝列表绕过 + 在缺少审批模块时安全失败；在启动 cua-driver MCP 前清理操作员环境；为 cron 作业脚本子进程清理环境；限制 TodoStore 内容长度/计数；扫描 REST cron 提示以确保与代理工具的奇偶性（[#40591](https://github.com/NousResearch/hermes-agent/pull/40591)、[#48423](https://github.com/NousResearch/hermes-agent/pull/48423)、[#49207](https://github.com/NousResearch/hermes-agent/pull/49207)、[#41648](https://github.com/NousResearch/hermes-agent/pull/41648)、[#41335](https://github.com/NousResearch/hermes-agent/pull/41335) — @teknium1、@kshitijk4poor） - 升级 urllib3 和 PyJWT 以清除 CVE；Langfuse 编辑 base64 数据 URI 而不是截断为无效的 base64（[#40179](https://github.com/NousResearch/hermes-agent/pull/40179)、[#43322](https://github.com/NousResearch/hermes-agent/pull/43322) — @teknium1） ## 🪟 Windows - 通过 ConPTY（`win_pty_bridge`）实现仪表板 `/chat` 标签页 + 测试；为 uv 安装解析 PowerShell 主机而不是裸 `powershell`；通过绝对路径解析 `powershell.exe` 以免桌面安装卡住（[#42251](https://github.com/NousResearch/hermes-agent/pull/42251)、[#48341](https://github.com/NousResearch/hermes-agent/pull/48341)、[#40927](https://github.com/NousResearch/hermes-agent/pull/40927) — @teknium1） - 修复过时的 winget 注册 + 刷新/合并 PATH；在重新创建 venv 前杀死 hermes 以释放 `_bcrypt.pyd` 锁；当环境变量过时时从注册表读取 HERMES_HOME；在更新修复期间隔离正在运行的 `hermes.exe`（[#44084](https://github.com/NousResearch/hermes-agent/pull/44084)、[#45120](https://github.com/NousResearch/hermes-agent/pull/45120)、[#46772](https://github.com/NousResearch/hermes-agent/pull/46772)、[#40409](https://github.com/NousResearch/hermes-agent/pull/40409) — @teknium1） - JOB 分离观察者可靠性 + 状态 --deep 探测；处理 Windows PTY 标准输入 + 分离的 WS 帧；将子进程输出解码为 UTF-8；原生 Windows 上的确认模态框（[#40909](https://github.com/NousResearch/hermes-agent/pull/40909)、[#41953](https://github.com/NousResearch/hermes-agent/pull/41953)、[#44328](https://github.com/NousResearch/hermes-agent/pull/44328)、[#42419](https://github.com/NousResearch/hermes-agent/pull/42419) — @teknium1） ## 🐛 值得注意的错误修复 - 百分比编码非 ASCII URL 组件；清理 FTS5 查询中的 `:` 以便冒号搜索不会静默返回空（[#41430](https://github.com/NousResearch/hermes-agent/pull/41430)、[#40653](https://github.com/NousResearch/hermes-agent/pull/40653) — @teknium1） - 通过崩溃恢复持久化保留多模态用户内容；在提供商同步前展平多模态内容；从压缩器输入中剥离 MEDIA 指令（[#47907](https://github.com/NousResearch/hermes-agent/pull/47907)、[#44738](https://github.com/NousResearch/hermes-agent/pull/44738)、[#44708](https://github.com/NousResearch/hermes-agent/pull/44708) — @teknium1） - 在真正的 Nous 429 上重新进入重试循环以便后备保护运行；将 Nous 标签限定到 Nous 辅助调用；在免费模型上抑制“信用访问暂停”通知（[#45136](https://github.com/NousResearch/hermes-agent/pull/45136)、[#45801](https://github.com/NousResearch/hermes-agent/pull/45801)、[#43669](https://github.com/NousResearch/hermes-agent/pull/43669) — @teknium1） - Cron：在无技能任务中不严格扫描脚本注入的输出；将每个任务的提供商“custom”解析为 `providers.custom` 而不是 codex；在容器重启后修复 cron 所有权（[#43223](https://github.com/NousResearch/hermes-agent/pull/43223)、[#43505](https://github.com/NousResearch/hermes-agent/pull/43505)、[#41976](https://github.com/NousResearch/hermes-agent/pull/41976) — @teknium1） - *（此窗口关闭了 300+ 个 issue；完整的按区域修复列表详尽——以上是影响最大的修复。）* ## ↩️ 此窗口已回滚（未发布） - `html-artifact` 技能 + 草图/架构图/概念图折叠（[#48899](https://github.com/NousResearch/hermes-agent/pull/48899)）——已回滚（[#49053](https://github.com/NousResearch/hermes-agent/pull/49053)）；在主干上不存在。 - Cron 每个任务配置文件支持已回滚（[#43956](https://github.com/NousResearch/hermes-agent/pull/43956)）；一个 nix patchPhase 解决方法已回滚（[#42151](https://github.com/NousResearch/hermes-agent/pull/42151)）。 ## 👥 贡献者 非常感谢所有为此次发布做出贡献的人——**245 位贡献者**，涵盖提交、共同作者和拯救的 PR。 ### 核心 @teknium1 ### 顶级社区贡献者（按合并 PR 数量） - @OutThisLife — 92 个 PR（桌面应用成熟度（快捷键、通知、监视窗口、主题）） - @kshitijk4poor — 60 个 PR（入门、模型选择器、cron 环境清理） - @xxxigm — 27 个 PR（桌面和网关修复） - @benbarclay — 23 个 PR（网关多路复用、Chronos cron、仪表板认证） - @helix4u — 21 个 PR（网关和安装器可靠性） - @austinpickett — 19 个 PR（仪表板和桌面 UX） - @alt-glitch — 14 个 PR（使用感知信用、Supermemory） - @ethernet8023 — 14 个 PR（桌面构建管道和 Linux/Windows） - @liuhao1024 — 5 个 PR（会话生命周期修复） ### 所有贡献者（按字母顺序） @0z1-ghb、@0xdany、@0xneobyte、@0xyg3n、@1960697431、@895252509、@achaljhawar、@Adolanium、@AhmetArif0、@AIalliAI、 @aimable100、@AJ、@ak2k、@alarcritty、@AlchemistChaos、@aldoeliacim、@alelpoan、@AlexanderBFoley、 @alfred-smith-0、@ali-nld、@alt-glitch、@am423、@ameobius、@AMIK-coorporations、@annguyenNous、@ArcanePivot、 @ARegalado1、@asdlem、@ashishpatel26、@austinpickett、@banditburai、@barronlroth、@Bartok9、@basilalshukaili、 @bbednarski9、@bcsmith528、@benbarclay、@benegessarit、@benfrank241、@bionicbutterfly13、@BlackishGreen33、 @blut-agent、@bmoore210、@bpasquini、@briandevans、@BROCCOLO1D、@capt-marbles、@ccook1963、@Cdddo、@channkim、 @ChasLui、@chimpera、@chromalinx、@CiarasClaws、@claytonchew、@cnfi、@colinwren-stripe、@cresslank、 @cyb0rgk1tty、@dangelo352、@davidgut1982、@deaneeth、@definitelynotguru、@Diyoncrz18、@draix、@dschnurbusch、 @Dusk1e、@dusterbloom、@ehz0ah、@emozilla、@enesilhaydin、@Erosika、@ethernet8023、@Evisolpxe、@firefly、 @flooryyyy、@flyinhigh、@foras910521-lab、@Frowtek、@ft-ioxcs、@fyzanshaik、@Ganesh0690、@gauravsaxena1997、 @giladbau、@glesperance、@GodsBoy、@goku94123、@H-Ali13381、@HaozheZhang6、@haran2001、@harshitAgr、@hbentel、 @helix4u、@HeLLGURD、@Hermes、@huangxun375-stack、@iamlukethedev、@ianculling、@IAvecilla、@iborazzi、 @infinitycrew39、@islam666、@ITheEqualizer、@itsflownium、@Jaaneek、@james47kjv、@jeeves-assistant、 @jeffrobodie-glitch、@JezzaHehn、@jiangkoumo、@jimjsong、@JimLiu、@JimStenstrom、@jmsunseri、@JoaoMarcos44、 @joel611、@JoelJJohnson、@joerj123、@johnjacobkenny、@jooray、@joshuadow、@jplew、@justinbao19、@Justlrnal4、 @Kailigithub、@kamonspecial、@kdunn926、@Kenmege、@Kewe63、@kmccammon、@konsisumer、@kristianvast、 @kshitijk4poor、@kyssta-exe、@l37525778-coder、@LaPhilosophie、@leo4226、@LeonSGP43、@liuhao1024、@Llugaes、 @loongfay、@loongzhao、@lsaether、@m4dni5、@manishbyatroy、@MaxFreedomPollard、@maxmilian、@maxtrigify、 @mnajafian-nv、@mohamedorigami-jpg、@mollusk、@MrDiamondBallz、@mssteuer、@mvanhorn、@naqerl、@Nea74、 @necoweb3、@nepenth、@nicoloboschi、@NormallyGaussian、@OmarB97、@omegazheng、@OndrejDrapalik、@OutThisLife、 @oxngon、@OYLFLMH、@Paperclip、@paulb26、@pengyuyanITYU、@PhilipAD、@pinguarmy、@plcunha、@ProgramCaiCai、 @psionic73、@qin-ctx、@qingshan89、@Que0x、@qWaitCrypto、@r266-tech、@randomsnowflake、@rbrtbn、@rewbs、 @rio-jeong、@Rivuza、@rob-maron、@rodboev、@ruangraung、@RyTsYdUp、@Sahil-SS9、@salesondemandio、@sanidhyasin、 @sarvesh1327、@sdyckjq-lab、@shannonsands、@SHL0MS、@simpolism、@sitkarev、@skyc1e、@skylarbpayne、@SNooZyy2、 @Spaceman-Spiffy、@srojk34、@sweetcornna、@synapsesx、@Tamaz-sujashvili、@tangtaizong666、@temalo、@tfournet、 @thedavidweng、@TheGardenGallery、@tim404x、@tomekpanek、@Tranquil-Flow、@tt-a1i、@tuancookiez-hub、 @underthestars-zhy、@Veritas-7、@victor-kyriazakos、@wesleysimplicio、@WolframRavenwolf、@WompaJango、@x1erra、 @xiaoxinova、@xtymac、@xushibo、@XVVH、@xxchan、@xxxigm、@xy200303、@y0shua1ee、@yanxue06、@yatesjalex、 @YLChen-007、@yoniebans、@youjunxiaji、@yubingz、@zakame、@zapabob、@zccyman、@zimigit2020、@ziwon、@zwcf5200、 @zxcasongs。 --- **完整更新日志：** [v2026.6.5...v2026.6.19](https://github.com/NousResearch/hermes-agent/compare/v2026.6.5...v2026.6.19) ```

# Hermes Agent v0.17.0 (v2026.6.19) **Release Date:** June 19, 2026 **Since v0.16.0:** ~1,475 commits · ~800 merged PRs · 1,693 files changed · 235,390 insertions · 50,730 deletions · 300+ issues closed · 245 community contributors > **The Reach Release.** v0.16.0 put Hermes on your desktop. v0.17.0 is about how far that reach extends — across new places to talk to it, deeper into the tools you already use, and out to the people running Hermes for a team. Hermes reached two new channels (iMessage via Photon, and the Raft agent network), the desktop app gained substantial new capability, subagents can now run in the background, image generation learned to edit, and Cursor's Composer model is reachable through an xAI Grok subscription. The dashboard got a full profile builder and secure login, the Skills Hub browser was rehauled, the `memory` tool got a major upgrade, and the curator stopped spending aux-model budget on every routine run. 300+ issues closed ride along, plus a security round. ## ✨ Highlights - **Hermes reaches iMessage — Photon Spectrum, no Mac relay required** — There's now an iMessage platform plugin built on Photon's managed line pool. Run `hermes photon login`, authenticate with a device code, and Hermes can send and receive iMessage — no Mac sitting in a closet running a relay, no BlueBubbles bridge to babysit. It's positioned as the successor to BlueBubbles: free to start, nothing to self-host. If your friends and family live in the blue bubbles, Hermes lives there now too. ([#32348](https://github.com/NousResearch/hermes-agent/pull/32348), [#42582](https://github.com/NousResearch/hermes-agent/pull/42582), [#44713](https://github.com/NousResearch/hermes-agent/pull/44713) — @teknium1) - **Raft — Hermes joins the Raft agent network as a gateway channel** — A new bundled Raft platform adapter lets Hermes connect to [Raft](https://raft.build) as an external agent through a wake-channel bridge. Set `RAFT_PROFILE`, run the bridge, and Raft can wake Hermes to handle messages — with a privacy-by-contract design where wake payloads carry only metadata (event IDs, timestamps), never message bodies. Another surface where Hermes can show up and do work. ([#48210](https://github.com/NousResearch/hermes-agent/pull/48210) — @xxchan, @teknium1) - **A substantially more capable desktop app** — v0.16.0 shipped the desktop app; v0.17.0 deepened it across dozens of PRs. Rebindable keyboard shortcuts, native OS notifications with per-type toggles, live subagent **watch-windows** that stream a delegated agent's activity into its own pane, a composer model selector with per-model presets, automatic RTL/bidi text direction, a resizable VS Code-themed terminal pane, per-thread composer drafts, and the ability to install **any VS Code Marketplace theme** directly into the app. The desktop is now a serious daily driver, not a preview. ([#45866](https://github.com/NousResearch/hermes-agent/pull/45866), [#40660](https://github.com/NousResearch/hermes-agent/pull/40660), [#47060](https://github.com/NousResearch/hermes-agent/pull/47060), [#46959](https://github.com/NousResearch/hermes-agent/pull/46959), [#43292](https://github.com/NousResearch/hermes-agent/pull/43292), [#44596](https://github.com/NousResearch/hermes-agent/pull/44596) — @OutThisLife, @teknium1) - **Background / async subagents — delegate work and keep going** — `delegate_task(background=true)` now dispatches a subagent that runs in the background and returns a handle immediately. You and the model keep working while it churns, and the full result re-enters the conversation as a new turn the moment it finishes. Kick off a long research dive or a multi-step build, then carry on with something else instead of sitting blocked waiting on it. ([#40946](https://github.com/NousResearch/hermes-agent/pull/40946), [#46968](https://github.com/NousResearch/hermes-agent/pull/46968) — @teknium1) - **Edit images, not just generate them — image-to-image in `image_generate`** — `image_generate` can now edit and transform a source image, not only create one from scratch. Pass an existing image and a prompt and it routes to the backend's edit endpoint (same tool, same pattern as `video_generate`), across every supported image provider. "Make this logo blue," "remove the background," "turn this sketch into a render" — all from the tool you already use. ([#48705](https://github.com/NousResearch/hermes-agent/pull/48705) — @teknium1) - **Automation Blueprints — schedule things without learning cron** — Pick an automation by name and Hermes asks you for what it needs — no cron syntax, no `slot=value` typing. One blueprint definition renders natively on every surface: a form in the dashboard, a slash command in the CLI/TUI/messenger, a conversation with the agent, an entry in the docs catalog. "Daily news briefing at 8am" becomes a thing you set up by answering questions, not by memorizing `0 8 * * *`. ([#41309](https://github.com/NousResearch/hermes-agent/pull/41309) — @teknium1) - **Cursor's Composer model, through your xAI Grok subscription** — `grok-composer-2.5-fast` is now in the xAI OAuth model picker, with its context window reconciled to the full 200k. Composer is the fast coding model behind Cursor — and if you have an xAI Grok subscription, you can now point Hermes at it directly over OAuth, no separate API key. Your Grok plan, Hermes's agent loop, Composer's coding speed. ([#47908](https://github.com/NousResearch/hermes-agent/pull/47908), [#6f89e17](https://github.com/NousResearch/hermes-agent/pull/47371) — @teknium1) - **Full profile builder in the dashboard** — Build a complete Hermes profile from the browser — pick its model, choose its skills, attach its MCP servers — without hand-editing `config.yaml`. The dashboard also unified multi-profile management into one machine-wide view with a global profile switcher, so you manage every profile from a single place. ([#39084](https://github.com/NousResearch/hermes-agent/pull/39084), [#44007](https://github.com/NousResearch/hermes-agent/pull/44007) — @teknium1) - **Skills Hub browser rehaul** — The dashboard's Skills Hub got a ground-up rework: connected hubs, a Featured section, full skill previews before you install, and a security scan on each skill. Browsing and installing skills from the trusted taps (OpenAI, Anthropic, HuggingFace, NVIDIA) is now a real browsing experience, not a flat list. ([#40384](https://github.com/NousResearch/hermes-agent/pull/40384), [#43398](https://github.com/NousResearch/hermes-agent/pull/43398) — @teknium1) - **The `memory` tool got a major upgrade — atomic batch operations** — The `memory` tool gained an `operations` array that applies a batch of add/replace/remove edits **atomically against the final character budget**. The model can free up space and add new entries in a single call — even when an add alone would overflow the budget — collapsing what used to be a fragile multi-turn dance into one reliable operation. Memory updates are now faster and far less likely to fail mid-edit. ([#48507](https://github.com/NousResearch/hermes-agent/pull/48507) — @teknium1) - **Secure dashboard login** — The dashboard's authentication was hardened: every token-required endpoint now correctly returns 401 behind the OAuth gate, websocket auth uses the served dashboard token, and a warning fires when a `public_url` override is silently rejected. Exposing your dashboard to the network is safer by default. ([#42578](https://github.com/NousResearch/hermes-agent/pull/42578), [#42578](https://github.com/NousResearch/hermes-agent/pull/43214) — @benbarclay, @teknium1) - **Official WhatsApp Business Cloud API adapter** — Alongside the existing Baileys bridge, Hermes now speaks the **official** WhatsApp Business Cloud API — Meta's first-party, hosted, no-bridge-process path. Point it at your Business API credentials and Hermes talks WhatsApp through the supported channel, with no QR-scanning bridge process to keep alive. ([#44331](https://github.com/NousResearch/hermes-agent/pull/44331), [#43921](https://github.com/NousResearch/hermes-agent/pull/43921) — @jquesnelle, @teknium1) - **Rich text for Telegram — Bot API 10.1 rich messages** — Telegram replies now render as proper rich messages via Bot API 10.1: better formatting, cleaner long-message handling, native markup instead of flattened text. It's on by default with an opt-out, so your Telegram conversations look the way they should without any configuration. ([#44829](https://github.com/NousResearch/hermes-agent/pull/44829), [#45584](https://github.com/NousResearch/hermes-agent/pull/45584), [#45953](https://github.com/NousResearch/hermes-agent/pull/45953) — @teknium1) - **Curator cost optimization — no aux-model spend on routine runs** — The skill curator now prunes stale skills by default but no longer runs its LLM-powered consolidation pass unless you opt in (`curator.consolidate: true` or `hermes curator run --consolidate`). The deterministic inactivity sweep keeps running for free; the opinionated, aux-model-spending "build umbrella skills" fork is now off by default. Routine background curation costs you **zero tokens**. ([#47840](https://github.com/NousResearch/hermes-agent/pull/47840) — @teknium1) ## 🖥️ Hermes Desktop App ### New surfaces & UX - Rebindable keyboard shortcuts panel; native OS notifications with per-type toggles; curated turn-completion cue + dismissable error banners ([#40660](https://github.com/NousResearch/hermes-agent/pull/40660), [#45866](https://github.com/NousResearch/hermes-agent/pull/45866), [#42480](https://github.com/NousResearch/hermes-agent/pull/42480), [#47985](https://github.com/NousResearch/hermes-agent/pull/47985) — @OutThisLife, @teknium1) - Live subagent **watch-windows** — stream a delegated agent's activity into its own pane; composer status stack + editable prompts; open any chat in its own window; new-session-in-compact-window hotkey ([#47060](https://github.com/NousResearch/hermes-agent/pull/47060), [#44630](https://github.com/NousResearch/hermes-agent/pull/44630), [#43219](https://github.com/NousResearch/hermes-agent/pull/43219), [#46951](https://github.com/NousResearch/hermes-agent/pull/46951) — @OutThisLife, @teknium1) - Composer model selector + per-model presets + external-provider disconnect; surface every provider/model from `hermes model` in the GUI; unify provider list to one source; warn when a main-model switch leaves auxiliary tasks pinned elsewhere ([#46959](https://github.com/NousResearch/hermes-agent/pull/46959), [#40563](https://github.com/NousResearch/hermes-agent/pull/40563), [#49080](https://github.com/NousResearch/hermes-agent/pull/49080), [#40286](https://github.com/NousResearch/hermes-agent/pull/40286) — @teknium1, @OutThisLife) - Install **any VS Code Marketplace theme**; assignable themes per profile; window translucency slider; unified overlay design system + BrandMark + onboarding redesign ([#43292](https://github.com/NousResearch/hermes-agent/pull/43292), [#42286](https://github.com/NousResearch/hermes-agent/pull/42286), [#45086](https://github.com/NousResearch/hermes-agent/pull/45086), [#40708](https://github.com/NousResearch/hermes-agent/pull/40708) — @teknium1, @OutThisLife) - Resizable VS Code-themed terminal pane + palette polish; auto-detect RTL/bidi text direction in chat; Mac-style session switcher (^Tab / ^1-9); worktree-aware sidebar grouping; hover-reveal collapsed sidebars; messaging source folders in sidebar ([#42521](https://github.com/NousResearch/hermes-agent/pull/42521), [#44596](https://github.com/NousResearch/hermes-agent/pull/44596), [#43111](https://github.com/NousResearch/hermes-agent/pull/43111), [#45273](https://github.com/NousResearch/hermes-agent/pull/45273), [#41670](https://github.com/NousResearch/hermes-agent/pull/41670), [#41751](https://github.com/NousResearch/hermes-agent/pull/41751) — @OutThisLife) - Arrow-key history + queue editing in composer; expand full command inline from the approval bar; follow-streaming-at-bottom + jump-to-bottom button; first-class cron jobs in the sidebar + dashboard scheduler ([#40234](https://github.com/NousResearch/hermes-agent/pull/40234), [#44864](https://github.com/NousResearch/hermes-agent/pull/44864), [#45263](https://github.com/NousResearch/hermes-agent/pull/45263), [#40684](https://github.com/NousResearch/hermes-agent/pull/40684) — @OutThisLife, @teknium1) - Desktop pets — pop-out overlay + notifications ([#47938](https://github.com/NousResearch/hermes-agent/pull/47938) — @teknium1) - Full tool-backend config (pickers + per-backend settings) in Settings; run tool-backend post-setup installs from the GUI; uninstall the Chat GUI without removing the agent; Shift+click status-bar zap to toggle YOLO globally; `/browser connect` on a local gateway ([#41232](https://github.com/NousResearch/hermes-agent/pull/41232), [#40559](https://github.com/NousResearch/hermes-agent/pull/40559), [#40355](https://github.com/NousResearch/hermes-agent/pull/40355), [#41666](https://github.com/NousResearch/hermes-agent/pull/41666), [#47245](https://github.com/NousResearch/hermes-agent/pull/47245) — @teknium1, @OutThisLife) - Japanese + Traditional Chinese language switching ([#40114](https://github.com/NousResearch/hermes-agent/pull/40114)) - "Restart gateway" action (renamed from "Restart messaging") surfaced in the statusbar + on messaging save/toggle toasts; rendered logs are selectable/copyable ([#49094](https://github.com/NousResearch/hermes-agent/pull/49094) — @OutThisLife) ### Remote-gateway & multi-profile - **Remote media relay** — attach images/PDFs and display agent-written images over the network for the first time; remote-gateway file attachments via `file.attach` ([#41336](https://github.com/NousResearch/hermes-agent/pull/41336), [#42634](https://github.com/NousResearch/hermes-agent/pull/42634) — @teknium1) - Client + backend version buttons + remote-backend update flow; browse remote backend files; route global-remote profile REST calls; recover chat after sleep/wake by revalidating a stale remote backend ([#42181](https://github.com/NousResearch/hermes-agent/pull/42181), [#44326](https://github.com/NousResearch/hermes-agent/pull/44326), [#47011](https://github.com/NousResearch/hermes-agent/pull/47011), [#41350](https://github.com/NousResearch/hermes-agent/pull/41350) — @OutThisLife, @teknium1) - Multi-profile fallout cleanup — WS auth + cross-profile session reads; release profile backends before delete; scope session list/model switch/timer per session ([#44529](https://github.com/NousResearch/hermes-agent/pull/44529), [#42613](https://github.com/NousResearch/hermes-agent/pull/42613), [#41103](https://github.com/NousResearch/hermes-agent/pull/41103), [#41120](https://github.com/NousResearch/hermes-agent/pull/41120), [#41182](https://github.com/NousResearch/hermes-agent/pull/41182) — @OutThisLife, @teknium1) - Stream subagent activity into watch windows; keep streaming painting in unfocused secondary chat windows; recover stranded session windows ([#47060](https://github.com/NousResearch/hermes-agent/pull/47060), [#47919](https://github.com/NousResearch/hermes-agent/pull/47919), [#47655](https://github.com/NousResearch/hermes-agent/pull/47655) — @OutThisLife) ## 📊 Web Dashboard - Full-featured profile builder (model + skills + MCPs); unify multi-profile management — one machine dashboard + global profile switcher; profile-scoped skills & toolsets; session switcher panel on the Chat tab ([#39084](https://github.com/NousResearch/hermes-agent/pull/39084), [#44007](https://github.com/NousResearch/hermes-agent/pull/44007), [#43808](https://github.com/NousResearch/hermes-agent/pull/43808), [#49077](https://github.com/NousResearch/hermes-agent/pull/49077) — @teknium1) - Skills hub browser rehaul — connected hubs, featured, preview + security scan; SKILL.md editor on Skills page + attach-skill selector in cron modals; full per-MCP catalog detail; full tool-backend config in the GUI ([#40384](https://github.com/NousResearch/hermes-agent/pull/40384), [#44231](https://github.com/NousResearch/hermes-agent/pull/44231), [#48520](https://github.com/NousResearch/hermes-agent/pull/48520), [#40418](https://github.com/NousResearch/hermes-agent/pull/40418) — @teknium1) - Enable webhooks from the Webhooks page; idempotent `hermes dashboard register`; auto-restart gateway after Telegram QR onboarding; file browser; change UI font from the theme picker; reasoning-effort picker in the chat sidebar ([#44021](https://github.com/NousResearch/hermes-agent/pull/44021), [#42455](https://github.com/NousResearch/hermes-agent/pull/42455), [#43424](https://github.com/NousResearch/hermes-agent/pull/43424), [#43512](https://github.com/NousResearch/hermes-agent/pull/43512), [#41145](https://github.com/NousResearch/hermes-agent/pull/41145), [#49141](https://github.com/NousResearch/hermes-agent/pull/49141) — @teknium1) ## 🏗️ Core Agent & Architecture ### God-file refactor wave (run_agent.py / cli.py / gateway/run.py) - **`cli.py` main() 3297 → 954 lines** — extracted 28 subcommand parsers into `hermes_cli/subcommands/`, then promoted 9 closure handlers; 32 slash-command handlers → `CLICommandsMixin`; 18 model-flow wizard functions → `model_setup_flows`; agent-construction cluster → `CLIAgentSetupMixin` ([#41798](https://github.com/NousResearch/hermes-agent/pull/41798), [#41835](https://github.com/NousResearch/hermes-agent/pull/41835), [#41942](https://github.com/NousResearch/hermes-agent/pull/41942), [#42174](https://github.com/NousResearch/hermes-agent/pull/42174), [#42153](https://github.com/NousResearch/hermes-agent/pull/42153) — @teknium1) - **`gateway/run.py` 19157 → 15870 lines** — 42 slash-command handlers → `GatewaySlashCommandsMixin`; authorization cluster → `GatewayAuthorizationMixin`; kanban watcher loops → `GatewayKanbanWatchersMixin` ([#41886](https://github.com/NousResearch/hermes-agent/pull/41886), [#42159](https://github.com/NousResearch/hermes-agent/pull/42159), [#41849](https://github.com/NousResearch/hermes-agent/pull/41849) — @teknium1) - **`run_agent.py` turn loop** — extracted prologue into `TurnContext`, post-loop tail into `finalize_turn`, consolidated inner-retry-loop recovery flags into `TurnRetryState` ([#41778](https://github.com/NousResearch/hermes-agent/pull/41778), [#42169](https://github.com/NousResearch/hermes-agent/pull/42169), [#41828](https://github.com/NousResearch/hermes-agent/pull/41828) — @teknium1) ### Agent loop, prompt & tools - **`memory` batch operations** — atomic add/replace/remove array against the final char budget, so a single call can free space and add entries ([#48507](https://github.com/NousResearch/hermes-agent/pull/48507) — @teknium1) - **`search_files` lossless densification** — headroom evaluation report + the one densification improvement worth shipping (fewer tokens per result, same matches) ([#47866](https://github.com/NousResearch/hermes-agent/pull/47866) — @teknium1) - Removed the agent-callable `send_message` tool; coding-context posture across CLI/TUI/desktop/ACP; `read_file` extracts `.ipynb`/`.docx`/`.xlsx` to text ([#47856](https://github.com/NousResearch/hermes-agent/pull/47856), [#43316](https://github.com/NousResearch/hermes-agent/pull/43316), [#37082](https://github.com/NousResearch/hermes-agent/pull/37082) — @teknium1) - Context-file handling: configurable truncation limit + warnings; scale context-file cap to model window + point agent at the truncated file ([#47251](https://github.com/NousResearch/hermes-agent/pull/47251), [#47846](https://github.com/NousResearch/hermes-agent/pull/47846) — @teknium1) - Compression: temporal anchoring in compaction summaries; raise compaction trigger to 85% for gpt-5.5 on Codex OAuth ([#41102](https://github.com/NousResearch/hermes-agent/pull/41102), [#40957](https://github.com/NousResearch/hermes-agent/pull/40957) — @teknium1) - Adaptive middleware (consumed by NeMo-Relay observer telemetry); usable mid-turn steer — desktop affordance + trusted injection ([#29724](https://github.com/NousResearch/hermes-agent/pull/29724), [#40240](https://github.com/NousResearch/hermes-agent/pull/40240) — @teknium1) ### Provider & model support - New models: `z-ai/glm-5.2` (verified 1M context, OpenRouter + Nous), `anthropic/claude-fable-5`, `laguna-m.1` + `nemotron-3-ultra`, xAI Composer 2.5 in the OAuth picker; default xAI to `grok-build-0.1` ([#47391](https://github.com/NousResearch/hermes-agent/pull/47391), [#45695](https://github.com/NousResearch/hermes-agent/pull/45695), [#42979](https://github.com/NousResearch/hermes-agent/pull/42979), [#42629](https://github.com/NousResearch/hermes-agent/pull/42629), [#47908](https://github.com/NousResearch/hermes-agent/pull/47908), [#47371](https://github.com/NousResearch/hermes-agent/pull/47371) — @teknium1) - Model picker: Refresh-Models control to bust stale cache; persist Nous recommended-models to disk + fall back on Portal failure; seed catalog disk cache from checkout on update; MiniMax-M3 reports true 1M context ([#48691](https://github.com/NousResearch/hermes-agent/pull/48691), [#42628](https://github.com/NousResearch/hermes-agent/pull/42628), [#42614](https://github.com/NousResearch/hermes-agent/pull/42614), [#43338](https://github.com/NousResearch/hermes-agent/pull/43338) — @teknium1) - Anthropic adaptive models: default to modern thinking contract; never send `reasoning` field; route `reasoning_effort` to verbosity; require confirmation for very expensive selections ([#42991](https://github.com/NousResearch/hermes-agent/pull/42991), [#43012](https://github.com/NousResearch/hermes-agent/pull/43012), [#43436](https://github.com/NousResearch/hermes-agent/pull/43436), [#43391](https://github.com/NousResearch/hermes-agent/pull/43391) — @teknium1) - Auth: auto-detect OpenRouter credential from the pool; keep Codex OAuth pool accounts distinct on add/re-auth; resolve xAI OAuth across profiles + write rotated tokens back to root; honor `model.default_headers` for custom OpenAI-compatible providers ([#42263](https://github.com/NousResearch/hermes-agent/pull/42263), [#42316](https://github.com/NousResearch/hermes-agent/pull/42316), [#46614](https://github.com/NousResearch/hermes-agent/pull/46614), [#41096](https://github.com/NousResearch/hermes-agent/pull/41096) — @teknium1) - Bedrock falls back to non-streaming `InvokeModel` when IAM denies the streaming variant; Ollama default `max_tokens=65536`; surface model refusals as `content_filter` ([#44293](https://github.com/NousResearch/hermes-agent/pull/44293), [#41694](https://github.com/NousResearch/hermes-agent/pull/41694), [#46013](https://github.com/NousResearch/hermes-agent/pull/46013) — @teknium1) ### Sessions, state & multi-agent - Optional **max session cap**; drop empty sessions on CLI exit and rotation; ACP session-provenance metadata for compression rotation ([#42389](https://github.com/NousResearch/hermes-agent/pull/42389), [#43855](https://github.com/NousResearch/hermes-agent/pull/43855), [#41724](https://github.com/NousResearch/hermes-agent/pull/41724) — @teknium1) - Delegation: resolve custom-endpoint subagent pools by endpoint identity; remove the default subagent wall-clock timeout; stop subagent completion lines leaking into parent CLI display ([#41730](https://github.com/NousResearch/hermes-agent/pull/41730), [#45149](https://github.com/NousResearch/hermes-agent/pull/45149), [#44223](https://github.com/NousResearch/hermes-agent/pull/44223) — @teknium1) - Kanban: config-gated auto-subscribe on `kanban_create`; machine-global singleton lock for the embedded dispatcher; pin assigned profile toolsets for workers; hold reclaim while worker still alive ([#48635](https://github.com/NousResearch/hermes-agent/pull/48635), [#49068](https://github.com/NousResearch/hermes-agent/pull/49068), [#45590](https://github.com/NousResearch/hermes-agent/pull/45590), [#49064](https://github.com/NousResearch/hermes-agent/pull/49064) — @teknium1) - Memory: configurable Hindsight retain observation scopes; OpenViking setup UX; Honcho gateway-gated identity tree; Supermemory session-level ingest ([#46611](https://github.com/NousResearch/hermes-agent/pull/46611), [#48262](https://github.com/NousResearch/hermes-agent/pull/48262), [#44431](https://github.com/NousResearch/hermes-agent/pull/44431), [#38756](https://github.com/NousResearch/hermes-agent/pull/38756) — @teknium1, @alt-glitch) ## 📱 Messaging Platforms (Gateway) ### New channels - **iMessage via Photon Spectrum** — `hermes photon login` (device-code OAuth), gRPC-native channel (no webhook), markdown rendering, emoji reactions, outbound media via spectrum-ts ([#32348](https://github.com/NousResearch/hermes-agent/pull/32348), [#42582](https://github.com/NousResearch/hermes-agent/pull/42582), [#44713](https://github.com/NousResearch/hermes-agent/pull/44713), [#42397](https://github.com/NousResearch/hermes-agent/pull/42397) — @teknium1) - **WhatsApp Business Cloud API** adapter (official, no bridge process) ([#44331](https://github.com/NousResearch/hermes-agent/pull/44331), [#43921](https://github.com/NousResearch/hermes-agent/pull/43921) — @jquesnelle, @teknium1) - **SimpleX** — groups, native attachments, text batching, auto-accept; **Raft** bundled platform plugin with activity hooks ([#42584](https://github.com/NousResearch/hermes-agent/pull/42584), [#48210](https://github.com/NousResearch/hermes-agent/pull/48210) — @teknium1) ### Gateway core & rendering - Render terminal tool calls as native bash code blocks on markdown platforms; bare fenced code blocks in chat; optional message timestamps for LLM context; configurable `tool_progress_grouping` ([#41215](https://github.com/NousResearch/hermes-agent/pull/41215), [#42576](https://github.com/NousResearch/hermes-agent/pull/42576), [#47253](https://github.com/NousResearch/hermes-agent/pull/47253), [#47228](https://github.com/NousResearch/hermes-agent/pull/47228) — @teknium1) - Telegram: Bot API 10.1 rich messages (now always-on with opt-out); opt-in Online/Offline bot status indicator; stop cutting long streamed responses; MarkdownV2 on progress edits; gate oversized voice/audio before download ([#44829](https://github.com/NousResearch/hermes-agent/pull/44829), [#45584](https://github.com/NousResearch/hermes-agent/pull/45584), [#49134](https://github.com/NousResearch/hermes-agent/pull/49134), [#43761](https://github.com/NousResearch/hermes-agent/pull/43761), [#44245](https://github.com/NousResearch/hermes-agent/pull/44245) — @teknium1) - Discord: propagate `role_authorized` so `DISCORD_ALLOWED_ROLES` works end-to-end; recover from runtime gateway task exits; cancel `_bot_task` on connect failure; stop typing after replies ([#43327](https://github.com/NousResearch/hermes-agent/pull/43327), [#44383](https://github.com/NousResearch/hermes-agent/pull/44383), [#44432](https://github.com/NousResearch/hermes-agent/pull/44432), [#44836](https://github.com/NousResearch/hermes-agent/pull/44836) — @teknium1) - Slack: scope top-level channel messages when `reply_in_thread=false`; thread approval UX (block-size overflow + typed-prefix); make video attachments available to agents; `register_slack_action_handler` plugin API ([#41703](https://github.com/NousResearch/hermes-agent/pull/41703), [#43444](https://github.com/NousResearch/hermes-agent/pull/43444), [#45512](https://github.com/NousResearch/hermes-agent/pull/45512), [#44664](https://github.com/NousResearch/hermes-agent/pull/44664) — @teknium1) - Replied-to media attachments included; document attachments classified as DOCUMENT on Signal/Email/SimpleX/Teams; WhatsApp restarts stale bridge processes; Matrix room-context isolation; QQbot CPU-spin fix; Weixin rate-limit circuit breaker ([#46107](https://github.com/NousResearch/hermes-agent/pull/46107), [#44695](https://github.com/NousResearch/hermes-agent/pull/44695), [#44205](https://github.com/NousResearch/hermes-agent/pull/44205), [#18505](https://github.com/NousResearch/hermes-agent/pull/18505), [#40574](https://github.com/NousResearch/hermes-agent/pull/40574), [#41718](https://github.com/NousResearch/hermes-agent/pull/41718) — @teknium1, @banditburai) ## 🖥️ CLI, TUI & Setup - `/version` slash command; `/billing` interactive terminal billing (TUI + CLI); show time since last final agent response on the status bar; persist resolved approval/clarify prompts in scrollback ([#40214](https://github.com/NousResearch/hermes-agent/pull/40214), [#45449](https://github.com/NousResearch/hermes-agent/pull/45449), [#44265](https://github.com/NousResearch/hermes-agent/pull/44265), [#44702](https://github.com/NousResearch/hermes-agent/pull/44702) — @teknium1) - Lock hermes worktrees so concurrent processes can't clobber them; display custom profile alias names in list/show; clone profiles from any source ([#48699](https://github.com/NousResearch/hermes-agent/pull/48699), [#40371](https://github.com/NousResearch/hermes-agent/pull/40371), [#45630](https://github.com/NousResearch/hermes-agent/pull/45630) — @teknium1) - Opt-in structured profile-build path on first contact; configurable per-platform system-prompt hints; configurable background memory/skill notifications ([#41114](https://github.com/NousResearch/hermes-agent/pull/41114), [#48630](https://github.com/NousResearch/hermes-agent/pull/48630), [#47226](https://github.com/NousResearch/hermes-agent/pull/47226) — @teknium1) - TUI: interactive Plugins Hub enable/disable overlay; session name in the terminal titlebar; paint approval/clarify/sudo/secret modals directly (not via throttle); wrap long approval commands instead of truncating ([#42965](https://github.com/NousResearch/hermes-agent/pull/42965), [#43188](https://github.com/NousResearch/hermes-agent/pull/43188), [#41155](https://github.com/NousResearch/hermes-agent/pull/41155), [#44691](https://github.com/NousResearch/hermes-agent/pull/44691) — @teknium1) - TTS: Gemini persona prompts + audio tags; xAI auto speech tags + speed/streaming knobs; Piper speaker_id; OGG for Telegram auto-TTS ([#43442](https://github.com/NousResearch/hermes-agent/pull/43442), [#49061](https://github.com/NousResearch/hermes-agent/pull/49061), [#49062](https://github.com/NousResearch/hermes-agent/pull/49062), [#49060](https://github.com/NousResearch/hermes-agent/pull/49060), [#41644](https://github.com/NousResearch/hermes-agent/pull/41644) — @teknium1) ## 🔧 Tool System, Skills & MCP - **image-to-image / editing** in `image_generate` across all backends; shrink images to provider dimension limit ([#48705](https://github.com/NousResearch/hermes-agent/pull/48705), [#45979](https://github.com/NousResearch/hermes-agent/pull/45979) — @teknium1) - MCP: official **Unreal Engine 5.8** MCP server in the catalog; **elicitation handler** so MCP servers can prompt for mid-tool-call confirmation (payment/OAuth) on whichever surface owns the session — CLI/TUI/Telegram/Slack; expose late-connecting MCP tools to the agent between turns (cache-safe); keepalive ping for short-TTL HTTP sessions; block exfil-shaped / suspicious stdio configs before probe; capability-gate `tools/list` so prompt-only servers connect; preserve stdio argv passthrough + Windows env vars ([#48397](https://github.com/NousResearch/hermes-agent/pull/48397), [#49203](https://github.com/NousResearch/hermes-agent/pull/49203), [#49208](https://github.com/NousResearch/hermes-agent/pull/49208), [#49221](https://github.com/NousResearch/hermes-agent/pull/49221), [#46083](https://github.com/NousResearch/hermes-agent/pull/46083), [#44550](https://github.com/NousResearch/hermes-agent/pull/44550), [#44324](https://github.com/NousResearch/hermes-agent/pull/44324) — @teknium1, @lgalabru) - Skills: `simplify-code` skill (parallel 3-agent code review & cleanup) + risk-tiered application with Chesterton's Fence; find & diff user-modified bundled skills; optional **payments** skills (Stripe Link, MPP, Projects); CLI-based shop skill; live per-source browse progress ([#41691](https://github.com/NousResearch/hermes-agent/pull/41691), [#49070](https://github.com/NousResearch/hermes-agent/pull/49070), [#48286](https://github.com/NousResearch/hermes-agent/pull/48286), [#31343](https://github.com/NousResearch/hermes-agent/pull/31343), [#47309](https://github.com/NousResearch/hermes-agent/pull/47309), [#43398](https://github.com/NousResearch/hermes-agent/pull/43398) — @teknium1, @colinwren-stripe) - Curator: make skill consolidation opt-in (prune stays default-on) ([#47840](https://github.com/NousResearch/hermes-agent/pull/47840) — @teknium1) - Plugins: install from a subdirectory within a repo; accept browser-pasted GitHub URLs in `hermes plugins install`; `session:compress` lifecycle event + `thread_id`/`chat_type` in agent:start/end context ([#42963](https://github.com/NousResearch/hermes-agent/pull/42963), [#33539](https://github.com/NousResearch/hermes-agent/pull/33539), [#47252](https://github.com/NousResearch/hermes-agent/pull/47252), [#41672](https://github.com/NousResearch/hermes-agent/pull/41672) — @teknium1) - Memory/skill **write approval** gate (default off) — boolean `write_approval` replaces the tri-state `write_mode` ([#38199](https://github.com/NousResearch/hermes-agent/pull/38199), [#43354](https://github.com/NousResearch/hermes-agent/pull/43354) — @teknium1) ## 🌐 Fleet, Relay & Automation - **Managed scope** — administrator-pinned, user-immutable config & secrets from a root-owned `/etc/hermes` ([#49098](https://github.com/NousResearch/hermes-agent/pull/49098) — @teknium1) - **Multiplex all profiles over one gateway process** (opt-in) ([#48273](https://github.com/NousResearch/hermes-agent/pull/48273) — @benbarclay) - **Pluggable CronScheduler** + Chronos managed-cron provider (scale-to-zero) ([#48275](https://github.com/NousResearch/hermes-agent/pull/48275) — @benbarclay) - **Automation Blueprints** — parameterized automation templates across every surface ([#41309](https://github.com/NousResearch/hermes-agent/pull/41309) — @teknium1) - Gateway-Gateway relay (phases 0-3): relay adapter + capability descriptor, connector⇄gateway channel auth + signed-HTTP inbound + enroll CLI, WS-only inbound, managed-boot self-provision client ([#48078](https://github.com/NousResearch/hermes-agent/pull/48078), [#48147](https://github.com/NousResearch/hermes-agent/pull/48147), [#48294](https://github.com/NousResearch/hermes-agent/pull/48294), [#48242](https://github.com/NousResearch/hermes-agent/pull/48242) — @teknium1) ## 🐳 Docker, Nix & Installer - s6: detect supervisor directly for gateway restart; register profile gateways without auto-starting; persist desired state; clear stale log locks ([#46290](https://github.com/NousResearch/hermes-agent/pull/46290), [#46266](https://github.com/NousResearch/hermes-agent/pull/46266), [#46292](https://github.com/NousResearch/hermes-agent/pull/46292), [#46289](https://github.com/NousResearch/hermes-agent/pull/46289) — @teknium1) - Docker: optimize image size (.dockerignore, drop dev deps, split layers); pre-install matrix deps; supervised gateway uses `--replace`; harden hosted install tree against self-modification ([#38749](https://github.com/NousResearch/hermes-agent/pull/38749), [#42413](https://github.com/NousResearch/hermes-agent/pull/42413), [#47555](https://github.com/NousResearch/hermes-agent/pull/47555), [#47490](https://github.com/NousResearch/hermes-agent/pull/47490) — @teknium1) - Nix: cold npm build fixes + auto-fix-lockfiles workflow; hashless npm deps via `importNpmLock`; refresh npmDepsHash after Electron 40.10.2 pin ([#41867](https://github.com/NousResearch/hermes-agent/pull/41867), [#48883](https://github.com/NousResearch/hermes-agent/pull/48883), [#48457](https://github.com/NousResearch/hermes-agent/pull/48457) — @teknium1) - Installer: clear unmerged git index before autostash; scope install-method stamp to the code tree ([#45515](https://github.com/NousResearch/hermes-agent/pull/45515), [#48188](https://github.com/NousResearch/hermes-agent/pull/48188) — @teknium1) ## 🔒 Security & Reliability - Fail closed on own-policy gateway adapters; fail closed for approval-button auth on Slack/Feishu/Discord when no allowlist is set ([#45634](https://github.com/NousResearch/hermes-agent/pull/45634), [#41226](https://github.com/NousResearch/hermes-agent/pull/41226) — @teknium1) - Redact secrets in request debug dumps; withhold host metadata from public status; block exfil-shaped / suspicious MCP stdio configs before probe ([#46637](https://github.com/NousResearch/hermes-agent/pull/46637), [#45642](https://github.com/NousResearch/hermes-agent/pull/45642), [#46083](https://github.com/NousResearch/hermes-agent/pull/46083) — @teknium1) - Close shell-escape denylist bypass + fail-closed on missing approval module; scrub operator environment before launching cua-driver MCP; sanitize env for cron job-script subprocesses; bound TodoStore content length/count; scan REST cron prompts for parity with the agent tool ([#40591](https://github.com/NousResearch/hermes-agent/pull/40591), [#48423](https://github.com/NousResearch/hermes-agent/pull/48423), [#49207](https://github.com/NousResearch/hermes-agent/pull/49207), [#41648](https://github.com/NousResearch/hermes-agent/pull/41648), [#41335](https://github.com/NousResearch/hermes-agent/pull/41335) — @teknium1, @kshitijk4poor) - Bump urllib3 and PyJWT to clear CVEs; Langfuse redacts base64 data URIs instead of truncating into invalid base64 ([#40179](https://github.com/NousResearch/hermes-agent/pull/40179), [#43322](https://github.com/NousResearch/hermes-agent/pull/43322) — @teknium1) ## 🪟 Windows - Dashboard `/chat` tab via ConPTY (`win_pty_bridge`) + tests; resolve PowerShell host instead of bare `powershell` for uv install; resolve `powershell.exe` by absolute path so Desktop install doesn't stall ([#42251](https://github.com/NousResearch/hermes-agent/pull/42251), [#48341](https://github.com/NousResearch/hermes-agent/pull/48341), [#40927](https://github.com/NousResearch/hermes-agent/pull/40927) — @teknium1) - Repair stale winget registration + refresh/merge PATH; kill hermes before recreating venv to release `_bcrypt.pyd` lock; read HERMES_HOME from the registry when env is stale; quarantine running `hermes.exe` during update repair ([#44084](https://github.com/NousResearch/hermes-agent/pull/44084), [#45120](https://github.com/NousResearch/hermes-agent/pull/45120), [#46772](https://github.com/NousResearch/hermes-agent/pull/46772), [#40409](https://github.com/NousResearch/hermes-agent/pull/40409) — @teknium1) - JOB-breakaway watcher reliability + status --deep probes; handle Windows PTY stdin + detached WS frames; decode subprocess output as UTF-8; confirm-modal on native Windows ([#40909](https://github.com/NousResearch/hermes-agent/pull/40909), [#41953](https://github.com/NousResearch/hermes-agent/pull/41953), [#44328](https://github.com/NousResearch/hermes-agent/pull/44328), [#42419](https://github.com/NousResearch/hermes-agent/pull/42419) — @teknium1) ## 🐛 Notable Bug Fixes - Percent-encode non-ascii URL components; sanitize `:` in FTS5 queries so colon searches don't silently return empty ([#41430](https://github.com/NousResearch/hermes-agent/pull/41430), [#40653](https://github.com/NousResearch/hermes-agent/pull/40653) — @teknium1) - Preserve multimodal user content through crash-resilience persist; flatten multimodal content before provider sync; strip MEDIA directives from compressor input ([#47907](https://github.com/NousResearch/hermes-agent/pull/47907), [#44738](https://github.com/NousResearch/hermes-agent/pull/44738), [#44708](https://github.com/NousResearch/hermes-agent/pull/44708) — @teknium1) - Re-enter retry loop on genuine Nous 429 so the fallback guard runs; scope Nous tags to Nous auxiliary calls; suppress "Credit access paused" notice on free models ([#45136](https://github.com/NousResearch/hermes-agent/pull/45136), [#45801](https://github.com/NousResearch/hermes-agent/pull/45801), [#43669](https://github.com/NousResearch/hermes-agent/pull/43669) — @teknium1) - Cron: don't strict-scan script-injected output in no-skills jobs; resolve per-job provider "custom" to `providers.custom` instead of codex; repair cron ownership on container restart ([#43223](https://github.com/NousResearch/hermes-agent/pull/43223), [#43505](https://github.com/NousResearch/hermes-agent/pull/43505), [#41976](https://github.com/NousResearch/hermes-agent/pull/41976) — @teknium1) - *(300+ issues closed this window; full per-area fix list is exhaustive — these are the highest-impact.)* ## ↩️ Reverted in this window (not shipping) - `html-artifact` skill + sketch/architecture-diagram/concept-diagrams fold ([#48899](https://github.com/NousResearch/hermes-agent/pull/48899)) — reverted ([#49053](https://github.com/NousResearch/hermes-agent/pull/49053)); absent on main. - Cron per-job profile support reverted ([#43956](https://github.com/NousResearch/hermes-agent/pull/43956)); a nix patchPhase workaround reverted ([#42151](https://github.com/NousResearch/hermes-agent/pull/42151)). ## 👥 Contributors A huge thank-you to everyone who contributed to this release — **245 contributors** across commits, co-author trailers, and salvaged PRs. ### Core @teknium1 ### Top community contributors (by merged PRs) - @OutThisLife — 92 PRs (desktop app maturity (shortcuts, notifications, watch-windows, themes)) - @kshitijk4poor — 60 PRs (onboarding, model picker, cron env sanitization) - @xxxigm — 27 PRs (desktop & gateway fixes) - @benbarclay — 23 PRs (gateway multiplex, Chronos cron, dashboard auth) - @helix4u — 21 PRs (gateway & installer reliability) - @austinpickett — 19 PRs (dashboard & desktop UX) - @alt-glitch — 14 PRs (usage-aware credits, Supermemory) - @ethernet8023 — 14 PRs (desktop build pipeline & Linux/Windows) - @liuhao1024 — 5 PRs (session lifecycle fixes) ### All contributors (alphabetical) @0z1-ghb, @0xdany, @0xneobyte, @0xyg3n, @1960697431, @895252509, @achaljhawar, @Adolanium, @AhmetArif0, @AIalliAI, @aimable100, @AJ, @ak2k, @alarcritty, @AlchemistChaos, @aldoeliacim, @alelpoan, @AlexanderBFoley, @alfred-smith-0, @ali-nld, @alt-glitch, @am423, @ameobius, @AMIK-coorporations, @annguyenNous, @ArcanePivot, @ARegalado1, @asdlem, @ashishpatel26, @austinpickett, @banditburai, @barronlroth, @Bartok9, @basilalshukaili, @bbednarski9, @bcsmith528, @benbarclay, @benegessarit, @benfrank241, @bionicbutterfly13, @BlackishGreen33, @blut-agent, @bmoore210, @bpasquini, @briandevans, @BROCCOLO1D, @capt-marbles, @ccook1963, @Cdddo, @channkim, @ChasLui, @chimpera, @chromalinx, @CiarasClaws, @claytonchew, @cnfi, @colinwren-stripe, @cresslank, @cyb0rgk1tty, @dangelo352, @davidgut1982, @deaneeth, @definitelynotguru, @Diyoncrz18, @draix, @dschnurbusch, @Dusk1e, @dusterbloom, @ehz0ah, @emozilla, @enesilhaydin, @Erosika, @ethernet8023, @Evisolpxe, @firefly, @flooryyyy, @flyinhigh, @foras910521-lab, @Frowtek, @ft-ioxcs, @fyzanshaik, @Ganesh0690, @gauravsaxena1997, @giladbau, @glesperance, @GodsBoy, @goku94123, @H-Ali13381, @HaozheZhang6, @haran2001, @harshitAgr, @hbentel, @helix4u, @HeLLGURD, @Hermes, @huangxun375-stack, @iamlukethedev, @ianculling, @IAvecilla, @iborazzi, @infinitycrew39, @islam666, @ITheEqualizer, @itsflownium, @Jaaneek, @james47kjv, @jeeves-assistant, @jeffrobodie-glitch, @JezzaHehn, @jiangkoumo, @jimjsong, @JimLiu, @JimStenstrom, @jmsunseri, @JoaoMarcos44, @joel611, @JoelJJohnson, @joerj123, @johnjacobkenny, @jooray, @joshuadow, @jplew, @justinbao19, @Justlrnal4, @Kailigithub, @kamonspecial, @kdunn926, @Kenmege, @Kewe63, @kmccammon, @konsisumer, @kristianvast, @kshitijk4poor, @kyssta-exe, @l37525778-coder, @LaPhilosophie, @leo4226, @LeonSGP43, @liuhao1024, @Llugaes, @loongfay, @loongzhao, @lsaether, @m4dni5, @manishbyatroy, @MaxFreedomPollard, @maxmilian, @maxtrigify, @mnajafian-nv, @mohamedorigami-jpg, @mollusk, @MrDiamondBallz, @mssteuer, @mvanhorn, @naqerl, @Nea74, @necoweb3, @nepenth, @nicoloboschi, @NormallyGaussian, @OmarB97, @omegazheng, @OndrejDrapalik, @OutThisLife, @oxngon, @OYLFLMH, @Paperclip, @paulb26, @pengyuyanITYU, @PhilipAD, @pinguarmy, @plcunha, @ProgramCaiCai, @psionic73, @qin-ctx, @qingshan89, @Que0x, @qWaitCrypto, @r266-tech, @randomsnowflake, @rbrtbn, @rewbs, @rio-jeong, @Rivuza, @rob-maron, @rodboev, @ruangraung, @RyTsYdUp, @Sahil-SS9, @salesondemandio, @sanidhyasin, @sarvesh1327, @sdyckjq-lab, @shannonsands, @SHL0MS, @simpolism, @sitkarev, @skyc1e, @skylarbpayne, @SNooZyy2, @Spaceman-Spiffy, @srojk34, @sweetcornna, @synapsesx, @Tamaz-sujashvili, @tangtaizong666, @temalo, @tfournet, @thedavidweng, @TheGardenGallery, @tim404x, @tomekpanek, @Tranquil-Flow, @tt-a1i, @tuancookiez-hub, @underthestars-zhy, @Veritas-7, @victor-kyriazakos, @wesleysimplicio, @WolframRavenwolf, @WompaJango, @x1erra, @xiaoxinova, @xtymac, @xushibo, @XVVH, @xxchan, @xxxigm, @xy200303, @y0shua1ee, @yanxue06, @yatesjalex, @YLChen-007, @yoniebans, @youjunxiaji, @yubingz, @zakame, @zapabob, @zccyman, @zimigit2020, @ziwon, @zwcf5200, @zxcasongs. --- **Full Changelog**: [v2026.6.5...v2026.6.19](https://github.com/NousResearch/hermes-agent/compare/v2026.6.5...v2026.6.19)

# Hermes Agent v0.16.0 (v2026.6.5) **发布日期：** 2026年6月5日 **自 v0.15.2 以来：** 874 个提交 · 542 个合并的 PR · 1,962 个文件变更 · 205,216 行新增 · 46,217 行删除 · 399 个已关闭议题（2 个 P0，62 个 P1，16 个安全标记） · 170 位社区贡献者（含合著者） > **「Surface」版本。** Hermes 在你工作的任何地方与你相遇。全新的原生桌面应用——一周内跨越 100 个 PR 和 159 个提交构建而成——让 Hermes 成为真正的 macOS/Linux/Windows 应用程序：一键安装、应用内自更新、拖拽文件到聊天中、状态栏中的内联模型选择器、并发的多配置文件会话、完整的简体中文翻译，以及通过 OAuth 或用户名/密码连接远程 Hermes 网关的能力。与此同时，Web 仪表盘新增了完整的基于浏览器的管理面板（MCP 目录、消息通道、凭据、Webhook、记忆、可插拔的 OIDC/用户名密码登录），首次设置增加了「通过 Nous Portal 快速设置」路径，让你从安装到第一条消息只需几秒钟。默认技能集精简为你真正需要的，NVIDIA/skills 加入了受信任的技能 Hub 源，模型选择器现在可以在所有地方（桌面、Web、TUI、CLI）进行模糊搜索，`/undo` 终于让你可以撤销最近的 N 轮对话。2 个 P0 和 62 个 P1 的关闭也一同包含在内，外加一轮安全修复（CVE-2026-48710 Starlette 锁定、SSRF 环路外加固、子进程凭据剥离）。 --- ## ✨ 亮点 - **Hermes Desktop——真正的原生应用，而非终端包装器**——这是头条新闻。现在有一个 `apps/desktop/` Electron 应用程序，可以像任何其他桌面应用一样安装在 macOS、Linux 和 Windows 上，在应用内部进行自更新，并为 Hermes 所做的一切提供精美的图形界面。你将拥有一个带有流式传输的聊天窗口、可以归档和搜索的会话列表、在聊天区域任何位置拖放文件、剪贴板图片粘贴、Cmd+K 命令面板，以及状态栏中的模型选择器。如果你一直告诉朋友「它是一个 CLI 代理」然后看到他们眼神呆滞——现在你只需发给他们一个安装程序。一周前这些东西都不存在。（[#20059](https://github.com/NousResearch/hermes-agent/pull/20059)、[#35607](https://github.com/NousResearch/hermes-agent/pull/35607)、[#37099](https://github.com/NousResearch/hermes-agent/pull/37099)、[#37379](https://github.com/NousResearch/hermes-agent/pull/37379)、[#38631](https://github.com/NousResearch/hermes-agent/pull/38631) —— @OutThisLife、@jquesnelle、@ethernet8023、@austinpickett、@benbarclay） - **将桌面应用连接到远程 Hermes——通过 OAuth 或用户名/密码登录**——桌面应用不必在本地运行 Hermes。将其指向远程 Hermes 网关（你的家庭实验室、托管服务器、队友的服务器），它会通过安全的 WebSocket 连接，使用 OAuth 或用户名/密码登录进行身份验证——无需折腾 `--insecure` 标志或手动复制的会话令牌。每个配置文件可以针对自己的远程主机，你可以在一个窗口中跨多个配置文件运行并发会话，并使用跨配置文件的 `@session` 链接。实际版本：你的笔记本电脑运行一个轻量 GUI，重型代理则运行在你的 API 密钥和计算资源所在之处。（[#37888](https://github.com/NousResearch/hermes-agent/pull/37888)、[#38851](https://github.com/NousResearch/hermes-agent/pull/38851)、[#39330](https://github.com/NousResearch/hermes-agent/pull/39330)、[#39778](https://github.com/NousResearch/hermes-agent/pull/39778) —— @benbarclay、@OutThisLife、@teknium1） - **Web 仪表盘现在是一个功能完整的管理面板——从浏览器配置一切**——仪表盘从「查看你的会话」发展成为一个完整的管理界面。新增了 Channels 页面，可以从浏览器设置每个网关消息平台（Telegram、Discord、Slack 等）；MCP 目录的管理面板带有启用/禁用开关；凭据管理；Webhook 和钩子创建；内存配置；网关控制；以及 System 页面，带有更新前检查和一键 Debug Share。你不再需要 SSH 登录并编辑 `config.yaml` 来配置新的消息通道或 MCP 服务器——现在全是点选操作。（[#36704](https://github.com/NousResearch/hermes-agent/pull/36704)、[#36736](https://github.com/NousResearch/hermes-agent/pull/36736)、[#37211](https://github.com/NousResearch/hermes-agent/pull/37211)、[#38205](https://github.com/NousResearch/hermes-agent/pull/38205)、[#38600](https://github.com/NousResearch/hermes-agent/pull/38600) —— @teknium1） - **Hermes Desktop 支持简体中文——聊天 GUI 全面汉化**——桌面应用现在在所有 UI 界面——聊天窗口本身、侧边栏、设置、命令中心、定时任务、消息、配置文件、技能、代理等等——都提供了完整的简体中文翻译。英文保持默认；在「外观」设置中切换语言，选择会持久化到你的配置中（`display.language`）。它基于一个合适的类型化 i18n 层构建，所以从这里添加更多语言会很简单。（[#38241](https://github.com/NousResearch/hermes-agent/pull/38241) —— @JimLiu） - **更精简的默认技能集——Hermes 只包含你真正需要的**——捆绑的技能集经过了有意的精简。冗余或已死的技能被移除（`spotify`——已被原生 Spotify 插件的 7 个工具取代；`linear`——已被 `hermes mcp install linear` 取代；`kanban-codex-lane`、`debugging-hermes-tui-commands`、一个过时的 `domain` 孤儿，以及几个空的分类标记）。较重或小众的技能从捆绑移到了可选（Baoyu 创意套装、`dspy`、`subagent-driven-development`、`minecraft-modpack-server`、`pokemon-player`、`hermes-s6-container-supervision`）——仍然只需 `hermes skills install` 一次，只是默认不加载。而一个新的 `environments:` 相关性门控可以让特定上下文的技能（看板、docker/s6）不进入技能索引，供永远不会使用它们的用户使用，同时仍可在明确请求时加载。结果：更小、更清晰的默认技能列表，选择器中噪音更少，提示更轻量。策展人现在也可以修剪未使用的**内置**技能（不仅限于代理创建的），每个技能的使用情况都会被跟踪。（[#39028](https://github.com/NousResearch/hermes-agent/pull/39028)、[#36701](https://github.com/NousResearch/hermes-agent/pull/36701)、[#36228](https://github.com/NousResearch/hermes-agent/pull/36228) —— @teknium1） - **NVIDIA/skills 现在是一个内置的受信任技能源**——`NVIDIA/skills` 加入了 OpenAI、Anthropic 和 HuggingFace，成为 Skills Hub 中的默认受信任源——可发现、可浏览、可搜索，并通过相同的管道自动更新。NVIDIA 针对 CUDA-X、AIQ、cuOpt 及其产品栈其他部分的已验证技能，只需一次安装，并带有来自仓库 `skills.sh.json` 侧车的真实分类标签。（[#34333](https://github.com/NousResearch/hermes-agent/pull/34333) —— @teknium1） - **通过 Nous Portal 快速设置——从安装到第一条消息只需几秒钟**——首次设置被精简为两个清晰的路径：快速设置（登录 Nous Portal，获取模型选择器，立即开始聊天）或完整设置（面向高级用户的详细向导）。`hermes portal` 现在是运行完整快速设置 Nous 流程的人类可读别名。首次运行菜单内联解释了区别，让新用户不用猜测。目标：全新用户无需阅读文档就能发送第一条消息。（[#35723](https://github.com/NousResearch/hermes-agent/pull/35723)、[#36227](https://github.com/NousResearch/hermes-agent/pull/36227)、[#38449](https://github.com/NousResearch/hermes-agent/pull/38449)、[#38465](https://github.com/NousResearch/hermes-agent/pull/38465) —— @teknium1、@kshitijk4poor） - **模糊模型选择器，无处不在——输入几个字母，找到你的模型**——模型选择器现在在桌面应用、Web 仪表盘、TUI 和 CLI 中都能进行模糊搜索。输入「v4fl」就会显示 `deepseek-v4-flash`；多端点提供商被归为一行，而不是用重复项混乱列表，每行还带有描述，让你知道自己选的是什么。目录每小时刷新一次（之前是每天），所以新模型在发布当天就能出现。本轮新增：`deepseek-v4-flash`、具有 100 万上下文的 `MiniMax-M3`、`qwen3.7-plus`。（[#36928](https://github.com/NousResearch/hermes-agent/pull/36928)、[#35227](https://github.com/NousResearch/hermes-agent/pull/35227)、[#35756](https://github.com/NousResearch/hermes-agent/pull/35756)、[#35659](https://github.com/NousResearch/hermes-agent/pull/35659)、[#36214](https://github.com/NousResearch/hermes-agent/pull/36214) —— @kshitijk4poor、@teknium1） - **`/undo [N]`——撤销最近 N 轮对话**——说错了话，或者把代理引向错误方向了？`/undo` 会回退 N 轮用户对话，预填你的最后一条消息以便编辑并重新发送，并软删除中间的对话轮次。它在 CLI、TUI 以及消息平台（Telegram、Discord 等）中都能完全一致地工作。关闭了一个长期存在的请求（[#21910](https://github.com/NousResearch/hermes-agent/issues/21910)）。（[#36229](https://github.com/NousResearch/hermes-agent/pull/36229)、[#36699](https://github.com/NousResearch/hermes-agent/pull/36699) —— @teknium1） - **选择你的默认界面——`cli` 或 `tui`**——你现在可以设置 `hermes chat` 默认是进入经典 CLI 还是 Ink TUI，并可以使用 `--cli` 标志覆盖每次调用。TUI 还获得了一个统一的 `/model` 命令和一个用于在实时会话间切换的 Sessions 覆盖层。使用你真正喜欢的界面。（[#37782](https://github.com/NousResearch/hermes-agent/pull/37782)、[#37112](https://github.com/NousResearch/hermes-agent/pull/37112) —— @OutThisLife） --- ## 🖥️ Hermes 桌面应用（全新） ### 安装与生命周期 - macOS 桌面安装 + 应用内自更新；在 macOS 上干净地重建并重启（[#35607](https://github.com/NousResearch/hermes-agent/pull/35607)、[#36198](https://github.com/NousResearch/hermes-agent/pull/36198)、[#38296](https://github.com/NousResearch/hermes-agent/pull/38296) —— @OutThisLife） - macOS 安装程序重命名为「Hermes」并变为启动器（[#37516](https://github.com/NousResearch/hermes-agent/pull/37516) —— @OutThisLife） - 在 macOS/Linux 上以独立的 `desktop` 阶段构建桌面，而不是静默跳过；内容哈希构建戳，`--build-only` / `--force-build` 标志（[#36134](https://github.com/NousResearch/hermes-agent/pull/36134) —— @OutThisLife，[#37597](https://github.com/NousResearch/hermes-agent/pull/37597) —— @ethernet8023） - 启动失败恢复 + 实时 API 密钥验证；可取消安装；从损坏的缓存 Electron 下载中恢复（[#35864](https://github.com/NousResearch/hermes-agent/pull/35864)、[#37379](https://github.com/NousResearch/hermes-agent/pull/37379)、[#39032](https://github.com/NousResearch/hermes-agent/pull/39032) —— @OutThisLife、@teknium1） - Windows：从引导安装中损坏的 Electron 缓存恢复；在应用内更新期间停止与自己的后端竞争（[#39465](https://github.com/NousResearch/hermes-agent/pull/39465)、[#39828](https://github.com/NousResearch/hermes-agent/pull/39828) —— @teknium1、@OutThisLife） - Linux：配置 Electron 沙箱辅助程序；检测 linux arm64 二进制文件；在远程显示器上禁用 GPU 加速以停止闪烁（[#37691](https://github.com/NousResearch/hermes-agent/pull/37691)、[#38594](https://github.com/NousResearch/hermes-agent/pull/38594)、[#37932](https://github.com/NousResearch/hermes-agent/pull/37932) —— @ethernet8023、@OutThisLife） - 要求 Node ≥20.19/22.12 用于桌面构建；零 eslint/typecheck 债务 + prettier 通扫（[#38255](https://github.com/NousResearch/hermes-agent/pull/38255)、[#39100](https://github.com/NousResearch/hermes-agent/pull/39100) —— @OutThisLife） ### 远程网关与多配置文件 - 连接到 OAuth 网关的远程网关；远程网关的用户名/密码登录；每配置文件远程网关主机（[#37888](https://github.com/NousResearch/hermes-agent/pull/37888)、[#38851](https://github.com/NousResearch/hermes-agent/pull/38851)、[#39778](https://github.com/NousResearch/hermes-agent/pull/39778) —— @benbarclay、@teknium1、@OutThisLife） - 并发多配置文件会话，跨配置文件 `@session` 链接；在网关重新连接时重新生成 OAuth WS 票据；在 AT 或 RT 上限制 OAuth 远程连接（[#39330](https://github.com/NousResearch/hermes-agent/pull/39330)、[#38886](https://github.com/NousResearch/hermes-agent/pull/38886)、[#39464](https://github.com/NousResearch/hermes-agent/pull/39464) —— @OutThisLife、@teknium1、@benbarclay） - 在受控网关启动失败时提供远程登录选项；在远程网关测试中验证实时 WebSocket（[#39402](https://github.com/NousResearch/hermes-agent/pull/39402)、[#39511](https://github.com/NousResearch/hermes-agent/pull/39511) —— @teknium1） ### 聊天 UX 与设置 - 在聊天区域任何位置拖放文件；剪贴板图片粘贴带去重；按 Enter 时附件处理，IME 组合输入处理（[#36262](https://github.com/NousResearch/hermes-agent/pull/36262)、[#38306](https://github.com/NousResearch/hermes-agent/pull/38306)、[#38677](https://github.com/NousResearch/hermes-agent/pull/38677) —— @OutThisLife、@teknium1） - 后台需要输入指示器，界面重新设计，Cmd+K 调色板和 UI 一致性通过；状态栏中的内联模型选择器；状态栏中的 YOLO 切换（TUI 同步）（[#38631](https://github.com/NousResearch/hermes-agent/pull/38631)、[#37738](https://github.com/NousResearch/hermes-agent/pull/37738)、[#38517](https://github.com/NousResearch/hermes-agent/pull/38517) —— @OutThisLife） - 会话列表改造，会话清理/归档，媒体流式传输 + 连接覆盖层；按 ID 搜索会话（SQL 限制）（[#37379](https://github.com/NousResearch/hermes-agent/pull/37379)、[#37099](https://github.com/NousResearch/hermes-agent/pull/37099)、[#39062](https://github.com/NousResearch/hermes-agent/pull/39062) —— @OutThisLife、@teknium1） - 专门的 Providers 设置 + 精炼的 Accounts/API-keys UX；将技能和工具管理合并到一个面板；将模型管理移到设置中（[#38551](https://github.com/NousResearch/hermes-agent/pull/38551) —— @austinpickett，[#37310](https://github.com/NousResearch/hermes-agent/pull/37310)、[#37330](https://github.com/NousResearch/hermes-agent/pull/37330) —— @jquesnelle） - 渲染审批/sudo/秘密提示，以便工具不再静默超时；在调色板中显示技能和快速命令的斜杠命令；在启动器中加入一流的 xAI Grok OAuth 提供者（[#38578](https://github.com/NousResearch/hermes-agent/pull/38578)、[#38531](https://github.com/NousResearch/hermes-agent/pull/38531)、[#37697](https://github.com/NousResearch/hermes-agent/pull/37697) —— @teknium1、@OutThisLife） - 首次运行入门中「稍后选择提供者」跳过选项；入门可以配置本地/自定义端点而无需 API 密钥；自定义缩放快捷键（[#39483](https://github.com/NousResearch/hermes-agent/pull/39483)、[#38572](https://github.com/NousResearch/hermes-agent/pull/38572)、[#37894](https://github.com/NousResearch/hermes-agent/pull/37894) —— @teknium1、@ethernet8023） - macOS 辅助程序麦克风权限；滚动跳跃修复（原生锚定、静止跳跃、上滚回弹）；思考块在流式传输中保持打开（[#37745](https://github.com/NousResearch/hermes-agent/pull/37745) —— @xxxigm，[#37866](https://github.com/NousResearch/hermes-agent/pull/37866)、[#38221](https://github.com/NousResearch/hermes-agent/pull/38221)、[#38809](https://github.com/NousResearch/hermes-agent/pull/38809) —— @OutThisLife、@teknium1、@stremtec） - GUI 生活质量批量修复；保留 AhmetArif0 桌面/仪表盘修复；通过 `session.title` RPC 重命名会话，使 `/title` 生效（[#37536](https://github.com/NousResearch/hermes-agent/pull/37536) —— @austinpickett，[#39070](https://github.com/NousResearch/hermes-agent/pull/39070)、[#39410](https://github.com/NousResearch/hermes-agent/pull/39410) —— @teknium1、@benbarclay） - `hermes debug share` / `/debug` / `hermes logs` 现在包含 `desktop.log`（[#38203](https://github.com/NousResearch/hermes-agent/pull/38203) —— @teknium1） - **简体中文翻译**涵盖所有桌面 UI 界面——类型化 i18n 层，在「外观」设置中切换，通过 `display.language` 持久化（英文保持默认）（[#38241](https://github.com/NousResearch/hermes-agent/pull/38241) —— @JimLiu） - 通过一个全局远程仪表盘支持完整的多配置文件；远程配置文件会话是一等公民（恢复、读取、重命名/归档/删除）；新聊天在全局远程模式下尊重其配置文件（[#39921](https://github.com/NousResearch/hermes-agent/pull/39921)、[#39894](https://github.com/NousResearch/hermes-agent/pull/39894)、[#39993](https://github.com/NousResearch/hermes-agent/pull/39993) —— @OutThisLife） ## 📊 Web 仪表盘 ### 管理面板（全新） - 完整的管理面板——MCP 目录带有启用/禁用开关、配对、Webhook、凭据、内存、网关、钩子创建、系统设置（[#36704](https://github.com/NousResearch/hermes-agent/pull/36704)、[#36736](https://github.com/NousResearch/hermes-agent/pull/36736) —— @teknium1） - Channels 页面——从浏览器设置每个网关消息通道（[#37211](https://github.com/NousResearch/hermes-agent/pull/37211) —— @teknium1） - System 页面：更新前检查流程 + Debug Share（[#38205](https://github.com/NousResearch/hermes-agent/pull/38205)、[#38600](https://github.com/NousResearch/hermes-agent/pull/38600) —— @teknium1） - nous-blue 主题、批量会话、调度选择器；丰富的配置文件仪表盘 + 去重通道环境变量；通过主题可配置的终端背景（[#37383](https://github.com/NousResearch/hermes-agent/pull/37383)、[#37872](https://github.com/NousResearch/hermes-agent/pull/37872) —— @austinpickett，[#37156](https://github.com/NousResearch/hermes-agent/pull/37156) —— @teknium1） - 始终启用嵌入式聊天；移除仪表盘 `--tui` 标志（[#38591](https://github.com/NousResearch/hermes-agent/pull/38591) —— @benbarclay） ### 认证 - 可插拔的用户名/密码登录（选项 B）；通用自托管 OIDC 提供者 + 多提供者验证修复；`hermes dashboard register` 用于自托管 OAuth 客户端（[#38819](https://github.com/NousResearch/hermes-agent/pull/38819) —— @benbarclay，[#38917](https://github.com/NousResearch/hermes-agent/pull/38917)、[#38802](https://github.com/NousResearch/hermes-agent/pull/38802) —— @teknium1） - 通过刷新令牌轮换仪表盘会话；在遗留网关和 OAuth 网关之间共享 `/api/*` 公共允许列表；从 OAuth `next=` 往返中移除 `/api/*` 路径（[#37247](https://github.com/NousResearch/hermes-agent/pull/37247)、[#34254](https://github.com/NousResearch/hermes-agent/pull/34254)、[#36244](https://github.com/NousResearch/hermes-agent/pull/36244) —— @benbarclay） - 聊天标签页在受限（OAuth）模式下工作；在 OAuth 限制绑定上信任经过票据认证的非 Web WS 来源；认证服务器生成的 PTY 子 WS；通过 SDK 辅助程序批准插件 WS/上传认证（[#34793](https://github.com/NousResearch/hermes-agent/pull/34793)、[#37870](https://github.com/NousResearch/hermes-agent/pull/37870)、[#37972](https://github.com/NousResearch/hermes-agent/pull/37972)、[#38549](https://github.com/NousResearch/hermes-agent/pull/38549) —— @teknium1、@benbarclay、@kshitijk4poor） - 允许远程绑定上的桌面 WebSocket 来源（[#37747](https://github.com/NousResearch/hermes-agent/pull/37747) —— @teknium1） ## 🏗️ 核心代理与架构 ### 提供者与模型支持 - 新模型：`deepseek-v4-flash`（+ 修剪变体，按制造者分组的精选列表）、具有 100 万上下文的 `MiniMax-M3`（原生 minimax 提供者）、`qwen3.7-plus`（Nous + OpenRouter）、`gemini-3.5-flash`（加入 Gemini OAuth + API 密钥选择器）（[#35659](https://github.com/NousResearch/hermes-agent/pull/35659)、[#36214](https://github.com/NousResearch/hermes-agent/pull/36214)、[#39409](https://github.com/NousResearch/hermes-agent/pull/39409)、[#37046](https://github.com/NousResearch/hermes-agent/pull/37046) —— @teknium1） - 模型选择器：在 WebUI/TUI/CLI 中模糊搜索；将多端点提供者归为一行；刷新提供者描述 + 描述分组行；每小时刷新目录；停止将 OpenAI 选择路由到 OpenRouter（[#36928](https://github.com/NousResearch/hermes-agent/pull/36928) —— @kshitijk4poor，[#35227](https://github.com/NousResearch/hermes-agent/pull/35227)、[#35773](https://github.com/NousResearch/hermes-agent/pull/35773)、[#35756](https://github.com/NousResearch/hermes-agent/pull/35756)、[#37175](https://github.com/NousResearch/hermes-agent/pull/37175)） - 始终显示 Nous Tool Gateway 后端，选择时登录；展示 Nous 免费工具池（授权 + 设置提示）；通过管理的 Nous 网关路由 FAL 视频生成（[#35792](https://github.com/NousResearch/hermes-agent/pull/35792) —— @teknium1，[#36153](https://github.com/NousResearch/hermes-agent/pull/36153)、[#33259](https://github.com/NousResearch/hermes-agent/pull/33259) —— @alt-glitch） - 将会话中的模型切换持久化到数据库；在中断后恢复轮次中恢复模型（[#35256](https://github.com/NousResearch/hermes-agent/pull/35256)、[#35381](https://github.com/NousResearch/hermes-agent/pull/35381) —— @teknium1） - 凭据池：为终端 OAuth 失败添加 `STATUS_DEAD`；隔离自定义提供者选择器凭据（[#34412](https://github.com/NousResearch/hermes-agent/pull/34412)、[#34810](https://github.com/NousResearch/hermes-agent/pull/34810) —— @teknium1） - 禁用 Nous Portal 遗留会话密钥推断回退——仅限 JWT（[#34508](https://github.com/NousResearch/hermes-agent/pull/34508) —— @kshitijk4poor） ### 代理循环、提示与工具 - `/undo [N]`——回退 N 轮用户对话，带预填和软删除（CLI/TUI + 消息平台同步）（[#36229](https://github.com/NousResearch/hermes-agent/pull/36229)、[#36699](https://github.com/NousResearch/hermes-agent/pull/36699) —— @teknium1） - 渐进式工具披露用于 MCP 和插件工具（限定范围）；用于系统提示的嵌入器环境提示钩子；通用任务完成指导 + 本地 Python 工具链探测（[#34493](https://github.com/NousResearch/hermes-agent/pull/34493)、[#34574](https://github.com/NousResearch/hermes-agent/pull/34574)、[#34340](https://github.com/NousResearch/hermes-agent/pull/34340) —— @teknium1） - 取消委派 `max_spawn_depth` 上限（下限为 1，无上限）；拓宽 Hermes 自我知识指针到文档 + 技能；`hermes prompt-size` 诊断（[#39772](https://github.com/NousResearch/hermes-agent/pull/39772)、[#38538](https://github.com/NousResearch/hermes-agent/pull/38538)、[#35276](https://github.com/NousResearch/hermes-agent/pull/35276) —— @teknium1） - `perf(read_file)`：紧凑行号边距——每次读取大约减少 14% 的令牌（现在是唯一格式）（[#35368](https://github.com/NousResearch/hermes-agent/pull/35368)、[#35532](https://github.com/NousResearch/hermes-agent/pull/35532) —— @teknium1） - 从 `TERMINAL_CWD` 通过一个读取器解析代理 cwd；对齐预填消息键处理；恢复时在会话的原始工作目录中重新启动（[#35028](https://github.com/NousResearch/hermes-agent/pull/35028) —— @banditburai，[#38760](https://github.com/NousResearch/hermes-agent/pull/38760) —— @helix4u，[#38562](https://github.com/NousResearch/hermes-agent/pull/38562) —— @teknium1） - 防止因并发压缩导致的会话 ID 分叉；观察者遥测钩子 + NeMo-Relay 插件（门控工具发送）（[#34351](https://github.com/NousResearch/hermes-agent/pull/34351)、[#38232](https://github.com/NousResearch/hermes-agent/pull/38232) —— @teknium1） ### 会话、状态与记忆 - `perf(state)`：在 VACUUM 时合并 FTS5 段 + `hermes sessions optimize`；在父会话重新打开后保持 `/branch` 会话可见；在缺少 FTS5 运行时幸存（[#34596](https://github.com/NousResearch/hermes-agent/pull/34596)、[#39214](https://github.com/NousResearch/hermes-agent/pull/39214) —— @teknium1、@kshitijk4poor，[#35452](https://github.com/NousResearch/hermes-agent/pull/35452) —— @helix4u） - Honcho：使启动时容错打开；加固自托管设置路径（[#24847](https://github.com/NousResearch/hermes-agent/pull/24847) —— @stephenschoettler，[#35170](https://github.com/NousResearch/hermes-agent/pull/35170) —— @kshitijk4poor） - Supermemory：会话级摄取 + kebab 别名（[#38756](https://github.com/NousResearch/hermes-agent/pull/38756) —— @alt-glitch） ## 🧩 多代理（看板）与技能 ### 看板 - `goal_mode` 卡片在 `/goal` 循环中运行工作节点；任务上的文件附件；将任务体中引用的图片附加到工作节点视觉（[#35710](https://github.com/NousResearch/hermes-agent/pull/35710)、[#35395](https://github.com/NousResearch/hermes-agent/pull/35395)、[#34210](https://github.com/NousResearch/hermes-agent/pull/34210) —— @teknium1） - `default_assignee` 回退 + 每配置文件并发上限；`POST /runs/{run_id}/terminate` 端点；通过 `dispatch_in_gateway` 门控通知器监视器（[#34244](https://github.com/NousResearch/hermes-agent/pull/34244)、[#34449](https://github.com/NousResearch/hermes-agent/pull/34449)、[#37174](https://github.com/NousResearch/hermes-agent/pull/37174) —— @teknium1） - CLI 调度配置传递 + 人性化技能交换（[#34337](https://github.com/NousResearch/hermes-agent/pull/34337) —— @teknium1） ### 技能 - **更精简的默认技能集**（[#39028](https://github.com/NousResearch/hermes-agent/pull/39028) —— @teknium1）： - 移除（冗余/已死）：`spotify`（→ Spotify 插件的 7 个原生工具）、`linear`（→ `hermes mcp install linear`）、`kanban-codex-lane`、`debugging-hermes-tui-commands`、过时的 `domain` 孤儿、空分类标记（`diagramming`、`gifs`、`inference-sh`、`mlops/training`、`mlops/vector-databases`） - 从捆绑变为可选：`baoyu-article-illustrator`、`baoyu-comic`、`creative-ideation`、`pixel-art`、`dspy`、`subagent-driven-development`、`minecraft-modpack-server`、`pokemon-player`、`hermes-s6-container-supervision` - 合并：`webhook-subscriptions` + `native-mcp` 被折叠到 `hermes-agent` 技能中作为按需引用；`writing-plans` 合并到 `plan`（v2.0.0） - 新的 `environments:` 前文相关性门控（`kanban` / `docker` / `s6`）——特定上下文的技能将不再出现在索引中供不使用它们的用户使用，但在明确请求时仍然加载 - 策展人现在可以修剪未使用的**内置**技能（不仅限于代理创建的），每个技能的使用情况都会被跟踪（[#36701](https://github.com/NousResearch/hermes-agent/pull/36701) —— @teknium1） - 空白技能支持——`install --no-skills` + 默认配置文件的 opt-out/opt-in（[#36228](https://github.com/NousResearch/hermes-agent/pull/36228) —— @teknium1） - **NVIDIA/skills 受信任源**——`NVIDIA/skills` 现在是 Skills Hub 中的默认受信任源，与 OpenAI/Anthropic/HuggingFace 并列；`skills.sh.json` 侧车为源提供真实的分类标签（[#34333](https://github.com/NousResearch/hermes-agent/pull/34333) —— @teknium1） - Skills Hub：修复浏览上限，添加来源链接 + 复制按钮 + 分类清理（[#37143](https://github.com/NousResearch/hermes-agent/pull/37143) —— @teknium1） - 新的可选技能：`grok`（xAI Grok Build CLI）、`antigravity-cli` 操作符（在 autonomous-ai-agents 下）（[#34582](https://github.com/NousResearch/hermes-agent/pull/34582)、[#34583](https://github.com/NousResearch/hermes-agent/pull/34583)、[#34604](https://github.com/NousResearch/hermes-agent/pull/34604) —— @teknium1） ## 📱 消息平台（网关） - 结构化流事件协议 + Telegram 草稿格式同步；按平台流式传输默认值（Telegram 开启，Discord 关闭）+ 仪表盘开关；在 `DEFAULT_CONFIG` 中显示网关流式传输块（[#37250](https://github.com/NousResearch/hermes-agent/pull/37250)、[#37303](https://github.com/NousResearch/hermes-agent/pull/37303)、[#37285](https://github.com/NousResearch/hermes-agent/pull/37285) —— @teknium1） - Discord 语音频道混音器——环境空闲背景音 + 与 TTS 重叠的口头确认；在单独切换时解释 `/voice` 用法（[#39659](https://github.com/NousResearch/hermes-agent/pull/39659)、[#39766](https://github.com/NousResearch/hermes-agent/pull/39766) —— @teknium1） - 处理飞书会议邀请；bluebubbles 群组提及门控；matrix bang-命令别名；matrix 关闭失败时的审批反应认证（[#39040](https://github.com/NousResearch/hermes-agent/pull/39040)、[#37091](https://github.com/NousResearch/hermes-agent/pull/37091)、[#38175](https://github.com/NousResearch/hermes-agent/pull/38175) —— @teknium1、@alt-glitch，[#34567](https://github.com/NousResearch/hermes-agent/pull/34567)） - 干净的服务重启流程；在重新连接失败时关闭 ResponseStore + 处置无主适配器；微信 `asyncio.wait_for` 超时（[#36188](https://github.com/NousResearch/hermes-agent/pull/36188) —— @helix4u，[#37679](https://github.com/NousResearch/hermes-agent/pull/37679) —— @Fearvox，[#35117](https://github.com/NousResearch/hermes-agent/pull/35117) —— @banditburai） ## 🖥️ CLI、TUI 与设置 - 可配置的默认界面（cli 或 tui）+ `--cli` 标志；TUI 单个 `/model` 命令 + 统一的 Sessions 覆盖层；在委派开始时引导用户至 `/agents` 仪表盘（[#37782](https://github.com/NousResearch/hermes-agent/pull/37782)、[#37112](https://github.com/NousResearch/hermes-agent/pull/37112) —— @OutThisLife，[#34704](https://github.com/NousResearch/hermes-agent/pull/34704) —— @kshitijk4poor） - 精简设置——通过 Nous Portal 快速设置 + 完整设置默认值；内联解释快速与完整设置；`hermes portal` 人类可读的 Portal 入门别名 + 完整的快速设置 Nous 流程（[#35723](https://github.com/NousResearch/hermes-agent/pull/35723)、[#36227](https://github.com/NousResearch/hermes-agent/pull/36227)、[#38449](https://github.com/NousResearch/hermes-agent/pull/38449)、[#38465](https://github.com/NousResearch/hermes-agent/pull/38465) —— @teknium1、@kshitijk4poor） - 在 ps/top/htop 中将进程标题设置为 `hermes`；警告不受支持的 pip 安装 + 修复过时的更新检查缓存（[#35143](https://github.com/NousResearch/hermes-agent/pull/35143)、[#34846](https://github.com/NousResearch/hermes-agent/pull/34846) —— @teknium1） - TUI 性能：停止慢/死的 MCP 服务器冻结启动；停止急切的 MCP 发现阻塞具备代理能力的启动；停止在跟踪行中持久化完整工具输出（静默 OOM）（[#35273](https://github.com/NousResearch/hermes-agent/pull/35273)、[#35397](https://github.com/NousResearch/hermes-agent/pull/35397)、[#38224](https://github.com/NousResearch/hermes-agent/pull/38224) —— @teknium1） - TUI 修复：在网关意外死亡时自动恢复会话；重新组装分割的 SGR 鼠标序列；保留 PowerShell 剪贴板中的 UTF-8；退出时重置终端输入模式；`/save` 快照保存在 Hermes 主目录下（[#35893](https://github.com/NousResearch/hermes-agent/pull/35893)、[#38564](https://github.com/NousResearch/hermes-agent/pull/38564) —— @OutThisLife，[#35222](https://github.com/NousResearch/hermes-agent/pull/35222) —— @teknium1，[#36864](https://github.com/NousResearch/hermes-agent/pull/36864) —— @maxmilian，[#38251](https://github.com/NousResearch/hermes-agent/pull/38251) —— @austinpickett） - 设置模型/提供者选择器从 `simple_term_menu` 迁移到 curses（ESC + 幻影行修复）；默认浏览器/TTS 选择器指向免费本地后端，而非付费 Nous（[#35806](https://github.com/NousResearch/hermes-agent/pull/35806)、[#37800](https://github.com/NousResearch/hermes-agent/pull/37800) —— @teknium1） ## 🔧 工具系统与安装程序 - `single managed-uv path`，删除 fts5 安装程序升级；安装程序提交锁定可选（默认跟随分支）；浅克隆（[#37660](https://github.com/NousResearch/hermes-agent/pull/37660) —— @ethernet8023，[#37123](https://github.com/NousResearch/hermes-agent/pull/37123) —— @jquesnelle，[#39423](https://github.com/NousResearch/hermes-agent/pull/39423) —— @ethernet8023） - `ensure_uv()` 在更新边界存活（首次运行不崩溃）；加固 venv 重建 + 安装后验证核心依赖；在执行破坏性清理前要求 managed 标记；默认在非交互式更新中 stash/restore（[#39780](https://github.com/NousResearch/hermes-agent/pull/39780) —— @OutThisLife，[#38887](https://github.com/NousResearch/hermes-agent/pull/38887)、[#39568](https://github.com/NousResearch/hermes-agent/pull/39568)、[#39645](https://github.com/NousResearch/hermes-agent/pull/39645) —— @teknium1、@helix4u） - MCP：当未获得令牌时停止报告错误的 OAuth 成功；视觉功能尊重 `model.supports_vision` 在 `vision_analyze` + `browser_vision` 中；MiniMax t2a_v2 TTS `raise_for_status`（[#34807](https://github.com/NousResearch/hermes-agent/pull/34807)、[#34562](https://github.com/NousResearch/hermes-agent/pull/34562)、[#39057](https://github.com/NousResearch/hermes-agent/pull/39057) —— @teknium1） - yuanbao：按 resourceId 缓存解析的媒体资源（[#34474](https://github.com/NousResearch/hermes-agent/pull/34474) —— @teknium1） ## 🐳 Docker 与部署 - 容器复用 + 有限同步清理 + 孤儿回收器；自动加入 Docker socket 组用于 docker-in-docker 后端；启动非 root 容器（当已经无特权时跳过 s6-setuidgid 降权）（[#33645](https://github.com/NousResearch/hermes-agent/pull/33645)、[#34407](https://github.com/NousResearch/hermes-agent/pull/34407) —— @benbarclay，[#34837](https://github.com/NousResearch/hermes-agent/pull/34837) —— @IAvecilla） - 当卷所有权与重新映射的 UID 匹配时跳过不必要的启动 chown；在首次启动时从 `HERMES_GATEWAY_BOOTSTRAP_STATE` 种子化 `gateway_state.json`；用 hermes-agent 标签标记容器以供识别（[#35027](https://github.com/NousResearch/hermes-agent/pull/35027) —— @Foldblade，[#37896](https://github.com/NousResearch/hermes-agent/pull/37896) —— @benbarclay） - 通过 `HERMES_TUI_DIR` 将 TUI 启动器指向预构建包；合并 node/nix 工作区锁定文件 + 更新所有消费者（[#37923](https://github.com/NousResearch/hermes-agent/pull/37923)、[#36171](https://github.com/NousResearch/hermes-agent/pull/36171) —— @benbarclay、@ethernet8023） ## 🔒 安全与可靠性 - **CVE-2026-48710（Starlette BadHost）**——锁定已修补的 Starlette ≥1.0.1（[#35118](https://github.com/NousResearch/hermes-agent/pull/35118) —— @teknium1） - 在异步路径中，在事件循环外运行 URL SSRF 检查；从子进程环境变量中剥离 Bedrock 推断 Bearer 令牌；将 `bws_cache.json` 添加到文件安全读取防护中；在变异验证器脚注中中和文件路径（[#39046](https://github.com/NousResearch/hermes-agent/pull/39046)、[#34498](https://github.com/NousResearch/hermes-agent/pull/34498)、[#34421](https://github.com/NousResearch/hermes-agent/pull/34421)、[#35684](https://github.com/NousResearch/hermes-agent/pull/35684) —— @teknium1） - 在 `execute_code` 中恢复审批/sudo 上下文 + 守卫入口点；将 docker restart/stop/kill 添加到 `DANGEROUS_PATTERNS`；在审查过的技能内容中净化不可见 unicode；在就地 xAI 突变前深拷贝工具（[#34497](https://github.com/NousResearch/hermes-agent/pull/34497)、[#33438](https://github.com/NousResearch/hermes-agent/pull/33438) —— @teknium1、@Sarbai，[#37245](https://github.com/NousResearch/hermes-agent/pull/37245)、[#34416](https://github.com/NousResearch/hermes-agent/pull/34416)） - 对每个任务的 `.hermes` 镜像的写入进行沙箱镜像软防护；Honcho 启动时容错打开（[#32213](https://github.com/NousResearch/hermes-agent/pull/32213) —— @briandevans，[#24847](https://github.com/NousResearch/hermes-agent/pull/24847) —— @stephenschoettler） ## 🐛 值得注意的错误修复 - **此轮关闭 399 个议题**（2 个 P0，62 个 P1，16 个安全标记，262 个 bug 标签）。 - 桌面：防止正在进行的对话在刷新时消失；当一轮对话已排队时，Stop 按钮实际中断；阻止后台会话消息渗入活动的对话记录；斜杠/@ 补全菜单可导航且 Esc 可关闭；IME 回车不再分割消息（[#37908](https://github.com/NousResearch/hermes-agent/pull/37908) —— @OutThisLife，[#37948](https://github.com/NousResearch/hermes-agent/pull/37948)、[#37975](https://github.com/NousResearch/hermes-agent/pull/37975)、[#37937](https://github.com/NousResearch/hermes-agent/pull/37937) —— @kshitijk4poor，[#38333](https://github.com/NousResearch/hermes-agent/pull/38333) —— @stremtec） - 更新：防止 stash/restore 在受管理克隆中覆盖桌面源码；不在重建中途令桌面重建/技能同步因 venv 而失败；将启动器虚拟环境导出到 uv（[#38542](https://github.com/NousResearch/hermes-agent/pull/38542)、[#38885](https://github.com/NousResearch/hermes-agent/pull/38885)、[#35224](https://github.com/NousResearch/hermes-agent/pull/35224) —— @teknium1） - 语音：在 PortAudio 回退中尊重 `PIPEWIRE_REMOTE`；当声音服务器可达时允许通过 SSH 使用 `/voice`；恢复 mistralai（2.4.8 干净，禁令解除）（[#33473](https://github.com/NousResearch/hermes-agent/pull/33473) —— @Dusk1e，[#35719](https://github.com/NousResearch/hermes-agent/pull/35719)、[#34841](https://github.com/NousResearch/hermes-agent/pull/34841) —— @teknium1） ## 📚 文档 - 新的桌面应用指南 + 远程后端部分（会话令牌、`--tui` 要求、用户名/密码连接、仪表盘/网关先决条件）（[#37457](https://github.com/NousResearch/hermes-agent/pull/37457)、[#38144](https://github.com/NousResearch/hermes-agent/pull/38144)、[#38180](https://github.com/NousResearch/hermes-agent/pull/38180)、[#38534](https://github.com/NousResearch/hermes-agent/pull/38534)、[#39128](https://github.com/NousResearch/hermes-agent/pull/39128) —— @teknium1） - 仪表盘认证提供者适用性 + 在仪表盘/Docker/桌面上的注册；Docker 的网络出口隔离指南（[#39633](https://github.com/NousResearch/hermes-agent/pull/39633) —— @benbarclay，[#26385](https://github.com/NousResearch/hermes-agent/pull/26385) —— @Manzela） ---  ## 👥 贡献者 衷心感谢在此版本中贡献了工作的 **170 位社区贡献者**（含合著者）。 ### 核心 - @teknium1 ### 顶级社区贡献者（按合并 PR 数量） - **@OutThisLife**（52）——从头到尾构建了桌面应用：安装、自更新、远程网关连接、多配置文件会话、聊天 UX、状态栏模型选择器 - **@benbarclay**（44）——Docker 加固、仪表盘认证（OIDC、用户名/密码、刷新令牌轮换）、桌面 OAuth 远程连接 - **@kshitijk4poor**（29）——模糊模型选择器、设置/Portal 入门、桌面补全菜单和 Stop 按钮修复、honcho 加固 - **@ethernet8023**（18）——桌面构建管道（内容哈希戳、构建标志）、Linux/arm64 桌面支持、managed-uv 合并 - **@austinpickett**（8）——仪表盘 nous-blue 主题 + 批量会话、桌面 Providers 设置、GUI 生活质量分类 - **@alt-glitch**（7）——Nous 工具池权限展示、FAL 视频生成管理网关、supermemory 会话摄取、matrix 别名 - **@helix4u**（6）——网关服务重启流程、更新破坏性清理防护、配置预填对齐 - **@jquesnelle**（4）——安装程序提交锁定可选、桌面技能/工具 + 模型管理合并 - **@JimLiu**——完整的简体中文桌面翻译 + 类型化 i18n 层 ### 所有贡献者 @0xharryriddle, @a1245582339, @adybag14-cyber, @AhmetArif0, @alaamohanad169-ship-it, @alelpoan, @alt-glitch, @Aman113114-IITD, @amathxbt, @aminvakil, @annguyenNous, @aqilaziz, @Archerouyang, @ashishpatel26, @asimons81, @austinpickett, @banditburai, @baofuen, @Bartok9, @bbednarski9, @beardthelion, @bedirhancode, @benbarclay, @benfrank241, @blackpilledsoftware-prog, @bluefishs, @brian-doherty, @briancl2, @briandevans, @BROCCOLO1D, @caojiguang, @CharZhou, @counterposition, @coygeek, @CryptoByz, @davetist, @davidgut1982, @dchenk, @dirtyren, @donovan-yohan, @dparikh79, @dskwe, @Dusk1e, @dvir-pashut, @EloquentBrush0x, @emozilla, @ErnestHysa, @Erosika, @ethernet8023, @f3rs3n, @faisfamilytravel, @Fearvox, @ferminquant, @fesalfayed, @firefly, @flooryyyy, @Foldblade, @Frowtek, @gbarany, @Glucksberg, @GodsBoy, @haran2001, @harjothkhara, @HashClawAI, @hayka-pacha, @helix4u, @HeLLGURD, @Hinotoi-agent, @hllqkb, @IAvecilla, @inchargeautomation-lab, @Interstellar-code, @isair, @ITheEqualizer, @jeffrobodie-glitch, @JezzaHehn, @JimLiu, @Julientalbot, @julio-cloudvisor, @karmeleon, @Kenmege, @Kewe63, @Kolektori, @konsisumer, @kshitijk4poor, @kurobaryo, @kweiner, @kyssta-exe, @Kyzcreig, @LengR, @leonardsellem, @LeonSGP43, @libre-7, @liuboacean, @liuhao1024, @loongzhao, @luyao618, @MaheshtheDev, @Manzela, @mathijsvandenhurk, @MattMaximo, @maxcz79, @maxmilian, @Moikapy, @MustafaKara7, @nateGeorge, @nepenth, @nielskaspers, @ninjmnky, @octavioturra, @OCWC22, @ohMyJason, @ousiaresearch, @outsourc-e, @OutThisLife, @Pluviobyte, @polnikale, @pxdsgnco, @Que0x, @r266-tech, @redpiggy-cyber, @rewbs, @rexdotsh, @rob-maron, @SaguaroDev, @sahibzada-allahyar, @Sarbai, @sarvesh1327, @scubamount, @SeaXen, @seppegadeyne, @shannonsands, @SHL0MS, @SimoKiihamaki, @SiTaggart, @solaitken, @sprmn24, @stephenschoettler, @steveonjava, @stremtec, @Subway2023, @sweetcornna, @Sylw3ster, @ThyFriendlyFox, @tillfalko, @tmchow, @TonyPepeBear, @Tranquil-Flow, @truenorth-lj, @tuancookiez-hub, @Twanislas, @tymrtn, @uzunkuyruk, @ViewWay, @VinciZhu, @vinoth12940, @vladkvlchk, @vynxevainglory-ai, @WadydX, @wenchengxucool, @whyhkzk, @worlldz, @Wysie, @x1am1, @xxxigm, @ygd58, @youngstar-eth, @zapabob, @zhaoleibd, @Zyrixtrex --- **完整更新日志**：[v2026.5.29.2...v2026.6.5](https://github.com/NousResearch/hermes-agent/compare/v2026.5.29.2...v2026.6.5)

# Hermes Agent v0.16.0 (v2026.6.5) **Release Date:** June 5, 2026 **Since v0.15.2:** 874 commits · 542 merged PRs · 1,962 files changed · 205,216 insertions · 46,217 deletions · 399 issues closed (2 P0, 62 P1, 16 security-tagged) · 170 community contributors (including co-authors) > **The Surface Release.** Hermes meets you wherever you work. A brand-new native desktop app — built across 100 PRs and 159 commits in a single week — gives you Hermes as a real macOS/Linux/Windows application: one-click install, in-app self-update, drag-and-drop files into chat, an inline model picker in the status bar, concurrent multi-profile sessions, a full Simplified Chinese translation, and the ability to connect to a remote Hermes gateway over OAuth or username/password. Alongside it, the web dashboard grew a full browser-based administration panel (MCP catalog, messaging channels, credentials, webhooks, memory, pluggable OIDC / username-password login), and first-time setup got a "Quick Setup via Nous Portal" path that gets you from install to first message in seconds. The default skill set was trimmed to what you actually need, NVIDIA/skills joined the trusted Skills Hub taps, the model picker is now fuzzy-searchable everywhere (desktop, web, TUI, CLI), and `/undo` finally lets you take back the last N turns. 2 P0 and 62 P1 closures ride along, plus a security round (CVE-2026-48710 Starlette pin, SSRF off-loop hardening, subprocess credential stripping). --- ## ✨ Highlights - **Hermes Desktop — a real native app, not a terminal wrapper** — This is the headline. There's now a `apps/desktop/` Electron application that installs like any other desktop app on macOS, Linux, and Windows, updates itself in place from inside the app, and gives you a polished GUI for everything Hermes does. You get a proper chat window with streaming, a session list you can archive and search, drag-and-drop files anywhere in the chat area, clipboard image paste, a Cmd+K command palette, and a model picker right in the status bar. If you've been telling friends "it's a CLI agent" and watching their eyes glaze over — now you can just send them an installer. None of this existed a week ago. ([#20059](https://github.com/NousResearch/hermes-agent/pull/20059), [#35607](https://github.com/NousResearch/hermes-agent/pull/35607), [#37099](https://github.com/NousResearch/hermes-agent/pull/37099), [#37379](https://github.com/NousResearch/hermes-agent/pull/37379), [#38631](https://github.com/NousResearch/hermes-agent/pull/38631) — @OutThisLife, @jquesnelle, @ethernet8023, @austinpickett, @benbarclay) - **Run the desktop app against a remote Hermes — sign in with OAuth or username/password** — The desktop app doesn't have to run Hermes locally. Point it at a remote Hermes gateway (your homelab, a hosted box, a teammate's server) and it connects over a secure WebSocket, authenticating with OAuth or a username/password login — no fiddling with `--insecure` flags or hand-copied session tokens. Each profile can target its own remote host, and you can run concurrent sessions across multiple profiles in one window with cross-profile `@session` links. The practical version: your laptop runs a thin GUI, the heavy agent runs wherever your API keys and compute live. ([#37888](https://github.com/NousResearch/hermes-agent/pull/37888), [#38851](https://github.com/NousResearch/hermes-agent/pull/38851), [#39330](https://github.com/NousResearch/hermes-agent/pull/39330), [#39778](https://github.com/NousResearch/hermes-agent/pull/39778) — @benbarclay, @OutThisLife, @teknium1) - **The web dashboard is now a full admin panel — configure everything from the browser** — The dashboard grew from "view your sessions" into a complete administration surface. There's a Channels page that sets up every gateway messaging platform (Telegram, Discord, Slack, etc.) from the browser, an admin panel for the MCP catalog with enable/disable toggles, credential management, webhook and hook creation, memory configuration, gateway controls, and a System page with check-before-update and one-click Debug Share. You no longer have to SSH in and edit `config.yaml` to wire up a new messaging channel or MCP server — it's all point-and-click now. ([#36704](https://github.com/NousResearch/hermes-agent/pull/36704), [#36736](https://github.com/NousResearch/hermes-agent/pull/36736), [#37211](https://github.com/NousResearch/hermes-agent/pull/37211), [#38205](https://github.com/NousResearch/hermes-agent/pull/38205), [#38600](https://github.com/NousResearch/hermes-agent/pull/38600) — @teknium1) - **Hermes Desktop speaks Simplified Chinese — full 简体中文 in the chat GUI** — The desktop app now ships a complete Simplified Chinese (简体中文) translation across every UI surface — the chat window itself, sidebar, settings, command center, cron, messaging, profiles, skills, agents, the lot. English stays the default; switch language in Appearance settings and the choice persists to your config (`display.language`). It's built on a proper typed i18n layer, so adding more languages from here is straightforward. ([#38241](https://github.com/NousResearch/hermes-agent/pull/38241) — @JimLiu) - **Leaner default skill set — Hermes ships only what you actually need** — The bundled skill set got a deliberate trim. Skills that were redundant or dead are gone (`spotify` — superseded by the native Spotify plugin's 7 tools; `linear` — superseded by `hermes mcp install linear`; `kanban-codex-lane`, `debugging-hermes-tui-commands`, a stale `domain` orphan, and several empty category markers). Heavier or niche skills moved from bundled to optional (the Baoyu creative set, `dspy`, `subagent-driven-development`, `minecraft-modpack-server`, `pokemon-player`, `hermes-s6-container-supervision`) — still one `hermes skills install` away, just not loaded by default. And a new `environments:` relevance gate keeps context-specific skills (kanban, docker/s6) out of the skills index for users who'll never use them, while still loading them on explicit request. The result: a smaller, sharper default skill list, less noise in the picker, and a lighter prompt. The curator can now prune unused **built-in** skills too (not just agent-created ones), with usage tracked for every skill. ([#39028](https://github.com/NousResearch/hermes-agent/pull/39028), [#36701](https://github.com/NousResearch/hermes-agent/pull/36701), [#36228](https://github.com/NousResearch/hermes-agent/pull/36228) — @teknium1) - **NVIDIA/skills is now a built-in trusted skills tap** — `NVIDIA/skills` joins OpenAI, Anthropic, and HuggingFace as a default trusted tap in the Skills Hub — discoverable, browsable, searchable, and auto-updating through the same pipeline. NVIDIA's verified skills for CUDA-X, AIQ, cuOpt and the rest of their product stack are one install away, with real category labels from the repo's `skills.sh.json` sidecar. ([#34333](https://github.com/NousResearch/hermes-agent/pull/34333) — @teknium1) - **Quick Setup via Nous Portal — from install to first message in seconds** — First-time setup was thinned down to two clear paths: Quick Setup (sign in with Nous Portal, get a model picker, start chatting immediately) or Full Setup (the detailed wizard for power users). `hermes portal` is now the human-readable alias that runs the full quick-setup Nous flow. The first-run menu explains the difference inline so newcomers aren't guessing. The goal: a brand-new user shouldn't need to read docs to send their first message. ([#35723](https://github.com/NousResearch/hermes-agent/pull/35723), [#36227](https://github.com/NousResearch/hermes-agent/pull/36227), [#38449](https://github.com/NousResearch/hermes-agent/pull/38449), [#38465](https://github.com/NousResearch/hermes-agent/pull/38465) — @teknium1, @kshitijk4poor) - **Fuzzy model picker, everywhere — type a few letters, find your model** — The model picker now does fuzzy search across the desktop app, web dashboard, TUI, and CLI. Type "v4fl" and `deepseek-v4-flash` surfaces; multi-endpoint providers are grouped under one row instead of cluttering the list with duplicates, and each row carries a description so you know what you're picking. The catalog refreshes hourly instead of daily, so new models show up the same day they launch. New this window: `deepseek-v4-flash`, `MiniMax-M3` with 1M context, `qwen3.7-plus`. ([#36928](https://github.com/NousResearch/hermes-agent/pull/36928), [#35227](https://github.com/NousResearch/hermes-agent/pull/35227), [#35756](https://github.com/NousResearch/hermes-agent/pull/35756), [#35659](https://github.com/NousResearch/hermes-agent/pull/35659), [#36214](https://github.com/NousResearch/hermes-agent/pull/36214) — @kshitijk4poor, @teknium1) - **`/undo [N]` — take back the last N turns** — Said the wrong thing, or sent the agent down a bad path? `/undo` backs up N user turns, prefills your last message so you can edit and resend, and soft-deletes the turns in between. It works in the CLI, the TUI, and across messaging platforms (Telegram, Discord, etc.) with full parity. Closes a long-standing request ([#21910](https://github.com/NousResearch/hermes-agent/issues/21910)). ([#36229](https://github.com/NousResearch/hermes-agent/pull/36229), [#36699](https://github.com/NousResearch/hermes-agent/pull/36699) — @teknium1) - **Choose your default interface — `cli` or `tui`** — You can now set whether `hermes chat` drops you into the classic CLI or the Ink TUI by default, with a `--cli` flag to override per-invocation. The TUI also got a single unified `/model` command and a Sessions overlay for switching between live sessions. Use the interface you actually like. ([#37782](https://github.com/NousResearch/hermes-agent/pull/37782), [#37112](https://github.com/NousResearch/hermes-agent/pull/37112) — @OutThisLife) --- ## 🖥️ Hermes Desktop App (NEW) ### Install & lifecycle - macOS desktop install + in-app self-update; rebuild-and-relaunch cleanly on macOS ([#35607](https://github.com/NousResearch/hermes-agent/pull/35607), [#36198](https://github.com/NousResearch/hermes-agent/pull/36198), [#38296](https://github.com/NousResearch/hermes-agent/pull/38296) — @OutThisLife) - macOS installer renamed to "Hermes" and turned into a launcher ([#37516](https://github.com/NousResearch/hermes-agent/pull/37516) — @OutThisLife) - Build desktop in its own `desktop` stage on macOS/Linux instead of silently skipping; content-hash build stamp, `--build-only` / `--force-build` flags ([#36134](https://github.com/NousResearch/hermes-agent/pull/36134) — @OutThisLife, [#37597](https://github.com/NousResearch/hermes-agent/pull/37597) — @ethernet8023) - Boot-failure recovery + live API-key validation; cancellable install; recover from corrupt cached Electron download ([#35864](https://github.com/NousResearch/hermes-agent/pull/35864), [#37379](https://github.com/NousResearch/hermes-agent/pull/37379), [#39032](https://github.com/NousResearch/hermes-agent/pull/39032) — @OutThisLife, @teknium1) - Windows: recover from corrupt Electron cache in bootstrap install; stop racing our own backend during in-app update ([#39465](https://github.com/NousResearch/hermes-agent/pull/39465), [#39828](https://github.com/NousResearch/hermes-agent/pull/39828) — @teknium1, @OutThisLife) - Linux: configure Electron sandbox helper; detect linux arm64 binary; disable GPU acceleration on remote displays to stop flicker ([#37691](https://github.com/NousResearch/hermes-agent/pull/37691), [#38594](https://github.com/NousResearch/hermes-agent/pull/38594), [#37932](https://github.com/NousResearch/hermes-agent/pull/37932) — @ethernet8023, @OutThisLife) - Require Node ≥20.19/22.12 for the desktop build; zero eslint/typecheck debt + prettier pass ([#38255](https://github.com/NousResearch/hermes-agent/pull/38255), [#39100](https://github.com/NousResearch/hermes-agent/pull/39100) — @OutThisLife) ### Remote-gateway & multi-profile - Connect to OAuth-gated remote gateways; username/password login for remote gateways; per-profile remote gateway hosts ([#37888](https://github.com/NousResearch/hermes-agent/pull/37888), [#38851](https://github.com/NousResearch/hermes-agent/pull/38851), [#39778](https://github.com/NousResearch/hermes-agent/pull/39778) — @benbarclay, @teknium1, @OutThisLife) - Concurrent multi-profile sessions, cross-profile `@session` links; re-mint OAuth WS ticket on gateway reconnect; gate OAuth remote connect on AT-or-RT ([#39330](https://github.com/NousResearch/hermes-agent/pull/39330), [#38886](https://github.com/NousResearch/hermes-agent/pull/38886), [#39464](https://github.com/NousResearch/hermes-agent/pull/39464) — @OutThisLife, @teknium1, @benbarclay) - Offer remote sign-in on a gated-gateway boot failure; validate live WebSocket in remote gateway test ([#39402](https://github.com/NousResearch/hermes-agent/pull/39402), [#39511](https://github.com/NousResearch/hermes-agent/pull/39511) — @teknium1) ### Chat UX & settings - Drop files anywhere in the chat area; clipboard image paste with dedupe; attachments on Enter, IME composition handling ([#36262](https://github.com/NousResearch/hermes-agent/pull/36262), [#38306](https://github.com/NousResearch/hermes-agent/pull/38306), [#38677](https://github.com/NousResearch/hermes-agent/pull/38677) — @OutThisLife, @teknium1) - Background needs-input indicator, clarify redesign, Cmd+K palette & UI consistency pass; inline model picker in the status bar; YOLO toggle in the status bar (TUI parity) ([#38631](https://github.com/NousResearch/hermes-agent/pull/38631), [#37738](https://github.com/NousResearch/hermes-agent/pull/37738), [#38517](https://github.com/NousResearch/hermes-agent/pull/38517) — @OutThisLife) - Session-list overhaul, session hygiene/archive, media streaming + connecting overlay; search sessions by id (SQL-bounded) ([#37379](https://github.com/NousResearch/hermes-agent/pull/37379), [#37099](https://github.com/NousResearch/hermes-agent/pull/37099), [#39062](https://github.com/NousResearch/hermes-agent/pull/39062) — @OutThisLife, @teknium1) - Dedicated Providers settings + polished Accounts/API-keys UX; consolidate skills + tools management into one pane; move model management into Settings ([#38551](https://github.com/NousResearch/hermes-agent/pull/38551) — @austinpickett, [#37310](https://github.com/NousResearch/hermes-agent/pull/37310), [#37330](https://github.com/NousResearch/hermes-agent/pull/37330) — @jquesnelle) - Render approval/sudo/secret prompts so tools stop silently timing out; surface skill & quick-command slash commands in the palette; first-class xAI Grok OAuth provider in the launcher ([#38578](https://github.com/NousResearch/hermes-agent/pull/38578), [#38531](https://github.com/NousResearch/hermes-agent/pull/38531), [#37697](https://github.com/NousResearch/hermes-agent/pull/37697) — @teknium1, @OutThisLife) - "Choose provider later" skip on first-run onboarding; onboarding can configure a local/custom endpoint without an API key; custom zoom shortcuts ([#39483](https://github.com/NousResearch/hermes-agent/pull/39483), [#38572](https://github.com/NousResearch/hermes-agent/pull/38572), [#37894](https://github.com/NousResearch/hermes-agent/pull/37894) — @teknium1, @ethernet8023) - macOS helper microphone entitlement; scroll-jump fixes (native anchoring, at-rest jump, wheel-up snap-back); thinking block stays open mid-streaming ([#37745](https://github.com/NousResearch/hermes-agent/pull/37745) — @xxxigm, [#37866](https://github.com/NousResearch/hermes-agent/pull/37866), [#38221](https://github.com/NousResearch/hermes-agent/pull/38221), [#38809](https://github.com/NousResearch/hermes-agent/pull/38809) — @OutThisLife, @teknium1, @stremtec) - GUI quality-of-life triage batch; salvaged AhmetArif0 desktop/dashboard fixes; rename session via `session.title` RPC so `/title` works ([#37536](https://github.com/NousResearch/hermes-agent/pull/37536) — @austinpickett, [#39070](https://github.com/NousResearch/hermes-agent/pull/39070), [#39410](https://github.com/NousResearch/hermes-agent/pull/39410) — @teknium1, @benbarclay) - `hermes debug share` / `/debug` / `hermes logs` now include `desktop.log` ([#38203](https://github.com/NousResearch/hermes-agent/pull/38203) — @teknium1) - **Simplified Chinese (简体中文) translation** across every desktop UI surface — typed i18n layer, switch in Appearance settings, persisted via `display.language` (English remains default) ([#38241](https://github.com/NousResearch/hermes-agent/pull/38241) — @JimLiu) - Full multi-profile support over one global-remote dashboard; remote-profile sessions are first-class (resume, read, rename/archive/delete); new chats honor their profile in global-remote mode ([#39921](https://github.com/NousResearch/hermes-agent/pull/39921), [#39894](https://github.com/NousResearch/hermes-agent/pull/39894), [#39993](https://github.com/NousResearch/hermes-agent/pull/39993) — @OutThisLife) ## 📊 Web Dashboard ### Administration panel (NEW) - Full administration panel — MCP catalog with enable/disable toggles, pairing, webhooks, credentials, memory, gateway, hook creation, system settings ([#36704](https://github.com/NousResearch/hermes-agent/pull/36704), [#36736](https://github.com/NousResearch/hermes-agent/pull/36736) — @teknium1) - Channels page — set up every gateway messaging channel from the browser ([#37211](https://github.com/NousResearch/hermes-agent/pull/37211) — @teknium1) - System page: check-before-update flow + Debug Share ([#38205](https://github.com/NousResearch/hermes-agent/pull/38205), [#38600](https://github.com/NousResearch/hermes-agent/pull/38600) — @teknium1) - nous-blue theme, bulk sessions, schedule picker; enriched profiles dashboard + de-dupe channel env vars; configurable terminal background via theme ([#37383](https://github.com/NousResearch/hermes-agent/pull/37383), [#37872](https://github.com/NousResearch/hermes-agent/pull/37872) — @austinpickett, [#37156](https://github.com/NousResearch/hermes-agent/pull/37156) — @teknium1) - Always enable embedded chat; remove dashboard `--tui` flag ([#38591](https://github.com/NousResearch/hermes-agent/pull/38591) — @benbarclay) ### Auth - Pluggable username/password login (Option B); generic self-hosted OIDC provider + multi-provider verify fix; `hermes dashboard register` for self-hosted OAuth client ([#38819](https://github.com/NousResearch/hermes-agent/pull/38819) — @benbarclay, [#38917](https://github.com/NousResearch/hermes-agent/pull/38917), [#38802](https://github.com/NousResearch/hermes-agent/pull/38802) — @teknium1) - Rotate dashboard sessions via refresh token; share `/api/*` public allowlist between legacy and OAuth gates; drop `/api/*` paths from OAuth `next=` round trip ([#37247](https://github.com/NousResearch/hermes-agent/pull/37247), [#34254](https://github.com/NousResearch/hermes-agent/pull/34254), [#36244](https://github.com/NousResearch/hermes-agent/pull/36244) — @benbarclay) - Chat tab works in gated (OAuth) mode; trust non-web WS origins on OAuth-gated binds after ticket auth; authenticate server-spawned PTY child WS; sanction plugin WS/upload auth via SDK helpers ([#34793](https://github.com/NousResearch/hermes-agent/pull/34793), [#37870](https://github.com/NousResearch/hermes-agent/pull/37870), [#37972](https://github.com/NousResearch/hermes-agent/pull/37972), [#38549](https://github.com/NousResearch/hermes-agent/pull/38549) — @teknium1, @benbarclay, @kshitijk4poor) - Allow desktop websocket origins on remote binds ([#37747](https://github.com/NousResearch/hermes-agent/pull/37747) — @teknium1) ## 🏗️ Core Agent & Architecture ### Provider & model support - New models: `deepseek-v4-flash` (+ trimmed variants, maker-grouped curated lists), `MiniMax-M3` with 1M context on native minimax providers, `qwen3.7-plus` (Nous + OpenRouter), `gemini-3.5-flash` to Gemini OAuth + API-key pickers ([#35659](https://github.com/NousResearch/hermes-agent/pull/35659), [#36214](https://github.com/NousResearch/hermes-agent/pull/36214), [#39409](https://github.com/NousResearch/hermes-agent/pull/39409), [#37046](https://github.com/NousResearch/hermes-agent/pull/37046) — @teknium1) - Model picker: fuzzy search across WebUI/TUI/CLI; group multi-endpoint providers under one row; refresh provider descriptions + describe grouped rows; refresh catalog hourly; stop routing OpenAI selection to OpenRouter ([#36928](https://github.com/NousResearch/hermes-agent/pull/36928) — @kshitijk4poor, [#35227](https://github.com/NousResearch/hermes-agent/pull/35227), [#35773](https://github.com/NousResearch/hermes-agent/pull/35773), [#35756](https://github.com/NousResearch/hermes-agent/pull/35756), [#37175](https://github.com/NousResearch/hermes-agent/pull/37175)) - Always show Nous Tool Gateway backends, login on select; surface the Nous free tool pool (entitlement + setup prompt); route FAL video gen through managed Nous gateway ([#35792](https://github.com/NousResearch/hermes-agent/pull/35792) — @teknium1, [#36153](https://github.com/NousResearch/hermes-agent/pull/36153), [#33259](https://github.com/NousResearch/hermes-agent/pull/33259) — @alt-glitch) - Persist mid-session model switch to database; recover model on post-interrupt recovery turn ([#35256](https://github.com/NousResearch/hermes-agent/pull/35256), [#35381](https://github.com/NousResearch/hermes-agent/pull/35381) — @teknium1) - Credential pool: `STATUS_DEAD` for terminal OAuth failures; isolate custom provider picker credentials ([#34412](https://github.com/NousResearch/hermes-agent/pull/34412), [#34810](https://github.com/NousResearch/hermes-agent/pull/34810) — @teknium1) - Disable Nous Portal legacy session-key inference fallback — JWT-only ([#34508](https://github.com/NousResearch/hermes-agent/pull/34508) — @kshitijk4poor) ### Agent loop, prompt & tools - `/undo [N]` — backs up N user turns with prefill + soft-delete (CLI/TUI + messaging-platform parity) ([#36229](https://github.com/NousResearch/hermes-agent/pull/36229), [#36699](https://github.com/NousResearch/hermes-agent/pull/36699) — @teknium1) - Progressive tool disclosure for MCP and plugin tools (scoped); embedder environment-hint hook for the system prompt; universal task-completion guidance + local Python toolchain probe ([#34493](https://github.com/NousResearch/hermes-agent/pull/34493), [#34574](https://github.com/NousResearch/hermes-agent/pull/34574), [#34340](https://github.com/NousResearch/hermes-agent/pull/34340) — @teknium1) - Uncap delegation `max_spawn_depth` (floor 1, no ceiling); broaden Hermes self-knowledge pointer to docs + skill; `hermes prompt-size` diagnostic ([#39772](https://github.com/NousResearch/hermes-agent/pull/39772), [#38538](https://github.com/NousResearch/hermes-agent/pull/38538), [#35276](https://github.com/NousResearch/hermes-agent/pull/35276) — @teknium1) - `perf(read_file)`: compact line-number gutter — ~14% fewer tokens per read (now the only format) ([#35368](https://github.com/NousResearch/hermes-agent/pull/35368), [#35532](https://github.com/NousResearch/hermes-agent/pull/35532) — @teknium1) - Resolve agent cwd from `TERMINAL_CWD` via one reader; align prefill messages key handling; resume relaunches in the session's original working directory ([#35028](https://github.com/NousResearch/hermes-agent/pull/35028) — @banditburai, [#38760](https://github.com/NousResearch/hermes-agent/pull/38760) — @helix4u, [#38562](https://github.com/NousResearch/hermes-agent/pull/38562) — @teknium1) - Prevent session-id fork from concurrent compressions; observer telemetry hooks + NeMo-Relay plugin (gated tool emit) ([#34351](https://github.com/NousResearch/hermes-agent/pull/34351), [#38232](https://github.com/NousResearch/hermes-agent/pull/38232) — @teknium1) ### Sessions, state & memory - `perf(state)`: merge FTS5 segments on VACUUM + `hermes sessions optimize`; keep `/branch` sessions visible after parent reopen; survive missing FTS5 runtimes ([#34596](https://github.com/NousResearch/hermes-agent/pull/34596), [#39214](https://github.com/NousResearch/hermes-agent/pull/39214) — @teknium1, @kshitijk4poor, [#35452](https://github.com/NousResearch/hermes-agent/pull/35452) — @helix4u) - Honcho: make startup fail open; harden self-hosted setup paths ([#24847](https://github.com/NousResearch/hermes-agent/pull/24847) — @stephenschoettler, [#35170](https://github.com/NousResearch/hermes-agent/pull/35170) — @kshitijk4poor) - Supermemory: session-level ingest + kebab aliases ([#38756](https://github.com/NousResearch/hermes-agent/pull/38756) — @alt-glitch) ## 🧩 Multi-Agent (Kanban) & Skills ### Kanban - `goal_mode` cards run workers in a `/goal` loop; file attachments on tasks; attach images referenced in task bodies to worker vision ([#35710](https://github.com/NousResearch/hermes-agent/pull/35710), [#35395](https://github.com/NousResearch/hermes-agent/pull/35395), [#34210](https://github.com/NousResearch/hermes-agent/pull/34210) — @teknium1) - `default_assignee` fallback + per-profile concurrency cap; `POST /runs/{run_id}/terminate` endpoint; gate notifier watcher on `dispatch_in_gateway` ([#34244](https://github.com/NousResearch/hermes-agent/pull/34244), [#34449](https://github.com/NousResearch/hermes-agent/pull/34449), [#37174](https://github.com/NousResearch/hermes-agent/pull/37174) — @teknium1) - CLI dispatch config passthrough + humanizer skill swap ([#34337](https://github.com/NousResearch/hermes-agent/pull/34337) — @teknium1) ### Skills - **Leaner default skill set** ([#39028](https://github.com/NousResearch/hermes-agent/pull/39028) — @teknium1): - Removed (redundant / dead): `spotify` (→ Spotify plugin's 7 native tools), `linear` (→ `hermes mcp install linear`), `kanban-codex-lane`, `debugging-hermes-tui-commands`, stale `domain` orphan, empty category markers (`diagramming`, `gifs`, `inference-sh`, `mlops/training`, `mlops/vector-databases`) - Bundled → optional: `baoyu-article-illustrator`, `baoyu-comic`, `creative-ideation`, `pixel-art`, `dspy`, `subagent-driven-development`, `minecraft-modpack-server`, `pokemon-player`, `hermes-s6-container-supervision` - Consolidated: `webhook-subscriptions` + `native-mcp` folded into the `hermes-agent` skill as on-demand references; `writing-plans` merged into `plan` (v2.0.0) - New `environments:` frontmatter relevance gate (`kanban` / `docker` / `s6`) — context-specific skills stop appearing in the index for users who won't use them, still load on explicit request - Curator can now prune unused **built-in** skills (not just agent-created), with usage tracked for every skill ([#36701](https://github.com/NousResearch/hermes-agent/pull/36701) — @teknium1) - Blank-slate skills — `install --no-skills` + opt-out/opt-in for the default profile ([#36228](https://github.com/NousResearch/hermes-agent/pull/36228) — @teknium1) - **NVIDIA/skills trusted tap** — `NVIDIA/skills` is now a default trusted Skills Hub tap alongside OpenAI/Anthropic/HuggingFace; `skills.sh.json` sidecar gives taps real category labels ([#34333](https://github.com/NousResearch/hermes-agent/pull/34333) — @teknium1) - Skills Hub: fix browse cap, add source links + copy buttons + category cleanup ([#37143](https://github.com/NousResearch/hermes-agent/pull/37143) — @teknium1) - New optional skills: `grok` (xAI Grok Build CLI), `antigravity-cli` operator (under autonomous-ai-agents) ([#34582](https://github.com/NousResearch/hermes-agent/pull/34582), [#34583](https://github.com/NousResearch/hermes-agent/pull/34583), [#34604](https://github.com/NousResearch/hermes-agent/pull/34604) — @teknium1) ## 📱 Messaging Platforms (Gateway) - Structured stream-event protocol + Telegram draft formatting parity; per-platform streaming defaults (Telegram on, Discord off) + dashboard toggles; surface gateway streaming block in `DEFAULT_CONFIG` ([#37250](https://github.com/NousResearch/hermes-agent/pull/37250), [#37303](https://github.com/NousResearch/hermes-agent/pull/37303), [#37285](https://github.com/NousResearch/hermes-agent/pull/37285) — @teknium1) - Discord voice-channel mixer — ambient idle bed + verbal acks that overlap TTS; explain `/voice` usage when toggled bare ([#39659](https://github.com/NousResearch/hermes-agent/pull/39659), [#39766](https://github.com/NousResearch/hermes-agent/pull/39766) — @teknium1) - Handle Feishu meeting invitations; bluebubbles group mention gating; matrix bang-command aliases; matrix fail-closed approval reaction auth ([#39040](https://github.com/NousResearch/hermes-agent/pull/39040), [#37091](https://github.com/NousResearch/hermes-agent/pull/37091), [#38175](https://github.com/NousResearch/hermes-agent/pull/38175) — @teknium1, @alt-glitch, [#34567](https://github.com/NousResearch/hermes-agent/pull/34567)) - Clean service restart flow; close ResponseStore + dispose unowned adapter on reconnect failure; weixin `asyncio.wait_for` timeouts ([#36188](https://github.com/NousResearch/hermes-agent/pull/36188) — @helix4u, [#37679](https://github.com/NousResearch/hermes-agent/pull/37679) — @Fearvox, [#35117](https://github.com/NousResearch/hermes-agent/pull/35117) — @banditburai) ## 🖥️ CLI, TUI & Setup - Configurable default interface (cli vs tui) + `--cli` flag; TUI single `/model` command + unified Sessions overlay; nudge toward `/agents` dashboard when delegation starts ([#37782](https://github.com/NousResearch/hermes-agent/pull/37782), [#37112](https://github.com/NousResearch/hermes-agent/pull/37112) — @OutThisLife, [#34704](https://github.com/NousResearch/hermes-agent/pull/34704) — @kshitijk4poor) - Thin out setup — Quick Setup via Nous Portal + Full Setup defaults; explain Quick vs Full inline; `hermes portal` human-readable Portal onboarding alias + full quick-setup Nous flow ([#35723](https://github.com/NousResearch/hermes-agent/pull/35723), [#36227](https://github.com/NousResearch/hermes-agent/pull/36227), [#38449](https://github.com/NousResearch/hermes-agent/pull/38449), [#38465](https://github.com/NousResearch/hermes-agent/pull/38465) — @teknium1, @kshitijk4poor) - Set process title to `hermes` in ps/top/htop; warn on unsupported pip installs + fix stale update-check cache ([#35143](https://github.com/NousResearch/hermes-agent/pull/35143), [#34846](https://github.com/NousResearch/hermes-agent/pull/34846) — @teknium1) - TUI perf: stop slow/dead MCP servers from freezing startup; stop eager MCP discovery from blocking agent-capable startup; stop persisting full tool output in trail lines (silent OOM) ([#35273](https://github.com/NousResearch/hermes-agent/pull/35273), [#35397](https://github.com/NousResearch/hermes-agent/pull/35397), [#38224](https://github.com/NousResearch/hermes-agent/pull/38224) — @teknium1) - TUI fixes: auto-recover session on unexpected gateway death; reassemble split SGR mouse sequences; preserve UTF-8 in PowerShell clipboard; reset terminal input modes on exit; `/save` snapshots under Hermes home ([#35893](https://github.com/NousResearch/hermes-agent/pull/35893), [#38564](https://github.com/NousResearch/hermes-agent/pull/38564) — @OutThisLife, [#35222](https://github.com/NousResearch/hermes-agent/pull/35222) — @teknium1, [#36864](https://github.com/NousResearch/hermes-agent/pull/36864) — @maxmilian, [#38251](https://github.com/NousResearch/hermes-agent/pull/38251) — @austinpickett) - Setup model/provider pickers migrated off `simple_term_menu` to curses (ESC + ghost-row fixes); default browser/TTS picker to free local backend, not paid Nous ([#35806](https://github.com/NousResearch/hermes-agent/pull/35806), [#37800](https://github.com/NousResearch/hermes-agent/pull/37800) — @teknium1) ## 🔧 Tool System & Installer - `single managed-uv path`, delete fts5 installer escalation; installer commit pinning opt-in (default branch-follow); shallow clones ([#37660](https://github.com/NousResearch/hermes-agent/pull/37660) — @ethernet8023, [#37123](https://github.com/NousResearch/hermes-agent/pull/37123) — @jquesnelle, [#39423](https://github.com/NousResearch/hermes-agent/pull/39423) — @ethernet8023) - `ensure_uv()` survives the update boundary (no first-run crash); harden venv rebuild + verify core deps after install; require managed marker before destructive clean; stash/restore by default for non-interactive updates ([#39780](https://github.com/NousResearch/hermes-agent/pull/39780) — @OutThisLife, [#38887](https://github.com/NousResearch/hermes-agent/pull/38887), [#39568](https://github.com/NousResearch/hermes-agent/pull/39568), [#39645](https://github.com/NousResearch/hermes-agent/pull/39645) — @teknium1, @helix4u) - MCP: stop reporting false OAuth success when no token was obtained; vision honors `model.supports_vision` in `vision_analyze` + `browser_vision`; MiniMax t2a_v2 TTS `raise_for_status` ([#34807](https://github.com/NousResearch/hermes-agent/pull/34807), [#34562](https://github.com/NousResearch/hermes-agent/pull/34562), [#39057](https://github.com/NousResearch/hermes-agent/pull/39057) — @teknium1) - yuanbao: cache resolved media resources by resourceId ([#34474](https://github.com/NousResearch/hermes-agent/pull/34474) — @teknium1) ## 🐳 Docker & Deployment - Container reuse + bounded-sync cleanup + orphan reaper; auto-join Docker socket group for docker-in-docker backend; boot non-root containers (skip s6-setuidgid drop when already unprivileged) ([#33645](https://github.com/NousResearch/hermes-agent/pull/33645), [#34407](https://github.com/NousResearch/hermes-agent/pull/34407) — @benbarclay, [#34837](https://github.com/NousResearch/hermes-agent/pull/34837) — @IAvecilla) - Skip unnecessary boot chown when volume ownership matches remapped UID; seed `gateway_state.json` from `HERMES_GATEWAY_BOOTSTRAP_STATE` on first boot; tag containers with hermes-agent labels for identification ([#35027](https://github.com/NousResearch/hermes-agent/pull/35027) — @Foldblade, [#37896](https://github.com/NousResearch/hermes-agent/pull/37896) — @benbarclay) - Point TUI launcher at prebuilt bundle via `HERMES_TUI_DIR`; consolidate node/nix workspace lockfile + update all consumers ([#37923](https://github.com/NousResearch/hermes-agent/pull/37923), [#36171](https://github.com/NousResearch/hermes-agent/pull/36171) — @benbarclay, @ethernet8023) ## 🔒 Security & Reliability - **CVE-2026-48710 (Starlette BadHost)** — pin patched Starlette ≥1.0.1 ([#35118](https://github.com/NousResearch/hermes-agent/pull/35118) — @teknium1) - Run URL SSRF checks off the event loop in async paths; strip Bedrock inference bearer token from subprocess env; add `bws_cache.json` to file-safety read guard; neutralize file paths in mutation-verifier footer ([#39046](https://github.com/NousResearch/hermes-agent/pull/39046), [#34498](https://github.com/NousResearch/hermes-agent/pull/34498), [#34421](https://github.com/NousResearch/hermes-agent/pull/34421), [#35684](https://github.com/NousResearch/hermes-agent/pull/35684) — @teknium1) - Restore approval/sudo context in `execute_code` + guard entry points; add docker restart/stop/kill to `DANGEROUS_PATTERNS`; sanitize invisible unicode in vetted skill content; deepcopy tools before in-place xAI mutation ([#34497](https://github.com/NousResearch/hermes-agent/pull/34497), [#33438](https://github.com/NousResearch/hermes-agent/pull/33438) — @teknium1, @Sarbai, [#37245](https://github.com/NousResearch/hermes-agent/pull/37245), [#34416](https://github.com/NousResearch/hermes-agent/pull/34416)) - Sandbox-mirror soft guard for writes to per-task `.hermes` mirrors; honcho fail-open on startup ([#32213](https://github.com/NousResearch/hermes-agent/pull/32213) — @briandevans, [#24847](https://github.com/NousResearch/hermes-agent/pull/24847) — @stephenschoettler) ## 🐛 Notable Bug Fixes - **399 issues closed** this window (2 P0, 62 P1, 16 security-tagged, 262 bug-labeled). - Desktop: keep in-flight new chats from vanishing on refresh; Stop button actually interrupts when a turn is queued; stop background session messages bleeding into the active transcript; slash/@ completion menu navigable & Esc-dismissable; IME Enter no longer splits messages ([#37908](https://github.com/NousResearch/hermes-agent/pull/37908) — @OutThisLife, [#37948](https://github.com/NousResearch/hermes-agent/pull/37948), [#37975](https://github.com/NousResearch/hermes-agent/pull/37975), [#37937](https://github.com/NousResearch/hermes-agent/pull/37937) — @kshitijk4poor, [#38333](https://github.com/NousResearch/hermes-agent/pull/38333) — @stremtec) - Update: stop stash/restore from clobbering desktop source on managed clones; don't fail desktop rebuild/skills sync on mid-rebuild venv; export launcher virtualenv to uv ([#38542](https://github.com/NousResearch/hermes-agent/pull/38542), [#38885](https://github.com/NousResearch/hermes-agent/pull/38885), [#35224](https://github.com/NousResearch/hermes-agent/pull/35224) — @teknium1) - Voice: honor `PIPEWIRE_REMOTE` in PortAudio fallback; allow `/voice` over SSH when a sound server is reachable; restore mistralai (2.4.8 clean, ban lifted) ([#33473](https://github.com/NousResearch/hermes-agent/pull/33473) — @Dusk1e, [#35719](https://github.com/NousResearch/hermes-agent/pull/35719), [#34841](https://github.com/NousResearch/hermes-agent/pull/34841) — @teknium1) ## 📚 Documentation - New Desktop App guide + remote-backend sections (session token, `--tui` requirement, username/password connect, dashboard/gateway prerequisites) ([#37457](https://github.com/NousResearch/hermes-agent/pull/37457), [#38144](https://github.com/NousResearch/hermes-agent/pull/38144), [#38180](https://github.com/NousResearch/hermes-agent/pull/38180), [#38534](https://github.com/NousResearch/hermes-agent/pull/38534), [#39128](https://github.com/NousResearch/hermes-agent/pull/39128) — @teknium1) - Dashboard auth-provider suitability + registration across dashboard/Docker/Desktop; network egress isolation guide for Docker ([#39633](https://github.com/NousResearch/hermes-agent/pull/39633) — @benbarclay, [#26385](https://github.com/NousResearch/hermes-agent/pull/26385) — @Manzela) ---  ## 👥 Contributors A huge thank-you to the **170 community contributors** (including co-authors) who shipped work in this release. ### Core - @teknium1 ### Top community contributors (by merged-PR count) - **@OutThisLife** (52) — built the desktop app end to end: install, self-update, remote-gateway connect, multi-profile sessions, chat UX, status-bar model picker - **@benbarclay** (44) — Docker hardening, dashboard auth (OIDC, username/password, refresh-token rotation), desktop OAuth remote connect - **@kshitijk4poor** (29) — fuzzy model picker, setup/portal onboarding, desktop completion-menu & Stop-button fixes, honcho hardening - **@ethernet8023** (18) — desktop build pipeline (content-hash stamp, build flags), Linux/arm64 desktop support, managed-uv consolidation - **@austinpickett** (8) — dashboard nous-blue theme + bulk sessions, desktop Providers settings, GUI QoL triage - **@alt-glitch** (7) — Nous tool-pool entitlement surfacing, FAL video-gen managed gateway, supermemory session ingest, matrix aliases - **@helix4u** (6) — gateway service-restart flow, update destructive-clean guard, config prefill alignment - **@jquesnelle** (4) — installer commit-pinning opt-in, desktop skills/tools + model-management consolidation - **@JimLiu** — full Simplified Chinese (简体中文) desktop translation + typed i18n layer ### All Contributors @0xharryriddle, @a1245582339, @adybag14-cyber, @AhmetArif0, @alaamohanad169-ship-it, @alelpoan, @alt-glitch, @Aman113114-IITD, @amathxbt, @aminvakil, @annguyenNous, @aqilaziz, @Archerouyang, @ashishpatel26, @asimons81, @austinpickett, @banditburai, @baofuen, @Bartok9, @bbednarski9, @beardthelion, @bedirhancode, @benbarclay, @benfrank241, @blackpilledsoftware-prog, @bluefishs, @brian-doherty, @briancl2, @briandevans, @BROCCOLO1D, @caojiguang, @CharZhou, @counterposition, @coygeek, @CryptoByz, @davetist, @davidgut1982, @dchenk, @dirtyren, @donovan-yohan, @dparikh79, @dskwe, @Dusk1e, @dvir-pashut, @EloquentBrush0x, @emozilla, @ErnestHysa, @Erosika, @ethernet8023, @f3rs3n, @faisfamilytravel, @Fearvox, @ferminquant, @fesalfayed, @firefly, @flooryyyy, @Foldblade, @Frowtek, @gbarany, @Glucksberg, @GodsBoy, @haran2001, @harjothkhara, @HashClawAI, @hayka-pacha, @helix4u, @HeLLGURD, @Hinotoi-agent, @hllqkb, @IAvecilla, @inchargeautomation-lab, @Interstellar-code, @isair, @ITheEqualizer, @jeffrobodie-glitch, @JezzaHehn, @JimLiu, @Julientalbot, @julio-cloudvisor, @karmeleon, @Kenmege, @Kewe63, @Kolektori, @konsisumer, @kshitijk4poor, @kurobaryo, @kweiner, @kyssta-exe, @Kyzcreig, @LengR, @leonardsellem, @LeonSGP43, @libre-7, @liuboacean, @liuhao1024, @loongzhao, @luyao618, @MaheshtheDev, @Manzela, @mathijsvandenhurk, @MattMaximo, @maxcz79, @maxmilian, @Moikapy, @MustafaKara7, @nateGeorge, @nepenth, @nielskaspers, @ninjmnky, @octavioturra, @OCWC22, @ohMyJason, @ousiaresearch, @outsourc-e, @OutThisLife, @Pluviobyte, @polnikale, @pxdsgnco, @Que0x, @r266-tech, @redpiggy-cyber, @rewbs, @rexdotsh, @rob-maron, @SaguaroDev, @sahibzada-allahyar, @Sarbai, @sarvesh1327, @scubamount, @SeaXen, @seppegadeyne, @shannonsands, @SHL0MS, @SimoKiihamaki, @SiTaggart, @solaitken, @sprmn24, @stephenschoettler, @steveonjava, @stremtec, @Subway2023, @sweetcornna, @Sylw3ster, @ThyFriendlyFox, @tillfalko, @tmchow, @TonyPepeBear, @Tranquil-Flow, @truenorth-lj, @tuancookiez-hub, @Twanislas, @tymrtn, @uzunkuyruk, @ViewWay, @VinciZhu, @vinoth12940, @vladkvlchk, @vynxevainglory-ai, @WadydX, @wenchengxucool, @whyhkzk, @worlldz, @Wysie, @x1am1, @xxxigm, @ygd58, @youngstar-eth, @zapabob, @zhaoleibd, @Zyrixtrex --- **Full Changelog**: [v2026.5.29.2...v2026.6.5](https://github.com/NousResearch/hermes-agent/compare/v2026.5.29.2...v2026.6.5)

### Highlights - **OpenAI GPT-5.6 support:** OpenClaw now recognizes the GPT-5.6 model family across catalog, capability, and runtime selection paths. (#98333) Thanks @steipete-oai. - **External harness attachment:** `openclaw attach` launches an external harness against an existing Gateway session, making interactive Codex-style workflows easier to resume and inspect. (#96454) Thanks @anagnorisis2peripeteia and @obviyus. - **Telegram Codex workflows:** Telegram can now start Codex pairing with `/login`, steer active Codex runs, and recover final replies across transient API failures. (#98006, #98126, #98786) Thanks @100yenadmin, @Kyzcreig, and @obviyus. - **Event-driven cron runs:** the new `on-exit` schedule kind wakes an agent when a watched command exits, while session-targeted runs can detach cleanly. (#92037, #98755) Thanks @anagnorisis2peripeteia, @obviyus, and @EthanSK. - **Native app refresh:** iOS adopts the iOS 26 visual system with clearer Chat, Talk, onboarding, and reconnect flows, while native app localization expands across Apple and Android surfaces. (#98452, #98736, #99243, #97110, #97111, #97112, #97113) Thanks @jcooley8 and @vincentkoc. - **Richer messaging:** iMessage gains native poll creation, reading, and voting, and built-in usage footers provide clearer per-turn accounting in chat. (#98421, #92657, #92877) Thanks @omarshahine, @lobster, and @Marvinthebored. - **Safer scoped conversations:** capability profiles prepare per-conversation tool and access boundaries without weakening the existing default profile. (#98536) - **Mac local Gateway setup:** the macOS app can now install and start its local Gateway automatically, reducing the manual setup needed before first use. (#99767) - **Control UI navigation:** a session-first sidebar, compact context meter, warm light theme, reasoning-effort slider, streamlined composer, and slash-command picker make active conversations and commands easier to reach. (#99289, #99426, #99838) Thanks @VicZhang6 and @Solvely-Colin. ### Changes - **ClawRouter routing and quotas:** add the bundled ClawRouter provider plugin with credential-scoped dynamic model discovery, OpenAI-compatible and native Anthropic/Gemini transports, and managed budget reporting across OpenClaw usage surfaces. (#99658) - **Model and provider coverage:** add GPT-5.6 support, use Nemotron Super's 1M context window, and preserve explicit OpenRouter authentication headers. (#98333, #98726, #98187) Thanks @steipete-oai, @eleqtrizit, @sunlit-deng, and @laurencebrown. - **CLI and node workflows:** add `openclaw attach`, node context-path support, actionable device-approval recovery guidance, soft-resume CLI sessions when prompt metadata changes, and clearer plugin install exit diagnostics. (#96454, #97679, #98115, #98146, #98497, #99822) Thanks @anagnorisis2peripeteia, @obviyus, @wm0018, @welfo-beo, @RomneyDa, @Sanjays2402, and @vincentkoc. - **Cron and usage:** add exit-triggered schedules, detached session-targeted runs, an in-flight job doctor warning, and a built-in full usage footer. (#92037, #98755, #98620, #92657, #92877) Thanks @anagnorisis2peripeteia, @obviyus, @EthanSK, @masatohoshino, and @Marvinthebored. - **Native apps and localization:** modernize iOS presentation, Chat, Talk, onboarding, and reconnect flows; add Gateway speech providers; improve QR onboarding and protocol recovery; install the local Gateway from macOS; localize core Apple and Android surfaces; and add Swedish mobile localization. (#98452, #98736, #99243, #98376, #98302, #98385, #99767, #97110, #97111, #97112, #97113, #98043) Thanks @jcooley8, @Tony-ooo, @joelnishanth, @cursoragent, @joshavant, @vincentkoc, and @yeager. - **Messaging capabilities:** add native iMessage polls, Telegram Codex pairing and steering, Telegram multi-lane progress summaries, and Signal target aliases. (#98421, #98006, #98126, #98907, #95738) Thanks @omarshahine, @lobster, @100yenadmin, @Kyzcreig, @Marvinthebored, and @jesse-merhi. - **Local inference and chat controls:** auto-discover Ollama inference nodes, add Control UI session-first navigation, reasoning controls, and command picking, and keep OpenClaw control tools available when deferred tool search selects the wrong tool family. (#99234, #99289, #99426, #99838, #99561) Thanks @100yenadmin, @joshavant, @VicZhang6, and @Solvely-Colin. - **Doctor and diagnostics:** expose auth-profile, workspace, device-pairing, channel-plugin, memory-provider, systemd exhaustion, and Windows LAN firewall findings. (#97125, #97358, #97366, #97496, #97968, #98291, #98666) Thanks @giodl73-repo, @masatohoshino, and @joshavant. - **Conversation and review controls:** prepare scoped conversation capability profiles and add Cursor Agent as an autoreview engine. (#98536, #97348) Thanks @hxy91819. ### Fixes - **ClawRouter auth profiles:** resolve credential-scoped catalog models during agent runs when the proxy key is stored in an auth profile, and document plugin and model allowlists. - **Telegram durability:** recover stalled ingress claims, retry restart-dropped media, survive transient polling errors, dead-letter poison updates, preserve forwarded rich text, route plugin callbacks correctly, keep progress updates in one stable multi-line window, throttle reconnect queue drains so fast healthy polls cannot starve active replies, and fall back safely when Telegram rejects rich final replies. (#97118, #98102, #98735, #98775, #98776, #97174, #98907, #98786) Thanks @vincentkoc, @luoyanglang, @DaveArcher18, @obviyus, @goldmar, @Marvinthebored, and @shakkernerd. - **Agent and context reliability:** preserve runtime overrides, steered subagent tasks, fallback tool-call hints, and legacy reseed attachments; soft-resume CLI sessions across prompt-only drift; honor stop signals that arrive during eager session-lock acquisition before any provider prompt starts; improve harness-aware context estimation and compaction prechecks; time out silent local streams; recover mid-stream failures; and cap Gateway run-cache growth. (#92237, #77539, #99851, #99839, #99822, #97928, #97861, #98525, #95430, #77973) Thanks @sercada, @amittell, @obviyus, @liuhao1024, @yetval, @osolmaz, @lzyyzznl, @vincentkoc, @alexelgier, and @fede-kamel. - **Provider and network safety:** bound oversized or malformed responses across Moonshot, MiniMax, Anthropic OAuth, Discord, Matrix, SMS, browser, update, embeddings, Tlön, and Inworld paths. (#96502, #96322, #96644, #97693, #97662, #97999, #98455, #98508, #98554, #98496, #98660) Thanks @hugenshen, @cursoragent, @lsr911, @solodmd, @Alix-007, @wings1029, @lzyyzznl, @sunlit-deng, @vincentkoc, and @Pandah97. - **Channel delivery and routing:** keep Slack replies in the active thread, preserve account-bound delivery routes, apply response prefixes, suppress internal traces and unwanted fallback replies, and retain WeChat session routing for opaque account ids. (#97168, #98240, #89949, #93639, #97989, #80928, #93686) Thanks @LiuwqGit, @gorkem2020, @yetval, @wangwllu, @ZengWen-DT, @alexuser, @UnClouded77, @zhangguiping-xydt, @htkillermax-gif, and @vincentkoc. - **Cron correctness:** preserve provider and model selections on timeouts, retain startup catch-up deferrals, keep action-required output, clear blank thinking overrides, and preserve provider-owned daily-reset sessions. (#95943, #94022, #93810, #96393, #96293, #98356) Thanks @ZengWen-DT, @cursoragent, @luke-renjoy, @RichChen01, @vincentkoc, @yetval, @snowzlmbot, @nz365guy, and @takamasa-aiso. - **Memory and session recovery:** detect unindexed transcripts, preserve notes through transient reads, avoid cross-directory resumes, disambiguate reserved wiki index pages, and skip empty QMD sync work. (#97857, #98360, #97785, #94326, #90030) Thanks @zw-xysk, @CHE10X, @qingminglong, @yetval, @vincentkoc, @sahibzada-allahyar, and @ruben2000de. - **Windows and execution:** bind allowlisted execution to the validated Windows path, propagate `PATHEXT`, normalize inbound paths case-insensitively, and prevent cleanup crashes on Windows. (#98260, #98093, #97630, #97901) Thanks @eleqtrizit, @wendy-chsy, @VectorPeak, and @paulcam206. - **Mobile and UI stability:** preserve iOS chat line breaks and final replies, improve Android pairing and TLS recovery, hide expired pairing cards, keep workspace file rails scrollable, restore copy-path over plain HTTP, and stop rubber-band scrolling in the Mac app Control UI. (#98304, #98117, #98366, #98439, #98483, #98049, #98646, #98611, #98764, #99830) Thanks @joshavant, @Jabato01, @ooiuuii, @wuqxuan, @645648406-max, @zw-xysk, @ZengWen-DT, and @adinballew. - **Codex and approval flows:** report ChatGPT authentication correctly, rename destructive approval mode to `ask`, classify dynamic goal and session tool results accurately, and derive terminal-idle timeouts from the explicit run deadline. (#91240, #98501, #98659, #96856, #85296) Thanks @849261680, @ukstem, @kevinslin, @yetval, @nxmxbbd, @alkor2000, and @vincentkoc. - **Configuration and plugin health:** restrict config traversal to owned properties, preserve config-health recovery state, surface unloadable channel plugins, preserve defaulted provider base URLs during patches, validate bundled plugin updates by manifest contract, prefer installed launcher targets when resolving package roots so bundled channel setup entries load from nested installs, resolve public artifacts from installed plugin roots, and retain legacy ClawHub families where required. (#99846, #99728, #96397, #98396, #98010, #98819, #98249) Thanks @vincentkoc, @zenglingbiao, @joshavant, @jalehman, @ccbridle, @849261680, @momothemage, @weltmaister, @LiLan0125, @herove, @amknight, @KelTech-Services, and @Patrick-Erichsen. - **Runtime process safety:** prevent unhandled child-stream errors in SSH tunnels, supervisors, and MCP stdio transports; keep auto-replies from waiting on transcript mirroring; and avoid splitting Unicode characters in approval previews and LINE outbound fields. (#99800, #99802, #99803, #99549, #99566, #98994) Thanks @cxbAsDev, @vincentkoc, @Shagrat2, @mikasa0818, and @LEXES7. - **Node runtime compatibility:** installers, the CLI launcher, doctor, and the macOS app now reject incompatible Node 23 runtimes and guide users toward supported Node 22 or 24 releases. (#99832) Thanks @vincentkoc and @fuller-stack-dev. - **QQBot media delivery:** scope sandbox-generated media sends to the active session's workspace so `/workspace/...` and relative generated-file paths resolve safely across QQBot media tags, structured payloads, and streaming delivery. (#92872) Thanks @zhangguiping-xydt and @sliverp. ### Complete contribution record This audited record covers the complete 66e676d29b92d040716376a75aca32bad655cfac..1e20f15581f9fe9132768379bd80cc74c047b8cc history: 455 merged PRs. The generation manifest also supplies direct commits as editorial input; the grouped notes above prioritize user impact. #### Pull requests - **PR #96502** fix(moonshot): bound video description JSON response reads. Thanks @hugenshen and @cursoragent. - **PR #98249** Preserve legacy ClawHub family for selected plugins. Thanks @Patrick-Erichsen. - **PR #93767** fix(reasoning-tags): strip MiniMax `mm:` namespaced reasoning tags. Thanks @DrHack1. - **PR #93820** fix(imessage): recognize MiniMax mm: reasoning tags in reflection guard (completes #93767). Thanks @Alix-007. - **PR #94096** fix(usage): reject inverted startDate-endDate range in usage.cost and sessions.usage. Thanks @Alix-007. - **PR #97125** Doctor: expose auth profile findings. Thanks @giodl73-repo. - **PR #98256** fix(mcp): require owner for Claude permission replies. Thanks @eleqtrizit. - **PR #98142** fix(cli): stop `pairing list` crashing with empty channel enum. Thanks @RomneyDa. - **PR #98260** fix(exec): bind Windows allowlist execution path. Thanks @eleqtrizit. - **PR #97118** fix(telegram): recover stalled ingress spool claims. Thanks @vincentkoc. - **PR #97168** fix(slack): prefer current thread session for inherited outbound replies. Related #96535. Thanks @LiuwqGit and @gorkem2020. - **PR #97769** fix(plugins): apply output text transforms to toolcall_delta and toolcall_end events. Related #97761. Thanks @ZOOWH and @get-viti. - **PR #96544** fix(doctor): merge colliding model-ref map keys instead of dropping. Thanks @yetval and @vincentkoc. - **PR #97177** fix(memory-wiki): gracefully handle unparsable YAML frontmatter in vault scans (#96125). Thanks @SunnyShu0925 and @cow11023. - **PR #97167** fix #96840: [Bug]: Targetless message.send fails with 'Action send requires a target' in WebChat despite docs stating source-reply sink should handle it. Thanks @zhangguiping-xydt and @MantisCartography. - **PR #98302** fix(ios): advance onboarding step after QR scan. Related #98297. Thanks @joelnishanth and @cursoragent. - **PR #96644** fix(anthropic-oauth): bound OAuth token endpoint response reads. Thanks @solodmd. - **PR #96397** fix: warn when configured channel plugins cannot load. Thanks @849261680. - **PR #96359** test: migrate src/commands tests to shared temp dir helpers. Thanks @xialonglee. - **PR #96293** fix(cron): clear agentTurn thinking override by blanking the field. Related #96287. Thanks @ZengWen-DT and @takamasa-aiso. - **PR #96058** test: prefer shared temp dir helpers in auto-reply and install-fallback tests. Thanks @xialonglee. - **PR #87298** test: add temp directory helper guidance. Thanks @hxy91819. - **PR #97785** fix(sessions): avoid cross-cwd recent resumes. Related #96542. Thanks @qingminglong and @yetval. - **PR #97698** fix(pdf): reject empty parsed page ranges before native analysis. Thanks @zhangguiping-xydt. - **PR #97693** fix(discord): bound requestDiscord happy-path response reads to prevent OOM. Thanks @Alix-007. - **PR #97683** fix(irc): guard surrogate-range codepoints in \u literal-escape decoder. Thanks @llagy009. - **PR #96938** fix(utils): keep reply directive ids unicode-safe. Thanks @ly-wang19. - **PR #97857** fix(memory): detect unindexed session transcripts in status mode (fixes #97814). Thanks @zw-xysk and @CHE10X. - **PR #98094** fix(android): clarify gateway auth recovery states. Thanks @qingminglong. - **PR #98205** test(gateway): add unit tests for node wake state tracking and testing seam. Thanks @zenglingbiao. - **PR #98115** fix: surface node approval guidance from devices CLI. Thanks @welfo-beo. - **PR #97898** docs: clarify source checkout Node floor. Related #97792. Thanks @lin-hongkuan and @aniruddhaadak80. - **PR #94526** test(telegram): add regression test for forum topic message_thread_id with streamed reasoning. Related #89352. Thanks @xialonglee and @pmika. - **PR #98145** fix(device-pairing): don't churn requestId on subset re-requests. Thanks @RomneyDa. - **PR #98267** fix(system-prompt): move exec-approval + Authorized Senders below cache boundary. Related #98261. Thanks @headbouyJB. - **PR #98304** fix: preserve iOS chat line breaks. Related #98028. Thanks @joshavant and @Jabato01. - **PR #98187** fix(openrouter): send explicit auth headers. Related #97934. Thanks @sunlit-deng and @laurencebrown. - **PR #95708** fix: show WebChat preamble progress during tool activity. Thanks @ragesaq. - **PR #98210** fix(gateway): iOS Talk treats SecretRef-backed API keys as missing. Related #98209. Thanks @ooiuuii. - **PR #98009** test(infra): add unit tests for SQLite number normalization. Thanks @dwc1997. - **PR #98087** test(config): add unit tests for resolveExecCommandHighlighting. Thanks @solodmd. - **PR #98219** test(utils): add unit tests for chunkItems. Thanks @zenglingbiao. - **PR #98093** fix(core): propagate caller env PATHEXT through isExecutableFile on Windows. Thanks @wendy-chsy. - **PR #97973** fix(matrix): guard JSON.parse against malformed homeserver response bodies. Thanks @lsr911. - **PR #97999** fix(sms): guard Twilio JSON.parse against malformed API response bodies. Thanks @lsr911. - **PR #98043** Add Swedish mobile app localization. Thanks @yeager. - **PR #98144** fix(tui): correct disconnect copy for device scope upgrades. Thanks @RomneyDa. - **PR #98240** fix(agents): keep merged delivery routes account-bound. Thanks @yetval. - **PR #89949** fix(media): pin requester delivery route when task starts. Thanks @wangwllu. - **PR #98226** Redact bare Fireworks API keys. Related #98225. Thanks @ooiuuii. - **PR #98319** docs: publish release notes for v2026.6.11. Thanks @hannesrudolph. - **PR #98257** fix: show in-progress status for channel runs. Thanks @scotthuang. - **PR #97931** fix(gateway): keep provider-owned CLI sessions across the daily default reset. Thanks @yetval. - **PR #98325** docs: refresh docs map for v2026.6.11. Thanks @hannesrudolph. - **PR #97929** fix(auto-reply): stop level directives from eating the next message word. Thanks @yetval. - **PR #97928** fix(agents): estimate harness role sizes in context guard char estimator (fixes #97927). Thanks @liuhao1024 and @yetval. - **PR #97861** fix(compaction): count bashExecution and summary turns in pre-prompt overflow precheck. Thanks @yetval. - **PR #97137** doctor: add memory search lint findings. Thanks @giodl73-repo. - **PR #97358** Doctor: expose workspace status findings. Thanks @giodl73-repo. - **PR #95622** test(qa-lab): harden whatsapp qa scenarios. Thanks @mcaxtr. - **PR #98346** fix: prevent skill-creator from bypassing workshop proposals. Related #96054. Thanks @momothemage and @xianshishan. - **PR #98169** fix(heartbeat): scope commitment fan-out prompts. Thanks @bdjben. - **PR #97366** Doctor: expose device pairing findings. Thanks @giodl73-repo. - **PR #98366** fix: Android TLS fingerprint verification times out on slow handshakes. Related #98365. Thanks @joshavant. - **PR #98353** fix(ios): open app on Chat by default. Thanks @BsnizND. - **PR #98352** fix(security): warn on agent skill MCP boundary drift. Thanks @momothemage. - **PR #98347** fix: retry image describe fallback models. Thanks @momothemage. - **PR #98117** fix(ios): avoid transient duplicate final replies. Related #98116. Thanks @ooiuuii and @joshavant. - **PR #98293** fix(gateway): emit stale exec approval followup diagnostics. Thanks @BsnizND. - **PR #98376** fix(ios): use Gateway speech providers in Talk. Related #98153. Thanks @Tony-ooo. - **PR #66685** Suppress expired exec approval followup warnings. Thanks @pfrederiksen. - **PR #98385** fix: show actionable mobile protocol mismatch recovery. Related #98384. Thanks @joshavant. - **PR #98146** fix(cli): explain how to recover from device approve deadlock. Thanks @RomneyDa. - **PR #98423** improve(ios): clarify Control and Talk visual hierarchy. Related #98397. - **PR #98217** fix(doctor): recover legacy cron archive across devices. Thanks @masatohoshino. - **PR #98333** feat(openai): add GPT-5.6 series support. Related #98296. Thanks @steipete-oai. - **PR #96393** fix(cron): preserve action-required command output. Related #96346. Thanks @snowzlmbot and @nz365guy. - **PR #98429** fix(ios): classify TLS fingerprint timeouts. Thanks @joshavant. - **PR #98439** fix: Android setup codes accept local mDNS gateway hosts. Thanks @joshavant. - **PR #98443** fix(ios): improve light and dark appearance contrast. Related #98440. - **PR #97742** fix(llm): preserve structured tool result text across providers. Thanks @snowzlmbot. - **PR #97968** fix(status): surface unregistered memory embedding providers. Thanks @masatohoshino. - **PR #92237** fix(agents): preserve runtime settings overrides [AI-assisted]. Thanks @sercada. - **PR #95888** fix(active-memory): caveat mutable ops facts; mark truncated recall as incomplete. Thanks @spencer2211. - **PR #98291** fix(gateway): surface systemd start-limit exhaustion. Thanks @masatohoshino. - **PR #90517** fix(gateway): hint missing external plugin for web login. Related #83277. Thanks @TUARAN and @carol-iung. - **PR #98369** test(infra): add unit tests for SQLite user_version pragma helper. Thanks @dwc1997. - **PR #98340** fix: extension api.exec leaves child processes after timeout. Related #98335. Thanks @ooiuuii. - **PR #92063** fix(ui): collapse duplicate assistant groups during segmented streaming. Related #63956. Thanks @harjothkhara and @contentfree. - **PR #98354** fix(infra): guard delivery queue inflate against corrupted entry_json. Thanks @Pick-cat. - **PR #90566** fix(agents): warn on cron announce skip. Related #68561. Thanks @sahibzada-allahyar and @Mibslee. - **PR #98371** fix(ports): validate lsof PID parsing before assignment. Thanks @lzyyzznl. - **PR #98356** fix(cron): keep provider-owned CLI sessions across the daily default reset. Thanks @yetval. - **PR #98395** test(shared): add unit tests for account enabled guard. Thanks @dwc1997. - **PR #98411** fix(agents): recover thinking errors from provider body. Related #98308. Thanks @sunlit-deng and @clearhorizoninvestments. - **PR #98494** docs(skills): support variable landable sweep batches. Thanks @vincentkoc. - **PR #91240** fix: report Codex ChatGPT status auth. Related #91099. Thanks @849261680 and @ukstem. - **PR #98370** test(agents): add unit tests for thinking block detection. Thanks @dwc1997. - **PR #96711** test: prefer shared temp dir helpers in config, gateway, cron, crestodian, and state tests. Thanks @xialonglee. - **PR #98483** fix: Android QR scan starts gateway pairing. Thanks @joshavant. - **PR #95230** fix docs-list-mdx-pages. Thanks @hugenshen. - **PR #96322** fix(minimax): bound JSON response reads to prevent OOM. Thanks @lsr911. - **PR #95348** fix config-chmod-warning. Thanks @hugenshen and @cursoragent. - **PR #95229** fix(copilot): guard against undefined runtime.state during cli-metadata registration. Related #94516. Thanks @sunlit-deng and @cuihaijun. - **PR #94636** fix(memory): skip raw snippets during promotion. Thanks @tayoun. - **PR #94013** [AI] fix(feishu): guard partial channelRuntime in monitor startup. Thanks @xydt-tanshanshan. - **PR #93466** [AI] fix(feishu): guard against missing inbound in channelRuntime fallback. Thanks @xydt-tanshanshan. - **PR #98049** fix: hide expired pairing QR cards in Control UI. Related #98039. Thanks @ooiuuii. - **PR #96094** fix(memory): prove live manager recovery after CLI reindex. Related #91167. Thanks @849261680 and @kiagentkronos-cell. - **PR #98482** fix: advertise route-aware LAN Control UI links. Thanks @joshavant. - **PR #71537** Recover archived (.reset) session transcripts in memory hook + session-logs skill. Thanks @injinj. - **PR #96375** docs(config-agents): correct built-in alias table for opus and gpt. Thanks @niks999. - **PR #98453** docs(gateway): fix Telegram streaming default in config-channels.md. Thanks @solodmd. - **PR #98533** fix: repair hosted CI baseline assertions. - **PR #98421** feat(imessage): native poll support — create, read, vote. Thanks @omarshahine and @lobster. - **PR #98318** docs(matrix): document missing streaming.progress mode, progress sub-fields, and mentionPatterns config. Thanks @wm0018 and @vincentkoc. - **PR #97753** docs(onboard): document 11 missing non-interactive CLI flags. Thanks @wm0018 and @vincentkoc. - **PR #97851** fix(mattermost): bound null-body error response reads. Thanks @Pick-cat. - **PR #98360** fix(memory-wiki): preserve notes after transient page reads. Related #98345. Thanks @qingminglong and @vincentkoc and @yetval. - **PR #98551** test: fix stale core test type failures. Thanks @RomneyDa. - **PR #98455** fix(browser): bound error body read in fetchHttpJson to prevent OOM. Thanks @wings1029. - **PR #95906** fix(code-mode): surface QuickJS error name and message to the model. Thanks @ZengWen-DT and @vincentkoc. - **PR #97901** fix(agents): stop copilot autoreview cleanup crash on Windows. Thanks @paulcam206. - **PR #97923** fix(slack): truncate served arg-menu option labels on a surrogate boundary. Thanks @LEXES7. - **PR #98010** fix(update): validate bundle plugin payloads by manifest contract. Related #97985. Thanks @LiLan0125 and @herove. - **PR #85296** fix(codex): derive terminal-idle watchdog from explicit run timeout. Thanks @alkor2000 and @vincentkoc. - **PR #97110** feat(i18n): add native app locale inventory. Thanks @vincentkoc. - **PR #98396** fix: allow config.patch with defaulted provider baseUrl. Related #98270. Thanks @momothemage and @weltmaister. - **PR #98503** fix(usage-bar): use Object.hasOwn instead of in operator to avoid prototype chain pollution. Related #98466. Thanks @chenyangjun-xy and @zhangLei99586. - **PR #97111** feat(android): localize core gateway surfaces. Thanks @vincentkoc. - **PR #97630** fix(media): normalize Windows inbound paths case-insensitively. Thanks @VectorPeak. - **PR #82638** fix(agents): skip implicit provider discovery when models.mode is 'replace' [AI-assisted]. Related #66957. Thanks @eldar702 and @wangzhengshu. - **PR #87917** fix sessions json lineage metadata. Related #80286. Thanks @zhangguiping-xydt and @islandpreneur007. - **PR #93639** fix(message-tool): apply messages.responsePrefix to outbound sends. Thanks @ZengWen-DT. - **PR #94440** fix: #94432 classify Cloudflare challenge 403 as upstream_html instead of auth_html. Thanks @lzyyzznl and @pbm9z95m6z-hue. - **PR #98119** fix: reduce Docker build memory pressure. Related #98118. Thanks @zyzo. - **PR #97679** feat(node): add --context-path flag to node run/install for reverse-p…. Related #97678. Thanks @wm0018. - **PR #98339** fix(irc): classify host-less nick!user allowlist entries as mutable. Thanks @yetval. - **PR #97662** fix(matrix): bound raw transport response reads to prevent OOM. Thanks @Alix-007. - **PR #98137** fix: hoist timer declaration to avoid TDZ ReferenceError in abortable delay. Thanks @zhangLei99586. - **PR #98134** fix: clear timeout timer in Tailscale binary probe Promise.race. Thanks @zhangLei99586. - **PR #97989** fix(sms): stop internal tool-trace banners from reaching SMS replies. Thanks @ZengWen-DT. - **PR #97972** fix(browser): CDP auth fails with percent-encoded credentials. Thanks @VectorPeak. - **PR #98063** fix(reply): suppress tool-error progress delivery when messages.suppressToolErrors is set. Thanks @moeedahmed and @amittell. - **PR #94964** fix(reload): cancel deferred channel reload on in-process restart. Related #79487. Thanks @lzyyzznl and @tseller. - **PR #98598** fix: restore main lint after timer repairs. Related #98462, #98464. Thanks @zhangLei99586. - **PR #98587** fix(slack): guard relay WebSocket frame JSON.parse against malformed input. Thanks @lsr911 and @vincentkoc. - **PR #90030** fix(memory-core): skip qmd zero-hit search sync. Related #90023. Thanks @sahibzada-allahyar and @ruben2000de. - **PR #98493** fix(transcripts): close readline interface and destroy read stream on error exit. Related #98467. Thanks @wangmiao0668000666 and @zhangLei99586. - **PR #98497** fix(cli): show exit code when plugin npm install returns empty output. Thanks @Sanjays2402 and @vincentkoc. - **PR #97112** feat(apple): localize core native app surfaces. Thanks @vincentkoc. - **PR #98610** fix: restore tooling CI after transcript test addition. - **PR #77539** fix(subagent): preserve steered task text on restart redispatch. Thanks @amittell. - **PR #97113** feat(i18n): refresh all native locale artifacts. Thanks @vincentkoc. - **PR #98620** feat(doctor): warn about in-flight cron jobs. Thanks @masatohoshino. - **PR #98605** test(shared): add unit tests for human-readable list formatting. Thanks @dwc1997. - **PR #97348** feat(autoreview): support cursor-agent engine. Thanks @hxy91819. - **PR #95943** fix(cron): preserve provider/model on isolated-run timeout row. Related #95873. Thanks @ZengWen-DT and @cursoragent and @luke-renjoy. - **PR #94149** fix(status): bound systemd service probes so status cannot hang on a wedged systemctl (#84698). Thanks @ZengWen-DT and @cursoragent and @zus-assistant. - **PR #88159** fix(cli): retry logs.tail after journal fallback in logs follow. Thanks @anyech and @vincentkoc. - **PR #98508** fix(update-check): bound npm registry JSON response read to prevent OOM. Thanks @lzyyzznl. - **PR #98496** fix(tlon): bound error response body reads to prevent OOM. Thanks @Pandah97. - **PR #98554** fix(openai): bound embedding batch file downloads. Thanks @sunlit-deng and @vincentkoc. - **PR #98652** fix: stop invalid message timeouts from stalling. - **PR #77973** fix(gateway): cap agentRunCache to prevent unbounded growth under run fan-out. Related #77976. Thanks @fede-kamel and @vincentkoc. - **PR #98525** fix(agents): time out local streams without first event. Thanks @osolmaz. - **PR #94022** fix(cron): persist startup catch-up deferral ids in service state to prevent read-RPC clobber. Related #93935. Thanks @RichChen01 and @vincentkoc and @yetval. - **PR #93810** fix(cron): preserve startup overflow catch-up deferrals in start() maintenance pass. Thanks @yetval and @vincentkoc. - **PR #98623** fix: media tools skip env-key provider plugins when auto-selecting models. Thanks @medns. - **PR #98665** fix(claude-cli): return updatedInput in can_use_tool allow response for Claude Code 2.1. Related #95171. Thanks @yetval and @carterdawson. - **PR #94250** fix(feishu): send blocks as independent messages when blockStreaming is enabled. Related #55027. Thanks @xialonglee and @vincentkoc and @ZichaoLong. - **PR #93379** fix(whatsapp): thread authDir through command authorization and owner bypass for LID JID resolution. Related #77755. Thanks @xialonglee and @jiveshkalra. - **PR #98646** fix: keep workspace rail file sections scrollable. Related #98566. Thanks @wuqxuan and @645648406-max. - **PR #98602** fix: iOS Talk fallback settings opens Voice & Talk. Related #98593. Thanks @PollyBot13. - **PR #98611** fix(ui): add overflow-y:auto to workspace rail sections to prevent file list overflow (fixes #98566). Thanks @zw-xysk and @645648406-max. - **PR #98619** fix(qa-lab): credential lease requests fail on oversized Convex broker responses. Thanks @ZengWen-DT. - **PR #94326** fix(memory-wiki): disambiguate the reserved index page stem for synthesis and ingest. Thanks @yetval and @vincentkoc. - **PR #98659** fix(codex): classify get_goal read statuses as successful dynamic tool calls. Thanks @yetval. - **PR #96856** fix(codex): successful sessions_spawn and goal tool results recorded as failures. Thanks @nxmxbbd. - **PR #98660** fix(inworld): guard voices JSON.parse against malformed API response bodies. Thanks @solodmd. - **PR #95430** fix(embedded-agent-runner): pump async streamFn through pumpStreamWithRecovery for mid-stream error recovery. Related #95429. Thanks @lzyyzznl and @vincentkoc and @alexelgier. - **PR #98644** fix: tool summaries preserve emoji truncation boundaries. Thanks @ZengWen-DT. - **PR #80928** fix(telegram): suppress fallback reply when plugin command returns suppressReply: true. Related #80756. Thanks @alexuser and @UnClouded77. - **PR #98701** fix: prevent agents-tools message test timeouts. - **PR #92657** feat(usage): ship built-in /usage full footer. Thanks @Marvinthebored. - **PR #92877** fix(usage): make built-in footer easier to wrap on Telegram. Thanks @Marvinthebored. - **PR #98126** Restore Telegram /steer for active Codex runs. Related #81594. Thanks @100yenadmin and @Kyzcreig. - **PR #92037** feat(cron): on-exit schedule — wake on a watched command's exit. Thanks @anagnorisis2peripeteia. - **PR #98452** feat(ios): modernize the app with iOS 26 Liquid Glass. - **PR #98006** Add Telegram /login Codex pairing flow. Thanks @100yenadmin. - **PR #98735** fix(telegram): preserve rich forwarded message text. Thanks @obviyus. - **PR #97962** refactor(qa): use transport-native actions in flow scenarios. Thanks @RomneyDa. - **PR #98726** fix(nvidia): use Nemotron Super 1M context. Thanks @eleqtrizit. - **PR #98691** fix(imessage): shed emoji anywhere in poll-vote echo match. Thanks @omarshahine. - **PR #97174** Fix Telegram plugin callback routing. Thanks @goldmar. - **PR #89597** fix: migrate QQBot credential backups to SQLite KV. - **PR #98536** feat: prepare scoped conversation capability profiles. - **PR #92274** fix(agents): classify embedded prompt lock error as permanent announce failure. Related #91527. Thanks @fsdwen and @zackchiutw. - **PR #98102** fix(telegram): durably retry inbound media dropped during restart (#98076). Thanks @luoyanglang and @DaveArcher18. - **PR #98755** fix(cron): detach session-targeted runs. Related #98121. Thanks @obviyus and @EthanSK. - **PR #96065** fix(install): manage config-secretref env refs via OPENCLAW_SERVICE_MANAGED_ENV_KEYS. Thanks @Darren2030 and @obviyus. - **PR #98666** fix: diagnose Windows LAN Gateway firewall blocks. Thanks @joshavant. - **PR #98501** fix(codex): rename destructive approval mode to ask. Related #98499. Thanks @kevinslin. - **PR #98775** fix(telegram): survive transient getUpdates errors and stop per-send cache rewrites. Related #98772, #98773. Thanks @obviyus. - **PR #98776** fix(telegram): back off, dead-letter, and tombstone spooled updates so poison messages cannot block or duplicate. Related #98774. Thanks @obviyus. - **PR #96454** feat(cli): openclaw attach — launch an external harness bound to a gateway session. Thanks @anagnorisis2peripeteia and @obviyus. - **PR #98786** fix(telegram): final replies no longer drop on rejected rich entities, captions, quotes, or long flood waits. Related #98778. Thanks @obviyus. - **PR #97496** Doctor: expose channel plugin blocker findings. Thanks @giodl73-repo. - **PR #98792** fix(ci): restore docs and test type checks. - **PR #98736** improve(ios): simplify Talk controls and composer alignment. - **PR #98183** fix(gateway): distinguish reachable gateway from failed status probe. Thanks @masatohoshino. - **PR #98808** docs(telegram): move maintainer decisions into scoped AGENTS.md with reliability invariants. Thanks @obviyus. - **PR #98138** fix: guard setDeep against empty keys array in Chrome profile decoration. Thanks @zhangLei99586. - **PR #92283** fix(agents): don't inject A2A turns into isolated-cron sessions_send (#92257). Thanks @harjothkhara and @vincentkoc and @nailujac. - **PR #98812** fix(codex): preserve plugin app approvals in side conversations. - **PR #97889** fix(discord): guard JSON.parse against malformed API response bodies. Thanks @lsr911. - **PR #98689** fix(wizard): reject loose gateway port input. Related #98681. Thanks @qingminglong. - **PR #98720** fix(nostr): clear per-relay publish timeout timer to prevent dangling handles. Related #98463. Thanks @wangmiao0668000666 and @zhangLei99586. - **PR #98787** fix(memory-wiki): retry transient existing-page reads in wiki_apply and chatgpt import. Thanks @yetval and @vincentkoc. - **PR #98818** fix(ci): recover incomplete Swift build caches. - **PR #98811** feat(ios): modernize navigation and settings. Related #98803. - **PR #98843** docs: update mobile app release messaging. Thanks @joshavant. - **PR #93209** test: prefer auto-cleaning temp dir helper. Thanks @hxy91819. - **PR #98789** fix(telegram): sends and actions without an account id ignore the configured defaultAccount. Thanks @yetval. - **PR #98806** fix(telegram): webhook updates survive crashes and restarts via durable spooling. Related #98777. Thanks @obviyus. - **PR #98688** fix(fal): route grok-imagine and nano-banana-2-lite edits to correct endpoints. Thanks @davenicoll and @vincentkoc. - **PR #98891** fix(agents): normalize non-array tool-result content at transcript ingest. Related #98825. Thanks @obviyus and @snowzlmbot. - **PR #98781** fix(imessage): poll render vote-cue, cross-run echo suppression, and comment fold. Thanks @omarshahine. - **PR #97500** Doctor: expose tool result cap findings. Thanks @giodl73-repo. - **PR #98769** fix: Telegram replies duplicate recent context after sent replies. Related #98767. Thanks @rabsef-bicrym. - **PR #98933** fix(agents): stop gateway crash from wedged claude-cli turns and persist heartbeat session bindings. Related #98894, #98895. Thanks @obviyus. - **PR #98934** fix(agents): recover claude-cli context-overflow sessions and keep retry artifacts alive. Related #98897. Thanks @obviyus. - **PR #98908** refactor(agents): fold assistant string normalization into transcript ingest. Thanks @obviyus. - **PR #98738** fix(agents): fail fast with attributable reason after MCP stdio session dies mid-run. Thanks @masatohoshino and @vincentkoc. - **PR #98879** fix: backup skips volatile cache paths. Related #98865. Thanks @ZengWen-DT and @vincentkoc and @carterstebbins23-spec. - **PR #98942** fix(agents): unify claude-cli output classification across live and one-shot paths. Related #98896. Thanks @obviyus. - **PR #98932** fix(anthropic): restore Fable 5 Vertex simple completions. - **PR #98947** fix(cron): restore persistent session targets. - **PR #96523** fix(agents): preserve embedded OpenAI completions usage. Thanks @ly85206559 and @vincentkoc. - **PR #98758** perf(build): reduce plugin SDK declaration package size. Related #98757. Thanks @RomneyDa. - **PR #98877** fix(mattermost): include later team members in peer directory. Related #98871. Thanks @qingminglong. - **PR #98953** feat(ios): refine the chat experience. Related #98929. - **PR #98876** fix(terminal): preserve sibling home-prefix paths. Related #98872. Thanks @qingminglong. - **PR #98930** feat(ios): PR1 brand color palette overhaul. Thanks @joelnishanth. - **PR #94566** fix(android): make offline chat actionable. Thanks @Tosko4. - **PR #98955** fix(agents): preserve fresh tool result text under aggregate cap. Related #98874. Thanks @momothemage and @lamkan0210. - **PR #98059** [codex] Support Android selected photo access. Thanks @NianJiuZst. - **PR #98914** fix(android): return settings details to their originating tab on Back. Thanks @Lokimorty. - **PR #98898** fix(ios): back from settings details returns to the originating screen. Thanks @Lokimorty. - **PR #98235** fix(feishu): include video upload duration. Thanks @areslp. - **PR #98966** fix(discord): gate guild metadata reads [AI]. Thanks @pgondhi987. - **PR #98985** fix: clean up iOS About page copy. Related #98943. Thanks @sahilsatralkar. - **PR #98856** fix(ios): gateway error shows twice on the Settings Gateway page. Thanks @Lokimorty. - **PR #98936** fix: Control row icons use inconsistent row styling (iOS). Related #98916. Thanks @sahilsatralkar. - **PR #98040** [codex] Fix Android camera snap cleanup. Thanks @NianJiuZst. - **PR #99039** fix(macos): stop runtime config-health sidecar access. Related #98917. Thanks @momothemage and @P51moustache. - **PR #92667** ci: add process exec CodeQL security shard. Thanks @hxy91819. - **PR #98055** [codex] Gate Android Talk capture starts in background. Thanks @NianJiuZst. - **PR #98067** [codex] Cancel Android gateway pending RPCs on close. Thanks @NianJiuZst. - **PR #98698** fix(android): show specific gateway auth-recovery reason instead of generic label. Related #98046. Thanks @masatohoshino and @ccaprani. - **PR #83826** test(android): poll for stale TLS probe cleanup in auth test. Thanks @NeatGuyCoding. - **PR #98983** fix(agents): handle variadic claude --mcp-config and serialize gemini credential staging. Related #98944, #98945. Thanks @obviyus. - **PR #99145** fix(auto-reply): suppress room-event notice leaks. Thanks @obviyus. - **PR #99144** fix(auto-reply): default room events to silence. Thanks @obviyus. - **PR #98608** fix: Mattermost fails to load after configured plugin repair. Related #98564. Thanks @jacobtomlinson. - **PR #99143** fix(telegram): keep group history always on. Related #99142. Thanks @obviyus. - **PR #99159** fix(agents): claude-cli lifecycle cleanup — loopback fail-loud, exit-0 failover, bounded reseed, image sweep, one prepare cleanup owner. Related #98946. Thanks @obviyus. - **PR #98391** Expose disk space doctor lint findings. Thanks @giodl73-repo. - **PR #98835** fix(config/sessions): narrow reply-session initialization revision to identity fields. Related #98672. Thanks @moguangyu5-design and @jalehman and @AaronFaby. - **PR #99123** fix(android): ignore chat events with missing assistant role in voice text extraction. Thanks @ly85206559 and @cursoragent. - **PR #99147** fix(android): preserve split SMS permission grants. Thanks @NianJiuZst. - **PR #99107** fix(android): bracket IPv6 hosts in manual gateway URL composition. Thanks @ly85206559 and @cursoragent. - **PR #99158** fix: require Android contact and calendar write permissions in onboarding. Thanks @NianJiuZst. - **PR #99110** fix(android): strip ws scheme prefix from manual gateway host input. Related #87216. Thanks @ly85206559 and @cursoragent and @ruben2000de. - **PR #99212** fix(ci): session concurrency test flakes during child handshake. - **PR #94385** fix(feishu): preserve button command values in fallback text and add Feishu comment guidance with callback privacy. Related #69754. Thanks @xialonglee and @1yihui. - **PR #98563** fix: route iOS OpenAI realtime Talk through WebRTC. Thanks @PollyBot13. - **PR #99204** fix: require Android contact and calendar write permissions. Thanks @NianJiuZst. - **PR #99134** fix: OAuth refresh failures report reauth instead of stale success. Related #99120. Thanks @100yenadmin. - **PR #99153** fix: clean up Android camera clips on cancellation. Thanks @NianJiuZst. - **PR #99118** [codex] fix(memory-lancedb): align apache arrow peer dependency. Related #90295. Thanks @allenhurff and @joshavant. - **PR #98066** fix: keep iOS LAN QR pairing authenticated after bootstrap. Related #98064. Thanks @ooiuuii. - **PR #99155** fix: stop iOS screen recording after cancellation. Thanks @NianJiuZst. - **PR #95973** fix(telegram): explain disabled plugin approval failures. Related #95800. Thanks @MonkeyLeeT and @ChrisBot2026. - **PR #99233** fix: ignore test-only network CI guard lines. Thanks @joshavant. - **PR #98951** fix: strict guarded fetch fails before managed proxy DNS. Related #98925. Thanks @momothemage and @sandl99. - **PR #99137** fix: prevent Voice Wake crash after Talk audio capture. Thanks @PollyBot13. - **PR #99052** fix: Update Dark/Light mode UI control appearance. Related #98995. Thanks @sahilsatralkar. - **PR #99245** fix(ios): return chat to originating control detail. Thanks @Solvely-Colin. - **PR #92602** fix(android): queue node events until gateway connect. Related #79552. Thanks @ashishpatel26 and @hectorrp13. - **PR #98277** fix: keep Android gateway settings save idempotent. Thanks @Solvely-Colin. - **PR #99256** fix(auto-reply): single canonical group history and deduped turn metadata. Related #99218. Thanks @obviyus. - **PR #99259** fix(android): use Bluetooth microphones for voice capture. Related #96241. Thanks @gwtaylor. - **PR #98751** test(qa): prove native command targeting across QA transports. Thanks @RomneyDa. - **PR #98779** test(qa): cover expanded Crabline bindings. Thanks @RomneyDa. - **PR #99262** test(qa): cover Crabline Signal sends. Thanks @RomneyDa. - **PR #99261** refactor(shared): establish lazy runtime loader foundation. Thanks @RomneyDa. - **PR #99264** test(qa): cover Crabline Mattermost sends. Thanks @RomneyDa. - **PR #99265** test(qa): cover Crabline Matrix sends. Thanks @RomneyDa. - **PR #98400** Expose heartbeat template doctor lint findings. Thanks @giodl73-repo. - **PR #98695** Expose legacy plugin manifest doctor lint findings. Thanks @giodl73-repo. - **PR #99278** refactor(shared): consolidate core leaf lazy loaders. Thanks @RomneyDa. - **PR #99274** fix(zalo): match native bot identity fields. Thanks @RomneyDa. - **PR #99126** test(discord): clarify and guardrail gateway proxy selection. Related #98266. Thanks @svuppala2006 and @joshavant and @sallyom. - **PR #99290** feat(ios): add licenses settings screen. Thanks @joshavant. - **PR #98907** fix(telegram): distinguish and render streamed reasoning/commentary progress lanes. Thanks @Marvinthebored. - **PR #99294** fix(qa): stagger isolated worker startup. Thanks @RomneyDa. - **PR #99276** fix(memory-wiki): source imports crash on unreadable pages. Thanks @obviyus. - **PR #99296** refactor(shared): consolidate gateway and stateful runtime lazy loaders. Thanks @RomneyDa. - **PR #98768** Allow alternate Zalo Bot API roots. Thanks @RomneyDa. - **PR #99298** refactor(shared): consolidate Discord Slack and Telegram lazy loaders. Thanks @RomneyDa. - **PR #99307** fix(memory-wiki): avoid implicit error coercion. Thanks @RomneyDa. - **PR #99306** feat(auto-reply): persist ambient room events as transcript rows. Related #99257. Thanks @obviyus. - **PR #99302** refactor(shared): consolidate remaining channel lazy loaders. Thanks @RomneyDa. - **PR #99303** test(qa): cover Crabline Zalo transport. Thanks @RomneyDa. - **PR #99299** feat(android): add licenses settings screen. Thanks @joshavant. - **PR #99220** fix(ios): onboarding Retry Connection does nothing after editing gateway details. Related #99219. Thanks @abdullahtas0. - **PR #98749** refactor(shared): consolidate provider and utility lazy loaders. Thanks @RomneyDa. - **PR #99355** fix(ci): restore Telegram and plugin SDK guard checks. Thanks @RomneyDa. - **PR #88899** fix(android): render chat content through Markdown. Related #88014. Thanks @Pluviobyte and @Iman-Sharif. - **PR #99310** test(qa): migrate channel streaming evidence to transport flow. Thanks @RomneyDa. - **PR #99350** fix(ios): add Photos permission controls. Related #99046. Thanks @Tony-ooo. - **PR #99361** refactor(plugins): consolidate record guards. Thanks @RomneyDa. - **PR #99359** refactor(shared): consolidate core record guards. Thanks @RomneyDa. - **PR #97208** fix: avoid DeepSeek-native thinking on OpenRouter V4. Related #97196. Thanks @NianJiuZst and @patelmm79. - **PR #99385** fix(sessions): scope ambient transcript watermark to session id. Related #99373. Thanks @obviyus. - **PR #99389** fix(auto-reply): restore per-turn message-tool delivery contract. Related #99371. Thanks @obviyus. - **PR #98269** fix(android): derive Voice readiness from Gateway catalog. Related #98268. Thanks @Solvely-Colin. - **PR #92872** fix(qqbot): allow scoped sandbox media sends. Thanks @zhangguiping-xydt and @sliverp. - **PR #99414** fix(android): expose exact gateway recovery actions. Related #98045, #98046. Thanks @ccaprani. - **PR #99289** feat: session-first sidebar, compact context ring, and warm light theme for the Control UI. Related #99288. - **PR #99234** feat(nodes): add auto-discovered Ollama inference. Related #99228. - **PR #97095** fix: memory_search honors generic embedding providers. Thanks @849261680. - **PR #98841** fix(gateway): include session label in deriveSessionTitle fallback chain. Related #98742. Thanks @SunnyShu0925 and @BSG2000. - **PR #99301** fix(feishu): catch unhandled promise rejection in streaming card flush timer. Thanks @lwy-2. - **PR #99391** fix(compaction): count nested tool result content. Related #99375. Thanks @LZY3538 and @imchloe92. - **PR #99407** fix(daemon): avoid loading full gateway logs during diagnostics. Thanks @sunlit-deng. - **PR #99291** Fix/issue 98958 gateway lock fd leak. Related #98958. Thanks @chenyangjun-xy and @zhangLei99586. - **PR #99475** fix(ios): contacts.add crashes the app via unfetched CNContactFormatter keys. Thanks @abdullahtas0. - **PR #98003** fix(anthropic): wire buildGuardedModelFetch into the Cloudflare createClient branch. Thanks @wangmiao0668000666. - **PR #99425** fix: strip conda env marker from host tool runs. Related #99424. Thanks @ooiuuii and @krissding. - **PR #99398** fix(cli): reject unsafe sessions tail counts. Thanks @qingminglong. - **PR #98855** fix: chat.send no reply when thinking metadata is set. Thanks @jesse-merhi. - **PR #98752** Rework Android gateway onboarding setup. Thanks @jesse-merhi. - **PR #99446** fix(agents): preserve fd find failures. Thanks @zhangguiping-xydt. - **PR #99152** fix(config): use Object.hasOwn instead of in operator in restoreOriginalValueOrThrow. Thanks @zenglingbiao. - **PR #99460** fix: redact dotted API key activity previews. Related #99459. Thanks @ooiuuii. - **PR #99455** fix: long mobile media recordings time out. Thanks @NianJiuZst. - **PR #99410** fix(subagents): match requesterSessionKey when controllerSessionKey differs in list filter. Related #75593. Thanks @sheyanmin and @aaajiao. - **PR #98791** feat(signal): show status reactions during inbound replies. Thanks @jesse-merhi. - **PR #98683** fix(ui): keep landscape composer compact. Related #98615. Thanks @qingminglong and @jin-li. - **PR #99428** fix(logging): redact Telegram bot tokens from timeout URLs. Related #96982. Thanks @xialonglee and @liuhaiyang14. - **PR #99217** Preserve Codex output after missing turn completion. Thanks @100yenadmin and @Sedrak-Hovhannisyan and @fuller-stack-dev. - **PR #99520** fix(gateway): declare the dev agent required by the gateway e2e session key. Related #99513. Thanks @masatohoshino. - **PR #95738** feat(signal): add target aliases. Thanks @jesse-merhi. - **PR #98258** improve: make native chat scrolling reader-managed. Related #98255. Thanks @christopheraaronhogg. - **PR #99506** fix: keep always-on group fallback messages in dispatch. Related #99457. Thanks @LZY3538 and @zqchris. - **PR #89671** fix(google-meet): force English Meet UI via hl=en so automation works on any locale. Thanks @Unayung. - **PR #98130** fix(infra): bound jsonl-socket response buffer to prevent OOM. Thanks @Pick-cat. - **PR #99526** fix(agents): preserve primitive tool result output. Related #99523. Thanks @snowzlm. - **PR #99525** fix(imessage): recognize bare hex group chat identifiers as chat targets. Related #89235. Thanks @MatthewDelprado. - **PR #99098** fix: harden native i18n identifier filtering. Thanks @hxy91819. - **PR #99099** fix: harden docs map heading rendering. Thanks @hxy91819. - **PR #98725** Expose legacy plugin dependency doctor lint findings. Thanks @giodl73-repo. - **PR #99595** fix(agents): keep cli session binding facts session-stable. Related #99372. Thanks @obviyus. - **PR #90152** fix(telegram): stop duplicate fallback when dispatch fails after final reply. Thanks @zhangguiping-xydt. - **PR #98729** Expose stale plugin runtime symlink doctor lint findings. Thanks @giodl73-repo. - **PR #99591** fix(android): preserve numeric invoke error codes. Thanks @ly85206559. - **PR #98406** Expose WhatsApp responsiveness doctor lint findings. Thanks @giodl73-repo. - **PR #99570** fix(android): reject IPv6 zone IDs in gateway endpoint URLs. Thanks @ly85206559 and @cursoragent. - **PR #99557** fix(android): filter device and internal sessions from thread picker. Thanks @ly85206559 and @cursoragent. - **PR #99568** fix(android): block self-package notification forwarding in allowlist mode. Thanks @ly85206559 and @cursoragent. - **PR #99592** fix(android): parse talk directive aliases case-insensitively. Thanks @ly85206559. - **PR #99477** fix: avoid iOS node permission prompts. Thanks @NianJiuZst. - **PR #99374** improve(qa): standardize script evidence output. Thanks @RomneyDa. - **PR #99468** improve: tighten iOS Control row density. Related #99439. Thanks @sahilsatralkar. - **PR #99642** test: avoid cross-OS socket close event race. Thanks @RomneyDa. - **PR #89967** fix(macos): LaunchAgent starts gateway on external home volumes. Related #87199. Thanks @zhangguiping-xydt and @joshdaynard. - **PR #98613** fix(media): guard ffprobe JSON parse against malformed output. Thanks @Pick-cat. - **PR #97839** fix: log terminal session persistence failures. Related #97795. Thanks @LZY3538 and @aniruddhaadak80. - **PR #99247** feat: clarify iOS Location Always permission flow. Thanks @PollyBot13. - **PR #98224** fix(auto-reply): strip stray punctuation before silent-reply token detection. Thanks @SunnyShu0925. - **PR #97328** fix(google): rotate Gemini API keys for LLM requests. Thanks @MonkeyLeeT. - **PR #99661** fix(macos): remote mode fails with managed SSH aliases. - **PR #99649** fix(qa): defer partial Crabline recorder rows. Related #99648. Thanks @RomneyDa. - **PR #99211** Expose legacy cron store doctor lint findings. Thanks @giodl73-repo. - **PR #99629** test(qa): redact script evidence diagnostics. Thanks @RomneyDa. - **PR #99647** Fix Slack retry for session init conflicts. Thanks @steipete-oai. - **PR #99628** improve: enforce canonical QA scenario ownership. Related #99627. Thanks @RomneyDa. - **PR #99632** refactor(qa): simplify transport adapter contracts. Related #99622. Thanks @RomneyDa. - **PR #99656** test(qa): use full-turn budget for native stop recovery. Related #99655. Thanks @RomneyDa. - **PR #99605** fix(google): bound OAuth token error response reads. Thanks @mushuiyu886. - **PR #99679** fix(qa): consume Crabline events without recorder polling. Related #99664. Thanks @RomneyDa. - **PR #99687** refactor(infra): consolidate SHA-256 digest helpers. Related #99675. Thanks @RomneyDa. - **PR #98796** [AI-assisted] feat(android): add chat command controls. Thanks @IWhatsskill. - **PR #99671** refactor: consolidate number coercion callers. Related #99667. Thanks @RomneyDa. - **PR #99640** fix: CLI agent session resume churns when owner and non-owner alternate in a group. Related #99633. Thanks @obviyus. - **PR #99682** refactor(models): consolidate catalog ref parsing. Related #99674. Thanks @RomneyDa. - **PR #99566** fix(exec): avoid splitting surrogate pairs in approval display. Thanks @mikasa0818. - **PR #99702** refactor: remove redundant unique-list aliases. Related #99697. Thanks @RomneyDa. - **PR #99246** feat(ios): implement branded typography design system. Thanks @joelnishanth and @cursoragent. - **PR #99710** fix(build): Docker package preparation misses plugin SDK declarations. Thanks @RomneyDa. - **PR #99715** refactor: consolidate image data URL formatting. Thanks @RomneyDa. - **PR #98764** fix(ui): copy workspace file paths over plain HTTP. Related #98759. Thanks @ZengWen-DT and @adinballew. - **PR #99678** fix(build): forward default exports through stable runtime aliases. Related #99677. Thanks @headbouyJB and @vincentkoc. - **PR #99370** fix(file-transfer): don't inline zero-byte files as image content blocks. Thanks @2loch-ness6 and @vincentkoc. - **PR #99540** fix(doctor): shell completion install fails doctor when profile is read-only. Related #99237. Thanks @rballiance and @hunglp6d. - **PR #99718** refactor(text): consolidate cleanup owners. Thanks @RomneyDa. - **PR #98819** fix(plugins): resolve public artifacts from installed plugin roots. Related #98740. Thanks @amknight and @KelTech-Services. - **PR #99721** refactor: consolidate async timing helpers. Thanks @RomneyDa. - **PR #99722** fix: group agent session resume churns when messages toggle between @-mention and plain. Related #99696. Thanks @obviyus. - **PR #99676** refactor: consolidate string reader mechanics. Related #99663. Thanks @RomneyDa. - **PR #99705** improve(qa): execute runtime scenarios through Docker. Thanks @RomneyDa. - **PR #99231** improve: native iOS look with stock SwiftUI navigation, forms, chat, and talk visualizer. Related #99195. Thanks @marvkr. - **PR #99549** fix(auto-reply): don't block reply completion on transcript mirror. Thanks @Shagrat2. - **PR #99736** fix(qa): prevent smoke gateways from losing built files. Related #99734. Thanks @RomneyDa. - **PR #99658** feat(providers): add ClawRouter routing and quotas. Related #99657. - **PR #99238** Expose channel preview warning doctor findings. Thanks @giodl73-repo. - **PR #99759** fix(providers): resolve ClawRouter auth-profile models. - **PR #99561** fix: keep OpenClaw control tools available when tool_search misroutes. Related #99464. Thanks @100yenadmin and @joshavant. - **PR #99750** refactor: consolidate exact boolean coercion. Thanks @RomneyDa. - **PR #99753** refactor: consolidate abort primitives. Thanks @RomneyDa. - **PR #99426** feat: add slash command picker in chat composer. Thanks @VicZhang6 and @Solvely-Colin. - **PR #99771** refactor: consolidate free-port test helpers. Thanks @RomneyDa. - **PR #99755** refactor: consolidate policy-free deferred promises. Thanks @RomneyDa. - **PR #99719** refactor(net): consolidate URL protocol predicates. Thanks @RomneyDa. - **PR #99743** fix: avoid native command QA timeout under CI contention. Thanks @RomneyDa. - **PR #99368** fix(qa): prevent qa smoke ci timeouts under gateway concurrency. Thanks @RomneyDa. - **PR #99778** refactor(scripts): share regexp literal escaping. Thanks @RomneyDa. - **PR #99737** test: add executable runtime fixture canaries. Thanks @RomneyDa. - **PR #99735** test(qa): exercise Gateway and MCP scenarios over real transports. Thanks @RomneyDa. - **PR #99784** fix(qa): stabilize primary smoke runtime evidence. Thanks @RomneyDa. - **PR #99726** fix(onboard): skip unavailable skill installers via lifecycle readiness preflight. Thanks @fuller-stack-dev and @Sedrak-Hovhannisyan. - **PR #99793** ci: reuse one package in QA smoke. - **PR #99767** feat(macos): install and run the local Gateway automatically. Related #99764. - **PR #99820** ci: increase artifact Testbox memory. - **PR #99822** feat: soft-resume CLI sessions on prompt drift instead of hard invalidation. Related #99729. Thanks @obviyus. - **PR #99129** fix(markdown-core): use Object.hasOwn instead of in operator in parseFrontmatterBlock. Thanks @zenglingbiao and @vincentkoc. - **PR #99803** fix(mcp): suppress unhandled error on stderr pipe in stdio transport. Thanks @cxbAsDev and @vincentkoc. - **PR #99802** fix(supervisor): suppress unhandled stream errors on child stdout/stderr. Thanks @cxbAsDev and @vincentkoc. - **PR #99800** fix(ssh-tunnel): handle spawn error to prevent unhandled rejection crash. Thanks @cxbAsDev and @vincentkoc. - **PR #99653** fix(cli): hide synthetic Claude reseed prompts. Related #99646. Thanks @ZOOWH and @vincentkoc and @Jeehut. - **PR #99728** fix(config): preserve recovery state during config-health migration. Related #99280. Thanks @joshavant and @jalehman and @ccbridle. - **PR #99839** fix(gateway): preserve legacy reseed attachments. Thanks @vincentkoc. - **PR #99830** fix: stop rubber-band bounce of the Control UI shell in the Mac app web view. - **PR #99851** fix(agents): preserve fallback tool-call hints. Thanks @vincentkoc. - **PR #98994** fix(line): truncate outbound altText, location, menu, and code fields on code point boundaries. Thanks @LEXES7 and @vincentkoc. - **PR #99846** fix(config): restrict config paths to own properties. Thanks @vincentkoc and @zenglingbiao. - **PR #99861** fix(telegram): one outbound rich-HTML normalizer and one rich-to-plain fallback policy. Related #99833. Thanks @obviyus. - **PR #99866** fix(telegram): classify inbound events from the canonical mention decision so direct mentions stop lurking. Related #99854. Thanks @obviyus. - **PR #99832** fix: reject incompatible Node 23 runtimes. Thanks @fuller-stack-dev. - **PR #99243** Polish iOS onboarding and chat critique fixes. Thanks @jcooley8. - **PR #99714** perf(usage): shrink durable usage cache entries. Related #99511. Thanks @dexhunter and @wayne524. - **PR #99838** feat: declutter the Control UI shell — reasoning effort slider, borderless composer controls, version out of the sidebar. Related #99837. - **PR #93686** fix(weixin): startAccount preserves session routing. Related #93556. Thanks @zhangguiping-xydt and @htkillermax-gif. ### Release verification - npm package: https://www.npmjs.com/package/openclaw/v/2026.7.1-beta.2 - registry tarball: https://registry.npmjs.org/openclaw/-/openclaw-2026.7.1-beta.2.tgz - integrity: `sha512-KYPBQnAfEb/9qrxlw/96a90mMQeKdAZdUABMROOue9Ph2oFbnDGezZjd5Bmw4WhRyzgyvHOHqHje/swGipC4xA==` - release SHA: `a580a7fe3fbd1b3329c978d58ca2f70e8ca37aee` - full release CI report: https://github.com/openclaw/releases/blob/main/evidence/2026.7.1-beta.2/release-evidence.md - release publish: https://github.com/openclaw/openclaw/actions/runs/28735224348 - npm preflight: https://github.com/openclaw/openclaw/actions/runs/28717730132 - full release validation: https://github.com/openclaw/openclaw/actions/runs/28717729503 - plugin npm publish: https://github.com/openclaw/openclaw/actions/runs/28735371120 - plugin ClawHub publish: dispatched separately, not awaited by this proof: https://github.com/openclaw/openclaw/actions/runs/28735371597 - plugin ClawHub bootstrap: not needed - OpenClaw npm publish: https://github.com/openclaw/openclaw/actions/runs/28735575588 - npm Telegram beta E2E: not supplied

### 亮点 - **OpenAI GPT-5.6 支持：** OpenClaw 现已在目录、能力和运行时选择路径中识别 GPT-5.6 模型家族。(#98333) 感谢 @steipete-oai。 - **外部测试工具挂载：** `openclaw attach` 针对现有 Gateway 会话启动外部测试工具，使交互式 Codex 风格的工作流更易于恢复和审查。(#96454) 感谢 @anagnorisis2peripeteia 和 @obviyus。 - **Telegram Codex 工作流：** Telegram 现可通过 `/login` 启动 Codex 配对，引导活跃的 Codex 运行，并在临时 API 故障中恢复最终回复。(#98006, #98126, #98786) 感谢 @100yenadmin, @Kyzcreig 和 @obviyus。 - **事件驱动的 cron 运行：** 新增的 `on-exit` 调度类型可在被监控命令退出时唤醒代理，而针对会话的运行可干净分离。(#92037, #98755) 感谢 @anagnorisis2peripeteia, @obviyus 和 @EthanSK。 - **原生应用刷新：** iOS 采用 iOS 26 视觉系统，提供更清晰的导航、设置、Chat、Talk 和入门引导流程，同时原生应用本地化覆盖 Apple 和 Android 平台。(#98452, #98736, #98811, #97110, #97111, #97112, #97113) 感谢 @vincentkoc。 - **更丰富的消息功能：** iMessage 获得原生投票创建、阅读和投票功能，内置的使用量页脚为聊天中的每轮对话提供更清晰的计数。(#98421, #92657, #92877) 感谢 @omarshahine, @lobster 和 @Marvinthebored。 - **更安全的作用域对话：** 能力配置文件为每个对话设置工具和访问边界，而不削弱现有的默认配置文件。(#98536) ### 变更 - **模型和提供商覆盖：** 添加 GPT-5.6 支持，使用 Nemotron Super 的 1M 上下文窗口，并保留显式的 OpenRouter 认证头。(#98333, #98726, #98187) 感谢 @steipete-oai, @eleqtrizit, @sunlit-deng 和 @laurencebrown。 - **CLI 和节点工作流：** 添加 `openclaw attach`、节点上下文路径支持、可操作的设备审批恢复指南以及更清晰的插件安装退出诊断。(#96454, #97679, #98115, #98146, #98497) 感谢 @anagnorisis2peripeteia, @obviyus, @wm0018, @welfo-beo, @RomneyDa, @Sanjays2402 和 @vincentkoc。 - **Cron 和用量：** 添加退出触发调度、分离的会话定向运行、飞行中任务医生警告以及内置完整用量页脚。(#92037, #98755, #98620, #92657, #92877) 感谢 @anagnorisis2peripeteia, @obviyus, @EthanSK, @masatohoshino 和 @Marvinthebored。 - **原生应用和本地化：** 现代化 iOS 导航、设置、呈现和 Talk 控制，添加 Gateway 语音提供商，改进 QR 入门引导和协议恢复，本地化核心 Apple 和 Android 界面，并添加瑞典移动端本地化。(#98452, #98736, #98811, #98376, #98302, #98385, #97110, #97111, #97112, #97113, #98043) 感谢 @Tony-ooo, @joelnishanth, @cursoragent, @joshavant, @vincentkoc 和 @yeager。 - **消息能力：** 添加原生 iMessage 投票以及 Telegram Codex 配对和引导流程。(#98421, #98006, #98126) 感谢 @omarshahine, @lobster, @100yenadmin 和 @Kyzcreig。 - **医生和诊断：** 暴露认证配置文件、工作区、设备配对、频道插件、内存提供商、systemd 耗尽和 Windows LAN 防火墙发现。(#97125, #97358, #97366, #97496, #97968, #98291, #98666) 感谢 @giodl73-repo, @masatohoshino 和 @joshavant。 - **对话和审查控制：** 准备作用域对话能力配置文件，并添加 Cursor Agent 作为自动审查引擎。(#98536, #97348) 感谢 @hxy91819。 ### 修复 - **Telegram 持久性：** 恢复停滞的入站请求，重试重启时丢失的媒体，在临时轮询错误中存活，对毒化更新进行死信处理，保留转发的富文本，正确路由插件回调，并在富最终回复被拒绝时安全回退。(#97118, #98102, #98735, #98775, #98776, #97174, #98786) 感谢 @vincentkoc, @luoyanglang, @DaveArcher18, @obviyus 和 @goldmar。 - **代理和上下文可靠性：** 保留运行时覆盖和引导的子代理任务，防止隔离的 cron `sessions_send` 回复反馈回请求者，改进测试工具感知的上下文估计和压缩预检查，超时静默的本地流，恢复流中故障，并限制 Gateway 运行缓存增长。(#92237, #92283, #77539, #97928, #97861, #98525, #95430, #77973) 感谢 @sercada, @amittell, @liuhao1024, @yetval, @harjothkhara, @nailujac, @osolmaz, @lzyyzznl, @vincentkoc, @alexelgier 和 @fede-kamel。 - **提供商和网络安全：** 限制 Moonshot、MiniMax、Anthropic OAuth、Discord、Matrix、SMS、浏览器、更新、嵌入、Tlön 和 Inworld 路径中的超大或格式错误响应。(#96502, #96322, #96644, #97693, #97889, #97662, #97999, #98455, #98508, #98554, #98496, #98660) 感谢 @hugenshen, @cursoragent, @lsr911, @solodmd, @Alix-007, @wings1029, @lzyyzznl, @sunlit-deng, @vincentkoc 和 @Pandah97。 - **频道投递和路由：** 将 Slack 回复保留在活跃线程中，保留账户绑定的投递路由，应用响应前缀，抑制内部追踪和不必要的回退回复，清除 Nostr 中继发布定时器，并为不透明账户 ID 保留微信会话路由。(#97168, #98240, #89949, #93639, #97989, #80928, #98720, #93686) 感谢 @LiuwqGit, @gorkem2020, @yetval, @wangwllu, @ZengWen-DT, @alexuser, @UnClouded77, @wangmiao0668000666, @zhangLei99586, @zhangguiping-xydt 和 @htkillermax-gif。 - **Cron 正确性：** 在超时时保留提供商和模型选择，保留启动追赶延迟，保留需要操作的操作输出，清除空白的思考覆盖，并保留提供商拥有的每日重置会话。(#95943, #94022, #93810, #96393, #96293, #98356) 感谢 @ZengWen-DT, @cursoragent, @luke-renjoy, @RichChen01, @vincentkoc, @yetval, @snowzlmbot, @nz365guy 和 @takamasa-aiso。 - **内存和会话恢复：** 检测未索引的转录，在笔记更新和 ChatGPT 导入期间通过临时 Wiki 读取保留笔记和手动编辑的前置元数据，而不削弱目录和硬链接冲突防护，避免跨目录恢复，消除保留的 Wiki 索引页歧义，并跳过空的 QMD 同步工作。(#97857, #98360, #98787, #97785, #94326, #90030) 感谢 @zw-xysk, @CHE10X, @qingminglong, @yetval, @vincentkoc, @sahibzada-allahyar 和 @ruben2000de。 - **Windows 和执行：** 将白名单执行绑定到验证的 Windows 路径，传播 `PATHEXT`，对入站路径进行不区分大小写规范化，并防止 Windows 上的清理崩溃。(#98260, #98093, #97630, #97901) 感谢 @eleqtrizit, @wendy-chsy, @VectorPeak 和 @paulcam206。 - **移动端和 UI 稳定性：** 保留 iOS 聊天换行和最终回复，改进 Android 配对和 TLS 恢复，隐藏过期的配对卡片，并保持工作区文件导轨可滚动。(#98304, #98117, #98366, #98439, #98483, #98049, #98646, #98611) 感谢 @joshavant, @Jabato01, @ooiuuii, @wuqxuan, @645648406-max 和 @zw-xysk。 - **Codex 和审批流程：** 正确报告 ChatGPT 认证，在侧边对话中保留插件应用审批，将破坏性审批模式重命名为 `ask`，准确分类动态目标和会话工具结果，并从显式运行截止时间推导终端空闲超时。(#91240, #98812, #98501, #98659, #96856, #85296) 感谢 @849261680, @ukstem, @kevinslin, @yetval, @nxmxbbd, @alkor2000 和 @vincentkoc。 - **配置和插件健康：** 暴露无法加载的频道插件，在补丁期间保留默认的提供商基础 URL，通过清单合约验证捆绑的插件更新，并在需要时保留旧版 ClawHub 家族。(#96397, #98396, #98010, #98249) 感谢 @849261680, @momothemage, @weltmaister, @LiLan0125, @herove 和 @Patrick-Erichsen。 - **Gateway、浏览器和设置诊断：** 区分可达的 Gateway 和失败的状态探测，在设置期间拒绝松散解析的 Gateway 端口，并避免从空路径创建杂散的 Chrome 配置文件配置键。(#98183, #98689, #98138) 感谢 @masatohoshino, @qingminglong 和 @zhangLei99586。 ### 完整的贡献记录 这份经过审计的记录涵盖了完整的 66e676d29b92d040716376a75aca32bad655cfac..HEAD 历史：222 个合并的 PR。生成清单也提供了直接提交作为编辑输入；上面的分组注释优先考虑用户影响。 #### 拉取请求 - **PR #96502** fix(moonshot): 限制视频描述 JSON 响应读取。感谢 @hugenshen 和 @cursoragent。 - **PR #98249** 为选定的插件保留旧版 ClawHub 家族。感谢 @Patrick-Erichsen。 - **PR #93767** fix(reasoning-tags): 剥离 MiniMax `mm:` 命名空间的推理标签。感谢 @DrHack1。 - **PR #93820** fix(imessage): 在反射防护中识别 MiniMax mm: 推理标签（完成 #93767）。感谢 @Alix-007。 - **PR #94096** fix(usage): 拒绝 usage.cost 和 sessions.usage 中反转的 startDate-endDate 范围。感谢 @Alix-007。 - **PR #97125** Doctor: 暴露认证配置文件发现。感谢 @giodl73-repo。 - **PR #98256** fix(mcp): 要求 Claude 权限回复的所有者。感谢 @eleqtrizit。 - **PR #98142** fix(cli): 防止 `pairing list` 在空频道枚举时崩溃。感谢 @RomneyDa。 - **PR #98260** fix(exec): 绑定 Windows 白名单执行路径。感谢 @eleqtrizit。 - **PR #97118** fix(telegram): 恢复停滞的入站请求。感谢 @vincentkoc。 - **PR #97168** fix(slack): 为继承的出站回复优先选择当前线程会话。相关 #96535。感谢 @LiuwqGit 和 @gorkem2020。 - **PR #97769** fix(plugins): 将输出文本转换应用于 toolcall_delta 和 toolcall_end 事件。相关 #97761。感谢 @ZOOWH 和 @get-viti。 - **PR #96544** fix(doctor): 合并冲突的模型引用映射键而不是丢弃。感谢 @yetval 和 @vincentkoc。 - **PR #97177** fix(memory-wiki): 优雅处理 vault 扫描中无法解析的 YAML 前置元数据 (#96125)。感谢 @SunnyShu0925 和 @cow11023。 - **PR #97167** 修复 #96840：[Bug]: 在 WebChat 中，无目标的消息发送失败，提示 'Action send requires a target'，尽管文档说明源回复接收器应处理。感谢 @zhangguiping-xydt 和 @MantisCartography。 - **PR #98302** fix(ios): 在 QR 扫描后推进入门引导步骤。相关 #98297。感谢 @joelnishanth 和 @cursoragent。 - **PR #96644** fix(anthropic-oauth): 限制 OAuth 令牌端点响应读取。感谢 @solodmd。 - **PR #96397** fix: 当配置的频道插件无法加载时发出警告。感谢 @849261680。 - **PR #96359** test: 将 src/commands 测试迁移到共享临时目录助手。感谢 @xialonglee。 - **PR #96293** fix(cron): 通过清空字段清除 agentTurn 思考覆盖。相关 #96287。感谢 @ZengWen-DT 和 @takamasa-aiso。 - **PR #96058** test: 在自动回复和安装回退测试中优先使用共享临时目录助手。感谢 @xialonglee。 - **PR #87298** test: 添加临时目录助手指南。感谢 @hxy91819。 - **PR #97785** fix(sessions): 避免跨工作目录的近期恢复。相关 #96542。感谢 @qingminglong 和 @yetval。 - **PR #97698** fix(pdf): 在本机分析前拒绝空的解析页面范围。感谢 @zhangguiping-xydt。 - **PR #97693** fix(discord): 限制 requestDiscord 正常路径响应读取以防止 OOM。感谢 @Alix-007。 - **PR #97683** fix(irc): 在 \u 字面量转义解码器中保护替代范围码点。感谢 @llagy009。 - **PR #96938** fix(utils): 保持回复指令 ID 为 Unicode 安全。感谢 @ly-wang19。 - **PR #97857** fix(memory): 在状态模式下检测未索引的会话转录（修复 #97814）。感谢 @zw-xysk 和 @CHE10X。 - **PR #98094** fix(android): 澄清网关认证恢复状态。感谢 @qingminglong。 - **PR #98205** test(gateway): 为节点唤醒状态跟踪和测试接缝添加单元测试。感谢 @zenglingbiao。 - **PR #98115** fix: 从设备 CLI 表面节点审批指南。感谢 @welfo-beo。 - **PR #97898** docs: 澄清源检出 Node 最低版本。相关 #97792。感谢 @lin-hongkuan 和 @aniruddhaadak80。 - **PR #94526** test(telegram): 为带有流式推理的论坛主题 message_thread_id 添加回归测试。相关 #89352。感谢 @xialonglee 和 @pmika。 - **PR #98145** fix(device-pairing): 不在子集重新请求时搅动 requestId。感谢 @RomneyDa。 - **PR #98267** fix(system-prompt): 将执行审批 + 授权发送者移动到缓存边界下方。相关 #98261。感谢 @headbouyJB。 - **PR #98304** fix: 保留 iOS 聊天换行。相关 #98028。感谢 @joshavant 和 @Jabato01。 - **PR #98187** fix(openrouter): 发送显式认证头。相关 #97934。感谢 @sunlit-deng 和 @laurencebrown。 - **PR #95708** fix: 在工具活动期间显示 WebChat 前言进度。感谢 @ragesaq。 - **PR #98210** fix(gateway): iOS Talk 将 SecretRef 支持的 API 密钥视为缺失。相关 #98209。感谢 @ooiuuii。 - **PR #98009** test(infra): 为 SQLite 数字规范化添加单元测试。感谢 @dwc1997。 - **PR #98087** test(config): 为 resolveExecCommandHighlighting 添加单元测试。感谢 @solodmd。 - **PR #98219** test(utils): 为 chunkItems 添加单元测试。感谢 @zenglingbiao。 - **PR #98093** fix(core): 在 Windows 上通过 isExecutableFile 传播调用者环境 PATHEXT。感谢 @wendy-chsy。 - **PR #97973** fix(matrix): 防护 JSON.parse 免受畸形 homeserver 响应体影响。感谢 @lsr911。 - **PR #97999** fix(sms): 防护 Twilio JSON.parse 免受畸形 API 响应体影响。感谢 @lsr911。 - **PR #98043** 添加瑞典移动应用本地化。感谢 @yeager。 - **PR #98144** fix(tui): 更正设备范围升级的断开连接文本。感谢 @RomneyDa。 - **PR #98240** fix(agents): 保持合并的投递路由为账户绑定。感谢 @yetval。 - **PR #89949** fix(media): 在任务启动时固定请求者投递路由。感谢 @wangwllu。 - **PR #98226** 编辑裸 Fireworks API 密钥。相关 #98225。感谢 @ooiuuii。 - **PR #98319** docs: 发布 v2026.6.11 版本说明。感谢 @hannesrudolph。 - **PR #98257** fix: 显示频道运行中的状态。感谢 @scotthuang。 - **PR #97931** fix(gateway): 在每日默认重置中保持提供商拥有的 CLI 会话。感谢 @yetval。 - **PR #98325** docs: 刷新 v2026.6.11 文档地图。感谢 @hannesrudolph。 - **PR #97929** fix(auto-reply): 阻止级别指令吃掉下一个消息单词。感谢 @yetval。 - **PR #97928** fix(agents): 在上下文防护字符估算器中估算测试工具角色大小（修复 #97927）。感谢 @liuhao1024 和 @yetval。 - **PR #97861** fix(compaction): 在提示前溢出预检查中计数 bashExecution 和 summary 轮次。感谢 @yetval。 - **PR #97137** doctor: 添加内存搜索 lint 发现。感谢 @giodl73-repo。 - **PR #97358** Doctor: 暴露工作区状态发现。感谢 @giodl73-repo。 - **PR #95622** test(qa-lab): 加强 whatsapp qa 场景。感谢 @mcaxtr。 - **PR #98346** fix: 防止技能创建者绕过工作坊提案。相关 #96054。感谢 @momothemage 和 @xianshishan。 - **PR #98169** fix(heartbeat): 限制承诺扇出提示。感谢 @bdjben。 - **PR #97366** Doctor: 暴露设备配对发现。感谢 @giodl73-repo。 - **PR #98366** fix: Android TLS 指纹验证在慢速握手中超时。相关 #98365。感谢 @joshavant。 - **PR #98353** fix(ios): 默认在 Chat 中打开应用。感谢 @BsnizND。 - **PR #98352** fix(security): 警告代理技能 MCP 边界漂移。感谢 @momothemage。 - **PR #98347** fix: 重试图像描述回退模型。感谢 @momothemage。 - **PR #98117** fix(ios): 避免临时重复的最终回复。相关 #98116。感谢 @ooiuuii 和 @joshavant。 - **PR #98293** fix(gateway): 发出过期的执行审批后续诊断。感谢 @BsnizND。 - **PR #98376** fix(ios): 在 Talk 中使用 Gateway 语音提供商。相关 #98153。感谢 @Tony-ooo。 - **PR #66685** 抑制过期的执行审批后续警告。感谢 @pfrederiksen。 - **PR #98385** fix: 显示可操作的移动设备协议不匹配恢复。相关 #98384。感谢 @joshavant。 - **PR #98146** fix(cli): 解释如何从设备审批死锁中恢复。感谢 @RomneyDa。 - **PR #98423** improve(ios): 澄清 Control 和 Talk 视觉层次。相关 #98397。 - **PR #98217** fix(doctor): 跨设备恢复旧版 cron 归档。感谢 @masatohoshino。 - **PR #98333** feat(openai): 添加 GPT-5.6 系列支持。相关 #98296。感谢 @steipete-oai。 - **PR #96393** fix(cron): 保留需要操作的命令输出。相关 #96346。感谢 @snowzlmbot 和 @nz365guy。 - **PR #98429** fix(ios): 分类 TLS 指纹超时。感谢 @joshavant。 - **PR #98439** fix: Android 设置代码接受本地 mDNS 网关主机。感谢 @joshavant。 - **PR #98443** fix(ios): 改进浅色和深色外观对比度。相关 #98440。 - **PR #97742** fix(llm): 跨提供商保留结构化工具结果文本。感谢 @snowzlmbot。 - **PR #97968** fix(status): 表面未注册的内存嵌入提供商。感谢 @masatohoshino。 - **PR #92237** fix(agents): 保留运行时设置覆盖 [AI 辅助]。感谢 @sercada。 - **PR #95888** fix(active-memory): 警告可变操作事实；将截断的回忆标记为不完整。感谢 @spencer2211。 - **PR #98291** fix(gateway): 表面 systemd 启动限制耗尽。感谢 @masatohoshino。 - **PR #90517** fix(gateway): 提示 web 登录缺少外部插件。相关 #83277。感谢 @TUARAN 和 @carol-iung。 - **PR #98369** test(infra): 为 SQLite user_version pragma 助手添加单元测试。感谢 @dwc1997。 - **PR #98340** fix: 扩展 api.exec 在超时后留下子进程。相关 #98335。感谢 @ooiuuii。 - **PR #92063** fix(ui): 在分段流式传输期间折叠重复的助手组。相关 #63956。感谢 @harjothkhara 和 @contentfree。 - **PR #98354** fix(infra): 防护投递队列膨胀免受损坏的 entry_json 影响。感谢 @Pick-cat。 - **PR #90566** fix(agents): 警告 cron 公告跳过。相关 #68561。感谢 @sahibzada-allahyar 和 @Mibslee。 - **PR #98371** fix(ports): 在赋值前验证 lsof PID 解析。感谢 @lzyyzznl。 - **PR #98356** fix(cron): 在每日默认重置中保持提供商拥有的 CLI 会话。感谢 @yetval。 - **PR #98395** test(shared): 为账户启用防护添加单元测试。感谢 @dwc1997。 - **PR #98411** fix(agents): 从提供商主体中恢复思考错误。相关 #98308。感谢 @sunlit-deng 和 @clearhorizoninvestments。 - **PR #98494** docs(skills): 支持变量可着陆批次扫描。感谢 @vincentkoc。 - **PR #91240** fix: 报告 Codex ChatGPT 状态认证。相关 #91099。感谢 @849261680 和 @ukstem。 - **PR #98370** test(agents): 为思考块检测添加单元测试。感谢 @dwc1997。 - **PR #96711** test: 在配置、网关、cron、crestodian 和状态测试中优先使用共享临时目录助手。感谢 @xialonglee。 - **PR #98483** fix: Android QR 扫描启动网关配对。感谢 @joshavant。 - **PR #95230** 修复 docs-list-mdx-pages。感谢 @hugenshen。 - **PR #96322** fix(minimax): 限制 JSON 响应读取以防止 OOM。感谢 @lsr911。 - **PR #95348** 修复 config-chmod-warning。感谢 @hugenshen 和 @cursoragent。 - **PR #95229** fix(copilot): 在 cli-metadata 注册期间防护未定义的 runtime.state。相关 #94516。感谢 @sunlit-deng 和 @cuihaijun。 - **PR #94636** fix(memory): 在提升期间跳过原始片段。感谢 @tayoun。 - **PR #94013** [AI] fix(feishu): 在监视器启动中防护部分 channelRuntime。感谢 @xydt-tanshanshan。 - **PR #93466** [AI] fix(feishu): 在 channelRuntime 回退中防护缺失的入站。感谢 @xydt-tanshanshan。 - **PR #98049** fix: 在 Control UI 中隐藏过期的配对 QR 卡片。相关 #98039。感谢 @ooiuuii。 - **PR #96094** fix(memory): 在 CLI 重新索引后证明实时管理器恢复。相关 #91167。感谢 @849261680 和 @kiagentkronos-cell。 - **PR #98482** fix: 广告路由感知的 LAN Control UI 链接。感谢 @joshavant。 - **PR #71537** 在内存钩子和 session-logs 技能中恢复已归档（.reset）的会话转录。感谢 @injinj。 - **PR #96375** docs(config-agents): 更正 opus 和 gpt 的内置别名表。感谢 @niks999。 - **PR #98453** docs(gateway): 修正 config-channels.md 中 Telegram 流式传输默认值。感谢 @solodmd。 - **PR #98533** fix: 修复托管 CI 基线断言。 - **PR #98421** feat(imessage): 原生投票支持——创建、阅读、投票。感谢 @omarshahine 和 @lobster。 - **PR #98318** docs(matrix): 记录缺失的 streaming.progress 模式、进度子字段和 mentionPatterns 配置。感谢 @wm0018 和 @vincentkoc。 - **PR #97753** docs(onboard): 记录 11 个缺失的非交互式 CLI 标志。感谢 @wm0018 和 @vincentkoc。 - **PR #97851** fix(mattermost): 限制空主体错误响应读取。感谢 @Pick-cat。 - **PR #98360** fix(memory-wiki): 在临时页面读取后保留笔记。相关 #98345。感谢 @qingminglong 和 @yetval。 - **PR #98551** test: 修复陈旧的核心测试类型失败。感谢 @RomneyDa。 - **PR #98455** fix(browser): 在 fetchHttpJson 中限制错误主体读取以防止 OOM。感谢 @wings1029。 - **PR #95906** fix(code-mode): 将 QuickJS 错误名称和消息呈现给模型。感谢 @ZengWen-DT 和 @vincentkoc。 - **PR #97901** fix(agents): 停止 copilot 自动审查在 Windows 上的清理崩溃。感谢 @paulcam206。 - **PR #97923** fix(slack): 在替代边界上截断服务的 arg-menu 选项标签。感谢 @LEXES7。 - **PR #98010** fix(update): 通过清单合约验证捆绑的插件有效载荷。相关 #97985。感谢 @LiLan0125 和 @herove。 - **PR #85296** fix(codex): 从显式运行超时推导终端空闲看门狗。感谢 @alkor2000 和 @vincentkoc。 - **PR #97110** feat(i18n): 添加原生应用语言环境清单。感谢 @vincentkoc。 - **PR #98396** fix: 允许 config.patch 带有默认的提供商 baseUrl。相关 #98270。感谢 @momothemage 和 @weltmaister。 - **PR #98503** fix(usage-bar): 使用 Object.hasOwn 而不是 in 运算符以避免原型链污染。相关 #98466。感谢 @chenyangjun-xy 和 @zhangLei99586。 - **PR #97111** feat(android): 本地化核心网关界面。感谢 @vincentkoc。 - **PR #97630** fix(media): 对 Windows 入站路径进行不区分大小写规范化。感谢 @VectorPeak。 - **PR #82638** fix(agents): 当 models.mode 为 'replace' 时跳过隐式提供商发现 [AI 辅助]。相关 #66957。感谢 @eldar702 和 @wangzhengshu。 - **PR #87917** 修复 sessions json 谱系元数据。相关 #80286。感谢 @zhangguiping-xydt 和 @islandpreneur007。 - **PR #93639** fix(message-tool): 将 messages.responsePrefix 应用于出站发送。感谢 @ZengWen-DT。 - **PR #94440** fix: #94432 将 Cloudflare 挑战 403 分类为 upstream_html 而不是 auth_html。感谢 @lzyyzznl 和 @pbm9z95m6z-hue。 - **PR #98119** fix: 减少 Docker 构建内存压力。相关 #98118。感谢 @zyzo。 - **PR #97679** feat(node): 为 node run/install 添加 --context-path 标志以支持反向 p…。相关 #97678。感谢 @wm0018。 - **PR #98339** fix(irc): 将无主机 nick!user 白名单条目分类为可变。感谢 @yetval。 - **PR #97662** fix(matrix): 限制原始传输响应读取以防止 OOM。感谢 @Alix-007。 - **PR #98137** fix: 提升定时器声明以避免在可中止延迟中出现 TDZ ReferenceError。感谢 @zhangLei99586。 - **PR #98134** fix: 在 Tailscale 二进制探测 Promise.race 中清除超时定时器。感谢 @zhangLei99586。 - **PR #97989** fix(sms): 阻止内部工具追踪横幅到达 SMS 回复。感谢 @ZengWen-DT。 - **PR #97972** fix(browser): CDP 认证在百分比编码的凭据下失败。感谢 @VectorPeak。 - **PR #98063** fix(reply): 当 messages.suppressToolErrors 设置时抑制工具错误进度投递。感谢 @moeedahmed 和 @amittell。 - **PR #94964** fix(reload): 在进程内重启时取消延迟的频道重载。相关 #79487。感谢 @lzyyzznl 和 @tseller。 - **PR #98598** fix: 在定时器修复后恢复主 lint。相关 #98462, #98464。感谢 @zhangLei99586。 - **PR #98587** fix(slack): 防护中继 WebSocket 帧 JSON.parse 免受畸形输入影响。感谢 @lsr911 和 @vincentkoc。 - **PR #90030** fix(memory-core): 跳过 qmd 零命中搜索同步。相关 #90023。感谢 @sahibzada-allahyar 和 @ruben2000de。 - **PR #98493** fix(transcripts): 在错误退出时关闭 readline 接口并销毁读取流。相关 #98467。感谢 @wangmiao0668000666 和 @zhangLei99586。 - **PR #98497** fix(cli): 当插件 npm install 返回空输出时显示退出代码。感谢 @Sanjays2402 和 @vincentkoc。 - **PR #97112** feat(apple): 本地化核心原生应用界面。感谢 @vincentkoc。 - **PR #98610** fix: 在转录测试添加后恢复工具 CI。 - **PR #77539** fix(subagent): 在重启重新分发时保留引导的任务文本。感谢 @amittell。 - **PR #97113** feat(i18n): 刷新所有原生语言环境工件。感谢 @vincentkoc。 - **PR #98620** feat(doctor): 警告飞行中的 cron 作业。感谢 @masatohoshino。 - **PR #98605** test(shared): 为人类可读列表格式化添加单元测试。感谢 @dwc1997。 - **PR #97348** feat(autoreview): 支持 cursor-agent 引擎。感谢 @hxy91819。 - **PR #95943** fix(cron): 在隔离运行超时行中保留提供商/模型。相关 #95873。感谢 @ZengWen-DT, @cursoragent 和 @luke-renjoy。 - **PR #94149** fix(status): 限制 systemd 服务探测，使状态不会在卡住的 systemctl 上挂起 (#84698)。感谢 @ZengWen-DT, @cursoragent 和 @zus-assistant。 - **PR #88159** fix(cli): 在日志跟踪中 journal 回退后重试 logs.tail。感谢 @anyech 和 @vincentkoc。 - **PR #98508** fix(update-check): 限制 npm 注册表 JSON 响应读取以防止 OOM。感谢 @lzyyzznl。 - **PR #98496** fix(tlon): 限制错误响应体读取以防止 OOM。感谢 @Pandah97。 - **PR #98554** fix(openai): 限制嵌入批处理文件下载。感谢 @sunlit-deng 和 @vincentkoc。 - **PR #98652** fix: 停止无效消息超时卡住。 - **PR #77973** fix(gateway): 限制 agentRunCache 以防止在运行扇出时无限增长。相关 #77976。感谢 @fede-kamel 和 @vincentkoc。 - **PR #98525** fix(agents): 在没有第一个事件时超时本地流。感谢 @osolmaz。 - **PR #94022** fix(cron): 在服务状态中持久化启动追赶延迟 ID 以防止读取-RPC 覆盖。相关 #93935。感谢 @RichChen01, @vincentkoc 和 @yetval。 - **PR #93810** fix(cron): 在 start() 维护过程中保留启动溢出追赶延迟。感谢 @yetval 和 @vincentkoc。 - **PR #98623** fix: 媒体工具在自动选择模型时跳过 env-key 提供商插件。感谢 @medns。 - **PR #98665** fix(claude-cli): 在 can_use_tool 允许响应中为 Claude Code 2.1 返回 updatedInput。相关 #95171。感谢 @yetval 和 @carterdawson。 - **PR #94250** fix(feishu): 当 blockStreaming 启用时，将块作为独立消息发送。相关 #55027。感谢 @xialonglee, @vincentkoc 和 @ZichaoLong。 - **PR #93379** fix(whatsapp): 通过命令授权和所有者绕过 LID JID 解析线程化 authDir。相关 #77755。感谢 @xialonglee 和 @jiveshkalra。 - **PR #98646** fix: 保持工作区导轨文件部分可滚动。相关 #98566。感谢 @wuqxuan 和 @645648406-max。 - **PR #98602** fix: iOS Talk 回退设置打开 Voice & Talk。相关 #98593。感谢 @PollyBot13。 - **PR #98611** fix(ui): 为工作区导轨部分添加 overflow-y:auto 以防止文件列表溢出（修复 #98566）。感谢 @zw-xysk 和 @645648406-max。 - **PR #98619** fix(qa-lab): 凭据租约请求在过大的 Convex 代理响应上失败。感谢 @ZengWen-DT。 - **PR #94326** fix(memory-wiki): 消除保留索引页分叉用于合成和摄取。感谢 @yetval 和 @vincentkoc。 - **PR #98659** fix(codex): 将 get_goal 读取状态分类为成功的动态工具调用。感谢 @yetval。 - **PR #96856** fix(codex): 成功的 sessions_spawn 和 goal 工具结果被记录为失败。感谢 @nxmxbbd。 - **PR #98660** fix(inworld): 防护 voices JSON.parse 免受畸形 API 响应体影响。感谢 @solodmd。 - **PR #95430** fix(embedded-agent-runner): 通过 pumpStreamWithRecovery 泵送异步 streamFn 以进行流中错误恢复。相关 #95429。感谢 @lzyyzznl, @vincentkoc 和 @alexelgier。 - **PR #98644** fix: 工具摘要保留表情符号截断边界。感谢 @ZengWen-DT。 - **PR #80928** fix(telegram): 当插件命令返回 suppressReply: true 时抑制回退回复。相关 #80756。感谢 @alexuser 和 @UnClouded77。 - **PR #98701** fix: 防止 agents-tools 消息测试超时。 - **PR #92657** feat(usage): 提供内置的 /usage 完整页脚。感谢 @Marvinthebored。 - **PR #92877** fix(usage): 使内置页脚在 Telegram 上更易换行。感谢 @Marvinthebored。 - **PR #98126** 恢复 Telegram /steer 用于活跃的 Codex 运行。相关 #81594。感谢 @100yenadmin 和 @Kyzcreig。 - **PR #92037** feat(cron): on-exit 调度——在被监控命令退出时唤醒。感谢 @anagnorisis2peripeteia。 - **PR #98452** feat(ios): 使用 iOS 26 Liquid Glass 现代化应用。 - **PR #98006** 添加 Telegram /login Codex 配对流程。感谢 @100yenadmin。 - **PR #98735** fix(telegram): 保留富文本转发的消息。感谢 @obviyus。 - **PR #97962** refactor(qa): 在流程场景中使用传输原生动作。感谢 @RomneyDa。 - **PR #98726** fix(nvidia): 使用 Nemotron Super 1M 上下文。感谢 @eleqtrizit。 - **PR #98691** fix(imessage): 在投票回显匹配中移除任何位置的 emoji。感谢 @omarshahine。 - **PR #97174** 修复 Telegram 插件回调路由。感谢 @goldmar。 - **PR #89597** fix: 将 QQBot 凭据备份迁移到 SQLite KV。 - **PR #98536** feat: 准备作用域对话能力配置文件。 - **PR #92274** fix(agents): 将嵌入式提示锁定错误分类为永久性公告失败。相关 #91527。感谢 @fsdwen 和 @zackchiutw。 - **PR #98102** fix(telegram): 持久重试重启期间丢失的入站媒体 (#98076)。感谢 @luoyanglang 和 @DaveArcher18。 - **PR #98755** fix(cron): 分离会话定向运行。相关 #98121。感谢 @obviyus 和 @EthanSK。 - **PR #96065** fix(install): 通过 OPENCLAW_SERVICE_MANAGED_ENV_KEYS 管理 config-secretref 环境引用。感谢 @Darren2030 和 @obviyus。 - **PR #98666** fix: 诊断 Windows LAN Gateway 防火墙阻止。感谢 @joshavant。 - **PR #98501** fix(codex): 将破坏性审批模式重命名为 ask。相关 #98499。感谢 @kevinslin。 - **PR #98775** fix(telegram): 在临时 getUpdates 错误中存活并停止每次发送的缓存重写。相关 #98772, #98773。感谢 @obviyus。 - **PR #98776** fix(telegram): 退避、死信和墓碑化排队的更新，以便毒化消息不会阻塞或重复。相关 #98774。感谢 @obviyus。 - **PR #96454** feat(cli): openclaw attach——启动绑定到网关会话的外部测试工具。感谢 @anagnorisis2peripeteia 和 @obviyus。 - **PR #98786** fix(telegram): 最终回复不再因被拒绝的富实体、标题、引用或长洪水等待而丢失。相关 #98778。感谢 @obviyus。 - **PR #97496** Doctor: 暴露频道插件阻塞器发现。感谢 @giodl73-repo。 - **PR #98792** fix(ci): 恢复文档和测试类型检查。 - **PR #98736** improve(ios): 简化 Talk 控制和作曲家对齐。 - **PR #97889** fix(discord): 防护 JSON.parse 免受畸形 API 响应体影响。感谢 @lsr911。 - **PR #98812** fix(codex): 在侧边对话中保留插件应用审批。 - **PR #92283** fix(agents): 不将 A2A 轮次注入隔离的 cron sessions_send (#92257)。感谢 @harjothkhara 和 @nailujac。 - **PR #98138** fix: 在 Chrome 配置文件装饰中防护 setDeep 免受空键数组影响。感谢 @zhangLei99586。 - **PR #98183** fix(gateway): 区分可达网关和失败的状态探测。感谢 @masatohoshino。 - **PR #98689** fix(wizard): 拒绝松散网关端口输入。相关 #98681。感谢 @qingminglong。 - **PR #98720** fix(nostr): 清除每个中继的发布超时定时器以防止悬垂句柄。相关 #98463。感谢 @wangmiao0668000666 和 @zhangLei99586。 - **PR #98818** fix(ci): 恢复不完整的 Swift 构建缓存。 - **PR #98787** fix(memory-wiki): 在 wiki_apply 和 chatgpt 导入中重试临时现有页面读取。感谢 @yetval。 - **PR #98811** feat(ios): 现代化导航和设置。相关 #98803。 - **PR #98843** docs: 更新移动应用发布消息。感谢 @joshavant。 - **PR #93686** fix(weixin): startAccount 保留会话路由。相关 #93556。感谢 @zhangguiping-xydt 和 @htkillermax-gif。 ### 发布验证 - npm 包：https://www.npmjs.com/package/openclaw/v/2026.7.1-beta.1 - 注册表 tarball：https://registry.npmjs.org/openclaw/-/openclaw-2026.7.1-beta.1.tgz - 完整性：`sha512-Yu/ELLje9mxvFTlaxVGHnIkKvHcLnoEj3AQhzmhpP4k8Fi7/ln0NvYnBgaZ2BJ8tdAAsq3iZ8uroRUuXiMB0dg==` - 发布 SHA：`4eb1d333cfeca440b796b6a3f70d3c2bef996243` - 完整发布 CI 报告：https://github.com/openclaw/releases/blob/main/evidence/2026.7.1-beta.1/release-evidence.md - 发布发布：https://github.com/openclaw/openclaw/actions/runs/28571485937 - npm 预检：https://github.com/openclaw/openclaw/actions/runs/28569693832 - 完整发布验证：https://github.com/openclaw/openclaw/actions/runs/28569693812 - 插件 npm 发布：https://github.com/openclaw/openclaw/actions/runs/28571834366 - 插件 ClawHub 发布：成功：https://github.com/openclaw/openclaw/actions/runs/28571835541 - 插件 ClawHub 引导：不需要 - OpenClaw npm 发布：https://github.com/openclaw/openclaw/actions/runs/28572195585 - 包 Telegram E2E：在发布检查中通过：https://github.com/openclaw/openclaw/actions/runs/28569815816/job/84705725486 - 咨询 WhatsApp 实时 QA：在两次尝试后因 Convex 凭据池耗尽而在场景前被阻止：https://github.com/openclaw/openclaw/actions/runs/28569693831

### Highlights - **OpenAI GPT-5.6 support:** OpenClaw now recognizes the GPT-5.6 model family across catalog, capability, and runtime selection paths. (#98333) Thanks @steipete-oai. - **External harness attachment:** `openclaw attach` launches an external harness against an existing Gateway session, making interactive Codex-style workflows easier to resume and inspect. (#96454) Thanks @anagnorisis2peripeteia and @obviyus. - **Telegram Codex workflows:** Telegram can now start Codex pairing with `/login`, steer active Codex runs, and recover final replies across transient API failures. (#98006, #98126, #98786) Thanks @100yenadmin, @Kyzcreig, and @obviyus. - **Event-driven cron runs:** the new `on-exit` schedule kind wakes an agent when a watched command exits, while session-targeted runs can detach cleanly. (#92037, #98755) Thanks @anagnorisis2peripeteia, @obviyus, and @EthanSK. - **Native app refresh:** iOS adopts the iOS 26 visual system with clearer navigation, settings, Chat, Talk, and onboarding flows, while native app localization expands across Apple and Android surfaces. (#98452, #98736, #98811, #97110, #97111, #97112, #97113) Thanks @vincentkoc. - **Richer messaging:** iMessage gains native poll creation, reading, and voting, and built-in usage footers provide clearer per-turn accounting in chat. (#98421, #92657, #92877) Thanks @omarshahine, @lobster, and @Marvinthebored. - **Safer scoped conversations:** capability profiles prepare per-conversation tool and access boundaries without weakening the existing default profile. (#98536) ### Changes - **Model and provider coverage:** add GPT-5.6 support, use Nemotron Super's 1M context window, and preserve explicit OpenRouter authentication headers. (#98333, #98726, #98187) Thanks @steipete-oai, @eleqtrizit, @sunlit-deng, and @laurencebrown. - **CLI and node workflows:** add `openclaw attach`, node context-path support, actionable device-approval recovery guidance, and clearer plugin install exit diagnostics. (#96454, #97679, #98115, #98146, #98497) Thanks @anagnorisis2peripeteia, @obviyus, @wm0018, @welfo-beo, @RomneyDa, @Sanjays2402, and @vincentkoc. - **Cron and usage:** add exit-triggered schedules, detached session-targeted runs, an in-flight job doctor warning, and a built-in full usage footer. (#92037, #98755, #98620, #92657, #92877) Thanks @anagnorisis2peripeteia, @obviyus, @EthanSK, @masatohoshino, and @Marvinthebored. - **Native apps and localization:** modernize iOS navigation, settings, presentation, and Talk controls, add Gateway speech providers, improve QR onboarding and protocol recovery, localize core Apple and Android surfaces, and add Swedish mobile localization. (#98452, #98736, #98811, #98376, #98302, #98385, #97110, #97111, #97112, #97113, #98043) Thanks @Tony-ooo, @joelnishanth, @cursoragent, @joshavant, @vincentkoc, and @yeager. - **Messaging capabilities:** add native iMessage polls and Telegram Codex pairing and steering flows. (#98421, #98006, #98126) Thanks @omarshahine, @lobster, @100yenadmin, and @Kyzcreig. - **Doctor and diagnostics:** expose auth-profile, workspace, device-pairing, channel-plugin, memory-provider, systemd exhaustion, and Windows LAN firewall findings. (#97125, #97358, #97366, #97496, #97968, #98291, #98666) Thanks @giodl73-repo, @masatohoshino, and @joshavant. - **Conversation and review controls:** prepare scoped conversation capability profiles and add Cursor Agent as an autoreview engine. (#98536, #97348) Thanks @hxy91819. ### Fixes - **Telegram durability:** recover stalled ingress claims, retry restart-dropped media, survive transient polling errors, dead-letter poison updates, preserve forwarded rich text, route plugin callbacks correctly, and fall back safely when rich final replies are rejected. (#97118, #98102, #98735, #98775, #98776, #97174, #98786) Thanks @vincentkoc, @luoyanglang, @DaveArcher18, @obviyus, and @goldmar. - **Agent and context reliability:** preserve runtime overrides and steered subagent tasks, keep isolated cron `sessions_send` replies from feeding back into the requester, improve harness-aware context estimation and compaction prechecks, time out silent local streams, recover mid-stream failures, and cap Gateway run-cache growth. (#92237, #92283, #77539, #97928, #97861, #98525, #95430, #77973) Thanks @sercada, @amittell, @liuhao1024, @yetval, @harjothkhara, @nailujac, @osolmaz, @lzyyzznl, @vincentkoc, @alexelgier, and @fede-kamel. - **Provider and network safety:** bound oversized or malformed responses across Moonshot, MiniMax, Anthropic OAuth, Discord, Matrix, SMS, browser, update, embeddings, Tlön, and Inworld paths. (#96502, #96322, #96644, #97693, #97889, #97662, #97999, #98455, #98508, #98554, #98496, #98660) Thanks @hugenshen, @cursoragent, @lsr911, @solodmd, @Alix-007, @wings1029, @lzyyzznl, @sunlit-deng, @vincentkoc, and @Pandah97. - **Channel delivery and routing:** keep Slack replies in the active thread, preserve account-bound delivery routes, apply response prefixes, suppress internal traces and unwanted fallback replies, clear Nostr relay publish timers, and retain WeChat session routing for opaque account ids. (#97168, #98240, #89949, #93639, #97989, #80928, #98720, #93686) Thanks @LiuwqGit, @gorkem2020, @yetval, @wangwllu, @ZengWen-DT, @alexuser, @UnClouded77, @wangmiao0668000666, @zhangLei99586, @zhangguiping-xydt, and @htkillermax-gif. - **Cron correctness:** preserve provider and model selections on timeouts, retain startup catch-up deferrals, keep action-required output, clear blank thinking overrides, and preserve provider-owned daily-reset sessions. (#95943, #94022, #93810, #96393, #96293, #98356) Thanks @ZengWen-DT, @cursoragent, @luke-renjoy, @RichChen01, @vincentkoc, @yetval, @snowzlmbot, @nz365guy, and @takamasa-aiso. - **Memory and session recovery:** detect unindexed transcripts, preserve notes and hand-edited frontmatter through transient wiki reads during note updates and ChatGPT imports without weakening directory and hardlink collision guards, avoid cross-directory resumes, disambiguate reserved wiki index pages, and skip empty QMD sync work. (#97857, #98360, #98787, #97785, #94326, #90030) Thanks @zw-xysk, @CHE10X, @qingminglong, @yetval, @vincentkoc, @sahibzada-allahyar, and @ruben2000de. - **Windows and execution:** bind allowlisted execution to the validated Windows path, propagate `PATHEXT`, normalize inbound paths case-insensitively, and prevent cleanup crashes on Windows. (#98260, #98093, #97630, #97901) Thanks @eleqtrizit, @wendy-chsy, @VectorPeak, and @paulcam206. - **Mobile and UI stability:** preserve iOS chat line breaks and final replies, improve Android pairing and TLS recovery, hide expired pairing cards, and keep workspace file rails scrollable. (#98304, #98117, #98366, #98439, #98483, #98049, #98646, #98611) Thanks @joshavant, @Jabato01, @ooiuuii, @wuqxuan, @645648406-max, and @zw-xysk. - **Codex and approval flows:** report ChatGPT authentication correctly, preserve plugin app approvals in side conversations, rename destructive approval mode to `ask`, classify dynamic goal and session tool results accurately, and derive terminal-idle timeouts from the explicit run deadline. (#91240, #98812, #98501, #98659, #96856, #85296) Thanks @849261680, @ukstem, @kevinslin, @yetval, @nxmxbbd, @alkor2000, and @vincentkoc. - **Configuration and plugin health:** surface unloadable channel plugins, preserve defaulted provider base URLs during patches, validate bundled plugin updates by manifest contract, and retain legacy ClawHub families where required. (#96397, #98396, #98010, #98249) Thanks @849261680, @momothemage, @weltmaister, @LiLan0125, @herove, and @Patrick-Erichsen. - **Gateway, browser, and setup diagnostics:** distinguish a reachable Gateway from a failed status probe, reject loosely parsed Gateway ports during setup, and avoid creating stray Chrome profile configuration keys from empty paths. (#98183, #98689, #98138) Thanks @masatohoshino, @qingminglong, and @zhangLei99586. ### Complete contribution record This audited record covers the complete 66e676d29b92d040716376a75aca32bad655cfac..HEAD history: 222 merged PRs. The generation manifest also supplies direct commits as editorial input; the grouped notes above prioritize user impact. #### Pull requests - **PR #96502** fix(moonshot): bound video description JSON response reads. Thanks @hugenshen and @cursoragent. - **PR #98249** Preserve legacy ClawHub family for selected plugins. Thanks @Patrick-Erichsen. - **PR #93767** fix(reasoning-tags): strip MiniMax `mm:` namespaced reasoning tags. Thanks @DrHack1. - **PR #93820** fix(imessage): recognize MiniMax mm: reasoning tags in reflection guard (completes #93767). Thanks @Alix-007. - **PR #94096** fix(usage): reject inverted startDate-endDate range in usage.cost and sessions.usage. Thanks @Alix-007. - **PR #97125** Doctor: expose auth profile findings. Thanks @giodl73-repo. - **PR #98256** fix(mcp): require owner for Claude permission replies. Thanks @eleqtrizit. - **PR #98142** fix(cli): stop `pairing list` crashing with empty channel enum. Thanks @RomneyDa. - **PR #98260** fix(exec): bind Windows allowlist execution path. Thanks @eleqtrizit. - **PR #97118** fix(telegram): recover stalled ingress spool claims. Thanks @vincentkoc. - **PR #97168** fix(slack): prefer current thread session for inherited outbound replies. Related #96535. Thanks @LiuwqGit and @gorkem2020. - **PR #97769** fix(plugins): apply output text transforms to toolcall_delta and toolcall_end events. Related #97761. Thanks @ZOOWH and @get-viti. - **PR #96544** fix(doctor): merge colliding model-ref map keys instead of dropping. Thanks @yetval and @vincentkoc. - **PR #97177** fix(memory-wiki): gracefully handle unparsable YAML frontmatter in vault scans (#96125). Thanks @SunnyShu0925 and @cow11023. - **PR #97167** fix #96840: [Bug]: Targetless message.send fails with 'Action send requires a target' in WebChat despite docs stating source-reply sink should handle it. Thanks @zhangguiping-xydt and @MantisCartography. - **PR #98302** fix(ios): advance onboarding step after QR scan. Related #98297. Thanks @joelnishanth and @cursoragent. - **PR #96644** fix(anthropic-oauth): bound OAuth token endpoint response reads. Thanks @solodmd. - **PR #96397** fix: warn when configured channel plugins cannot load. Thanks @849261680. - **PR #96359** test: migrate src/commands tests to shared temp dir helpers. Thanks @xialonglee. - **PR #96293** fix(cron): clear agentTurn thinking override by blanking the field. Related #96287. Thanks @ZengWen-DT and @takamasa-aiso. - **PR #96058** test: prefer shared temp dir helpers in auto-reply and install-fallback tests. Thanks @xialonglee. - **PR #87298** test: add temp directory helper guidance. Thanks @hxy91819. - **PR #97785** fix(sessions): avoid cross-cwd recent resumes. Related #96542. Thanks @qingminglong and @yetval. - **PR #97698** fix(pdf): reject empty parsed page ranges before native analysis. Thanks @zhangguiping-xydt. - **PR #97693** fix(discord): bound requestDiscord happy-path response reads to prevent OOM. Thanks @Alix-007. - **PR #97683** fix(irc): guard surrogate-range codepoints in \u literal-escape decoder. Thanks @llagy009. - **PR #96938** fix(utils): keep reply directive ids unicode-safe. Thanks @ly-wang19. - **PR #97857** fix(memory): detect unindexed session transcripts in status mode (fixes #97814). Thanks @zw-xysk and @CHE10X. - **PR #98094** fix(android): clarify gateway auth recovery states. Thanks @qingminglong. - **PR #98205** test(gateway): add unit tests for node wake state tracking and testing seam. Thanks @zenglingbiao. - **PR #98115** fix: surface node approval guidance from devices CLI. Thanks @welfo-beo. - **PR #97898** docs: clarify source checkout Node floor. Related #97792. Thanks @lin-hongkuan and @aniruddhaadak80. - **PR #94526** test(telegram): add regression test for forum topic message_thread_id with streamed reasoning. Related #89352. Thanks @xialonglee and @pmika. - **PR #98145** fix(device-pairing): don't churn requestId on subset re-requests. Thanks @RomneyDa. - **PR #98267** fix(system-prompt): move exec-approval + Authorized Senders below cache boundary. Related #98261. Thanks @headbouyJB. - **PR #98304** fix: preserve iOS chat line breaks. Related #98028. Thanks @joshavant and @Jabato01. - **PR #98187** fix(openrouter): send explicit auth headers. Related #97934. Thanks @sunlit-deng and @laurencebrown. - **PR #95708** fix: show WebChat preamble progress during tool activity. Thanks @ragesaq. - **PR #98210** fix(gateway): iOS Talk treats SecretRef-backed API keys as missing. Related #98209. Thanks @ooiuuii. - **PR #98009** test(infra): add unit tests for SQLite number normalization. Thanks @dwc1997. - **PR #98087** test(config): add unit tests for resolveExecCommandHighlighting. Thanks @solodmd. - **PR #98219** test(utils): add unit tests for chunkItems. Thanks @zenglingbiao. - **PR #98093** fix(core): propagate caller env PATHEXT through isExecutableFile on Windows. Thanks @wendy-chsy. - **PR #97973** fix(matrix): guard JSON.parse against malformed homeserver response bodies. Thanks @lsr911. - **PR #97999** fix(sms): guard Twilio JSON.parse against malformed API response bodies. Thanks @lsr911. - **PR #98043** Add Swedish mobile app localization. Thanks @yeager. - **PR #98144** fix(tui): correct disconnect copy for device scope upgrades. Thanks @RomneyDa. - **PR #98240** fix(agents): keep merged delivery routes account-bound. Thanks @yetval. - **PR #89949** fix(media): pin requester delivery route when task starts. Thanks @wangwllu. - **PR #98226** Redact bare Fireworks API keys. Related #98225. Thanks @ooiuuii. - **PR #98319** docs: publish release notes for v2026.6.11. Thanks @hannesrudolph. - **PR #98257** fix: show in-progress status for channel runs. Thanks @scotthuang. - **PR #97931** fix(gateway): keep provider-owned CLI sessions across the daily default reset. Thanks @yetval. - **PR #98325** docs: refresh docs map for v2026.6.11. Thanks @hannesrudolph. - **PR #97929** fix(auto-reply): stop level directives from eating the next message word. Thanks @yetval. - **PR #97928** fix(agents): estimate harness role sizes in context guard char estimator (fixes #97927). Thanks @liuhao1024 and @yetval. - **PR #97861** fix(compaction): count bashExecution and summary turns in pre-prompt overflow precheck. Thanks @yetval. - **PR #97137** doctor: add memory search lint findings. Thanks @giodl73-repo. - **PR #97358** Doctor: expose workspace status findings. Thanks @giodl73-repo. - **PR #95622** test(qa-lab): harden whatsapp qa scenarios. Thanks @mcaxtr. - **PR #98346** fix: prevent skill-creator from bypassing workshop proposals. Related #96054. Thanks @momothemage and @xianshishan. - **PR #98169** fix(heartbeat): scope commitment fan-out prompts. Thanks @bdjben. - **PR #97366** Doctor: expose device pairing findings. Thanks @giodl73-repo. - **PR #98366** fix: Android TLS fingerprint verification times out on slow handshakes. Related #98365. Thanks @joshavant. - **PR #98353** fix(ios): open app on Chat by default. Thanks @BsnizND. - **PR #98352** fix(security): warn on agent skill MCP boundary drift. Thanks @momothemage. - **PR #98347** fix: retry image describe fallback models. Thanks @momothemage. - **PR #98117** fix(ios): avoid transient duplicate final replies. Related #98116. Thanks @ooiuuii and @joshavant. - **PR #98293** fix(gateway): emit stale exec approval followup diagnostics. Thanks @BsnizND. - **PR #98376** fix(ios): use Gateway speech providers in Talk. Related #98153. Thanks @Tony-ooo. - **PR #66685** Suppress expired exec approval followup warnings. Thanks @pfrederiksen. - **PR #98385** fix: show actionable mobile protocol mismatch recovery. Related #98384. Thanks @joshavant. - **PR #98146** fix(cli): explain how to recover from device approve deadlock. Thanks @RomneyDa. - **PR #98423** improve(ios): clarify Control and Talk visual hierarchy. Related #98397. - **PR #98217** fix(doctor): recover legacy cron archive across devices. Thanks @masatohoshino. - **PR #98333** feat(openai): add GPT-5.6 series support. Related #98296. Thanks @steipete-oai. - **PR #96393** fix(cron): preserve action-required command output. Related #96346. Thanks @snowzlmbot and @nz365guy. - **PR #98429** fix(ios): classify TLS fingerprint timeouts. Thanks @joshavant. - **PR #98439** fix: Android setup codes accept local mDNS gateway hosts. Thanks @joshavant. - **PR #98443** fix(ios): improve light and dark appearance contrast. Related #98440. - **PR #97742** fix(llm): preserve structured tool result text across providers. Thanks @snowzlmbot. - **PR #97968** fix(status): surface unregistered memory embedding providers. Thanks @masatohoshino. - **PR #92237** fix(agents): preserve runtime settings overrides [AI-assisted]. Thanks @sercada. - **PR #95888** fix(active-memory): caveat mutable ops facts; mark truncated recall as incomplete. Thanks @spencer2211. - **PR #98291** fix(gateway): surface systemd start-limit exhaustion. Thanks @masatohoshino. - **PR #90517** fix(gateway): hint missing external plugin for web login. Related #83277. Thanks @TUARAN and @carol-iung. - **PR #98369** test(infra): add unit tests for SQLite user_version pragma helper. Thanks @dwc1997. - **PR #98340** fix: extension api.exec leaves child processes after timeout. Related #98335. Thanks @ooiuuii. - **PR #92063** fix(ui): collapse duplicate assistant groups during segmented streaming. Related #63956. Thanks @harjothkhara and @contentfree. - **PR #98354** fix(infra): guard delivery queue inflate against corrupted entry_json. Thanks @Pick-cat. - **PR #90566** fix(agents): warn on cron announce skip. Related #68561. Thanks @sahibzada-allahyar and @Mibslee. - **PR #98371** fix(ports): validate lsof PID parsing before assignment. Thanks @lzyyzznl. - **PR #98356** fix(cron): keep provider-owned CLI sessions across the daily default reset. Thanks @yetval. - **PR #98395** test(shared): add unit tests for account enabled guard. Thanks @dwc1997. - **PR #98411** fix(agents): recover thinking errors from provider body. Related #98308. Thanks @sunlit-deng and @clearhorizoninvestments. - **PR #98494** docs(skills): support variable landable sweep batches. Thanks @vincentkoc. - **PR #91240** fix: report Codex ChatGPT status auth. Related #91099. Thanks @849261680 and @ukstem. - **PR #98370** test(agents): add unit tests for thinking block detection. Thanks @dwc1997. - **PR #96711** test: prefer shared temp dir helpers in config, gateway, cron, crestodian, and state tests. Thanks @xialonglee. - **PR #98483** fix: Android QR scan starts gateway pairing. Thanks @joshavant. - **PR #95230** fix docs-list-mdx-pages. Thanks @hugenshen. - **PR #96322** fix(minimax): bound JSON response reads to prevent OOM. Thanks @lsr911. - **PR #95348** fix config-chmod-warning. Thanks @hugenshen and @cursoragent. - **PR #95229** fix(copilot): guard against undefined runtime.state during cli-metadata registration. Related #94516. Thanks @sunlit-deng and @cuihaijun. - **PR #94636** fix(memory): skip raw snippets during promotion. Thanks @tayoun. - **PR #94013** [AI] fix(feishu): guard partial channelRuntime in monitor startup. Thanks @xydt-tanshanshan. - **PR #93466** [AI] fix(feishu): guard against missing inbound in channelRuntime fallback. Thanks @xydt-tanshanshan. - **PR #98049** fix: hide expired pairing QR cards in Control UI. Related #98039. Thanks @ooiuuii. - **PR #96094** fix(memory): prove live manager recovery after CLI reindex. Related #91167. Thanks @849261680 and @kiagentkronos-cell. - **PR #98482** fix: advertise route-aware LAN Control UI links. Thanks @joshavant. - **PR #71537** Recover archived (.reset) session transcripts in memory hook + session-logs skill. Thanks @injinj. - **PR #96375** docs(config-agents): correct built-in alias table for opus and gpt. Thanks @niks999. - **PR #98453** docs(gateway): fix Telegram streaming default in config-channels.md. Thanks @solodmd. - **PR #98533** fix: repair hosted CI baseline assertions. - **PR #98421** feat(imessage): native poll support — create, read, vote. Thanks @omarshahine and @lobster. - **PR #98318** docs(matrix): document missing streaming.progress mode, progress sub-fields, and mentionPatterns config. Thanks @wm0018 and @vincentkoc. - **PR #97753** docs(onboard): document 11 missing non-interactive CLI flags. Thanks @wm0018 and @vincentkoc. - **PR #97851** fix(mattermost): bound null-body error response reads. Thanks @Pick-cat. - **PR #98360** fix(memory-wiki): preserve notes after transient page reads. Related #98345. Thanks @qingminglong and @yetval. - **PR #98551** test: fix stale core test type failures. Thanks @RomneyDa. - **PR #98455** fix(browser): bound error body read in fetchHttpJson to prevent OOM. Thanks @wings1029. - **PR #95906** fix(code-mode): surface QuickJS error name and message to the model. Thanks @ZengWen-DT and @vincentkoc. - **PR #97901** fix(agents): stop copilot autoreview cleanup crash on Windows. Thanks @paulcam206. - **PR #97923** fix(slack): truncate served arg-menu option labels on a surrogate boundary. Thanks @LEXES7. - **PR #98010** fix(update): validate bundle plugin payloads by manifest contract. Related #97985. Thanks @LiLan0125 and @herove. - **PR #85296** fix(codex): derive terminal-idle watchdog from explicit run timeout. Thanks @alkor2000 and @vincentkoc. - **PR #97110** feat(i18n): add native app locale inventory. Thanks @vincentkoc. - **PR #98396** fix: allow config.patch with defaulted provider baseUrl. Related #98270. Thanks @momothemage and @weltmaister. - **PR #98503** fix(usage-bar): use Object.hasOwn instead of in operator to avoid prototype chain pollution. Related #98466. Thanks @chenyangjun-xy and @zhangLei99586. - **PR #97111** feat(android): localize core gateway surfaces. Thanks @vincentkoc. - **PR #97630** fix(media): normalize Windows inbound paths case-insensitively. Thanks @VectorPeak. - **PR #82638** fix(agents): skip implicit provider discovery when models.mode is 'replace' [AI-assisted]. Related #66957. Thanks @eldar702 and @wangzhengshu. - **PR #87917** fix sessions json lineage metadata. Related #80286. Thanks @zhangguiping-xydt and @islandpreneur007. - **PR #93639** fix(message-tool): apply messages.responsePrefix to outbound sends. Thanks @ZengWen-DT. - **PR #94440** fix: #94432 classify Cloudflare challenge 403 as upstream_html instead of auth_html. Thanks @lzyyzznl and @pbm9z95m6z-hue. - **PR #98119** fix: reduce Docker build memory pressure. Related #98118. Thanks @zyzo. - **PR #97679** feat(node): add --context-path flag to node run/install for reverse-p…. Related #97678. Thanks @wm0018. - **PR #98339** fix(irc): classify host-less nick!user allowlist entries as mutable. Thanks @yetval. - **PR #97662** fix(matrix): bound raw transport response reads to prevent OOM. Thanks @Alix-007. - **PR #98137** fix: hoist timer declaration to avoid TDZ ReferenceError in abortable delay. Thanks @zhangLei99586. - **PR #98134** fix: clear timeout timer in Tailscale binary probe Promise.race. Thanks @zhangLei99586. - **PR #97989** fix(sms): stop internal tool-trace banners from reaching SMS replies. Thanks @ZengWen-DT. - **PR #97972** fix(browser): CDP auth fails with percent-encoded credentials. Thanks @VectorPeak. - **PR #98063** fix(reply): suppress tool-error progress delivery when messages.suppressToolErrors is set. Thanks @moeedahmed and @amittell. - **PR #94964** fix(reload): cancel deferred channel reload on in-process restart. Related #79487. Thanks @lzyyzznl and @tseller. - **PR #98598** fix: restore main lint after timer repairs. Related #98462, #98464. Thanks @zhangLei99586. - **PR #98587** fix(slack): guard relay WebSocket frame JSON.parse against malformed input. Thanks @lsr911 and @vincentkoc. - **PR #90030** fix(memory-core): skip qmd zero-hit search sync. Related #90023. Thanks @sahibzada-allahyar and @ruben2000de. - **PR #98493** fix(transcripts): close readline interface and destroy read stream on error exit. Related #98467. Thanks @wangmiao0668000666 and @zhangLei99586. - **PR #98497** fix(cli): show exit code when plugin npm install returns empty output. Thanks @Sanjays2402 and @vincentkoc. - **PR #97112** feat(apple): localize core native app surfaces. Thanks @vincentkoc. - **PR #98610** fix: restore tooling CI after transcript test addition. - **PR #77539** fix(subagent): preserve steered task text on restart redispatch. Thanks @amittell. - **PR #97113** feat(i18n): refresh all native locale artifacts. Thanks @vincentkoc. - **PR #98620** feat(doctor): warn about in-flight cron jobs. Thanks @masatohoshino. - **PR #98605** test(shared): add unit tests for human-readable list formatting. Thanks @dwc1997. - **PR #97348** feat(autoreview): support cursor-agent engine. Thanks @hxy91819. - **PR #95943** fix(cron): preserve provider/model on isolated-run timeout row. Related #95873. Thanks @ZengWen-DT and @cursoragent and @luke-renjoy. - **PR #94149** fix(status): bound systemd service probes so status cannot hang on a wedged systemctl (#84698). Thanks @ZengWen-DT and @cursoragent and @zus-assistant. - **PR #88159** fix(cli): retry logs.tail after journal fallback in logs follow. Thanks @anyech and @vincentkoc. - **PR #98508** fix(update-check): bound npm registry JSON response read to prevent OOM. Thanks @lzyyzznl. - **PR #98496** fix(tlon): bound error response body reads to prevent OOM. Thanks @Pandah97. - **PR #98554** fix(openai): bound embedding batch file downloads. Thanks @sunlit-deng and @vincentkoc. - **PR #98652** fix: stop invalid message timeouts from stalling. - **PR #77973** fix(gateway): cap agentRunCache to prevent unbounded growth under run fan-out. Related #77976. Thanks @fede-kamel and @vincentkoc. - **PR #98525** fix(agents): time out local streams without first event. Thanks @osolmaz. - **PR #94022** fix(cron): persist startup catch-up deferral ids in service state to prevent read-RPC clobber. Related #93935. Thanks @RichChen01 and @vincentkoc and @yetval. - **PR #93810** fix(cron): preserve startup overflow catch-up deferrals in start() maintenance pass. Thanks @yetval and @vincentkoc. - **PR #98623** fix: media tools skip env-key provider plugins when auto-selecting models. Thanks @medns. - **PR #98665** fix(claude-cli): return updatedInput in can_use_tool allow response for Claude Code 2.1. Related #95171. Thanks @yetval and @carterdawson. - **PR #94250** fix(feishu): send blocks as independent messages when blockStreaming is enabled. Related #55027. Thanks @xialonglee and @vincentkoc and @ZichaoLong. - **PR #93379** fix(whatsapp): thread authDir through command authorization and owner bypass for LID JID resolution. Related #77755. Thanks @xialonglee and @jiveshkalra. - **PR #98646** fix: keep workspace rail file sections scrollable. Related #98566. Thanks @wuqxuan and @645648406-max. - **PR #98602** fix: iOS Talk fallback settings opens Voice & Talk. Related #98593. Thanks @PollyBot13. - **PR #98611** fix(ui): add overflow-y:auto to workspace rail sections to prevent file list overflow (fixes #98566). Thanks @zw-xysk and @645648406-max. - **PR #98619** fix(qa-lab): credential lease requests fail on oversized Convex broker responses. Thanks @ZengWen-DT. - **PR #94326** fix(memory-wiki): disambiguate the reserved index page stem for synthesis and ingest. Thanks @yetval and @vincentkoc. - **PR #98659** fix(codex): classify get_goal read statuses as successful dynamic tool calls. Thanks @yetval. - **PR #96856** fix(codex): successful sessions_spawn and goal tool results recorded as failures. Thanks @nxmxbbd. - **PR #98660** fix(inworld): guard voices JSON.parse against malformed API response bodies. Thanks @solodmd. - **PR #95430** fix(embedded-agent-runner): pump async streamFn through pumpStreamWithRecovery for mid-stream error recovery. Related #95429. Thanks @lzyyzznl and @vincentkoc and @alexelgier. - **PR #98644** fix: tool summaries preserve emoji truncation boundaries. Thanks @ZengWen-DT. - **PR #80928** fix(telegram): suppress fallback reply when plugin command returns suppressReply: true. Related #80756. Thanks @alexuser and @UnClouded77. - **PR #98701** fix: prevent agents-tools message test timeouts. - **PR #92657** feat(usage): ship built-in /usage full footer. Thanks @Marvinthebored. - **PR #92877** fix(usage): make built-in footer easier to wrap on Telegram. Thanks @Marvinthebored. - **PR #98126** Restore Telegram /steer for active Codex runs. Related #81594. Thanks @100yenadmin and @Kyzcreig. - **PR #92037** feat(cron): on-exit schedule — wake on a watched command's exit. Thanks @anagnorisis2peripeteia. - **PR #98452** feat(ios): modernize the app with iOS 26 Liquid Glass. - **PR #98006** Add Telegram /login Codex pairing flow. Thanks @100yenadmin. - **PR #98735** fix(telegram): preserve rich forwarded message text. Thanks @obviyus. - **PR #97962** refactor(qa): use transport-native actions in flow scenarios. Thanks @RomneyDa. - **PR #98726** fix(nvidia): use Nemotron Super 1M context. Thanks @eleqtrizit. - **PR #98691** fix(imessage): shed emoji anywhere in poll-vote echo match. Thanks @omarshahine. - **PR #97174** Fix Telegram plugin callback routing. Thanks @goldmar. - **PR #89597** fix: migrate QQBot credential backups to SQLite KV. - **PR #98536** feat: prepare scoped conversation capability profiles. - **PR #92274** fix(agents): classify embedded prompt lock error as permanent announce failure. Related #91527. Thanks @fsdwen and @zackchiutw. - **PR #98102** fix(telegram): durably retry inbound media dropped during restart (#98076). Thanks @luoyanglang and @DaveArcher18. - **PR #98755** fix(cron): detach session-targeted runs. Related #98121. Thanks @obviyus and @EthanSK. - **PR #96065** fix(install): manage config-secretref env refs via OPENCLAW_SERVICE_MANAGED_ENV_KEYS. Thanks @Darren2030 and @obviyus. - **PR #98666** fix: diagnose Windows LAN Gateway firewall blocks. Thanks @joshavant. - **PR #98501** fix(codex): rename destructive approval mode to ask. Related #98499. Thanks @kevinslin. - **PR #98775** fix(telegram): survive transient getUpdates errors and stop per-send cache rewrites. Related #98772, #98773. Thanks @obviyus. - **PR #98776** fix(telegram): back off, dead-letter, and tombstone spooled updates so poison messages cannot block or duplicate. Related #98774. Thanks @obviyus. - **PR #96454** feat(cli): openclaw attach — launch an external harness bound to a gateway session. Thanks @anagnorisis2peripeteia and @obviyus. - **PR #98786** fix(telegram): final replies no longer drop on rejected rich entities, captions, quotes, or long flood waits. Related #98778. Thanks @obviyus. - **PR #97496** Doctor: expose channel plugin blocker findings. Thanks @giodl73-repo. - **PR #98792** fix(ci): restore docs and test type checks. - **PR #98736** improve(ios): simplify Talk controls and composer alignment. - **PR #97889** fix(discord): guard JSON.parse against malformed API response bodies. Thanks @lsr911. - **PR #98812** fix(codex): preserve plugin app approvals in side conversations. - **PR #92283** fix(agents): don't inject A2A turns into isolated-cron sessions_send (#92257). Thanks @harjothkhara and @nailujac. - **PR #98138** fix: guard setDeep against empty keys array in Chrome profile decoration. Thanks @zhangLei99586. - **PR #98183** fix(gateway): distinguish reachable gateway from failed status probe. Thanks @masatohoshino. - **PR #98689** fix(wizard): reject loose gateway port input. Related #98681. Thanks @qingminglong. - **PR #98720** fix(nostr): clear per-relay publish timeout timer to prevent dangling handles. Related #98463. Thanks @wangmiao0668000666 and @zhangLei99586. - **PR #98818** fix(ci): recover incomplete Swift build caches. - **PR #98787** fix(memory-wiki): retry transient existing-page reads in wiki_apply and chatgpt import. Thanks @yetval. - **PR #98811** feat(ios): modernize navigation and settings. Related #98803. - **PR #98843** docs: update mobile app release messaging. Thanks @joshavant. - **PR #93686** fix(weixin): startAccount preserves session routing. Related #93556. Thanks @zhangguiping-xydt and @htkillermax-gif. ### Release verification - npm package: https://www.npmjs.com/package/openclaw/v/2026.7.1-beta.1 - registry tarball: https://registry.npmjs.org/openclaw/-/openclaw-2026.7.1-beta.1.tgz - integrity: `sha512-Yu/ELLje9mxvFTlaxVGHnIkKvHcLnoEj3AQhzmhpP4k8Fi7/ln0NvYnBgaZ2BJ8tdAAsq3iZ8uroRUuXiMB0dg==` - release SHA: `4eb1d333cfeca440b796b6a3f70d3c2bef996243` - full release CI report: https://github.com/openclaw/releases/blob/main/evidence/2026.7.1-beta.1/release-evidence.md - release publish: https://github.com/openclaw/openclaw/actions/runs/28571485937 - npm preflight: https://github.com/openclaw/openclaw/actions/runs/28569693832 - full release validation: https://github.com/openclaw/openclaw/actions/runs/28569693812 - plugin npm publish: https://github.com/openclaw/openclaw/actions/runs/28571834366 - plugin ClawHub publish: success: https://github.com/openclaw/openclaw/actions/runs/28571835541 - plugin ClawHub bootstrap: not needed - OpenClaw npm publish: https://github.com/openclaw/openclaw/actions/runs/28572195585 - package Telegram E2E: passed in release checks: https://github.com/openclaw/openclaw/actions/runs/28569815816/job/84705725486 - advisory WhatsApp live QA: blocked before scenarios by exhausted Convex credential pool after two attempts: https://github.com/openclaw/openclaw/actions/runs/28569693831

我们听到了反馈。v2026.6.11 版本专注于打磨那些让 OpenClaw 感觉不够可靠的粗糙边缘，修复了回复错位、发送卡住、重连、模型设置失败等问题，并提供了更安全的管理员默认设置。 [完整发布说明](https://docs.openclaw.ai/releases/2026.6.11) ### 亮点 #### 渠道投递可靠性 投递和重连修复覆盖了 [Telegram](https://docs.openclaw.ai/channels/telegram)、[WhatsApp](https://docs.openclaw.ai/channels/whatsapp)、[Matrix](https://docs.openclaw.ai/channels/matrix)、[Google Chat](https://docs.openclaw.ai/channels/googlechat)、[iMessage](https://docs.openclaw.ai/channels/imessage)、[飞书](https://docs.openclaw.ai/channels/feishu)、[Mattermost](https://docs.openclaw.ai/channels/mattermost)、[WebChat](https://docs.openclaw.ai/web/webchat)、[控制界面](https://docs.openclaw.ai/web/control-ui)以及[终端界面](https://docs.openclaw.ai/cli/tui)。 - 修复了较新的 Google Chat 私信有时被当作群组对话处理的问题，现在它们能正确到达一对一聊天，而 Space 和群聊消息则保持原有的路由。[#58993](https://github.com/openclaw/openclaw/pull/58993) 感谢 @starhappysh，@vincentkoc。 - OpenClaw 发出的飞书语音回复现在会在聊天气泡中显示其时长，这样接收方在播放前就能看到音频有多长。[#89172](https://github.com/openclaw/openclaw/pull/89172) 相关 [#53798](https://github.com/openclaw/openclaw/issues/53798)。感谢 @areslp，@fxz26284407，@kinrocw。 - Discord 和 Telegram 的回复以及镜像聊天历史现在能更一致地保持与预期对话的关联，包括在重复的 Telegram 回复和会话变更中也是如此。[#89911](https://github.com/openclaw/openclaw/pull/89911) 感谢 @jalehman。 - 当任务开始时没有完整的对话目标时，背景图片、视频和音乐结果现在会返回到请求它们的聊天中，而不是在创建后看似失败，或者随着会话移动而被发送到错误的对话方。[#89949](https://github.com/openclaw/openclaw/pull/89949) 相关 [#86034](https://github.com/openclaw/openclaw/issues/86034)。感谢 @tianxiaochannel-oss88，@wangwllu。 - 当用户引用机器人之前的消息时，Telegram 的答案现在会保持关联到用户当前的提问，而引用他人消息的情况仍然会回复所选引用。[#90475](https://github.com/openclaw/openclaw/pull/90475) 感谢 @moeedahmed。 - QQBot 群组管理员现在可以选择斜杠命令的可用范围，私密命令现在会引导用户进入私聊，而不是在群组中暴露或被静默忽略。[#92154](https://github.com/openclaw/openclaw/pull/92154) 感谢 @sliverp。 - 使用推理能力模型的心跳检查现在会显示助手意图中的回复，而不是在 Telegram、WhatsApp 和其他渠道中暴露内部推理过程，同时选择加入的思考消息仍然有效。[#92356](https://github.com/openclaw/openclaw/pull/92356) 相关 [#92260](https://github.com/openclaw/openclaw/issues/92260)。感谢 @jmpei，@tangtaizong666，@vincentkoc。 - Telegram 的进度模式聊天现在会在新的工具输出或工件出现之前清除旧的进度气泡，保持对话整洁、可读的顺序。[#93002](https://github.com/openclaw/openclaw/pull/93002) 相关 [#90753](https://github.com/openclaw/openclaw/issues/90753)。感谢 @shadow-enthusiast，@zhangguiping-xydt。 - iMessage 的命令和链接消息现在会在延迟的链接预览到达时保持为一个 OpenClaw 回合，而对于已启用相同发送者私信合并功能的用户，不相关的快速消息则保持分开。[#93143](https://github.com/openclaw/openclaw/pull/93143) 感谢 @omarshahine。 - 通过消息工具发送的成功 Discord 回复不再在受影响的 `message_tool_only` 源渠道回合中触发误导性的失败警告。[#94072](https://github.com/openclaw/openclaw/pull/94072) 相关 [#93875](https://github.com/openclaw/openclaw/issues/93875)。感谢 @chenyangjun-xy，@hoyanhan，@vincentkoc。 - WhatsApp 群组对话现在在重试、重连和群组变更时更可靠地保留了正确的消息和群组上下文。[#94338](https://github.com/openclaw/openclaw/pull/94338) 相关 [#7433](https://github.com/openclaw/openclaw/issues/7433)。感谢 @mcaxtr，@octopuslabs-fl，@xialonglee。 - 修复了 OpenClaw 有时会回复自己延迟的 iMessage 回音的问题，这是由于多余的前导字符导致已发送消息无法被识别。[#94442](https://github.com/openclaw/openclaw/pull/94442) 感谢 @ly-wang19。 - Telegram webhook 用户可以在短暂的渠道重启、配置重新加载和恢复周期内继续接收私信和群组消息，而不会出现临时消息中断。[#94506](https://github.com/openclaw/openclaw/pull/94506) 相关 [#90254](https://github.com/openclaw/openclaw/issues/90254)。感谢 @obviyus，@travellingsoldier85，@xialonglee。 - Matrix E2EE 网关可以在长时间运行期间保持在线，而不会逐渐消耗内存直到崩溃导致渠道和进行中的工作丢失。[#94942](https://github.com/openclaw/openclaw/pull/94942) 相关 [#90455](https://github.com/openclaw/openclaw/issues/90455)。感谢 @xzh-icenter，@yar-sh。 - Telegram 用户现在会看到预期的原生反应，而不是泄露的指令或仅回复反应的消息，只有在 Telegram 接受后才会记录成功。[#94977](https://github.com/openclaw/openclaw/pull/94977) 相关 [#71140](https://github.com/openclaw/openclaw/issues/71140)。感谢 @cuttingwater，@hugenshen。 - Telegram 上关于命令、搜索、更新和 API 活动的进度更新现在保持可读性，而不是暴露嘈杂的 HTML 或代码样式行，并且在 Telegram 无法解析格式时会回退为纯文本。[#95007](https://github.com/openclaw/openclaw/pull/95007) 相关 [#95002](https://github.com/openclaw/openclaw/issues/95002)。 - 在 WebChat 中继续的 Telegram 对话现在每回合只显示一条助手回复，并将后续回复保留在活跃对话中，而不是重复答案或将它们发回 Telegram。[#95069](https://github.com/openclaw/openclaw/pull/95069) 相关 [#94930](https://github.com/openclaw/openclaw/issues/94930)。感谢 @heichaowo。 - Google Chat 现在会在工具结果无害时隐藏误导性的内部失败横幅，让用户看到完整的答案，同时正常的助手文本保持不变。[#95084](https://github.com/openclaw/openclaw/pull/95084) 相关 [#90684](https://github.com/openclaw/openclaw/issues/90684)。感谢 @jailbirt，@studentzhou-svg。 - 绑定的多智能体渠道对话现在会加载为配置的智能体设置的工作空间文件，而不是默认智能体，不过之前归档错误的对话可能会在纠正后的智能体存储中重新开始。[#95118](https://github.com/openclaw/openclaw/pull/95118) 相关 [#92903](https://github.com/openclaw/openclaw/issues/92903)。感谢 @849261680，@axjing。 - 共享同一个 OpenClaw 网关的用户现在可以在支持的聊天渠道中为单个私信联系人分配不同的模型，同时现有的群组和通配符模型选择继续保持原样。[#95120](https://github.com/openclaw/openclaw/pull/95120) 相关 [#53638](https://github.com/openclaw/openclaw/issues/53638)。感谢 @gandalf-at-lerian，@thomaszta，@xydigit-zt。 - Telegram 现在会在短暂的初始预览或进度模式回复期间显示 OpenClaw 仍在工作，而不是在最终消息到达之前保持聊天静默。[#95183](https://github.com/openclaw/openclaw/pull/95183) 相关 [#95004](https://github.com/openclaw/openclaw/issues/95004)。感谢 @obviyus。 - Matrix 用户和操作员现在会在家服务器发送超大或停滞的响应时收到明确的失败提示，而不是让 OpenClaw 继续缓冲并冒无限制内存使用的风险。[#95240](https://github.com/openclaw/openclaw/pull/95240) 感谢 @alix-007。 - 修复了在注入大量环境变量的 Kubernetes 风格部署中，Telegram 和其他排队渠道回复延迟或丢失的问题，因为打开队列数据库可能会阻塞网关。[#95278](https://github.com/openclaw/openclaw/pull/95278) 相关 [#94571](https://github.com/openclaw/openclaw/issues/94571)。感谢 @kaka-srp。 - Telegram 聊天在一条消息超时后可以恢复，允许同一聊天或主题中的后续消息在无需重启网关的情况下到达智能体。[#95299](https://github.com/openclaw/openclaw/pull/95299) 相关 [#95248](https://github.com/openclaw/openclaw/issues/95248)。感谢 @kriegerbangerz-ship-it，@mikasa0818，@obviyus。 - 当用户在共享的直接对话中切换 Telegram 和另一个 OpenClaw 客户端时，简短的 Telegram 回复现在会跟随最新的对话，而不是回应一个更早的、不相关的 Telegram 提议。[#95390](https://github.com/openclaw/openclaw/pull/95390) 相关 [#95378](https://github.com/openclaw/openclaw/issues/95378)。感谢 @maiduy708，@mikasa0818，@obviyus。 - 修复了在多次消息回复后，已完成的助手消息在 Telegram、Discord、Slack 和其他流式聊天中出现两次的问题。[#95432](https://github.com/openclaw/openclaw/pull/95432) 感谢 @vincentkoc，@yetval。 - WhatsApp 回复现在会保持关联到被回复的直接或群组消息，而不是作为一条丢失对话上下文的独立消息出现。[#95483](https://github.com/openclaw/openclaw/pull/95483) 感谢 @mcaxtr。 - Telegram 富文本消息回复现在会保持段落、项目符号和状态行分开，而不是将多行内容合并成一个连续的文本块，无需任何配置更改。[#95532](https://github.com/openclaw/openclaw/pull/95532) 相关 [#95409](https://github.com/openclaw/openclaw/issues/95409)。感谢 @amknight。 - 启用原生斜杠命令的 Mattermost 操作员现在可以直接在 Mattermost 中使用 `/oc_queue` 来调整正在运行的任务的排队行为，包括其模式、去抖动时间、上限和丢弃处理。[#95546](https://github.com/openclaw/openclaw/pull/95546) 感谢 @amknight。 - 在传统的多账户渠道升级后，之前允许的消息仍能到达指定账户，继承的私信和群组访问规则在 Mattermost、Discord、Slack、Telegram、Signal、WhatsApp、iMessage 和 IRC 中得到保留。[#95550](https://github.com/openclaw/openclaw/pull/95550) 感谢 @amknight。 - Mattermost 用户在机器人回复后可以继续在话题中发言而无需再次提及机器人，并且这种参与在网关重启后仍能保持，直到话题闲置 7 天。[#95552](https://github.com/openclaw/openclaw/pull/95552) 感谢 @amknight。 - 入站 Telegram 消息现在能及时到达配置的 OpenClaw 会话，而不是在下一个轮询间隔、网关重启或手动干预之前一直无人应答。[#95577](https://github.com/openclaw/openclaw/pull/95577) 相关 [#86957](https://github.com/openclaw/openclaw/issues/86957)。感谢 @freidrich-goldenflow，@liuwqgit。 - QQBot 用户现在可以收到完整的 markdown 表格，前提是有效的分隔符每列使用一个或两个连字符，而不是丢失标题行并且只保留最后一行。[#95637](https://github.com/openclaw/openclaw/pull/95637) 感谢 @ly-wang19。 - Synology Chat 用户现在可以接收需要超过 120 秒的智能体回复，只要配置的核心超时允许，而不是被渠道提前拒绝。[#95707](https://github.com/openclaw/openclaw/pull/95707) 感谢 @sahibzada-allahyar，@vincentkoc。 - Telegram 论坛话题的定时任务现在会保持单独配置的失败警报发送到预期目的地，即使主公告使用了同一聊天中的话题。[#95794](https://github.com/openclaw/openclaw/pull/95794) 感谢 @vincentkoc。 - 修复了 WhatsApp 群组回复可能引用较旧的 OpenClaw 消息而不是用户的触发消息的问题，现在最终答案在有回复目标可用时能正确关联到预期消息，并在无法关联时避免指回过时的机器人上下文。[#95914](https://github.com/openclaw/openclaw/pull/95914) 感谢 @mcaxtr。 - WhatsApp 用户可以通过反应来批准或拒绝提示，而不会在 WhatsApp 以不同方式识别同一直接聊天时导致提示卡住，同时群组批准仍然关联到正确的群组和人。[#95935](https://github.com/openclaw/openclaw/pull/95935) 感谢 @mcaxtr。 - 最终回复处理现在在 OpenClaw 检查块文本是否已发送时使用更少的 CPU，不会改变到达聊天的回复或重复文本的抑制方式。[#96087](https://github.com/openclaw/openclaw/pull/96087) 感谢 @vincentkoc。 - 来自外部渠道插件的执行批准结果现在会返回到命令启动的渠道或私信，而不是回退到 WebChat 或在批准后消失。[#96140](https://github.com/openclaw/openclaw/pull/96140) 相关 [#96103](https://github.com/openclaw/openclaw/issues/96103)。感谢 @lansenger-pm，@vincentkoc，@yetval。 - 当用户回复一条较旧的 OpenClaw 消息时，WhatsApp 的最终答案现在会被引用到用户刚刚发送的后续消息，而不是未引用地到达或指回旧的机器人回复。[#96220](https://github.com/openclaw/openclaw/pull/96220) 感谢 @mcaxtr。 - Nextcloud Talk 机器人现在会忽略普通的文件分享和生命周期事件，不会将其记录为机器人错误或冒禁用投递的风险，同时格式错误的聊天载荷仍然会返回错误。[#96243](https://github.com/openclaw/openclaw/pull/96243) 相关 [#81566](https://github.com/openclaw/openclaw/issues/81566)。感谢 @arkyu2077，@rafaelmgbh，@vincentkoc。 - Mattermost 渠道中的回复和消息工具投递现在使用渠道和话题指导，因为智能体将这些对话识别为渠道而非群聊，而现有的群聊行为保持不变。[#96244](https://github.com/openclaw/openclaw/pull/96244) 相关 [#95645](https://github.com/openclaw/openclaw/issues/95645)。感谢 @arkyu2077，@iloveleon19，@vincentkoc。 - MCP 渠道集成现在将对话列表、消息读取、事件轮询和等待保持在可预测的范围内，即使客户端请求过高的限制或超时。[a39e548](https://github.com/openclaw/openclaw/commit/a39e548ede228aa1978bf9d509613cbed6db0c99) 感谢 @vincentkoc。 - 长时间运行的流式自动回复不太可能在配置了异常大的超时时提前停止或不一致地中止。[6c85b90](https://github.com/openclaw/openclaw/commit/6c85b90469f94955ef00c1609e1f1d6fd2cf4ca8) 感谢 @vincentkoc。 - 渠道进度现在在工作真正在另一次更新后返回时会显示重复的状态，而不是将有用的上下文隐藏为重复项。[8a75c4d](https://github.com/openclaw/openclaw/commit/8a75c4dd5f3e625a22a7a08c6e1f368798c48111) - 已完成的渠道回复不再获得延迟的进度通知，防止过期的状态文本在答案完成后出现。[a594d2c](https://github.com/openclaw/openclaw/commit/a594d2ce73257326b7ab78adb3c4643245ec9431) 感谢 @vincentkoc。 - 在流式渠道回复期间，进度消息现在会持续显示最新状态，而不是卡在较早的更新上。[e114001](https://github.com/openclaw/openclaw/commit/e114001ccafa83b8b366e095a9d7748dfc50c082) 感谢 @vincentkoc。 - Matrix 强制重置现在可以在秘密存储不可用的情况下处理而不会出现运行时错误，将恢复访问视为不可用，以便重置路径可以安全继续。[5c5a8a4](https://github.com/openclaw/openclaw/commit/5c5a8a49d76954b53fefc2463bc7b1d6b960e8fc) 感谢 @vincentkoc。 - 配置的渠道现在仍然在 `openclaw channels status --json` 中可见，同时预定的公告会拒绝没有活跃插件投递它们的过期条目。[a641c0d](https://github.com/openclaw/openclaw/commit/a641c0d560fd15373e462829facf15fd6a466aeb) - Discord 语音对话现在可以连续处理助手的连续回复，因此排队的回复会在前一个音频流关闭后播放，而不是保持卡住。[88b64e4](https://github.com/openclaw/openclaw/commit/88b64e4b869e696d99de7417fb52425e9ed67cbf) 感谢 @vincentkoc。 - 当智能体响应已经开始到达时，Discord 进度预览不太可能在最终编辑之前停止。[86ea382](https://github.com/openclaw/openclaw/commit/86ea382121b00e73af4b4c329d0a2447592e4071) - 当助手打算保持沉默时，聊天不再显示杂散的 `NO_REPLY` 文本，同时合法的媒体回复仍然会到达，没有占位符。[96c6f80](https://github.com/openclaw/openclaw/commit/96c6f8022c2420826830b11f4353ce855ab2ac5c) - Telegram 流式回复现在每个进度标题只显示一次，使工具和搜索更新更易于浏览。[013e33c](https://github.com/openclaw/openclaw/commit/013e33c6d3672a980550912442bb1ac5505918aa) 感谢 @vincentkoc。 - 在长时间运行的任务、崩溃或网关重启后卡住的 Telegram 消息现在会自动恢复处理，因此后续消息不再需要静默等待或要求操作员手动修复队列。[#97543](https://github.com/openclaw/openclaw/pull/97543) 感谢 @romneyda，@vincentkoc。 #### 提供商和模型恢复 [模型设置](https://docs.openclaw.ai/providers/models)、[OpenAI](https://docs.openclaw.ai/providers/openai)、[OpenRouter](https://docs.openclaw.ai/providers/openrouter)、[opencode-go](https://docs.openclaw.ai/providers/opencode-go)和[快速模式](https://docs.openclaw.ai/tools/thinking#fast-mode-fast)在受影响配置中的后续行为现在恢复得更清晰。 - MiniMax 文本转语音和语音笔记不太可能失败，因为 OpenClaw 现在会明确请求它可以解码的音频格式，而不是依赖提供商默认设置。[#73079](https://github.com/openclaw/openclaw/pull/73079) 感谢 @efe-arv。 - 网关操作员现在可以在正常日志中再次看到提供商、模型、请求状态和计时详细信息，这使得模型路由和传输问题更容易诊断，而无需启用额外的调试日志。[#89648](https://github.com/openclaw/openclaw/pull/89648) 相关 [#89300](https://github.com/openclaw/openclaw/issues/89300)。感谢 @enominera，@xiaobao-k8s。 - 通过 Google、Mistral、OpenAI Responses、Azure OpenAI Responses 和 ChatGPT/Codex Responses 访问的模型现在会收到干净的系统指令，而不会将 OpenClaw 的内部缓存边界标记泄漏到提示中。[#89716](https://github.com/openclaw/openclaw/pull/89716) 感谢 @enominera，@masatohoshino。 - 通过兼容 OpenAI 的提供商使用 Gemini 模型的 Cron 工具调用现在可以正常运行，而不会因可空字段触发提供商模式拒绝。[#91559](https://github.com/openclaw/openclaw/pull/91559) 相关 [#91542](https://github.com/openclaw/openclaw/issues/91542)。感谢 @pick-cat，@qiukui666。 - 提供商限定的模型 ID 现在会遵守其配置的智能体运行时策略和 CLI 别名，而不是意外地回退到 OpenClaw 的默认运行时。[#91724](https://github.com/openclaw/openclaw/pull/91724) 感谢 @vincentkoc，@yu-xin-c。 - 聊天 `/models` 列表和其他识别插件的模型或提供商选择路径现在会快速响应，而不是停滞数秒并通过重复的设置扫描消耗一个 CPU 核心，同时插件更改仍会正常刷新。[#93356](https://github.com/openclaw/openclaw/pull/93356) 感谢 @obuchowski。 - 托管的 Ollama Cloud 用户在重启后可以只保留他们明确配置的模型，而不会将完整的共享目录添加回来，同时本地和自托管 Ollama 服务器的自动发现继续工作。[#93956](https://github.com/openclaw/openclaw/pull/93956) 感谢 @jason-allen-oneal。 - 当本地提供商返回通用的 `LLM request failed.` 错误时，Cron 作业现在可以重试或切换到配置的备用模型，而不是失败且备用模型未被使用。[#94062](https://github.com/openclaw/openclaw/pull/94062) 相关 [#93931](https://github.com/openclaw/openclaw/issues/93931)。感谢 @hugenshen。 - 过期的提供商令牌不再用重复的回退警告淹没有用的操作员日志，同时第一个警告和后续的重复摘要仍然可用于诊断。[#94233](https://github.com/openclaw/openclaw/pull/94233) 相关 [#56979](https://github.com/openclaw/openclaw/issues/56979)。感谢 @goutamadwant，@yanan1991。 - Google Gemini 3.5 Flash 现在可以使用其完整的 1，048，576 token 上下文窗口来选择，避免出现模型缺失错误和不必要的提示大小拒绝。[#94726](https://github.com/openclaw/openclaw/pull/94726) 相关 [#94723](https://github.com/openclaw/openclaw/issues/94723)。感谢 @ajwan8998，@anguslogan01，@kevinat。 - 仪表盘子会话现在一致地处理允许的提供商限定模型选择，并在保存的模型状态过时时提供准确的恢复指导。[#94752](https://github.com/openclaw/openclaw/pull/94752) 相关 [#94713](https://github.com/openclaw/openclaw/issues/94713)。感谢 @gr4via。 - Claude CLI 用户不再收到可能永远不会到达的完成更新承诺，因为 OpenClaw 现在会在不支持的本地后台工作可能妨碍进度之前阻止它。[#95008](https://github.com/openclaw/openclaw/pull/95008) 感谢 @anagnorisis2peripeteia。 - OpenClaw 现在会在将整个响应缓冲到内存之前，用明确的错误拒绝过大的提供商目录或 JSON 响应。[#95218](https://github.com/openclaw/openclaw/pull/95218) 感谢 @alix-007。 - OpenRouter 用户现在可以选择并运行广告中较短的 DeepSeek V4 模型 ID，而不会因为 OpenClaw 发送了重复的提供商前缀而导致请求失败并返回 `model_not_found`。[#95268](https://github.com/openclaw/openclaw/pull/95268) 相关 [#95198](https://github.com/openclaw/openclaw/issues/95198)。感谢 @daniel-alejandro-t，@darren2030。 - 启用 `/reasoning on` 后，DeepSeek 风格的兼容 OpenAI 模型现在会将最终答案与推理过程分开显示，而不是将其折叠到推理块中，无需任何配置更改。[#95283](https://github.com/openclaw/openclaw/pull/95283) 相关 [#95280](https://github.com/openclaw/openclaw/issues/95280)。感谢 @marvinthebored，@vincentkoc，@zengwen-dt。 - 当 Codex 订阅达到使用限制时，OpenClaw 现在会转移到配置的备用模型，而不是在失败结果上停止，并且不会重试已经产生可见输出的运行。[#95400](https://github.com/openclaw/openclaw/pull/95400) 感谢 @jason-allen-oneal，@sallyom。 - LM Studio 用户现在可以运行量化或多变体本地模型，而不会出现虚假的助手回合失败或由模型键不匹配导致的幽灵后缀模型条目。[#95401](https://github.com/openclaw/openclaw/pull/95401) 感谢 @monkeyleet。 - Google 支持的内嵌智能体运行现在会在它们耗尽内存或导致运行停滞之前停止读取过大或永不结束的提示缓存响应。[#95417](https://github.com/openclaw/openclaw/pull/95417) 感谢 @alix-007。 - OpenRouter 模型扫描在遇到过大或格式错误的目录时会安全失败，而不是冒可能导致 OpenClaw 不稳定的过度内存使用风险。[#95418](https://github.com/openclaw/openclaw/pull/95418) 感谢 @alix-007。 - OpenRouter 设置现在会在可能耗尽 OpenClaw 内存之前拒绝过大的模型目录，而不会缓存或立即重新获取失败的响应。[#95420](https://github.com/openclaw/openclaw/pull/95420) 感谢 @alix-007，@sallyom。 - 配置的备用模型现在可以在 Claude CLI 用完积分或遇到通用运行器故障时提供回答，而不是让用户将失败消息作为最终响应。[#95508](https://github.com/openclaw/openclaw/pull/95508) 相关 [#95489](https://github.com/openclaw/openclaw/issues/95489)。感谢 @mikasa0818，@riazrahaman，@sallyom。 - 使用 `freshness: "day"` 或 `pd` 的 Gemini 支持的网页搜索现在可以完成，而不会因提供商 400 错误而失败，同时更广泛的时效性选择和明确的日期范围保留更严格的过滤。[#95682](https://github.com/openclaw/openclaw/pull/95682) 感谢 @sunjae-k，@vincentkoc。 - 来自 xAI 推理模型（如 Grok Composer）的后续答案现在更可靠地保留了早期的推理上下文，即使配置的推理工作量不受支持。[#95686](https://github.com/openclaw/openclaw/pull/95686) 感谢 @fuller-stack-dev，@geraint0923。 - Vercel AI Gateway 用户现在可以运行从实时目录中选择的模型，包括仅存在于实时目录中但不在 OpenClaw 捆绑列表中的模型 ID。[#95710](https://github.com/openclaw/openclaw/pull/95710) 感谢 @vincentkoc。 - 修复了清单定义的提供商在 `stripPrefixes` 条目包含多余空格或不同大小写时将有效模型 ID 变成无效的问题，现在操作员和插件作者能得到预期的提供商模型。[#95744](https://github.com/openclaw/openclaw/pull/95744) 相关 [#95743](https://github.com/openclaw/openclaw/issues/95743)。感谢 @parveshsaini。 - 首次运行设置现在会为新安装的外部提供商打开凭据提示，而不是看似循环并保留 OpenAI 被选中。[#95792](https://github.com/openclaw/openclaw/pull/95792) 相关 [#95765](https://github.com/openclaw/openclaw/issues/95765)。 - 过大或停滞的提供商目录现在会快速失败并显示明确的错误，而不是挂起 OpenClaw 或消耗无限制的内存，同时正常目录继续加载。[#95827](https://github.com/openclaw/openclaw/pull/95827) 感谢 @alix-007。 - 小米 Token Plan 用户现在可以使用 `mimo-v2.5` 和 `mimo-v2.5-pro` 的输出 token 上限达到 128K，而不是被限制在过时的 32，000 个 token。[#95934](https://github.com/openclaw/openclaw/pull/95934) 感谢 @idootop。 - 工具密集型的模型响应可以以更少的开销进行流式传输，同时重复的工具调用 ID 和加密的推理细节在 Google 和兼容 OpenAI 的提供商之间保持与正确调用的匹配。[#95957](https://github.com/openclaw/openclaw/pull/95957) 感谢 @vincentkoc。 - 对于捆绑的 ACPX 用户，token 使用量统计更加可靠，因为 OpenClaw 现在默认包含 ACPX 0.11.2 的持久性修复，无需单独的软件包覆盖或手动客户端更新。[#96124](https://github.com/openclaw/openclaw/pull/96124) 感谢 @vincentkoc。 - Ollama Cloud 用户现在可以找到并选择 `glm-5.2:cloud`，拥有其 1，000，000 token 上下文窗口、推理和工具支持，即使它不在公共模型列表中。[11484f8](https://github.com/openclaw/openclaw/commit/11484f8a1483b7c42aa2971de2d88289fcef7046) - MiniMax 图像理解请求不再在超时为零、负数或极大时在到达提供商之前失败；无效值现在使用正常或安全的最大等待时间。[4b6182e](https://github.com/openclaw/openclaw/commit/4b6182ee2a250005e0c25edfeae4db6ec59b7cb8) 感谢 @vincentkoc。 - Codex 运行现在遵循当前快速模式的选择，而不是沿用旧的速度层级，并且状态行会清楚显示何时快速模式是自动的。[77012f9](https://github.com/openclaw/openclaw/commit/77012f9807851c662e064d05097497a25ab13505) 感谢 @vincentkoc。 - Codex 支持的对话在自动快速模式被清除后会恢复正常路由，防止后续回合或模型更改重用旧的优先级层级。[8afc1f7](https://github.com/openclaw/openclaw/commit/8afc1f770bbef30a4d2d9957ef26a685c508448c) - 备用智能体运行现在会遵守每个模型配置的自动快速模式截止点，即使快速模式在本次运行中被覆盖，使备用行为与所选的模型策略保持一致。[efd3172](https://github.com/openclaw/openclaw/commit/efd3172662ce023eb8d6568b689361536edf06dd) - 实时模型切换重试现在会保留长时间运行会话的原始快速模式截止点，同时显式快速模式避免误导性的自动截止点进度消息。[d990115](https://github.com/openclaw/openclaw/commit/d990115d1972fdf4361884a29bbf8396f33e5cba) 感谢 @vincentkoc。 - 内嵌智能体运行现在在重试和进度更新期间保持一致地工作，而不会将其与手动选择的快速模式设置混淆。[cf1b6fe](https://github.com/openclaw/openclaw/commit/cf1b6fef4403bee7c206299efc4385a7fcb74375) 感谢 @vincentkoc。 - 快速模式运行现在在模型回退重试期间保持其速度设置，并在状态中显示配置的自动阈值，避免不一致的重试行为和无帮助的通用标签。[aa3797c](https://github.com/openclaw/openclaw/commit/aa3797c8d0d74b4502d24852ce6baa70286f2f06) 感谢 @vincentkoc。 - 智能体回复和计划的 cron 运行现在更可靠地处理快速模式回退重试，保持最终尝试完成或正确报告进度所需的状态。[14e448e](https://github.com/openclaw/openclaw/commit/14e448e0e13db9f194ea16bb98e0f846a67769fd) 感谢 @vincentkoc。 - 用户不再会在模型回退尝试仍在运行时看到快速模式重置通知；它只会在运行达到其最终回退尝试时出现。[6eb72a8](https://github.com/openclaw/openclaw/commit/6eb72a830ece3e2b4c6c85e5a9c2b72b59e0dae9) - 用户和操作员现在在配置的实时模型变得不可用时得到更清晰的处理，因为 OpenClaw 会将提供商的“所选模型未找到”响应识别为模型未找到失败，而不是通用错误。[2405d02](https://github.com/openclaw/openclaw/commit/2405d029d437ee58ab94da800a5b213bc6bf4628) 感谢 @vincentkoc。 - Qwen 和 vLLM 现在在思考模式切换时一致地保留现有的聊天模板设置，并且提供商插件可以使用相同的测试辅助函数。[2ba9d6e](https://github.com/openclaw/openclaw/commit/2ba9d6eabef9427a950bacc47f077200446cb865) 感谢 @vincentkoc。 - 兼容 OpenAI 的代理提供商可以更一致地处理思考级别和传统的 `reasoning_effort` 字段，插件开发者和提供商维护者可以通过 OpenRouter、Kilocode 和 SDK 使用一个文档化的规范化辅助函数。[35bafea](https://github.com/openclaw/openclaw/commit/35bafea757fab0386292951a3dc2a2d3514f370e) - 浏览器和 Vite 构建现在可以加载 OpenAI ChatGPT Responses 提供商，而不会因为仅服务器依赖导致打包失败，同时 WebSocket 失败仍然正常出现。[8c8eb86](https://github.com/openclaw/openclaw/commit/8c8eb86fff6e843bd391808ceee249ac8c7f5fa5) 感谢 @vincentkoc。 - OpenRouter 模型扫描现在接受与运行时发现相同大小的有效目录，同时仍然在过大响应可能消耗无限制内存之前拒绝它们。[ad3b2f4](https://github.com/openclaw/openclaw/commit/ad3b2f4b8827cd73b4c1a6c8288569c0966276fe) 感谢 @vincentkoc。 - OpenAI Responses 用户，包括受影响的 Bedrock Mantle GPT-5.x 推理设置，现在会得到一个干净的最终答案，附带对齐的保存记录和重播上下文，而不是数十个重复的累积副本。[#92399](https://github.com/openclaw/openclaw/pull/92399) 相关 [#91959](https://github.com/openclaw/openclaw/issues/91959)。感谢 @amersheeny，@daimingnj，@phoenixyy，@pigfoot。 - 使用 opencode-go 模型的预定作业和隔离会话现在会将停滞的请求移入配置的超时或回退处理，而不是挂起几分钟然后以通用的 `LLM request failed` 错误结束。[#93965](https://github.com/openclaw/openclaw/pull/93965) 相关 [#93610](https://github.com/openclaw/openclaw/issues/93610)。感谢 @forceconstant，@zhangguiping-xydt。 #### 会话、记忆和信任连续性 [会话](https://docs.openclaw.ai/concepts/session)、[压缩](https://docs.openclaw.ai/concepts/compaction)、[记忆](https://docs.openclaw.ai/concepts/memory)、[基于 QMD 的记忆](https://docs.openclaw.ai/concepts/memory-qmd)和[工具搜索](https://docs.openclaw.ai/tools/tool-search)现在更一致地保留有用状态，同时 [Matrix](https://docs.openclaw.ai/channels/matrix) 恢复、[工具策略](https://docs.openclaw.ai/gateway/sandbox-vs-tool-policy-vs-elevated)和[批准](https://docs.openclaw.ai/cli/approvals)保持与预期信任边界的关联。 - 使用 OpenAI Responses 的受影响智能体对话现在可以在可见渠道响应使其保存的历史记录不完整后恢复并继续回复，而不是在每个后续回合失败之前无回复出现。[#84708](https://github.com/openclaw/openclaw/pull/84708) 感谢 @anyech。 - 当 Codex 支持的智能体产生异常大的工具输出时，保存和重播的对话现在会将其文本保持在通常的大小限制内，同时非文本内容保持不变。[#87912](https://github.com/openclaw/openclaw/pull/87912) 感谢 @adrianip0204。 - 控制界面中的对话现在在睡眠、网络断开或网关重新连接后仍然可见，并在同一会话中继续，而不是在发送下一条消息时消失。[#89017](https://github.com/openclaw/openclaw/pull/89017) 相关 [#87700](https://github.com/openclaw/openclaw/issues/87700)。感谢 @zhangguiping-xydt，@asicoe。 - 捆绑的 Codex 和 Copilot 集成现在会在存储演变时保持镜像聊天历史和记录更新与正确的 OpenClaw 会话关联，同时现有的基于文件的活跃记录在迁移期间继续工作。[#89518](https://github.com/openclaw/openclaw/pull/89518) 感谢 @jalehman。 - WebChat 的当前会话状态现在与您实际使用的对话匹配，因此会话身份、思考级别、token 上下文和费用详细信息不再来自回退的 `main` 会话。[#89800](https://github.com/openclaw/openclaw/pull/89800) 相关 [#89773](https://github.com/openclaw/openclaw/issues/89773)。感谢 @killo3967，@sweetcornna。 - 在自动压缩期间按下停止按钮后，您的对话不太可能丢失上下文，因为压缩请求现在也会被取消。[#89886](https://github.com/openclaw/openclaw/pull/89886) 相关 [#89868](https://github.com/openclaw/openclaw/issues/89868)。感谢 @lykeion-dev，@openperf，@vincentkoc。 - 当跨智能体会话访问被阻止时，OpenClaw 现在会列出所有必需的可见性、智能体间和允许列表设置，帮助操作员纠正策略配置，而不是追踪不存在的智能体故障。[#90489](https://github.com/openclaw/openclaw/pull/90489) 相关 [#90443](https://github.com/openclaw/openclaw/issues/90443)。感谢 @ramitrkar-hash，@sahibzada-allahyar，@vincentkoc。 - `openclaw memory status` 现在会显示活跃的光照或 REM 做梦阶段，而不是错误地报告 `Dreaming: off`，这样操作员可以看到有效的记忆配置已启用。[#93113](https://github.com/openclaw/openclaw/pull/93113) 相关 [#67868](https://github.com/openclaw/openclaw/issues/67868)。感谢 @agentarclab，@mrossit。 - 超时的 QMD 记忆搜索现在会在智能体继续前进时停止其后台工作，防止被放弃的进程继续消耗 CPU 和内存。[#93394](https://github.com/openclaw/openclaw/pull/93394) 感谢 @alix-007。 - 压缩后发送的重复指令现在会保留在对话中，防止丢失回合、无关联的回复以及某些提供商拒绝的格式错误历史记录。[#94328](https://github.com/openclaw/openclaw/pull/94328) 感谢 @vincentkoc，@yetval。 - 记忆维基的过时页面报告现在会将持久的概念和合成页面排除在新鲜度警告之外，将注意力集中在可能真正需要审查的源页面和实体页面上。[#94369](https://github.com/openclaw/openclaw/pull/94369) 感谢 @sunnyshu0925，@vincentkoc。 - 具有近期进度的长时间内嵌运行现在不太可能被过时会话恢复中断，而真正停滞的运行仍然可以被清除，以便排队的工作继续进行。[#94701](https://github.com/openclaw/openclaw/pull/94701) 感谢 @imadal1n，@mrclawfield。 - Ollama 记忆搜索现在会遵守配置的较小嵌入维度，并保持不同维度的索引分离，避免不兼容的向量被混合在一起。[#94811](https://github.com/openclaw/openclaw/pull/94811) 感谢 @mushuiyu886。 - 记忆搜索和定向刷新现在会保持连接到正确的 OpenClaw 会话，即使记录文件名更改或 QMD 导出使用不同的名称。[#95087](https://github.com/openclaw/openclaw/pull/95087) 感谢 @jalehman。 - 包含屏幕截图或其他图像的长时间运行对话现在会在 OpenClaw 为新消息腾出空间时更一致地保持其连续性，而不是重复填满却不推进保留的对话。[#95128](https://github.com/openclaw/openclaw/pull/95128) 感谢 @yetval。 - Windows 用户现在可以通过配置的绝对 `memory.qmd.command` 路径运行基于 QMD 的记忆索引和搜索，包括驱动器号和 UNC 路径，而 OpenClaw 不会在启动前去除路径分隔符。[#95274](https://github.com/openclaw/openclaw/pull/95274) 相关 [#92302](https://github.com/openclaw/openclaw/issues/92302)。感谢 @ardooken，@ly85206559。 - 使用 `/usage full` 或 `/usage tokens` 选择的使用情况页脚现在在每日或空闲会话滚动后仍然可见，因此用户不必再次打开它们。[#95322](https://github.com/openclaw/openclaw/pull/95322) 感谢 @litang9。 - 后续回复、反应、话题消息和状态检查在网页聊天或系统活动后仍然属于它们所属的聊天，而真正的渠道切换仍然会清除过时的路由细节。[#95467](https://github.com/openclaw/openclaw/pull/95467) 感谢 @yetval。 - 长时间运行的主对话现在会在用户经过一夜或延迟的后续操作后返回时保留其之前的上下文，而不是在正常完成后静默重新开始。[#95472](https://github.com/openclaw/openclaw/pull/95472) 感谢 @xydt-tanshanshan。 - 拥有大量会话历史的用户可以列出、预览和查找会话，而不会遇到数秒的冻结，同时较旧的大小写混合会话键仍在启动时进行迁移。[#95699](https://github.com/openclaw/openclaw/pull/95699) 感谢 @jalehman，@jzakirov。 - 修复了当操作员使用自定义或按智能体配置的 `session.store` 时，已投递的回复有时被保存到错误的对话历史或从中省略的问题，提高了预期会话的连续性和可审计性。[#95782](https://github.com/openclaw/openclaw/pull/95782) 相关 [#95781](https://github.com/openclaw/openclaw/issues/95781)。感谢 @youngting520。 - 保存的会话记忆摘要现在会排除原始模型 token、工具调用块、媒体占位符、角色标签和过时的 `NO_REPLY` 标记，以便未来的对话保留有用的上下文。[#95791](https://github.com/openclaw/openclaw/pull/95791) 感谢 @sweetsophia，@vincentkoc，@yb0y。 - 使用 Codex/ChatGPT OAuth 的长时间运行的 OpenAI 会话现在可以压缩而无需单独的 API 密钥，无论是手动运行 `/compact` 还是自动触发。[#95831](https://github.com/openclaw/openclaw/pull/95831) 相关 [#95693](https://github.com/openclaw/openclaw/issues/95693)。感谢 @sallyom，@yui-tien。 - 长时间、工具密集型的会话现在会压缩过大的对话，而不是在末尾出现大工具结果时卡住。[#95860](https://github.com/openclaw/openclaw/pull/95860) 相关 [#78478](https://github.com/openclaw/openclaw/issues/78478)。感谢 @jw8957，@wzhgba，@yetval。 - 当 `memory_search` 由于 Node 运行时缺少 `node:sqlite` 而不可用时，OpenClaw 现在会引导用户使用兼容的运行时，而不是将他们引向无关的嵌入提供商故障排除。[#95916](https://github.com/openclaw/openclaw/pull/95916) 感谢 @rrrrrredy，@vincentkoc。 - 检查压缩后的 Copilot 会话的开发者和管理员现在会得到其摘要、压缩前后的 token 计数以及会话详情，而不是不完整的结果。[#96049](https://github.com/openclaw/openclaw/pull/96049) 感谢 @vincentkoc。 - `/stop` 和 abort 命令现在即使需要规范化会话键或无法保存中止元数据，也能持续停止活跃运行、清除排队的后续操作并结束相关的子智能体。[#96201](https://github.com/openclaw/openclaw/pull/96201) 感谢 @jalehman。 - 语音唤醒升级现在会保持现有的触发短语和路由规则正常工作，因为 OpenClaw 将它们从已废弃的设置文件迁移到共享状态数据库中。[bdf81a8](https://github.com/openclaw/openclaw/commit/bdf81a825fa3ef66ad2c535c1eeb0bb4e31b6d1b) - 从较旧的 OpenClaw 状态布局升级现在会保留更新通知、检查节流、可用版本记录和自动更新尝试历史，因为该状态被移入 SQLite。[eb00d49](https://github.com/openclaw/openclaw/commit/eb00d499d16feea600fceef92d575fa30f005649) 感谢 @vincentkoc。 - 插件渠道的对话在启动、医生检查和状态修复期间更可靠地保持其预期会话，同时较旧的绑定记录被迁移到 OpenClaw 的共享数据库中。[9f888d9](https://github.com/openclaw/openclaw/commit/9f888d95e082d50380a66db18ee2e32683e688e0) - Windows 基于记忆的会话同步现在即使在路径格式不同时也能继续使用预期的记录文件。[b3b5b08](https://github.com/openclaw/openclaw/commit/b3b5b08e67a26efd648c7c7d879e5487223cd796) 感谢 @vincentkoc。 - 会话键缺失或为空的内嵌智能体运行现在会保持关联到预期的会话，而不是通过不一致的会话路由发送。[911f853](https://github.com/openclaw/openclaw/commit/911f853b7fc4d819e2175b001662a01eba30453d) 感谢 @vincentkoc。 - 当模型猜测错误的工具名称时，工具搜索和代码模式现在会显示如何找到并重试正确的工具，减少长时间运行会话卡住或压缩期间丢失持久记忆的风险。[#93374](https://github.com/openclaw/openclaw/pull/93374) 相关 [#92273](https://github.com/openclaw/openclaw/issues/92273)。感谢 @mushuiyu886，@poison，@vincentkoc。 - 修复了压缩后助手回复从 webchat、控制界面、飞书和其他内嵌对话中消失的问题，使刷新的聊天保持可读性，后续请求保持独立。[#95484](https://github.com/openclaw/openclaw/pull/95484) 相关 [#76729](https://github.com/openclaw/openclaw/issues/76729)。感谢 @maweibin，@njuboy11，@vincentkoc。 - OpenClaw 记忆功能现在会保持活跃、重置和已删除的记录覆盖范围与配置的会话存储和智能体所有权对齐，使做梦、QMD 导出、索引和同步不太可能遗漏或错误归属对话历史。[#96162](https://github.com/openclaw/openclaw/pull/96162) 感谢 @jalehman。 - 网关 TLS 设置现在会明确拒绝空的证书或密钥路径，或使用 OpenClaw 的默认值，避免令人困惑的启动和证书生成失败，同时保留有效路径。[#94054](https://github.com/openclaw/openclaw/pull/94054) 感谢 @miorbnli。 - 配置的插件策略在网关注册表更改、重新加载或后续钩子初始化后仍然会阻止或重写敏感的工具调用，而不是被静默跳过。[#94545](https://github.com/openclaw/openclaw/pull/94545) 感谢 @jesse-merhi。 - 拥有 `operator.approvals` 的移动操作员现在可以在发起请求的 iOS 设备上查看和解决聊天触发的执行批准，包括在应用打开时，而不仅仅依赖推送通知。[#95175](https://github.com/openclaw/openclaw/pull/95175) 感谢 @joshavant。 - 控制界面用户现在会获得修补后的 DOMPurify 版本，减少 GHSA-cmwh-pvxp-8882 清理器漏洞的暴露，而不会改变界面的行为。[#95691](https://github.com/openclaw/openclaw/pull/95691) 感谢 @vincentkoc。 - 插件对话绑定的“始终允许”批准现在会从旧的设置文件中继承，并且在多个 OpenClaw 进程运行时不太可能丢失或被覆盖。[ae41b00](https://github.com/openclaw/openclaw/commit/ae41b009224b0a8e3a990912503258d4478fb4d0) 感谢 @vincentkoc。 - Matrix 用户现在会看到在强制跨签名重置可以继续之前需要活跃的恢复密钥，防止第二次重置导致加密恢复和房间密钥备份不可用。[#95720](https://github.com/openclaw/openclaw/pull/95720) 相关 [#78396](https://github.com/openclaw/openclaw/issues/78396)。感谢 @jteddy，@vincentkoc，@xialonglee。 #### Slack 路由器中继模式 [Slack 路由器中继模式](https://docs.openclaw.ai/channels/slack)为托管和多网关部署提供了一种受支持的方式来集中传入的 Slack 流量，同时保留网关对 @提及、话题和回复的所有权。 - 托管 Slack 部署现在可以使用中央路由器将 @提及和进行中的话题发送到正确的 OpenClaw 网关，同时回复仍然通过 Slack 出现。[#94707](https://github.com/openclaw/openclaw/pull/94707) 感谢 @pash-openai，@sjf-oa。 #### Raft 外部智能体唤醒桥接 [Raft 渠道](https://docs.openclaw.ai/channels/raft)和 [Raft 插件](https://docs.openclaw.ai/plugins/reference/raft)现在支持用于外部智能体的本地 CLI 唤醒路径。 - Raft 外部智能体操作员现在可以在工作区有排队的待处理工作时通过支持的本地 CLI 桥接唤醒 OpenClaw 智能体，支持命名配置文件和检查缺少的 CLI 先决条件。[#95497](https://github.com/openclaw/openclaw/pull/95497) 感谢 @vincentkoc。 #### 官方插件安装和修复 [插件管理](https://docs.openclaw.ai/plugins/manage-plugins)和[插件清单](https://docs.openclaw.ai/plugins/plugin-inventory)现在通过正常的安装、更新和修复工作流涵盖了更多官方集成。 - 当 `plugins.allow` 使用渠道或软件包名称而不是真实插件 ID 时，启动指南现在会识别不匹配的条目并显示发现的插件 ID，以便纠正配置。[#68389](https://github.com/openclaw/openclaw/pull/68389) 相关 [#68352](https://github.com/openclaw/openclaw/issues/68352)。感谢 @aym9999，@jirboy，@lyfuci，@pahuchi-joe，@zmxccxy。 - 首次或全新安装的插件信任警告现在包含一个可直接复制的 `plugins.allow` 示例以及用于列出或检查插件 ID 的命令，以便用户可以在信任或重新安装插件代码之前解决警告。[#78105](https://github.com/openclaw/openclaw/pull/78105) 相关 [#68780](https://github.com/openclaw/openclaw/issues/68780)。感谢 @jirboy，@pahuchi-joe。 - Codex 迁移现在适用于标准的全局插件安装，因为 `openclaw migrate` 可以找到已安装的提供商，而不是因 `Unknown migration provider` 而失败。[#89612](https://github.com/openclaw/openclaw/pull/89612) 相关 [#89609](https://github.com/openclaw/openclaw/issues/89609)。感谢 @mugabuga，@zerone0x。 - 插件安装和更新现在可以从 OpenClaw 管理的过时依赖固定中恢复，而不是因 `npm EOVERRIDE` 而失败，同时后续的同步不会降级或移除用户显式安装的包。[#91786](https://github.com/openclaw/openclaw/pull/91786) 相关 [#91772](https://github.com/openclaw/openclaw/issues/91772)。感谢 @amknight，@mkdelta221。 - 渠道插件开发者现在可以通过钩子和选定的执行工作流携带原生发送者和对话标识符，为集成提供更精确的路由，而不会破坏现有的发送者和聊天字段。[#91903](https://github.com/openclaw/openclaw/pull/91903) 感谢 @lanzhi-lee，@vincentkoc。 - 插件发现现在在启动期间重复更少的阻塞文件系统检查，减少了捆绑插件树的可避免冷启动工作量，尤其是在较慢的 Windows 文件系统上，而不会改变捆绑发现行为。[#93919](https://github.com/openclaw/openclaw/pull/93919) 相关 [#76209](https://github.com/openclaw/openclaw/issues/76209)。感谢 @ml12580，@shenhonglong456-ai。 - 插件网关方法现在在注册后可以通过 `openclaw gateway call` 使用，因此插件作者可以从脚本和 cron 作业中使用它们，而不会遇到 `unknown method` 错误。[#94154](https://github.com/openclaw/openclaw/pull/94154) 相关 [#94127](https://github.com/openclaw/openclaw/issues/94127)。感谢 @brycemurray，@pick-cat，@vincentkoc。 - ClawHub 技能发现和安装检查不太可能停滞或崩溃 OpenClaw，因为过大或停滞的市场响应现在会在它们耗尽内存之前被停止。[#95226](https://github.com/openclaw/openclaw/pull/95226) 感谢 @alix-007。 - 当操作员在升级后遵循 `openclaw doctor` 或深层网关状态中的修复建议时，固定的官方插件不再停留在旧版本上。[#95541](https://github.com/openclaw/openclaw/pull/95541) 感谢 @ooiuuii，@vincentkoc。 - 托管的 npm 插件更新不太可能因缺失模块错误而破坏正在运行的网关上的工作，因为旧的插件文件仍然可用，直到后续的网关启动将它们清理掉。[#95589](https://github.com/openclaw/openclaw/pull/95589) 感谢 @ooiuuii，@vincentkoc。 - 受支持品牌的官方插件卡片现在在 ClawHub 和其他目录中显示可识别的图标，并且插件作者可以通过文档化的清单字段提供市场艺术品。[#95845](https://github.com/openclaw/openclaw/pull/95845) 感谢 @patrick-erichsen。 - ClawHub 和其他目录中的官方插件图标不再被强制使用相同的硬编码颜色，允许 Simple Icons 使用其默认艺术作品。[#95987](https://github.com/openclaw/openclaw/pull/95987) 感谢 @patrick-erichsen。 - Docker 用户现在在 GHCR 旁边有了一个官方的 `openclaw/openclaw` Docker Hub 镜像，带版本号的测试版发布不会移动稳定的 `latest` 和 `main` 别名。[#97122](https://github.com/openclaw/openclaw/pull/97122) 感谢 @vincentkoc。 - 基于 Git 的 OpenClaw 安装现在会使用仓库固定的 pnpm 版本，即使存在另一个全局 pnpm 或周围的项目包管理器，因此安装命令不会再针对错误的包管理器环境运行。[bd74a62](https://github.com/openclaw/openclaw/commit/bd74a62118aa4774706359d9494116ded8c1f6e3) 感谢 @vincentkoc。 - ClawHub 技能卡片和更新请求现在即使在接收到异常大的超时值时也能预测性地完成或超时。[8cd0c11](https://github.com/openclaw/openclaw/commit/8cd0c11227f6f4096d089cd6108d6f2ae31252b7) 感谢 @vincentkoc。 - Windows 用户可以完成源码安装，而无需 llama.cpp 设置步骤阻塞或减慢速度，并且安装程序会在之后恢复其现有的 shell 设置。[ea9065b](https://github.com/openclaw/openclaw/commit/ea9065bc68dd4ff94495b85a7dcb4491cf41b67a) 感谢 @vincentkoc。 - 更多官方渠道、提供商和网页搜索插件现在可以通过正常的外部包目录安装或修复，同时仍能从其现有凭据中被识别。[#95683](https://github.com/openclaw/openclaw/pull/95683) 感谢 @vincentkoc。 ### 渠道与消息 #### 其他渠道修复 额外的 [Telegram](https://docs.openclaw.ai/channels/telegram) 和渠道配置修复涵盖了较窄的投递和设置问题。 - Telegram 回复链在上下文更改后会保持缓存的回复关联，而不是在重用这些缓存的回复时失败。[#82909](https://github.com/openclaw/openclaw/pull/82909) 感谢 @lidge-jun。 - 修复了 Discord 在关闭的代码块接近 2，000 字符消息限制时丢弃包含代码块的整个长回复的问题。[#95661](https://github.com/openclaw/openclaw/pull/95661) 感谢 @ly-wang19。 - Slack 操作员现在可以将令牌和签名密钥存储为受支持的 SecretRef 输入，同时读取、写入、允许列表和目标查找以及设置检查使用解析后的凭据，而不是拒绝或误读引用。[7da955f](https://github.com/openclaw/openclaw/commit/7da955fae4ca2083599aa33a1f93dbfff53cb187) 感谢 @vincentkoc。 - 渠道能力检查现在会在集成停止响应时返回明确的超时，防止故障排查在终端或自动化运行中挂起。[8ecdb97](https://github.com/openclaw/openclaw/commit/8ecdb97b636e4c3fcc6d142d217327404ae06581) 感谢 @vincentkoc。 ### 提供商与模型 #### 其他提供商和模型修复 额外的[模型配置](https://docs.openclaw.ai/providers/models)修复改进了选择、使用量报告、状态输出和流式传输。 - 更改默认模型后，使用 `/new` 或 `/reset` 启动新的渠道会话现在会使用新的默认值，而不是静默重用之前缓存的模型，同时显式的 `/model` 覆盖保持不变。[#77339](https://github.com/openclaw/openclaw/pull/77339) 相关 [#77322](https://github.com/openclaw/openclaw/issues/77322)。感谢 @mjamiv，@zaynl。 - 在 HTTP 或 HTTPS 代理后面，`openclaw status --usage --json` 和控制界面中的 Codex/OpenAI 使用量配额检查现在会获取使用量窗口，而不是在 chatgpt.com 无法直接访问时失败。[#93943](https://github.com/openclaw/openclaw/pull/93943) 相关 [#78714](https://github.com/openclaw/openclaw/issues/78714)。感谢 @tnzgit，@turbotheturtle。 - `/status` 现在将活跃模型以及如何清除固定选择保持在一行紧凑显示中，因此 Discord 和其他聊天用户可以浏览模型状态而无需多行解释。[#95797](https://github.com/openclaw/openclaw/pull/95797) 感谢 @solvely-colin。 - Anthropic 流式响应现在将交错的文本、思考和工具调用更新保持关联到正确的响应块，而不是在多个块同时激活时混合它们。[#96013](https://github.com/openclaw/openclaw/pull/96013) 感谢 @vincentkoc。 ### 记忆、会话与状态 #### 其他会话和记忆修复 额外的[记忆](https://docs.openclaw.ai/concepts/memory)修复涵盖了恢复、索引、同步和清理边缘案例。 - 记忆维基状态卡片和基于桥接的源同步在同时进行页面重写时不太可能失败，因为 OpenClaw 现在会重试短暂的路径不匹配，同时仍然停止不安全或持久的文件系统写入。[#94443](https://github.com/openclaw/openclaw/pull/94443) 相关 [#92134](https://github.com/openclaw/openclaw/issues/92134)。感谢 @cknzraposo，@zengwen-dt。 - 修复了最近会话恢复为具有长工作区路径的用户打开新对话而不是返回其现有记录的问题。[#94578](https://github.com/openclaw/openclaw/pull/94578) 相关 [#94577](https://github.com/openclaw/openclaw/issues/94577)。感谢 @rohitjavvadi，@vincentkoc。 - 记忆维基现在会在现有源页面被重新摄取或同步时保留用户编写的笔记，同时仍然刷新其生成的内容。[#95614](https://github.com/openclaw/openclaw/pull/95614) 感谢 @yetval。 - 修复了记忆维基在保险库存储在工作区记忆文件夹内时反复将其自身生成的源页面复制回自身的问题，避免了重复文件、重复清理和不必要的记忆索引增长。[#95666](https://github.com/openclaw/openclaw/pull/95666) 相关 [#95657](https://github.com/openclaw/openclaw/issues/95657)。感谢 @johannes0402，@turbotheturtle，@vincentkoc。 - 对于使用记忆核心做梦的活跃记忆操作员，夜间做梦作业不再启动不必要的回忆工作并遇到 45 秒超时，同时常规网页聊天继续接收记忆回忆。[#95721](https://github.com/openclaw/openclaw/pull/95721) 相关 [#78500](https://github.com/openclaw/openclaw/issues/78500)。感谢 @vincentkoc，@vishutdhar，@xialonglee。 - 具有许多工具调用的智能体会话会用更少的重复工作修复乱序结果，同时保持每个结果与正确的工具调用配对。[#96014](https://github.com/openclaw/openclaw/pull/96014) 感谢 @vincentkoc。 - Windows 上基于 qmd 的记忆工作现在会在可用性探测和命令超时后停止所有相关进程，防止 qmd 子进程继续在后台运行。[830691b](https://github.com/openclaw/openclaw/commit/830691b2010bd0406399adb8a2e97e0b043e2ca8) ### 网关、安全与信任 #### 重启和就绪恢复 [网关健康](https://docs.openclaw.ai/gateway/health)、[故障排除](https://docs.openclaw.ai/gateway/troubleshooting)和 [`openclaw agent`](https://docs.openclaw.ai/cli/agent) 工作流现在能更清晰地报告更多运行时故障并从中恢复。 - 当配置或显式的远程网关缓慢但可达时，`openclaw gateway probe --timeout ...` 现在会等待请求的超时时间，而不是在较短的内部截止时间后报告它不可达。[#89859](https://github.com/openclaw/openclaw/pull/89859) 相关 [#65355](https://github.com/openclaw/openclaw/issues/65355)。感谢 @hellocli，@mushuiyu886。 - 长时间或并行的内部子智能体运行现在会避免不必要的实时预览处理，同时可见的子智能体会话仍然显示实时更新和最终响应。[#91906](https://github.com/openclaw/openclaw/pull/91906) 感谢 @lanzhi-lee，@vincentkoc。 - ACP 对话，特别是 Kiro 支持的话题，现在会在后端无法再恢复旧会话时通过启动新会话来继续第一条回复之后的内容。[#93547](https://github.com/openclaw/openclaw/pull/93547) 相关 [#87830](https://github.com/openclaw/openclaw/issues/87830)。感谢 @amersheeny，@chouzz。 - 当 Linux 内存压力杀死子命令或会话时，由 systemd 管理的 OpenClaw 网关现在会保持运行并保持渠道连接存活，同时报告子进程失败。[#93585](https://github.com/openclaw/openclaw/pull/93585) 感谢 @snowzlm。 - 在工具工作中取消 OpenClaw 运行现在会立即结束它，而不是开始另一个模型回合或保持会话锁定。[#94412](https://github.com/openclaw/openclaw/pull/94412) 感谢 @szsip239，@vincentkoc。 - 使用云模型的预定 OpenClaw 作业现在默认从静默、卡住的模型调用中恢复，帮助防止后续 cron 工作积压，同时本地或自托管提供商保持其现有的超时行为。[#94445](https://github.com/openclaw/openclaw/pull/94445) 感谢 @bek91。 - 网关就绪检查现在会在重启排空期间变为不健康，防止流量管理器向暂时拒绝请求的网关发送新工作。[#94915](https://github.com/openclaw/openclaw/pull/94915) 相关 [#78136](https://github.com/openclaw/openclaw/issues/78136)。感谢 @markoub，@maxschachere，@vincentkoc。 - Mac 用户现在可以在 OpenClaw 升级期间保持由 LaunchAgent 管理的网关运行，而不是在看到较旧的文本转换运行时代码仍被缓存时反复崩溃和重启循环。[#95081](https://github.com/openclaw/openclaw/pull/95081) 相关 [#95057](https://github.com/openclaw/openclaw/issues/95057)。感谢 @849261680，@yveslarose。 - Codex 驱动的对话在 TUI、WebChat 和兼容的流式 API 中现在会在写入时显示回复，同时干净地替换临时文本，使最终答案不包括过时的草稿。[#95404](https://github.com/openclaw/openclaw/pull/95404) 相关 [#95422](https://github.com/openclaw/openclaw/issues/95422)。感谢 @agonza1，@vincentkoc。 - 网关重启后，用户不再会在 OpenClaw 已经恢复中断的回复或报告实际恢复失败时看到误导性的重试通知，减少了不必要的重复尝试。[#95431](https://github.com/openclaw/openclaw/pull/95431) 感谢 @moeedahmed，@vincentkoc。 - 长时间、工具密集型的智能体会话现在会在结果积累时保留提示缓存节省，减少因在回合之间重写历史而产生的可避免延迟和成本。[#95624](https://github.com/openclaw/openclaw/pull/95624) 感谢 @vincentkoc。 - 网关重启不再使配置的 Codex、Copilot 或受信任的基于插件的智能体暂时不可用，并且不受信任的工作区插件仍然被阻止激活自身。[#95652](https://github.com/openclaw/openclaw/pull/95652) 感谢 @vincentkoc。 - 长响应、繁忙的工具流、图像密集型请求和记忆回忆现在会产生更少的 CPU 和文件系统开销，无需更改设置或工作流。[#95697](https://github.com/openclaw/openclaw/pull/95697) 感谢 @vincentkoc。 - 操作员现在可以在无配置设置中为已知渠道添加或更新预定公告，同时配置的环境仍然会在投递前拒绝已禁用、过期、无所有者或未知的目的地。[#95754](https://github.com/openclaw/openclaw/pull/95754) 感谢 @vincentkoc。 - macOS 用户在停止或更新受管网关时不太可能看到虚假的端口冲突失败，因为 OpenClaw 会短暂等待正常关机以释放端口，同时仍然报告持续存在的冲突。[#95886](https://github.com/openclaw/openclaw/pull/95886) 感谢 @fuller-stack-dev。 - Copilot 支持的智能体现在可以通过 OpenClaw 的正常聊天回复流程向用户提问并接受答案，同时紧凑的工具搜索和代码模式控制避免将完整工具目录加载到会话中。[#96005](https://github.com/openclaw/openclaw/pull/96005) 感谢 @vincentkoc。 - systemd 或容器设置上的网关重启不再留下旧的 Codex 或 Claude 适配器进程，帮助新的 ACPX 会话在多次重启后无需数分钟的清理停滞即可启动。[#96032](https://github.com/openclaw/openclaw/pull/96032) 感谢 @t2wei，@vincentkoc。 - Copilot 支持的会话现在会在工作展开时显示计划更新，并且其原生子任务在完成或失败时仍然可见，而不是从 OpenClaw 的任务视图中消失。[#96062](https://github.com/openclaw/openclaw/pull/96062) 感谢 @vincentkoc。 - 诸如 OpenCode 之类的连接智能体现在即使其框架无法选择请求的模型也能通过 OpenClaw 启动，而真正不支持的模型选择仍然返回原始错误。[#96068](https://github.com/openclaw/openclaw/pull/96068) 相关 [#95869](https://github.com/openclaw/openclaw/issues/95869)。感谢 @sabatech-dev，@vincentkoc。 - 使用 `heartbeat_prompt_contribution` 的插件现在会在智能体通过框架运行时（如 Codex 应用服务器）将其特定心跳上下文传递给模型，而不会影响普通的用户回合或不使用该钩子的插件。[#96233](https://github.com/openclaw/openclaw/pull/96233) 感谢 @azogheb，@vincentkoc。 - Windows 网关清理和监听器检查现在会一致地处理 UTF-16 WMIC 命令行数据，减少正在运行的网关进程的失败或冲突识别。[15c880a](https://github.com/openclaw/openclaw/commit/15c880aeff1f4e55964ad7204d14733a2d1362f7) 感谢 @vincentkoc。 - 长上下文、工具密集型的智能体会话现在在重复回合之间保持更稳定的提示缓存重用，而不会丢失每个结果的大小限制，同时高级操作员可以为大型上下文模型配置更大的工具结果上限，而无需配置拒绝。[a60947f](https://github.com/openclaw/openclaw/commit/a60947fb3e92f45ea7eb2581da8877b10a8bebb2) 感谢 @vincentkoc。 - 长、工具密集型的智能体会话随着工具输出的积累，不太可能使模型请求膨胀，同时重复回合保持稳定的提示缓存重用。[2f33999](https://github.com/openclaw/openclaw/commit/2f3399989893e6af18be49fb810e58941d7a4a45) 感谢 @vincentkoc。 - 网关重启现在使用 OpenClaw 的持久状态数据库进行交接，而过时的、格式错误的、错误进程的或被取代的请求在影响重启之前被丢弃。[0ad48da](https://github.com/openclaw/openclaw/commit/0ad48dad2c4747f255d5a156b94578f7d87386a1) - 网关状态、医生和重启诊断现在在 OpenClaw 的共享状态数据库中保留最近的重启详情，而过期或格式错误的记录仍然被丢弃。[a39a3b7](https://github.com/openclaw/openclaw/commit/a39a3b74de05f06227ede904a73c1b4687679d3e) 感谢 @vincentkoc。 - 网关重启和托管服务更新现在会保持正确的继续消息，避免重用过时的交接状态，并一致地标记失败的更新交接。[514b336](https://github.com/openclaw/openclaw/commit/514b3365b54c8b3493eaf8a94198b7c04ea34aec) 感谢 @vincentkoc。 - 网关启动的智能体不再在诸如实时 cron 检查之类的任务中丢失仅限所有者的 OpenClaw 工具，因此授权操作可以使用正确的请求上下文使用预期工具。[c2ee9b0](https://github.com/openclaw/openclaw/commit/c2ee9b0be8aeeadedffc8c6aaa9f5f291283fea5) 感谢 @vincentkoc。 - 格式错误的网关重启请求现在会明确失败而不会安排重启，防止错误或意外的集成调用意外重启网关。[108d6d7](https://github.com/openclaw/openclaw/commit/108d6d7eca0000a736e28d198e77100f4d7774e5) 感谢 @vincentkoc。 - 网关排队的过期节点请求现在会自动过期，因此旧工作不太可能存留并影响后续活动。[f6d432e](https://github.com/openclaw/openclaw/commit/f6d432e545e2e2be91d17badc48354a0135e5294) - 当集成请求过多的子智能体会话历史时，插件工作流不太可能停滞或过载网关，因为每次读取现在都被限制在安全上限内。[b66b450](https://github.com/openclaw/openclaw/commit/b66b4504f87205dd8ba0393e763e23a4a6158a79) 感谢 @vincentkoc。 - 图像描述现在通过将其上限限制在安全的运行时限制内来一致地处理极大超时设置，而不是冒定时器溢出的风险。[88b21fc](https://github.com/openclaw/openclaw/commit/88b21fc30b4ac615b3d3870e483190bc832f9846) - 内嵌智能体会话现在即使在极大超时下也能可靠地等待另一个会话释放其文件锁，而不是冒定时器溢出的风险。[4c736df](https://github.com/openclaw/openclaw/commit/4c736df975fed8e39f18db43bfaea9654b9ac0c7) 感谢 @vincentkoc。 - 具有极大任务超时设置排队的命令现在会可靠地超时，因为 OpenClaw 将等待时间上限限制在运行时的安全最大值内。[1f6ae32](https://github.com/openclaw/openclaw/commit/1f6ae32cabb9d5ed308bb30715c287936ef483f1) 感谢 @vincentkoc。 - 修复了规范化核心暴露错误的字符串强制入口点以及 ACP 会话显示过时快速模式值的问题，使集成获得预期的 API，用户看到实际生效的模式。[93ad397](https://github.com/openclaw/openclaw/commit/93ad39772590a34be3821d83709a5050c186fdf9) 感谢 @vincentkoc。 - 快速自动运行现在更一致地投递最终回复，进度重置处理仅限于自动模式，因此不会干扰响应或转发的回调。[9e8ab08](https://github.com/openclaw/openclaw/commit/9e8ab083dd6b9df4dadf40ee523ff18ac1472bd3) 感谢 @vincentkoc。 - 使用 OpenAI Responses 的智能体会话现在会在重播的历史包含不匹配的工具请求和结果时恢复基于工具的工作，而不会失败或丢失进度。[b4bc1f2](https://github.com/openclaw/openclaw/commit/b4bc1f20c9fca6d316561d42cbabdd793c67a6e7) 感谢 @vincentkoc。 - 完成的插件子智能体和 QA 运行不再在网关返回替代完成信封形状时被误报为失败，使成功交接对插件作者和操作员更可靠。[d1b268f](https://github.com/openclaw/openclaw/commit/d1b268f7f7f7309dd5db99728019218b4d453e18) 感谢 @vincentkoc。 - 修复了完成的子智能体任务有时无更新地结束的问题，因此用户会收到结果或父智能体的下一步。[68a1e00](https://github.com/openclaw/openclaw/commit/68a1e00b73bd746f6fb577f4127fecb0ade9e228) 感谢 @vincentkoc。 #### 远程结果和媒体投递 额外的[网关](https://docs.openclaw.ai/gateway/index)修复更可靠地将远程媒体和完成的子智能体结果返回给活跃对话。 - 来自远程 Codex 应用服务器生成的图像现在作为附件到达，而不是显示 `Media failed` 或在成功生成后仅返回文本。[#96212](https://github.com/openclaw/openclaw/pull/96212) 感谢 @sjf-oa。 - 当子智能体完成时，其结果现在更可靠地到达活跃的父运行，而不是看似静默。[7fc4bbc](https://github.com/openclaw/openclaw/commit/7fc4bbc0bcbabc2aa99b1fd51e77099f2b26f4e1) 感谢 @vincentkoc。 #### 其他安全和信任修复 额外的[秘密处理](https://docs.openclaw.ai/gateway/secrets)强化将受信任的包路径保持在预期边界内。 - 受信任的 OpenClaw 包源现在会拒绝外观相似的兄弟路径，因此信任 `/artifactory/openclaw` 不再同时允许 `/artifactory/openclaw-malicious` 等路径。[12c34fc](https://github.com/openclaw/openclaw/commit/12c34fc3a95121f5a15f01c3f971a1bc5b0fe6f9) 感谢 @vincentkoc。 ### 客户端与界面 #### 客户端发送和重连 [WebChat](https://docs.openclaw.ai/web/webchat)、[控制界面](https://docs.openclaw.ai/web/control-ui)、移动客户端和[终端界面](https://docs.openclaw.ai/cli/tui)现在可以恢复已完成、被拒绝或中断的发送和重连，而不会让对话看起来卡住。 - 当 WebChat 消息在智能体开始之前失败时，WebChat 和控制界面现在会将会话显示为失败，而不是让它看起来仍在运行。[#84352](https://github.com/openclaw/openclaw/pull/84352) 感谢 @jesse-merhi。 - 修复了控制界面会话选择器卡在隐藏的子智能体会话后面的问题，使“加载更多”到达下一个可用聊天，而不会显示误导性的总数。[#89323](https://github.com/openclaw/openclaw/pull/89323) 相关 [#89249](https://github.com/openclaw/openclaw/issues/89249)。感谢 @giodl73-repo，@originsecured-do。 - 当用户从历史记录中重新打开控制界面对话时，他们的提示现在会与助手的回复一起出现，保留了问答上下文，而不会在长记录中出现空白间隙。[#93841](https://github.com/openclaw/openclaw/pull/93841) 相关 [#90241](https://github.com/openclaw/openclaw/issues/90241)。感谢 @mushuiyu886，@pronzcw。 - 位于路径前缀后面的控制界面部署现在将清单、favicon 和服务工作者请求保持在该前缀下，避免登录后令人困惑的根级别 403 错误。[#94204](https://github.com/openclaw/openclaw/pull/94204) 相关 [#94157](https://github.com/openclaw/openclaw/issues/94157)。感谢 @hugenshen，@xrow。 - Android 用户现在可以在设置中打开健康日志和技能行以获取可读的详细信息，使故障排查网关活动、检查技能设置和状态以及了解如何与现有设置代码配对更加容易。[#95148](https://github.com/openclaw/openclaw/pull/95148) 感谢 @tosko4。 - 已发送的提示不再在发送后重新出现在控制界面编辑器中，因此用户可以切换会话或开始下一条消息，而不会冒重复发送或覆盖新草稿的风险，同时有意的重新输入仍然有效。[#95503](https://github.com/openclaw/openclaw/pull/95503) 相关 [#89466](https://github.com/openclaw/openclaw/issues/89466)。感谢 @vincentkoc，@zhangguiping-xydt，@zhong18804784882。 - Android 用户现在会得到一个更清晰的概览，连接状态、配置的智能体、节点健康、批准、最近会话以及聊天和通话操作一目了然。[#95557](https://github.com/openclaw/openclaw/pull/95557) 感谢 @joshavant，@solvely-colin。 - Android 用户现在可以从应用内的批准屏幕刷新和解决网关命令批准，在连接时选择允许一次、始终允许或拒绝。[#95593](https://github.com/openclaw/openclaw/pull/95593) 感谢 @solvely-colin。 - iOS 用户现在避免意外的通知提示，并在批准警报不可用时获得清晰指导，权限管理在一个可预测的设置屏幕中。[#95640](https://github.com/openclaw/openclaw/pull/95640) 感谢 @joshavant。 - 本地 TUI 关机现在即使在 `OPENCLAW_TUI_LOCAL_RUN_SHUTDOWN_GRACE_MS` 设置极高时也保持在安全的定时器限制内。[c21dcfc](https://github.com/openclaw/openclaw/commit/c21dcfc7c272201484514bbc096bd51ab112bd47) - Canvas A2UI 现在在每次构建后只提供当前的应用资产，因此过时的兼容性图像和遗留文件不太可能出现。[a89e65c](https://github.com/openclaw/openclaw/commit/a89e65c167f3280b047baff931f52e393875c892) 感谢 @vincentkoc。 - iOS 推送中继设置失败更容易定位，因为注册诊断会显示设置在哪里停止，同时将敏感的推送凭据排除在日志之外。[f2b8668](https://github.com/openclaw/openclaw/commit/f2b8668a549b50339a96e47020a12615734640a9) 感谢 @joshavant。 - iOS 设备现在只有在用户接受托管中继披露并允许通知后才注册推送通知，防止在同意之前发布注册数据。[8efed50](https://github.com/openclaw/openclaw/commit/8efed50c4ed33105cfed9f2f96532fc9fe2d7e6d) 感谢 @joshavant。 - 修复了聊天、语音、TUI 和转发的发送有时在网关已经完成或拒绝它们后看似卡住或消失的问题，现在受影响的客户端会清除待处理状态、恢复可重试的输入、刷新历史或显示有用的失败信息。[#91049](https://github.com/openclaw/openclaw/pull/91049) 相关 [#91048](https://github.com/openclaw/openclaw/issues/91048)。感谢 @nxmxbbd。 #### 其他客户端界面修复 额外的[控制界面](https://docs.openclaw.ai/web/control-ui)、移动和桌面修复改进了显示准确性、可访问性、入门和应用行为。 - 恢复了扩展的控制界面聊天侧边栏中的 OpenAI/Codex 使用量配额，因此用户无需离开对话即可检查其限制。[#94219](https://github.com/openclaw/openclaw/pull/94219) 相关 [#93041](https://github.com/openclaw/openclaw/issues/93041)。感谢 @jazzroutine，@pick-cat。 - iOS 屏幕现在在入门、设置、聊天、批准提示、语音权限、小部件和共享聊天视图中使用一致的 OpenClaw 强调色和状态颜色。[#94627](https://github.com/openclaw/openclaw/pull/94627) 感谢 @zats。 - 控制界面现在可以通过主会话中的“预设为静默”且不通知来创建早间作业，而不是让对话框保持打开且无明显结果。[#95459](https://github.com/openclaw/openclaw/pull/95459) 相关 [#95073](https://github.com/openclaw/openclaw/issues/95073)。感谢 @vincentkoc，@vporton，@zoowh。 - 在百万 token 边界处，控制界面徽章和使用量读数现在显示 "1M" 而不是令人困惑的 "1000k"，同时底层 token 计数保持不变。[#95485](https://github.com/openclaw/openclaw/pull/95485) 感谢 @narahariraghava，@vincentkoc。 - 控制界面概览现在只计数和标记需要关注的已启用 cron 作业，而禁用的作业保留其过去的失败详情而不作为当前问题出现。[#95723](https://github.com/openclaw/openclaw/pull/95723) 相关 [#95716](https://github.com/openclaw/openclaw/issues/95716)。感谢 @voytas75，@zengwen-dt。 - 控制界面用户现在会看到更短的系统、亮色和暗色主题工具提示，同时屏幕阅读器会宣布更少的重复标签，而不会丢失周围的颜色模式上下文。[#95837](https://github.com/openclaw/openclaw/pull/95837) 感谢 @hannesrudolph，@sannidhyasah。 - 从表单视图切换后，原始配置不再显示为缺失，因为 JSON 会被带回视图中，而不是保留之前的滚动位置。[#96145](https://github.com/openclaw/openclaw/pull/96145) 相关 [#94202](https://github.com/openclaw/openclaw/issues/94202)。感谢 @sunlit-deng，@vporton。 - 新的 iOS 用户在 iOS 请求本地网络访问之前会先到达 OpenClaw 的欢迎和入门页面，而现有用户在打开网关设置或需要局域网网关发现时仍然会收到请求。[#96181](https://github.com/openclaw/openclaw/pull/96181) 感谢 @joshavant。 ### 插件与打包 #### 其他插件和包修复 额外的[插件管理](https://docs.openclaw.ai/plugins/manage-plugins)修复改进了安装、验证、包解析和更新。 - 通过 OpenClaw 可信目录安装或更新官方 Yuanbao 渠道插件的用户现在会获得版本 2.15.0，具有预期的完整性检查和与该版本一致的缺失插件指导。[#94470](https://github.com/openclaw/openclaw/pull/94470) 感谢 @jase-283。 - 首次运行入门现在可以通过 Homebrew 安装捆绑的 `gog` 技能，而不会由于已删除的第三方 tap 公式而失败。[#95019](https://github.com/openclaw/openclaw/pull/95019) 相关 [#95017](https://github.com/openclaw/openclaw/issues/95017)。感谢 @sedrak-hovhannisyan，@vincentkoc，@zengwen-dt。 - Canvas、Discord、Slack、语音通话和 WhatsApp 用户在每个已安装或捆绑的插件中保持相同的技能指导，而对先前根目录 `skills/...` 路径的引用需要移到相关插件目录中。[#95664](https://github.com/openclaw/openclaw/pull/95664) 感谢 @vincentkoc。 - ClawHub 技能验证现在接受与安装和更新相同的 `@owner/<slug>` 引用，因此用户无需切换到模糊的裸 slug 即可检查预期的发布者。[#95992](https://github.com/openclaw/openclaw/pull/95992) 感谢 @patrick-erichsen。 - OpenClaw 安装时的包管理器警告现在正确识别 npm、Yarn、Yarn Berry 和 Corepack 风格的启动器，避免当这些工具通过替代可执行名称运行时出现误导性指导。[11a2e03](https://github.com/openclaw/openclaw/commit/11a2e03bd4deda748336553710cb6426d448d952) 感谢 @vincentkoc。 - 包 URL 安装现在可以处理过大的下载超时，而不会在可用的包数据解析之前失败。[c310f8c](https://github.com/openclaw/openclaw/commit/c310f8cfa4524453c7082bb5aab642c9decc6e99) - 插件和 CLI 开发者现在会看到更清晰的受支持命令格式化 API，同时设备配对、节点注册和医生指导继续生成相同的 shell 安全命令。[23b4f33](https://github.com/openclaw/openclaw/commit/23b4f33195933ff4def4609d970f073293760683) - Windows ARM64 用户现在在通过 x64 模拟的 shell 运行 PowerShell 安装程序时会获得匹配的 ARM64 Node 和 MinGit 下载。[fac091b](https://github.com/openclaw/openclaw/commit/fac091b39de230d9ed90e8412123126cf676004f) 感谢 @vincentkoc。 - 默认的 OpenClaw 安装不再花费时间构建可选的 llama.cpp 支持，避免了未启用它的用户遇到原生构建失败。[cc1b3a8](https://github.com/openclaw/openclaw/commit/cc1b3a8550dd9c29f581799934d140aaf5a84f1c) 感谢 @vincentkoc。 ### 文档和管理工具 #### 设置和命令可靠性 [OpenClaw CLI](https://docs.openclaw.ai/cli/index)、[Tab 补全](https://docs.openclaw.ai/cli/completion)和 [`openclaw doctor`](https://docs.openclaw.ai/cli/doctor) 现在能更干净地处理更多设置和修复案例。 - OpenClaw 的 zsh tab 补全菜单现在会原样显示包含 `$` 变量或反引号包裹示例的选项描述，而不是将其作为 shell 输入评估并损坏菜单。[#64490](https://github.com/openclaw/openclaw/pull/64490) 感谢 @edenkangdw。 - 从较旧的沙箱存储升级后，操作员现在会从 `openclaw doctor` 收到关于遗留注册表文件的明确警告，并可以使用 `openclaw doctor --fix` 迁移或清理它们。[#84326](https://github.com/openclaw/openclaw/pull/84326) 感谢 @giodl73-repo。 - 操作员现在可以使用 `doctor --lint` 发现过时的遗留网关服务并预览清理，而有意添加的额外服务仍然只是信息性提示，不会使默认检查失败。[#84340](https://github.com/openclaw/openclaw/pull/84340) 感谢 @giodl73-repo。 - macOS 网关操作员现在在重新安装、修复或重启覆盖自定义 LaunchAgent 包装器行为之前会收到警告，同时 `openclaw status` 区分仅 CLI 的缺失秘密检查和已安装的服务。[#90537](https://github.com/openclaw/openclaw/pull/90537) 相关 [#90518](https://github.com/openclaw/openclaw/issues/90518)。感谢 @turbotheturtle，@vincentkoc。 - 长、多行或代码密集型的提示现在可以使用 `--message-file` 发送到 `openclaw agent`，避免了脆弱的 shell 引号，并在发送前报告无效文件。[#93351](https://github.com/openclaw/openclaw/pull/93351) 感谢 @ooiuuii。 - 修复了预定的 `doctor --fix --non-interactive` 修复在临时健康检查失败后重启已经运行的网关的问题，因此无人值守维护不再中断实时服务。[#94148](https://github.com/openclaw/openclaw/pull/94148) 相关 [#78217](https://github.com/openclaw/openclaw/issues/78217)。感谢 @esqandil，@zhangguiping-xydt。 - `openclaw configure` 和裸 `openclaw config` 现在会在从脚本或管道运行时停止并提供清晰的子命令指导，而不是打开部分交互式向导并模糊退出。[#94238](https://github.com/openclaw/openclaw/pull/94238) 相关 [#93953](https://github.com/openclaw/openclaw/issues/93953)。感谢 @nianjiuzst，@ruomuxydt。 - 多智能体操作员现在可以使用 `openclaw gateway usage-cost` 查看一个配置的智能体或所有智能体的成本，同时现有的默认智能体命令保持不变。[#94483](https://github.com/openclaw/openclaw/pull/94483) 感谢 @ly-wang19。 - 已归档的工作板卡片不再杂乱默认的 `openclaw workboard list` 输出，而 `--include-archived` 和 JSON 输出在需要时仍然提供访问。[#94562](https://github.com/openclaw/openclaw/pull/94562) 相关 [#94555](https://github.com/openclaw/openclaw/issues/94555)。感谢 @ecican，@vincentkoc，@zengwen-dt。 - OpenClaw Doctor 现在为工作的隔离 shell 提示 cron 作业提供准确的指导，而不是反复建议一个无法清除警告的 `--fix` 命令。[#94784](https://github.com/openclaw/openclaw/pull/94784) 相关 [#94655](https://github.com/openclaw/openclaw/issues/94655)。感谢 @altaywtf，@geekoagent，@zengwen-dt。 - 修复了 `openclaw doctor` 在有意跳过的就绪检查后为健康的本地 GGUF 记忆设置显示需要修复警告的问题，同时在配置的本地模型确实丢失时保留警告。[#95393](https://github.com/openclaw/openclaw/pull/95393) 相关 [#92582](https://github.com/openclaw/openclaw/issues/92582)。感谢 @mikasa0818，@neekolascmd，@vincentkoc。 - 在 Windows 上，安装程序创建的网关任务现在在后台运行，没有控制台窗口，用户可能意外关闭并停止网关。[#95480](https://github.com/openclaw/openclaw/pull/95480) 相关 [#89231](https://github.com/openclaw/openclaw/issues/89231)。感谢 @cameronweller，@mikasa0818，@vincentkoc。 - 智能体渠道绑定现在会拒绝格式错误的账户规范（例如 `matrix:work:extra`）并显示明确的错误，而不是静默地将智能体路由到不同的账户。[#95572](https://github.com/openclaw/openclaw/pull/95572) 感谢 @ly-wang19。 - ClawHub 技能更新现在会遵守您配置的安装安全策略，并且 `openclaw skills update --all` 只更新已跟踪的 ClawHub 技能，而不是意外安装其他配置的技能。[#95684](https://github.com/openclaw/openclaw/pull/95684) 感谢 @vincentkoc。 - Windows 重启和网关启动工作流更可靠，因为 OpenClaw 现在将命令传递给受信任的系统 `cmd.exe` 路径，而不是依赖进程查找。[7dd01d1](https://github.com/openclaw/openclaw/commit/7dd01d15c56da2ee50f55746ba725d708682fca9) 感谢 @vincentkoc。 - 当 PATH 查找不完整时，Windows 网关清理和监听器检查更可靠，因此操作员可以识别网关进程并释放占用端口，而无需遗漏已安装的系统工具。[e9b694e](https://github.com/openclaw/openclaw/commit/e9b694ef9cd8b7528a76d85b8be2a830b296a5e1) 感谢 @vincentkoc。 - 在 Windows 上，OpenClaw 启动和 TUI Codex 交接现在即使在 PATH 上早期出现另一个 `where` 命令时也能通过受信任的系统定位器找到 `bun`、`codex` 和其他运行时二进制文件。[72b9bc7](https://github.com/openclaw/openclaw/commit/72b9bc730370e3b1155fb231621c2bc7b2c87b56) 感谢 @vincentkoc。 - Windows 端口诊断现在即使在 PATH 条目缺失或被遮蔽时也使用预期的系统工具，因此网关和服务端口冲突不太可能因命令解析失败而被掩盖。[c4facb2](https://github.com/openclaw/openclaw/commit/c4facb2bb372e99037f497e2640ca7bdc5cbc5f6) 感谢 @vincentkoc。 - 当 PATH 不完整、不常见或被遮蔽时，Windows 守护进程恢复不太可能错过进程检测或清理，因为计划任务回退现在在受信任的系统位置中找到 PowerShell 和 taskkill。[2a140e6](https://github.com/openclaw/openclaw/commit/2a140e6e6ae8c48edb5bd52d8d177915f0555d70) 感谢 @vincentkoc。 - OpenClaw 现在通过迁移将其配置恢复标记、最后已知良好的快照和可疑读取历史保留在共享状态中，而不会留下单独的配置健康日志文件。[6daabd2](https://github.com/openclaw/openclaw/commit/6daabd23f821c66154739de4b0f103e33343333c) 感谢 @vincentkoc。 - 在 Windows 上，通过 Node 包垫片启动的 Crabbox 命令现在会按输入接收提供商标志、shell 命令和特殊 shell 字符，而不是丢失或重新解释它们。[54d24cd](https://github.com/openclaw/openclaw/commit/54d24cd956ff91f4fa8c4924f17c06798c1e0359) 感谢 @vincentkoc。 - 针对 Windows 的 Crabbox 工作流在通过 `.cmd` 和 `.bat` 垫片启动 Node 工具时不太可能失败或回退到较慢的 shell 处理。[d48dcc6](https://github.com/openclaw/openclaw/commit/d48dcc664bc6e1106a61942a951745886f22d582) 感谢 @vincentkoc。 - Windows 用户现在可以通过 npm 安装的命令垫片运行 `crabbox`、`git` 和其他基于 Node 的工具，而不会在工具打开之前被 Crabbox 阻止。[77f4e45](https://github.com/openclaw/openclaw/commit/77f4e45c3518751b5f586eac193c4aee904f02d9) 感谢 @vincentkoc。 - `openclaw doctor` 现在会检查省略工具策略设置的配置文件，而不会将有意的省略视为错误。[03ba09b](https://github.com/openclaw/openclaw/commit/03ba09bfa8676832d55bdc7724e79d9980fdd2d7) - `openclaw doctor` 不再在无法评估自定义预览配置文件时显示误导性的工具部分警告。[420a0e6](https://github.com/openclaw/openclaw/commit/420a0e6fce4b2c5339e535e6b307f50df1c00bb2) 感谢 @vincentkoc。 - `openclaw doctor` 现在将预览警告限制为它可以评估的工具配置文件，避免对未知配置文件显示误导性的已配置授权警告。[541f7ff](https://github.com/openclaw/openclaw/commit/541f7ffc6558c0e59a8afca066a9f00884d39b65) 感谢 @vincentkoc。 - Windows 用户可以从源码安装 OpenClaw，而不会因安装程序强制通过 `cmd.exe` 运行 npm 或 pnpm 脚本而导致依赖设置被阻塞。[1252378](https://github.com/openclaw/openclaw/commit/1252378018f899bfc110914bac7cba94b65b8930) 感谢 @vincentkoc。 #### 工具和计划工作 [预定作业](https://docs.openclaw.ai/cli/cron)和[工具工作流](https://docs.openclaw.ai/tools/index)现在更一致地完成、验证和报告失败。 - 使用 `deleteAfterRun` 的隔离 cron 作业现在在完成后会删除其临时会话和记录，包括禁用了投递的运行，减少了过时文件、累积的上下文和手动清理。[#84794](https://github.com/openclaw/openclaw/pull/84794) 相关 [#84707](https://github.com/openclaw/openclaw/issues/84707)。感谢 @bottenbenny，@turbotheturtle。 - 单个预定作业现在可以使用自己的备用模型，禁用回退运行，或通过 CLI 返回正常的回退继承，而无需操作员编辑底层载荷数据。[#93369](https://github.com/openclaw/openclaw/pull/93369) 相关 [#90302](https://github.com/openclaw/openclaw/issues/90302)。感谢 @849261680，@walliiee。 - Cron 历史记录现在可靠地找到作业 ID 包含额外前后空格的条目，并在可能创建无法安全读回的日志记录之前拒绝嵌套或空白的 ID。[#93567](https://github.com/openclaw/openclaw/pull/93567) 感谢 @alix-007，@vincentkoc。 - 添加或删除 cron 作业不再导致另一个已经到期的定期作业丢失其待处理运行。[#94323](https://github.com/openclaw/openclaw/pull/94323) 感谢 @yetval。 - Word、PowerPoint 和 Excel 文档的读取和写入现在使用预期的 `.docx`、`.pptx` 或 `.xlsx` 路径，而不是针对虚构的扩展名失败。[#95805](https://github.com/openclaw/openclaw/pull/95805) 相关 [#93326](https://github.com/openclaw/openclaw/issues/93326)。感谢 @bhnan，@lzyyzznl，@vincentkoc，@xzh-icenter。 - 浏览器自动化用户保持相同的丰富引用快照，包括紧凑结果中的有用分支，在快照生成期间减少可避免的处理。[#96072](https://github.com/openclaw/openclaw/pull/96072) 感谢 @vincentkoc。 - 修复了在 Windows 上超时的命令和中断的核心更新有时留下子进程运行的问题，因此 OpenClaw 可以在取消、超时或更新清理后更可靠地停止整个进程树。[a192b2e](https://github.com/openclaw/openclaw/commit/a192b2ea52b3166a7d190bf5f60f3feb030306bb) 感谢 @vincentkoc。 - Windows 用户不太可能看到智能体管理的工具安装失败，同时解压 ZIP 下载（如 ripgrep），因为 OpenClaw 现在使用内置的 Windows 解压程序，而不是依赖 PATH 查找。[a5fde91](https://github.com/openclaw/openclaw/commit/a5fde9119c9c50685a392bff078a2f76a03d749d) - Windows 设置和运行时检查现在更可靠地使用受信任的 System32 解析器找到所需工具，而不是依赖意外的 PATH 条目。[d3b4444](https://github.com/openclaw/openclaw/commit/d3b44442f6c8bedd765dc20a06316d3420b5d854) 感谢 @vincentkoc。 - 使用 `timeoutMs: 0` 创建的 SDK 运行现在保持请求的零超时，而不会有不想要的客户端侧看门狗。[2bdcc83](https://github.com/openclaw/openclaw/commit/2bdcc8314d3fce9ee2d0300759cd6a1d9bb45a7d) 感谢 @vincentkoc。 - 卡住的 OpenClaw 命令现在即使调用者提供极大的执行或空闲输出超时也能可靠地停止。[1425bb3](https://github.com/openclaw/openclaw/commit/1425bb3a03189813787194c23b8a38518166005a) 感谢 @vincentkoc。 - 给定极大超时设置的命令现在使用安全最大值，而不是因为运行时无法调度请求的等待而失败。[66b94ba](https://github.com/openclaw/openclaw/commit/66b94ba577b8836a4afa7cb59cd1a749bf2a8d68) - 提供商特定的工具允许/拒绝设置现在与 OpenClaw 的医生警告更一致地对齐，包括具有提供商别名、模型特定键、OpenRouter 风格模型 ID 或格式错误的策略条目的配置。[8f2882f](https://github.com/openclaw/openclaw/commit/8f2882f94affbe5e89994ae175fdaf7304d7b392) #### 其他设置和 CLI 修复 额外的 [`openclaw config`](https://docs.openclaw.ai/cli/config) 和 CLI 修复改进了诊断、配对、启动和命令行为。 - 较大的 OpenClaw 配置现在可以更有效地初始化和生成 UI 提示，同时敏感字段继续以相同方式标记。[#55018](https://github.com/openclaw/openclaw/pull/55018) 感谢 @huangyandi-red，@vincentkoc，@xdhuangyandi。 - 仍然需要手动网关重启的配置更改现在会显示明确的需要重启通知，并保留原始提示，而不是看起来已完成并显示误导性的 `config-patch ok` 消息。[#83041](https://github.com/openclaw/openclaw/pull/83041) 相关 [#46797](https://github.com/openclaw/openclaw/issues/46797)。感谢 @stache73，@xuruiray。 - `doctor`、`gateway`、`models`、`plugins`、`sessions` 和 `tasks` 的帮助现在在数十毫秒内出现，而 `sessions --help` 和 `tasks --help` 等命令之前需要大约 1.6 到 1.8 秒才开始响应。[#89628](https://github.com/openclaw/openclaw/pull/89628) 感谢 @yyzquwu。 - 诸如 Langfuse 之类的 OpenTelemetry 跟踪后端现在会显示实际的提供商/模型名称，而不是对于带斜杠限定的模型 ID 显示 "unknown"。[#89981](https://github.com/openclaw/openclaw/pull/89981) 感谢 @mycarrysun，@vincentkoc。 - 格式错误或较旧的设备配对记录不再阻止 `openclaw devices list` 显示待处理的批准请求，同时有效的角色仍然正常显示。[#93504](https://github.com/openclaw/openclaw/pull/93504) 感谢 @ly-wang19。 - OpenClaw 现在会在 SSH 目标产生可能无效的 SSH 配置或隧道启动失败之前拒绝带有多余前导或尾部冒号的目标，适用于 SSH 支持的沙箱和网关。[#93887](https://github.com/openclaw/openclaw/pull/93887) 感谢 @miorbnli。 - 其非交互式设置因本地网关健康检查失败的用户现在会收到可运行的 `openclaw onboard --install-daemon` 或 `openclaw onboard --skip-health` 恢复命令，而不是不支持的 `setup` 标志。[#93994](https://github.com/openclaw/openclaw/pull/93994) 相关 [#93947](https://github.com/openclaw/openclaw/issues/93947)。感谢 @bk-z1，@nianjiuzst。 - 网关健康探测检查现在接受与启动本地网关相同的自定义 `--port`，提前拒绝无效端口，并在 JSON 输出中显示所选的环回目标。[#94687](https://github.com/openclaw/openclaw/pull/94687) 相关 [#79100](https://github.com/openclaw/openclaw/issues/79100)。感谢 @bryantegomoh，@ozthedivine。 - `gateway --force` 现在会检测仅 IPv4 的进程占用网关端口，并且在端口检查不确定时仍然尝试清理，而不是错误地将端口视为空闲。[#94949](https://github.com/openclaw/openclaw/pull/94949) 相关 [#94426](https://github.com/openclaw/openclaw/issues/94426)。感谢 @sunlit-deng，@vincentkoc，@wangwllu。 - `openclaw config validate` 现在接受显式使用 `transport: "stdio"` 的基于命令的 MCP 服务器设置，避免错误的验证错误，同时仍然拒绝无效的远程风格 stdio 配置。[#95102](https://github.com/openclaw/openclaw/pull/95102) 相关 [#95082](https://github.com/openclaw/openclaw/issues/95082)。感谢 @ken-jo，@lzyyzznl。 - CLI 图像编辑现在可以通过 `--count <n>` 在一次命令中返回多个变体，而不是受限于提供商的默认单一结果。[#95300](https://github.com/openclaw/openclaw/pull/95300) 感谢 @ly-wang19。 - `openclaw sessions export-trajectory` 现在可以找到其他会话命令已经可以看到的会话，当配置了自定义的、基于 `~` 的或基于 `{agentId}` 模板的存储时，无需再次指定存储路径。[#95570](https://github.com/openclaw/openclaw/pull/95570) 相关 [#95568](https://github.com/openclaw/openclaw/issues/95568)。感谢 @youngting520。 - 修复了 `infer inspect --name <id> --json` 显示匹配 CLI 命令不接受的标志的问题，因此开发者和操作员可以可靠地发现支持的模型、认证和转录选项。[#95719](https://github.com/openclaw/openclaw/pull/95719) 感谢 @ly-wang19，@vincentkoc。 - 检查非常大或乱序的会话的人可以打开使用详情并仍然获取最新的带时间戳日志条目，而无需 OpenClaw 在内存中保留整个解析的日志历史。[#96019](https://github.com/openclaw/openclaw/pull/96019) 感谢 @vincentkoc。 - 操作员现在可以为自己的 OpenClaw 主机设置认证监视器、systemd 定时器和 Termux 小部件，而无需首先替换维护者特定的主机名和文件系统路径。[af3e509](https://github.com/openclaw/openclaw/commit/af3e509ab823dac5f91b16915ee7067b369656a3) 感谢 @vincentkoc。 - 原生 Windows crabbox 水合现在自动选择所需的 Windows 守护进程作业，避免失败或路由错误的运行，同时 WSL2 和显式作业覆盖保持不变。[d5d9a82](https://github.com/openclaw/openclaw/commit/d5d9a8256d6bc2ff8d699152923357bd61c606c1) 感谢 @vincentkoc。 #### 用户操作文档 [网关配置](https://docs.openclaw.ai/gateway/configuration)和相关用户指南现在提供更清晰的设置和操作说明。 - 设置本地智能体头像的人可以通过保持工作区相关文件在 2 MB 以下来避免图像缺失，而 HTTP(S) 和数据 URI 头像不受该限制。[#78884](https://github.com/openclaw/openclaw/pull/78884) 相关 [#65312](https://github.com/openclaw/openclaw/issues/65312)。感谢 @wangjieweb3-design，@nyx-nocturna。 - OpenClaw 的默认智能体指令现在要求智能体在提议自定义构建之前检查是否有合适的免费或开源解决方案，同时在自定义工作更合适时仍然允许。[#86608](https://github.com/openclaw/openclaw/pull/86608) 感谢 @cablackmon。 - 插件作者现在可以在 `subagent_ended` 事件上使用文档化的 `targetSessionKey` 将它们与相应的生成匹配，而不是依赖未发出的 `agentId` 或 `childSessionKey` 字段。[#95191](https://github.com/openclaw/openclaw/pull/95191) 相关 [#95186](https://github.com/openclaw/openclaw/issues/95186)。感谢 @ken-jo，@mahaohao-ch。 - OpenClaw 文档和展示卡片中的 ClawHub 技能链接现在打开规范的拥有者合格页面，安装示例使用可复制的 `openclaw skills install @owner/<slug>` 引用，而不是较旧的裸 slug 路由。[#95972](https://github.com/openclaw/openclaw/pull/95972) 感谢 @patrick-erichsen。 #### 其他工具和自动化修复 额外的[插件 SDK](https://docs.openclaw.ai/plugins/sdk-runtime) 和自动化修复改进了工具事件、cron 输入和 Windows 交接。 - SDK 应用程序现在会在终端工具失败或被阻止时收到 `tool.call.failed` 事件，而不是误导性的完成事件，因此现有的失败处理可以正确反应。[#95383](https://github.com/openclaw/openclaw/pull/95383) 感谢 @ly-wang19。 - 修复了当已识别的作业字段带有无害的尾部空格时 cron 添加和更新请求被拒绝的问题，因此可以在不放松对歧义或不安全输入的检查的情况下保存计划。[#95674](https://github.com/openclaw/openclaw/pull/95674) 相关 [#95407](https://github.com/openclaw/openclaw/issues/95407)。感谢 @nassiel，@zw-xysk。 - Codex 子智能体监控以更少的不必要扫描处理大型子智能体和记录文件集，同时较旧的记录文件名格式继续像以前一样解析。[#96085](https://github.com/openclaw/openclaw/pull/96085) 感谢 @vincentkoc。 - 修复了原生 Windows crabbox 水合在运行程序和守护进程使用不同主目录时卡住或错过交接的问题，因此两者可以找到相同的作业状态和停止文件。[f354889](https://github.com/openclaw/openclaw/commit/f354889efa1c8bafca9304767afba2c270add549) 感谢 @vincentkoc。 ### 额外贡献 面向维护者的额外贡献：[#95308](https://github.com/openclaw/openclaw/pull/95308)，[#95465](https://github.com/openclaw/openclaw/pull/95465)，[#95625](https://github.com/openclaw/openclaw/pull/95625)，[#95649](https://github.com/openclaw/openclaw/pull/95649)，[#95681](https://github.com/openclaw/openclaw/pull/95681)，[#95857](https://github.com/openclaw/openclaw/pull/95857)，[#95870](https://github.com/openclaw/openclaw/pull/95870)，[#95872](https://github.com/openclaw/openclaw/pull/95872)，[#95879](https://github.com/openclaw/openclaw/pull/95879)，[#95890](https://github.com/openclaw/openclaw/pull/95890)，[#95909](https://github.com/openclaw/openclaw/pull/95909)，[#95922](https://github.com/openclaw/openclaw/pull/95922)，[#95946](https://github.com/openclaw/openclaw/pull/95946)，[#95967](https://github.com/openclaw/openclaw/pull/95967)，[#95983](https://github.com/openclaw/openclaw/pull/95983)，[#96258](https://github.com/openclaw/openclaw/pull/96258)，[#95094](https://github.com/openclaw/openclaw/pull/95094)，[#95466](https://github.com/openclaw/openclaw/pull/95466)，[#95876](https://github.com/openclaw/openclaw/pull/95876)，[#95880](https://github.com/openclaw/openclaw/pull/95880)，[#95919](https://github.com/openclaw/openclaw/pull/95919)，[#95928](https://github.com/openclaw/openclaw/pull/95928)，[#95991](https://github.com/openclaw/openclaw/pull/95991)，[#96235](https://github.com/openclaw/openclaw/pull/96235)，[#94272](https://github.com/openclaw/openclaw/pull/94272)，[#94622](https://github.com/openclaw/openclaw/pull/94622)，[#95898](https://github.com/openclaw/openclaw/pull/95898)，[#95901](https://github.com/openclaw/openclaw/pull/95901)，[#95999](https://github.com/openclaw/openclaw/pull/95999)，[#96055](https://github.com/openclaw/openclaw/pull/96055)，[#96226](https://github.com/openclaw/openclaw/pull/96226)，[#96271](https://github.com/openclaw/openclaw/pull/96271)，[#97909](https://github.com/openclaw/openclaw/pull/97909)，[#91502](https://github.com/openclaw/openclaw/pull/91502)，[#91506](https://github.com/openclaw/openclaw/pull/91506)，[#94700](https://github.com/openclaw/openclaw/pull/94700)，[#95406](https://github.com/openclaw/openclaw/pull/95406)，[#95858](https://github.com/openclaw/openclaw/pull/95858)，[#95933](https://github.com/openclaw/openclaw/pull/95933)，[#95944](https://github.com/openclaw/openclaw/pull/95944)，[#95947](https://github.com/openclaw/openclaw/pull/95947)，[#95952](https://github.com/openclaw/openclaw/pull/95952)，[#95961](https://github.com/openclaw/openclaw/pull/95961)，[#95971](https://github.com/openclaw/openclaw/pull/95971)，[#95975](https://github.com/openclaw/openclaw/pull/95975)，[#96003](https://github.com/openclaw/openclaw/pull/96003)，[#96017](https://github.com/openclaw/openclaw/pull/96017)，[#96030](https://github.com/openclaw/openclaw/pull/96030)，[#96246](https://github.com/openclaw/openclaw/pull/96246)，[#87121](https://github.com/openclaw/openclaw/pull/87121)，[#90223](https://github.com/openclaw/openclaw/pull/90223)，[#93378](https://github.com/openclaw/openclaw/pull/93378)，[#95475](https://github.com/openclaw/openclaw/pull/95475)，[#95499](https://github.com/openclaw/openclaw/pull/95499)，[#95602](https://github.com/openclaw/openclaw/pull/95602)，[#95653](https://github.com/openclaw/openclaw/pull/95653)，[#76668](https://github.com/openclaw/openclaw/pull/76668)，[#87861](https://github.com/openclaw/openclaw/pull/87861)，[#95243](https://github.com/openclaw/openclaw/pull/95243)，[#78715](https://github.com/openclaw/openclaw/pull/78715)，[#93502](https://github.com/openclaw/openclaw/pull/93502)，[#96044](https://github.com/openclaw/openclaw/pull/96044)，[#96057](https://github.com/openclaw/openclaw/pull/96057)，[#96061](https://github.com/openclaw/openclaw/pull/96061)，[#91193](https://github.com/openclaw/openclaw/pull/91193)，[#95706](https://github.com/openclaw/openclaw/pull/95706)，[#96179](https://github.com/openclaw/openclaw/pull/96179)，[#89912](https://github.com/openclaw/openclaw/pull/89912)，[#90439](https://github.com/openclaw/openclaw/pull/90439)，[#96182](https://github.com/openclaw/openclaw/pull/96182)，[#96191](https://github.com/openclaw/openclaw/pull/96191)，[#96193](https://github.com/openclaw/openclaw/pull/96193)，[#96195](https://github.com/openclaw/openclaw/pull/96195)，[#96204](https://github.com/openclaw/openclaw/pull/96204)，[#96206](https://github.com/openclaw/openclaw/pull/96206)，[#96213](https://github.com/openclaw/openclaw/pull/96213)，[#96218](https://github.com/openclaw/openclaw/pull/96218)，[#95893](https://github.com/openclaw/openclaw/pull/95893)，[#95930](https://github.com/openclaw/openclaw/pull/95930)。感谢 @aniruddhaadak80，@coder999999999，@davinci282828，@harjothkhara，@hugenshen，@jalehman，@jason-allen-oneal，@joshavant，@kklouzal，@lizuju，@mehrazmorshed，@mmyzwl，@patrick-erichsen，@romneyda，@rushindrasinha，@shuofengzhang，@tayoun，@vincentkoc，@wadydx，@wangmiao0668000666，@whiteyzy，@yachiyo1680，@zats。 ### 发布验证 - npm 包：https://www.npmjs.com/package/openclaw/v/2026.6.11 - 注册表 tarball：https://registry.npmjs.org/openclaw/-/openclaw-2026.6.11.tgz - 完整性：`sha512-T+P/g19IheeT1ckXMoPN61dYuE8vBF4MderI+kWkvpuFYxPkJxn8AXLpu9IXCnN9g36Acpm9+mMD/V+lsvOkyA==` - 发布 SHA：`e085fa1a3ffd32d0ea6917e1e6fb4ecbffbb77d2` - 完整发布 CI 报告：https://github.com/openclaw/releases/blob/main/evidence/2026.6.11/release-evidence.md - 发布发布：https://github.com/openclaw/openclaw/actions/runs/28456822852 - npm 预检：https://github.com/openclaw/openclaw/actions/runs/28417979026 - 完整发布验证：https://github.com/openclaw/openclaw/actions/runs/28415183795 - 插件 npm 发布：https://github.com/openclaw/openclaw/actions/runs/28457240090 - 插件 ClawHub 发布：已单独分派，此证明不等待：https://github.com/openclaw/openclaw/actions/runs/28457242470 - 插件 ClawHub 引导：不需要 - OpenClaw npm 发布：https://github.com/openclaw/openclaw/actions/runs/28458034111 - npm Telegram 测试版 E2E：https://github.com/openclaw/openclaw/actions/runs/28419920660 - Windows Hub 推广：https://github.com/openclaw/openclaw/actions/runs/28458452299 来自 openclaw/[email protected]

We heard the feedback. v2026.6.11 focuses on the rough edges that make OpenClaw feel less dependable, with fixes for misplaced replies, stuck sends, reconnects, model setup failures, and safer admin defaults. [Full release notes](https://docs.openclaw.ai/releases/2026.6.11) ### Highlights #### Channel delivery reliability Delivery and reconnect fixes span [Telegram](https://docs.openclaw.ai/channels/telegram), [WhatsApp](https://docs.openclaw.ai/channels/whatsapp), [Matrix](https://docs.openclaw.ai/channels/matrix), [Google Chat](https://docs.openclaw.ai/channels/googlechat), [iMessage](https://docs.openclaw.ai/channels/imessage), [Feishu](https://docs.openclaw.ai/channels/feishu), [Mattermost](https://docs.openclaw.ai/channels/mattermost), [WebChat](https://docs.openclaw.ai/web/webchat), the [Control UI](https://docs.openclaw.ai/web/control-ui), and the [terminal UI](https://docs.openclaw.ai/cli/tui). - Fixes newer Google Chat direct messages sometimes being treated like group conversations, so they reach the correct one-to-one chat while Space and group-chat messages keep their existing routing. [#58993](https://github.com/openclaw/openclaw/pull/58993) Thanks @starhappysh, @vincentkoc. - Feishu voice replies from OpenClaw now show their duration in the chat bubble, so recipients can see how long the audio is before playing it. [#89172](https://github.com/openclaw/openclaw/pull/89172) Related [#53798](https://github.com/openclaw/openclaw/issues/53798). Thanks @areslp, @fxz26284407, @kinrocw. - Discord and Telegram replies and mirrored chat history stay tied to the intended conversation more consistently, including across repeated Telegram replies and session changes. [#89911](https://github.com/openclaw/openclaw/pull/89911) Thanks @jalehman. - Background image, video, and music results now return to the chat that requested them when the task starts without a full conversation target, instead of appearing to fail after creation or being sent to the wrong peer as the session moves. [#89949](https://github.com/openclaw/openclaw/pull/89949) Related [#86034](https://github.com/openclaw/openclaw/issues/86034). Thanks @tianxiaochannel-oss88, @wangwllu. - Telegram answers now stay attached to the user's current question when they quote an earlier bot message, while quotes of other people's messages still reply to the selected quote. [#90475](https://github.com/openclaw/openclaw/pull/90475) Thanks @moeedahmed. - QQBot group admins can choose how broadly slash commands are available, and private-only commands now direct users to a private chat instead of being exposed or silently ignored in groups. [#92154](https://github.com/openclaw/openclaw/pull/92154) Thanks @sliverp. - Heartbeat checks using reasoning-capable models now show the assistant's intended reply instead of exposing internal reasoning in Telegram, WhatsApp, and other channels, while opt-in Thinking messages still work. [#92356](https://github.com/openclaw/openclaw/pull/92356) Related [#92260](https://github.com/openclaw/openclaw/issues/92260). Thanks @jmpei, @tangtaizong666, @vincentkoc. - Telegram progress-mode chats now clear an old progress bubble before newer tool output or artifacts appear, keeping the conversation in a clean, readable order. [#93002](https://github.com/openclaw/openclaw/pull/93002) Related [#90753](https://github.com/openclaw/openclaw/issues/90753). Thanks @shadow-enthusiast, @zhangguiping-xydt. - iMessage command-and-link messages now stay together as one OpenClaw turn when delayed link previews arrive, while unrelated quick messages remain separate for users who enabled same-sender DM coalescing. [#93143](https://github.com/openclaw/openclaw/pull/93143) Thanks @omarshahine. - Successful Discord replies sent through the message tool no longer trigger a misleading failure warning in affected `message_tool_only` source-channel turns. [#94072](https://github.com/openclaw/openclaw/pull/94072) Related [#93875](https://github.com/openclaw/openclaw/issues/93875). Thanks @chenyangjun-xy, @hoyanhan, @vincentkoc. - WhatsApp group conversations now preserve the right message and group context more reliably during retries, reconnects, and group changes. [#94338](https://github.com/openclaw/openclaw/pull/94338) Related [#7433](https://github.com/openclaw/openclaw/issues/7433). Thanks @mcaxtr, @octopuslabs-fl, @xialonglee. - Fixes OpenClaw sometimes replying to its own delayed iMessage echoes when stray leading characters keep the sent message from being recognized. [#94442](https://github.com/openclaw/openclaw/pull/94442) Thanks @ly-wang19. - Telegram webhook users can keep receiving DMs and group messages through brief channel restarts, configuration reloads, and recovery cycles without temporary message blackouts. [#94506](https://github.com/openclaw/openclaw/pull/94506) Related [#90254](https://github.com/openclaw/openclaw/issues/90254). Thanks @obviyus, @travellingsoldier85, @xialonglee. - Matrix E2EE gateways can stay online during long-running use instead of gradually consuming memory until a crash takes channels and in-flight work down. [#94942](https://github.com/openclaw/openclaw/pull/94942) Related [#90455](https://github.com/openclaw/openclaw/issues/90455). Thanks @xzh-icenter, @yar-sh. - Telegram users now see the intended native reaction instead of leaked instructions or a dropped reaction-only reply, with success recorded only after Telegram accepts it. [#94977](https://github.com/openclaw/openclaw/pull/94977) Related [#71140](https://github.com/openclaw/openclaw/issues/71140). Thanks @cuttingwater, @hugenshen. - Telegram progress updates for commands, searches, updates, and API activity now stay readable instead of exposing noisy HTML or code-style rows, with plain-text fallback when Telegram cannot parse the formatting. [#95007](https://github.com/openclaw/openclaw/pull/95007) Related [#95002](https://github.com/openclaw/openclaw/issues/95002). - Telegram conversations continued in WebChat now show one assistant reply per turn and keep later replies with the active conversation instead of duplicating answers or sending them back to Telegram. [#95069](https://github.com/openclaw/openclaw/pull/95069) Related [#94930](https://github.com/openclaw/openclaw/issues/94930). Thanks @heichaowo. - Google Chat now hides misleading internal failure banners when a tool result is harmless, leaving users with the completed answer while normal assistant text remains unchanged. [#95084](https://github.com/openclaw/openclaw/pull/95084) Related [#90684](https://github.com/openclaw/openclaw/issues/90684). Thanks @jailbirt, @studentzhou-svg. - Bound multi-agent channel conversations now load the workspace files for the configured agent instead of the default agent, though previously misfiled conversations may start fresh in the corrected agent store. [#95118](https://github.com/openclaw/openclaw/pull/95118) Related [#92903](https://github.com/openclaw/openclaw/issues/92903). Thanks @849261680, @axjing. - People sharing an OpenClaw gateway can now assign different models to individual direct-message contacts across supported chat channels, while existing group and wildcard model choices keep working as before. [#95120](https://github.com/openclaw/openclaw/pull/95120) Related [#53638](https://github.com/openclaw/openclaw/issues/53638). Thanks @gandalf-at-lerian, @thomaszta, @xydigit-zt. - Telegram now shows that OpenClaw is still working during short initial previews or progress-mode replies instead of leaving the chat silent until the final message arrives. [#95183](https://github.com/openclaw/openclaw/pull/95183) Related [#95004](https://github.com/openclaw/openclaw/issues/95004). Thanks @obviyus. - Matrix users and operators now get a clear failure when a homeserver sends an oversized or stalled response, instead of OpenClaw continuing to buffer it and risking unbounded memory use. [#95240](https://github.com/openclaw/openclaw/pull/95240) Thanks @alix-007. - Fixes delayed or missing Telegram and other queued channel replies in Kubernetes-style deployments with many injected environment variables, where opening the queue database could stall the gateway. [#95278](https://github.com/openclaw/openclaw/pull/95278) Related [#94571](https://github.com/openclaw/openclaw/issues/94571). Thanks @kaka-srp. - Telegram chats recover after one stuck message times out, allowing later messages in the same chat or topic to reach the agent without restarting the gateway. [#95299](https://github.com/openclaw/openclaw/pull/95299) Related [#95248](https://github.com/openclaw/openclaw/issues/95248). Thanks @kriegerbangerz-ship-it, @mikasa0818, @obviyus. - When people switch between Telegram and another OpenClaw client in a shared direct conversation, short Telegram replies now follow the latest conversation instead of responding to an older, unrelated Telegram proposal. [#95390](https://github.com/openclaw/openclaw/pull/95390) Related [#95378](https://github.com/openclaw/openclaw/issues/95378). Thanks @maiduy708, @mikasa0818, @obviyus. - Fixes completed assistant messages appearing twice in Telegram, Discord, Slack, and other streamed chats after a multi-message reply. [#95432](https://github.com/openclaw/openclaw/pull/95432) Thanks @vincentkoc, @yetval. - WhatsApp replies now stay attached to the direct or group message being answered instead of appearing as a separate message that loses the conversation context. [#95483](https://github.com/openclaw/openclaw/pull/95483) Thanks @mcaxtr. - Telegram rich-message replies now keep paragraphs, bullets, and status lines separated instead of collapsing multi-line content into one run-on block, with no configuration change required. [#95532](https://github.com/openclaw/openclaw/pull/95532) Related [#95409](https://github.com/openclaw/openclaw/issues/95409). Thanks @amknight. - Mattermost operators who enable native slash commands can now use `/oc_queue` directly in Mattermost to tune active-run queuing, including its mode, debounce timing, cap, and drop handling. [#95546](https://github.com/openclaw/openclaw/pull/95546) Thanks @amknight. - Previously allowed messages keep reaching named accounts after legacy multi-account channel upgrades, with inherited DM and group access rules preserved across Mattermost, Discord, Slack, Telegram, Signal, WhatsApp, iMessage, and IRC. [#95550](https://github.com/openclaw/openclaw/pull/95550) Thanks @amknight. - Mattermost users can keep talking in a thread without mentioning the bot again after it replies, and that participation survives gateway restarts until the thread has been idle for seven days. [#95552](https://github.com/openclaw/openclaw/pull/95552) Thanks @amknight. - Inbound Telegram messages now reach the configured OpenClaw session promptly instead of sitting unanswered until the next polling interval, a gateway restart, or manual intervention. [#95577](https://github.com/openclaw/openclaw/pull/95577) Related [#86957](https://github.com/openclaw/openclaw/issues/86957). Thanks @freidrich-goldenflow, @liuwqgit. - QQBot users now receive complete markdown tables when valid separators use one or two dashes per column, instead of losing the header and all but the final row. [#95637](https://github.com/openclaw/openclaw/pull/95637) Thanks @ly-wang19. - Synology Chat users can now receive agent replies that take more than 120 seconds when the configured core timeout allows it, instead of having the channel reject them early. [#95707](https://github.com/openclaw/openclaw/pull/95707) Thanks @sahibzada-allahyar, @vincentkoc. - Telegram forum-topic cron jobs now keep separately configured failure alerts going to their intended destination, even when the main announcement uses a topic in the same chat. [#95794](https://github.com/openclaw/openclaw/pull/95794) Thanks @vincentkoc. - Fixes WhatsApp group replies that could quote an older OpenClaw message instead of the user's triggering message, so final answers stay attached to the intended message when a reply target is available and avoid pointing back to stale bot context when it is not. [#95914](https://github.com/openclaw/openclaw/pull/95914) Thanks @mcaxtr. - WhatsApp users can approve or deny prompts by reaction without the prompt staying stuck when WhatsApp identifies the same direct chat differently, while group approvals remain tied to the correct group and person. [#95935](https://github.com/openclaw/openclaw/pull/95935) Thanks @mcaxtr. - Final reply processing now uses less CPU when OpenClaw checks whether block text was already sent, without changing which reply reaches the chat or how duplicate text is suppressed. [#96087](https://github.com/openclaw/openclaw/pull/96087) Thanks @vincentkoc. - Exec approval results from external channel plugins now return to the channel or DM where the command started instead of falling back to WebChat or seeming to disappear after approval. [#96140](https://github.com/openclaw/openclaw/pull/96140) Related [#96103](https://github.com/openclaw/openclaw/issues/96103). Thanks @lansenger-pm, @vincentkoc, @yetval. - WhatsApp's final answer now stays quoted to the follow-up message a user just sent when replying to an older OpenClaw message, instead of arriving unquoted or pointing back to the older bot reply. [#96220](https://github.com/openclaw/openclaw/pull/96220) Thanks @mcaxtr. - Nextcloud Talk bots now ignore ordinary file-share and lifecycle events without logging them as bot errors or risking disabled delivery, while malformed chat payloads still return an error. [#96243](https://github.com/openclaw/openclaw/pull/96243) Related [#81566](https://github.com/openclaw/openclaw/issues/81566). Thanks @arkyu2077, @rafaelmgbh, @vincentkoc. - Replies and message-tool delivery in Mattermost channels now use channel and thread guidance because the agent identifies those conversations as channels rather than group chats, while existing group-chat behavior remains unchanged. [#96244](https://github.com/openclaw/openclaw/pull/96244) Related [#95645](https://github.com/openclaw/openclaw/issues/95645). Thanks @arkyu2077, @iloveleon19, @vincentkoc. - MCP channel integrations now keep conversation lists, message reads, event polls, and waits within predictable bounds even when a client requests excessive limits or timeouts. [a39e548](https://github.com/openclaw/openclaw/commit/a39e548ede228aa1978bf9d509613cbed6db0c99) Thanks @vincentkoc. - Long-running streamed auto-replies are less likely to stop early or abort inconsistently when an unusually large timeout is configured. [6c85b90](https://github.com/openclaw/openclaw/commit/6c85b90469f94955ef00c1609e1f1d6fd2cf4ca8) Thanks @vincentkoc. - Channel progress now shows a repeated status when work genuinely returns to it after another update, instead of hiding useful context as a duplicate. [8a75c4d](https://github.com/openclaw/openclaw/commit/8a75c4dd5f3e625a22a7a08c6e1f368798c48111) - Completed channel replies no longer gain late progress notices, preventing stale status text from appearing after the answer is finished. [a594d2c](https://github.com/openclaw/openclaw/commit/a594d2ce73257326b7ab78adb3c4643245ec9431) Thanks @vincentkoc. - During streaming channel replies, progress messages now keep showing the latest state instead of getting stuck on an older update. [e114001](https://github.com/openclaw/openclaw/commit/e114001ccafa83b8b366e095a9d7748dfc50c082) Thanks @vincentkoc. - Matrix forced resets now handle unavailable secret storage without a runtime error, treating recovery access as unavailable so the reset path can continue safely. [5c5a8a4](https://github.com/openclaw/openclaw/commit/5c5a8a49d76954b53fefc2463bc7b1d6b960e8fc) Thanks @vincentkoc. - Configured channels now remain visible in `openclaw channels status --json`, while scheduled announcements reject stale entries that have no active plugin to deliver them. [a641c0d](https://github.com/openclaw/openclaw/commit/a641c0d560fd15373e462829facf15fd6a466aeb) - Discord voice conversations now keep back-to-back assistant responses moving, so a queued reply plays after the previous audio stream closes instead of remaining stuck. [88b64e4](https://github.com/openclaw/openclaw/commit/88b64e4b869e696d99de7417fb52425e9ed67cbf) Thanks @vincentkoc. - Discord progress previews are less likely to stop before the final edits when an agent response has already started arriving. [86ea382](https://github.com/openclaw/openclaw/commit/86ea382121b00e73af4b4c329d0a2447592e4071) - Chats no longer show stray `NO_REPLY` text when the assistant means to stay silent, while legitimate media responses still arrive without the placeholder. [96c6f80](https://github.com/openclaw/openclaw/commit/96c6f8022c2420826830b11f4353ce855ab2ac5c) - Telegram streaming replies now show each progress heading once, keeping tool and search updates easier to scan. [013e33c](https://github.com/openclaw/openclaw/commit/013e33c6d3672a980550912442bb1ac5505918aa) Thanks @vincentkoc. - Telegram messages that get stuck after a long-running task, crash, or gateway restart now resume processing automatically, so later messages no longer wait silently or require operators to repair the queue by hand. [#97543](https://github.com/openclaw/openclaw/pull/97543) Thanks @romneyda, @vincentkoc. #### Provider and model recovery [Model setup](https://docs.openclaw.ai/providers/models), [OpenAI](https://docs.openclaw.ai/providers/openai), [OpenRouter](https://docs.openclaw.ai/providers/openrouter), [opencode-go](https://docs.openclaw.ai/providers/opencode-go), and [fast-mode](https://docs.openclaw.ai/tools/thinking#fast-mode-fast) follow-up behavior now recover more clearly in affected configurations. - MiniMax text-to-speech and voice notes are less likely to fail because OpenClaw now explicitly requests the audio format it can decode instead of relying on provider defaults. [#73079](https://github.com/openclaw/openclaw/pull/73079) Thanks @efe-arv. - Gateway operators can again see provider, model, request status, and timing details in normal logs, making model-routing and transport problems easier to diagnose without enabling extra debug logging. [#89648](https://github.com/openclaw/openclaw/pull/89648) Related [#89300](https://github.com/openclaw/openclaw/issues/89300). Thanks @enominera, @xiaobao-k8s. - Models reached through Google, Mistral, OpenAI Responses, Azure OpenAI Responses, and ChatGPT/Codex Responses now receive clean system instructions without OpenClaw's internal cache-boundary marker leaking into the prompt. [#89716](https://github.com/openclaw/openclaw/pull/89716) Thanks @enominera, @masatohoshino. - Cron tool calls using Gemini models through OpenAI-compatible providers now run without nullable fields triggering provider schema rejections. [#91559](https://github.com/openclaw/openclaw/pull/91559) Related [#91542](https://github.com/openclaw/openclaw/issues/91542). Thanks @pick-cat, @qiukui666. - Provider-qualified model IDs now honor their configured agent runtime policies and CLI aliases instead of unexpectedly falling back to OpenClaw's default runtime. [#91724](https://github.com/openclaw/openclaw/pull/91724) Thanks @vincentkoc, @yu-xin-c. - The chat `/models` list and other plugin-aware model or provider selection paths now respond quickly instead of stalling for seconds and consuming a CPU core through repeated setup scans, while plugin changes still refresh normally. [#93356](https://github.com/openclaw/openclaw/pull/93356) Thanks @obuchowski. - Hosted Ollama Cloud users can keep only the models they explicitly configured after a restart, without the full shared catalog being added back, while automatic discovery continues for local and self-hosted Ollama servers. [#93956](https://github.com/openclaw/openclaw/pull/93956) Thanks @jason-allen-oneal. - Cron jobs can now retry or switch to a configured fallback model when a local provider returns the generic `LLM request failed.` error, instead of failing with the fallback unused. [#94062](https://github.com/openclaw/openclaw/pull/94062) Related [#93931](https://github.com/openclaw/openclaw/issues/93931). Thanks @hugenshen. - Expired provider tokens no longer bury useful operator logs under repeated fallback warnings, while the first warning and later duplicate summaries remain available for diagnosis. [#94233](https://github.com/openclaw/openclaw/pull/94233) Related [#56979](https://github.com/openclaw/openclaw/issues/56979). Thanks @goutamadwant, @yanan1991. - Google Gemini 3.5 Flash can now be selected with its full 1,048,576-token context window, avoiding missing-model errors and needless prompt-size rejections. [#94726](https://github.com/openclaw/openclaw/pull/94726) Related [#94723](https://github.com/openclaw/openclaw/issues/94723). Thanks @ajwan8998, @anguslogan01, @kevinat. - Dashboard child sessions now handle allowed provider-qualified model choices consistently and give accurate recovery guidance when saved model state is stale. [#94752](https://github.com/openclaw/openclaw/pull/94752) Related [#94713](https://github.com/openclaw/openclaw/issues/94713). Thanks @gr4via. - Claude CLI users no longer get promises of completion updates that may never arrive, because OpenClaw now blocks unsupported native background work before it can strand progress. [#95008](https://github.com/openclaw/openclaw/pull/95008) Thanks @anagnorisis2peripeteia. - OpenClaw now rejects oversized provider catalog or JSON responses with a clear error before buffering the entire response in memory. [#95218](https://github.com/openclaw/openclaw/pull/95218) Thanks @alix-007. - OpenRouter users can now select and run the advertised short DeepSeek V4 model IDs without requests failing with `model_not_found` because OpenClaw sent a duplicated provider prefix. [#95268](https://github.com/openclaw/openclaw/pull/95268) Related [#95198](https://github.com/openclaw/openclaw/issues/95198). Thanks @daniel-alejandro-t, @darren2030. - With `/reasoning on`, DeepSeek-style OpenAI-compatible models now show the final answer separately from their reasoning instead of folding it into the reasoning block, with no configuration change required. [#95283](https://github.com/openclaw/openclaw/pull/95283) Related [#95280](https://github.com/openclaw/openclaw/issues/95280). Thanks @marvinthebored, @vincentkoc, @zengwen-dt. - When a Codex subscription reaches its usage limit, OpenClaw now moves to configured fallback models instead of stopping on the failed result, and it does not retry runs that already produced visible output. [#95400](https://github.com/openclaw/openclaw/pull/95400) Thanks @jason-allen-oneal, @sallyom. - LM Studio users can now run quantized or multi-variant local models without false assistant-turn failures or phantom suffixed model entries caused by mismatched model keys. [#95401](https://github.com/openclaw/openclaw/pull/95401) Thanks @monkeyleet. - Google-backed embedded-agent runs now stop reading oversized or never-ending prompt-cache responses before they can exhaust memory or leave the run stalled. [#95417](https://github.com/openclaw/openclaw/pull/95417) Thanks @alix-007. - OpenRouter model scans fail safely on oversized or malformed catalogs instead of risking excessive memory use that can destabilize OpenClaw. [#95418](https://github.com/openclaw/openclaw/pull/95418) Thanks @alix-007. - OpenRouter setups now reject oversized model catalogs before they can exhaust OpenClaw's memory, without caching or immediately refetching the failed response. [#95420](https://github.com/openclaw/openclaw/pull/95420) Thanks @alix-007, @sallyom. - Configured fallback models can now answer when Claude CLI runs out of credits or hits a generic runner failure, instead of leaving users with the failure message as the final response. [#95508](https://github.com/openclaw/openclaw/pull/95508) Related [#95489](https://github.com/openclaw/openclaw/issues/95489). Thanks @mikasa0818, @riazrahaman, @sallyom. - Gemini-backed web searches using `freshness: "day"` or `pd` now complete instead of failing with a provider 400 error, while broader freshness choices and explicit date ranges retain stricter filtering. [#95682](https://github.com/openclaw/openclaw/pull/95682) Thanks @sunjae-k, @vincentkoc. - Follow-up answers from xAI reasoning models such as Grok Composer now preserve earlier reasoning context more reliably, even when configurable reasoning effort is unsupported. [#95686](https://github.com/openclaw/openclaw/pull/95686) Thanks @fuller-stack-dev, @geraint0923. - Vercel AI Gateway users can now run models chosen from the live catalog, including live-only model IDs that are absent from OpenClaw's bundled list. [#95710](https://github.com/openclaw/openclaw/pull/95710) Thanks @vincentkoc. - Fixes manifest-defined providers turning valid model IDs into broken ones when `stripPrefixes` entries have stray spaces or different casing, so operators and plugin authors get the intended provider model. [#95744](https://github.com/openclaw/openclaw/pull/95744) Related [#95743](https://github.com/openclaw/openclaw/issues/95743). Thanks @parveshsaini. - First-run setup now opens the credential prompt for a newly installed external provider instead of appearing to loop and leaving OpenAI selected. [#95792](https://github.com/openclaw/openclaw/pull/95792) Related [#95765](https://github.com/openclaw/openclaw/issues/95765). - Oversized or stalled provider catalogs now fail quickly with a clear error instead of hanging OpenClaw or consuming unbounded memory, while normal catalogs continue to load. [#95827](https://github.com/openclaw/openclaw/pull/95827) Thanks @alix-007. - Xiaomi Token Plan users can now use up to 128K output tokens with `mimo-v2.5` and `mimo-v2.5-pro` instead of being stopped at the outdated 32,000-token limit. [#95934](https://github.com/openclaw/openclaw/pull/95934) Thanks @idootop. - Tool-heavy model responses can stream with less overhead while repeated tool-call IDs and encrypted reasoning details stay matched to the correct call across Google and OpenAI-compatible providers. [#95957](https://github.com/openclaw/openclaw/pull/95957) Thanks @vincentkoc. - Token-usage accounting is more reliable for bundled ACPX users because OpenClaw now includes ACPX 0.11.2's persistence fix by default, without a separate package override or manual client update. [#96124](https://github.com/openclaw/openclaw/pull/96124) Thanks @vincentkoc. - Ollama Cloud users can now find and select `glm-5.2:cloud` with its 1,000,000-token context window, reasoning, and tool support even when it is absent from the public model list. [11484f8](https://github.com/openclaw/openclaw/commit/11484f8a1483b7c42aa2971de2d88289fcef7046) - MiniMax image-understanding requests no longer fail before reaching the provider when a timeout is zero, negative, or extremely large; invalid values now use a normal or safe maximum wait. [4b6182e](https://github.com/openclaw/openclaw/commit/4b6182ee2a250005e0c25edfeae4db6ec59b7cb8) Thanks @vincentkoc. - Codex runs now follow the current fast-mode choice instead of carrying over an old speed tier, and the status line clearly shows when fast mode is automatic. [77012f9](https://github.com/openclaw/openclaw/commit/77012f9807851c662e064d05097497a25ab13505) Thanks @vincentkoc. - Codex-backed conversations now return to normal routing after automatic fast mode is cleared, preventing later turns or model changes from reusing a stale priority tier. [8afc1f7](https://github.com/openclaw/openclaw/commit/8afc1f770bbef30a4d2d9957ef26a685c508448c) - Fallback agent runs now honor each model's configured automatic fast-mode cutoff even when fast mode is overridden for the run, keeping fallback behavior aligned with the selected model policy. [efd3172](https://github.com/openclaw/openclaw/commit/efd3172662ce023eb8d6568b689361536edf06dd) - Live model-switch retries now preserve the original fast-mode cutoff for long-running sessions, while explicit fast mode avoids misleading automatic-cutoff progress messages. [d990115](https://github.com/openclaw/openclaw/commit/d990115d1972fdf4361884a29bbf8396f33e5cba) Thanks @vincentkoc. - Embedded agent runs now keep automatic fast mode working consistently through retries and progress updates without confusing it with a manually selected fast-mode setting. [cf1b6fe](https://github.com/openclaw/openclaw/commit/cf1b6fef4403bee7c206299efc4385a7fcb74375) Thanks @vincentkoc. - Fast-mode runs now keep their speed setting through model fallback retries and show the configured automatic threshold in status, avoiding inconsistent retry behavior and an unhelpful generic label. [aa3797c](https://github.com/openclaw/openclaw/commit/aa3797c8d0d74b4502d24852ce6baa70286f2f06) Thanks @vincentkoc. - Agent replies and scheduled cron runs now handle fast-mode fallback retries more reliably, keeping the state needed for the final attempt to finish or report progress correctly. [14e448e](https://github.com/openclaw/openclaw/commit/14e448e0e13db9f194ea16bb98e0f846a67769fd) Thanks @vincentkoc. - Users no longer see a fast-mode reset notice while model fallback attempts are still running; it appears only when the run reaches its final fallback attempt. [6eb72a8](https://github.com/openclaw/openclaw/commit/6eb72a830ece3e2b4c6c85e5a9c2b72b59e0dae9) - Users and operators now get clearer handling when a configured live model becomes unavailable because OpenClaw recognizes the provider's "selected model was not found" response as a model-not-found failure instead of a generic error. [2405d02](https://github.com/openclaw/openclaw/commit/2405d029d437ee58ab94da800a5b213bc6bf4628) Thanks @vincentkoc. - Qwen and vLLM now preserve existing chat-template settings consistently when thinking is switched on or off, and provider plugins can use the same tested helper. [2ba9d6e](https://github.com/openclaw/openclaw/commit/2ba9d6eabef9427a950bacc47f077200446cb865) Thanks @vincentkoc. - OpenAI-compatible proxy providers can handle thinking levels and legacy `reasoning_effort` fields more consistently, with plugin developers and provider maintainers using one documented normalization helper across OpenRouter, Kilocode, and the SDK. [35bafea](https://github.com/openclaw/openclaw/commit/35bafea757fab0386292951a3dc2a2d3514f370e) - Browser and Vite builds can now load the OpenAI ChatGPT Responses provider without a server-only dependency breaking the bundle, while WebSocket failures still appear normally. [8c8eb86](https://github.com/openclaw/openclaw/commit/8c8eb86fff6e843bd391808ceee249ac8c7f5fa5) Thanks @vincentkoc. - OpenRouter model scans now accept the same larger valid catalogs as runtime discovery while still rejecting oversized responses before they can consume unbounded memory. [ad3b2f4](https://github.com/openclaw/openclaw/commit/ad3b2f4b8827cd73b4c1a6c8288569c0966276fe) Thanks @vincentkoc. - OpenAI Responses users, including affected Bedrock Mantle GPT-5.x reasoning setups, now get one clean final answer with aligned saved transcripts and replay context instead of dozens of repeated cumulative copies. [#92399](https://github.com/openclaw/openclaw/pull/92399) Related [#91959](https://github.com/openclaw/openclaw/issues/91959). Thanks @amersheeny, @daimingnj, @phoenixyy, @pigfoot. - Scheduled jobs and isolated sessions using opencode-go models now move stalled requests into configured timeout or fallback handling instead of hanging for minutes before ending with a generic `LLM request failed` error. [#93965](https://github.com/openclaw/openclaw/pull/93965) Related [#93610](https://github.com/openclaw/openclaw/issues/93610). Thanks @forceconstant, @zhangguiping-xydt. #### Session, memory, and trust continuity [Sessions](https://docs.openclaw.ai/concepts/session), [compaction](https://docs.openclaw.ai/concepts/compaction), [memory](https://docs.openclaw.ai/concepts/memory), [QMD-backed memory](https://docs.openclaw.ai/concepts/memory-qmd), and [Tool Search](https://docs.openclaw.ai/tools/tool-search) retain useful state more consistently, while [Matrix](https://docs.openclaw.ai/channels/matrix) recovery, [tool policies](https://docs.openclaw.ai/gateway/sandbox-vs-tool-policy-vs-elevated), and [approvals](https://docs.openclaw.ai/cli/approvals) stay attached to the intended trust boundary. - Affected agent conversations using OpenAI Responses can now recover and keep replying after a visible channel response leaves their saved history incomplete, instead of every later turn failing before a reply appears. [#84708](https://github.com/openclaw/openclaw/pull/84708) Thanks @anyech. - When a Codex-backed agent produces unusually large tool output, saved and replayed conversations now keep its text within the usual size limit while leaving non-text content unchanged. [#87912](https://github.com/openclaw/openclaw/pull/87912) Thanks @adrianip0204. - Control UI conversations now stay visible and continue in the same session after a sleep, network drop, or Gateway reconnect instead of disappearing when the next message is sent. [#89017](https://github.com/openclaw/openclaw/pull/89017) Related [#87700](https://github.com/openclaw/openclaw/issues/87700). Thanks @zhangguiping-xydt, @asicoe. - Bundled Codex and Copilot integrations now keep mirrored chat history and transcript updates tied to the correct OpenClaw session as storage evolves, while existing file-backed active transcripts continue working during the migration. [#89518](https://github.com/openclaw/openclaw/pull/89518) Thanks @jalehman. - WebChat's current-session status now matches the conversation you are actually using, so the session identity, thinking level, token context, and cost details no longer come from the fallback `main` session. [#89800](https://github.com/openclaw/openclaw/pull/89800) Related [#89773](https://github.com/openclaw/openclaw/issues/89773). Thanks @killo3967, @sweetcornna. - Your conversation is less likely to lose its context after you press stop during automatic compaction because the compaction request is now cancelled too. [#89886](https://github.com/openclaw/openclaw/pull/89886) Related [#89868](https://github.com/openclaw/openclaw/issues/89868). Thanks @lykeion-dev, @openperf, @vincentkoc. - When cross-agent session access is blocked, OpenClaw now lists all required visibility, agent-to-agent, and allow-list settings, helping operators correct policy configuration instead of chasing a nonexistent agent failure. [#90489](https://github.com/openclaw/openclaw/pull/90489) Related [#90443](https://github.com/openclaw/openclaw/issues/90443). Thanks @ramitrkar-hash, @sahibzada-allahyar, @vincentkoc. - `openclaw memory status` now shows an active light or REM dreaming phase instead of incorrectly reporting `Dreaming: off`, so operators can see that valid memory configurations are enabled. [#93113](https://github.com/openclaw/openclaw/pull/93113) Related [#67868](https://github.com/openclaw/openclaw/issues/67868). Thanks @agentarclab, @mrossit. - Timed-out QMD memory searches now stop their background work when the agent moves on, preventing abandoned processes from continuing to consume CPU and memory. [#93394](https://github.com/openclaw/openclaw/pull/93394) Thanks @alix-007. - Repeated instructions sent after compaction now remain in the conversation, preventing lost turns, orphaned replies, and malformed history that some providers reject. [#94328](https://github.com/openclaw/openclaw/pull/94328) Thanks @vincentkoc, @yetval. - Memory Wiki's Stale Pages report now leaves durable concept and synthesis pages out of freshness warnings, keeping attention on source and entity pages that may actually need review. [#94369](https://github.com/openclaw/openclaw/pull/94369) Thanks @sunnyshu0925, @vincentkoc. - Long embedded runs with recent progress are now less likely to be interrupted by stale-session recovery, while genuinely stalled runs can still be cleared so queued work continues. [#94701](https://github.com/openclaw/openclaw/pull/94701) Thanks @imadal1n, @mrclawfield. - Ollama memory search now respects a configured smaller embedding dimension and keeps indexes for different dimensions separate, avoiding incompatible vectors being mixed together. [#94811](https://github.com/openclaw/openclaw/pull/94811) Thanks @mushuiyu886. - Memory searches and targeted refreshes now stay connected to the correct OpenClaw session even when transcript filenames change or QMD exports use a different name. [#95087](https://github.com/openclaw/openclaw/pull/95087) Thanks @jalehman. - Long-running conversations with screenshots or other images now keep their continuity more consistently when OpenClaw makes room for new messages, instead of repeatedly filling up without moving the retained conversation forward. [#95128](https://github.com/openclaw/openclaw/pull/95128) Thanks @yetval. - Windows users can now run QMD-backed memory indexing and search through configured absolute `memory.qmd.command` paths, including drive-letter and UNC locations, without OpenClaw stripping the path separators before launch. [#95274](https://github.com/openclaw/openclaw/pull/95274) Related [#92302](https://github.com/openclaw/openclaw/issues/92302). Thanks @ardooken, @ly85206559. - Usage footers selected with `/usage full` or `/usage tokens` now remain visible after daily or idle session rollover, so users do not have to turn them on again. [#95322](https://github.com/openclaw/openclaw/pull/95322) Thanks @litang9. - Follow-up replies, reactions, threaded messages, and status checks stay with the chat they belong to after webchat or system activity, while real channel switches still clear outdated routing details. [#95467](https://github.com/openclaw/openclaw/pull/95467) Thanks @yetval. - Long-running main conversations now keep their prior context when users return after an overnight or delayed follow-up, rather than silently starting over after an otherwise normal completion. [#95472](https://github.com/openclaw/openclaw/pull/95472) Thanks @xydt-tanshanshan. - People with large session histories can list, preview, and find sessions without multi-second freezes, while older mixed-case session keys are still migrated at startup. [#95699](https://github.com/openclaw/openclaw/pull/95699) Thanks @jalehman, @jzakirov. - Fixes delivered replies sometimes being saved to the wrong conversation history, or omitted from it, when operators use a custom or per-agent `session.store`, improving continuity and auditability for the intended session. [#95782](https://github.com/openclaw/openclaw/pull/95782) Related [#95781](https://github.com/openclaw/openclaw/issues/95781). Thanks @youngting520. - Saved session-memory summaries now leave out raw model tokens, tool-call blocks, media placeholders, role tags, and stale `NO_REPLY` markers so future conversations keep useful context. [#95791](https://github.com/openclaw/openclaw/pull/95791) Thanks @sweetsophia, @vincentkoc, @yb0y. - Long-running OpenAI sessions using Codex/ChatGPT OAuth can now compact without a separate API key, whether `/compact` is run manually or triggered automatically. [#95831](https://github.com/openclaw/openclaw/pull/95831) Related [#95693](https://github.com/openclaw/openclaw/issues/95693). Thanks @sallyom, @yui-tien. - Long, tool-heavy sessions now compact oversized conversations instead of getting stuck when a large tool result appears at the end. [#95860](https://github.com/openclaw/openclaw/pull/95860) Related [#78478](https://github.com/openclaw/openclaw/issues/78478). Thanks @jw8957, @wzhgba, @yetval. - When `memory_search` is unavailable because the Node runtime lacks `node:sqlite`, OpenClaw now points users to a compatible runtime instead of sending them through unrelated embedding-provider troubleshooting. [#95916](https://github.com/openclaw/openclaw/pull/95916) Thanks @rrrrrredy, @vincentkoc. - Developers and operators inspecting a compacted Copilot session now get its summary, before-and-after token counts, and session details instead of an incomplete result. [#96049](https://github.com/openclaw/openclaw/pull/96049) Thanks @vincentkoc. - The `/stop` and abort commands now keep stopping active runs, clearing queued followups, and ending related subagents promptly even when session keys need canonicalizing or abort metadata cannot be saved. [#96201](https://github.com/openclaw/openclaw/pull/96201) Thanks @jalehman. - Voice Wake upgrades now keep existing trigger phrases and routing rules working as OpenClaw moves them from retired settings files into the shared state database. [bdf81a8](https://github.com/openclaw/openclaw/commit/bdf81a825fa3ef66ad2c535c1eeb0bb4e31b6d1b) - Upgrades from older OpenClaw state layouts now preserve update notifications, check throttling, available-version records, and automatic-update attempt history as that state moves into SQLite. [eb00d49](https://github.com/openclaw/openclaw/commit/eb00d499d16feea600fceef92d575fa30f005649) Thanks @vincentkoc. - Plugin-channel conversations keep their intended session more reliably through startup, doctor checks, and state repairs, with older binding records migrated into OpenClaw's shared database. [9f888d9](https://github.com/openclaw/openclaw/commit/9f888d95e082d50380a66db18ee2e32683e688e0) - Windows memory-backed session syncing now keeps using the intended transcript file even when path formatting differs. [b3b5b08](https://github.com/openclaw/openclaw/commit/b3b5b08e67a26efd648c7c7d879e5487223cd796) Thanks @vincentkoc. - Embedded agent runs with a missing or blank session key now stay attached to the intended session instead of being sent through inconsistent session routing. [911f853](https://github.com/openclaw/openclaw/commit/911f853b7fc4d819e2175b001662a01eba30453d) Thanks @vincentkoc. - When a model guesses the wrong tool name, Tool Search and Code Mode now show how to find and retry the correct tool, reducing the risk that long-running sessions get stuck or lose durable memory during compaction. [#93374](https://github.com/openclaw/openclaw/pull/93374) Related [#92273](https://github.com/openclaw/openclaw/issues/92273). Thanks @mushuiyu886, @poison, @vincentkoc. - Fixes assistant replies disappearing from webchat, Control UI, Feishu, and other embedded conversations after compaction, keeping refreshed chats readable and follow-up requests separate. [#95484](https://github.com/openclaw/openclaw/pull/95484) Related [#76729](https://github.com/openclaw/openclaw/issues/76729). Thanks @maweibin, @njuboy11, @vincentkoc. - OpenClaw memory features now keep active, reset, and deleted transcript coverage aligned with configured session stores and agent ownership, making dreaming, QMD exports, indexing, and sync less likely to miss or misattribute conversation history. [#96162](https://github.com/openclaw/openclaw/pull/96162) Thanks @jalehman. - Gateway TLS setup now rejects blank certificate or key paths clearly or uses OpenClaw's defaults, avoiding confusing startup and certificate-generation failures while preserving valid paths. [#94054](https://github.com/openclaw/openclaw/pull/94054) Thanks @miorbnli. - Configured plugin policies keep blocking or rewriting sensitive tool calls after Gateway registry changes, reloads, or later hook initialization instead of being silently skipped. [#94545](https://github.com/openclaw/openclaw/pull/94545) Thanks @jesse-merhi. - Mobile operators with `operator.approvals` can now see and resolve chat-triggered exec approvals on the iOS device that started the request, including while the app is open, without relying only on push notifications. [#95175](https://github.com/openclaw/openclaw/pull/95175) Thanks @joshavant. - Control UI users now get the patched DOMPurify release, reducing exposure to the GHSA-cmwh-pvxp-8882 sanitizer vulnerability without changing how the interface behaves. [#95691](https://github.com/openclaw/openclaw/pull/95691) Thanks @vincentkoc. - "Always allow" approvals for plugin conversation bindings now carry over from the old settings file and are less likely to be lost or overwritten when multiple OpenClaw processes are running. [ae41b00](https://github.com/openclaw/openclaw/commit/ae41b009224b0a8e3a990912503258d4478fb4d0) Thanks @vincentkoc. - Matrix users now see that the active recovery key is required before a forced cross-signing reset can proceed, preventing a second reset from leaving encryption recovery and room-key backups unusable. [#95720](https://github.com/openclaw/openclaw/pull/95720) Related [#78396](https://github.com/openclaw/openclaw/issues/78396). Thanks @jteddy, @vincentkoc, @xialonglee. #### Slack router relay mode [Slack router relay mode](https://docs.openclaw.ai/channels/slack) gives managed and multi-gateway deployments a supported way to centralize incoming Slack traffic while preserving gateway ownership of mentions, threads, and replies. - Managed Slack deployments can now use a central router to send mentions and ongoing threads to the right OpenClaw gateway while replies still appear through Slack. [#94707](https://github.com/openclaw/openclaw/pull/94707) Thanks @pash-openai, @sjf-oa. #### Raft External Agent wake bridge The [Raft channel](https://docs.openclaw.ai/channels/raft) and [Raft plugin](https://docs.openclaw.ai/plugins/reference/raft) now support the local CLI wake path for External Agents. - Raft External Agent operators can now wake an OpenClaw agent when a workspace has pending work through the supported local CLI bridge, with named profiles and checks for missing CLI prerequisites. [#95497](https://github.com/openclaw/openclaw/pull/95497) Thanks @vincentkoc. #### Official plugin installation and repair [Plugin management](https://docs.openclaw.ai/plugins/manage-plugins) and the [plugin inventory](https://docs.openclaw.ai/plugins/plugin-inventory) now cover more official integrations through normal installation, update, and repair workflows. - When `plugins.allow` uses a channel or package name instead of the real plugin id, startup guidance now identifies the unmatched entry and shows the discovered plugin ids needed to correct the configuration. [#68389](https://github.com/openclaw/openclaw/pull/68389) Related [#68352](https://github.com/openclaw/openclaw/issues/68352). Thanks @aym9999, @jirboy, @lyfuci, @pahuchi-joe, @zmxccxy. - Plugin trust warnings for first-time or fresh installs now include a ready-to-copy `plugins.allow` example and commands to list or inspect plugin ids, so users can resolve the warning before trusting or reinstalling plugin code. [#78105](https://github.com/openclaw/openclaw/pull/78105) Related [#68780](https://github.com/openclaw/openclaw/issues/68780). Thanks @jirboy, @pahuchi-joe. - Codex migrations now work with standard global plugin installs because `openclaw migrate` can find the installed provider instead of failing with `Unknown migration provider`. [#89612](https://github.com/openclaw/openclaw/pull/89612) Related [#89609](https://github.com/openclaw/openclaw/issues/89609). Thanks @mugabuga, @zerone0x. - Plugin installs and updates recover from stale OpenClaw-managed dependency pins instead of failing with `npm EOVERRIDE`, without later synchronization downgrading or removing packages users installed explicitly. [#91786](https://github.com/openclaw/openclaw/pull/91786) Related [#91772](https://github.com/openclaw/openclaw/issues/91772). Thanks @amknight, @mkdelta221. - Channel plugin developers can now carry native sender and conversation identifiers through hooks and selected exec workflows, giving integrations more precise routing without breaking existing sender and chat fields. [#91903](https://github.com/openclaw/openclaw/pull/91903) Thanks @lanzhi-lee, @vincentkoc. - Plugin discovery now repeats fewer blocking filesystem checks during startup, reducing avoidable cold-start work for bundled plugin trees, especially on slower Windows filesystems, without changing bundle discovery behavior. [#93919](https://github.com/openclaw/openclaw/pull/93919) Related [#76209](https://github.com/openclaw/openclaw/issues/76209). Thanks @ml12580, @shenhonglong456-ai. - Plugin Gateway methods now work through `openclaw gateway call` after registration, so plugin authors can use them from scripts and cron jobs instead of hitting an `unknown method` error. [#94154](https://github.com/openclaw/openclaw/pull/94154) Related [#94127](https://github.com/openclaw/openclaw/issues/94127). Thanks @brycemurray, @pick-cat, @vincentkoc. - ClawHub skill discovery and install checks are less likely to stall or crash OpenClaw because oversized or stalled marketplace responses are now stopped before they can exhaust memory. [#95226](https://github.com/openclaw/openclaw/pull/95226) Thanks @alix-007. - Pinned official plugins no longer stay on an old release when operators follow the repair advice from `openclaw doctor` or deep gateway status after an upgrade. [#95541](https://github.com/openclaw/openclaw/pull/95541) Thanks @ooiuuii, @vincentkoc. - Managed npm plugin updates are less likely to break work on a running gateway with missing-module errors, because the older plugin files remain available until a later gateway start cleans them up. [#95589](https://github.com/openclaw/openclaw/pull/95589) Thanks @ooiuuii, @vincentkoc. - Official plugin cards for supported brands now show recognizable icons in ClawHub and other catalogs, and plugin authors can provide marketplace artwork through the documented manifest field. [#95845](https://github.com/openclaw/openclaw/pull/95845) Thanks @patrick-erichsen. - Official plugin icons in ClawHub and other catalogs are no longer forced into the same hard-coded color, allowing Simple Icons to use its default artwork instead. [#95987](https://github.com/openclaw/openclaw/pull/95987) Thanks @patrick-erichsen. - Docker users now have an official `openclaw/openclaw` Docker Hub mirror alongside GHCR, with versioned beta releases kept from moving the stable `latest` and `main` aliases. [#97122](https://github.com/openclaw/openclaw/pull/97122) Thanks @vincentkoc. - Git-based OpenClaw installs now use the repository's pinned pnpm version even when another global pnpm or surrounding project package manager is present, so setup commands no longer run against the wrong package-manager environment. [bd74a62](https://github.com/openclaw/openclaw/commit/bd74a62118aa4774706359d9494116ded8c1f6e3) Thanks @vincentkoc. - ClawHub skill-card and update requests now complete or time out predictably even when they receive an unusually large timeout value. [8cd0c11](https://github.com/openclaw/openclaw/commit/8cd0c11227f6f4096d089cd6108d6f2ae31252b7) Thanks @vincentkoc. - Windows users can complete source installs without a llama.cpp setup step blocking or slowing them, and the installer restores their existing shell setting afterward. [ea9065b](https://github.com/openclaw/openclaw/commit/ea9065bc68dd4ff94495b85a7dcb4491cf41b67a) Thanks @vincentkoc. - More official channel, provider, and web-search plugins can now be installed or repaired through normal external package catalogs while still being recognized from their existing credentials. [#95683](https://github.com/openclaw/openclaw/pull/95683) Thanks @vincentkoc. ### Channels and Messaging #### Additional channel fixes Additional [Telegram](https://docs.openclaw.ai/channels/telegram) and channel configuration fixes cover narrower delivery and setup problems. - Telegram reply chains keep cached replies attached after context changes instead of failing when those cached replies are reused. [#82909](https://github.com/openclaw/openclaw/pull/82909) Thanks @lidge-jun. - Fixes Discord dropping an entire long reply with fenced code blocks when a closing code fence lands near the 2,000-character message limit. [#95661](https://github.com/openclaw/openclaw/pull/95661) Thanks @ly-wang19. - Slack operators can now store tokens and signing secrets as supported SecretRef inputs, while reads, writes, allowlist and target lookups, and setup checks use the resolved credentials instead of rejecting or misreading the references. [7da955f](https://github.com/openclaw/openclaw/commit/7da955fae4ca2083599aa33a1f93dbfff53cb187) Thanks @vincentkoc. - Channel capability checks now return a clear timeout when an integration stops responding, keeping troubleshooting from hanging in a terminal or automation run. [8ecdb97](https://github.com/openclaw/openclaw/commit/8ecdb97b636e4c3fcc6d142d217327404ae06581) Thanks @vincentkoc. ### Providers and Models #### Additional provider and model fixes Additional [model configuration](https://docs.openclaw.ai/providers/models) fixes improve selection, usage reporting, status output, and streaming. - After changing the default model, starting a fresh channel session with `/new` or `/reset` now uses the new default instead of silently reusing the previous cached model, while explicit `/model` overrides remain unchanged. [#77339](https://github.com/openclaw/openclaw/pull/77339) Related [#77322](https://github.com/openclaw/openclaw/issues/77322). Thanks @mjamiv, @zaynl. - Behind HTTP or HTTPS proxies, Codex/OpenAI usage and quota checks in `openclaw status --usage --json` and the Control UI now retrieve usage windows instead of failing when chatgpt.com is unreachable directly. [#93943](https://github.com/openclaw/openclaw/pull/93943) Related [#78714](https://github.com/openclaw/openclaw/issues/78714). Thanks @tnzgit, @turbotheturtle. - `/status` now keeps the active model and how to clear a pinned choice on one compact line, so Discord and other chat users can scan model status without a multi-line explanation. [#95797](https://github.com/openclaw/openclaw/pull/95797) Thanks @solvely-colin. - Anthropic streaming responses now keep interleaved text, thinking, and tool-call updates attached to the correct response block instead of mixing them when several blocks are active at once. [#96013](https://github.com/openclaw/openclaw/pull/96013) Thanks @vincentkoc. ### Memory, Sessions, and State #### Additional session and memory fixes Additional [memory](https://docs.openclaw.ai/concepts/memory) fixes cover resume, indexing, synchronization, and cleanup edge cases. - Memory-wiki status cards and bridge-backed source sync are less likely to fail during simultaneous page rewrites because OpenClaw now retries the transient path mismatch while still stopping unsafe or persistent filesystem writes. [#94443](https://github.com/openclaw/openclaw/pull/94443) Related [#92134](https://github.com/openclaw/openclaw/issues/92134). Thanks @cknzraposo, @zengwen-dt. - Fixes recent-session resume opening a fresh conversation for users with long workspace paths instead of returning to their existing transcript. [#94578](https://github.com/openclaw/openclaw/pull/94578) Related [#94577](https://github.com/openclaw/openclaw/issues/94577). Thanks @rohitjavvadi, @vincentkoc. - Memory Wiki now keeps user-written notes intact when an existing source page is re-ingested or synced, while still refreshing its generated content. [#95614](https://github.com/openclaw/openclaw/pull/95614) Thanks @yetval. - Fixes Memory Wiki repeatedly copying its own generated source pages back into itself when its vault is stored inside the workspace memory folder, avoiding duplicate files, repeated cleanup, and unnecessary memory index growth. [#95666](https://github.com/openclaw/openclaw/pull/95666) Related [#95657](https://github.com/openclaw/openclaw/issues/95657). Thanks @johannes0402, @turbotheturtle, @vincentkoc. - For operators using Active Memory with memory-core dreaming, nightly dreaming jobs no longer start unnecessary recall work and hit 45-second timeouts, while regular web chats continue to receive memory recall. [#95721](https://github.com/openclaw/openclaw/pull/95721) Related [#78500](https://github.com/openclaw/openclaw/issues/78500). Thanks @vincentkoc, @vishutdhar, @xialonglee. - Agent sessions with many tool calls repair out-of-order results with less repeated work while keeping each result paired with the right tool call. [#96014](https://github.com/openclaw/openclaw/pull/96014) Thanks @vincentkoc. - Windows qmd-backed memory work now stops all related processes after availability probes and command timeouts, preventing qmd children from continuing to run in the background. [830691b](https://github.com/openclaw/openclaw/commit/830691b2010bd0406399adb8a2e97e0b043e2ca8) ### Gateway, Security, and Trust #### Restart and readiness recovery [Gateway health](https://docs.openclaw.ai/gateway/health), [troubleshooting](https://docs.openclaw.ai/gateway/troubleshooting), and [`openclaw agent`](https://docs.openclaw.ai/cli/agent) workflows now report and recover from more runtime failures clearly. - When a configured or explicit remote gateway is slow but reachable, `openclaw gateway probe --timeout ...` now waits for the requested timeout instead of reporting it unreachable after a shorter internal cutoff. [#89859](https://github.com/openclaw/openclaw/pull/89859) Related [#65355](https://github.com/openclaw/openclaw/issues/65355). Thanks @hellocli, @mushuiyu886. - Long or parallel internal subagent runs now avoid unnecessary live-preview processing, while visible subagent sessions still show live updates and final responses. [#91906](https://github.com/openclaw/openclaw/pull/91906) Thanks @lanzhi-lee, @vincentkoc. - ACP conversations, especially Kiro-backed threads, now continue past the first reply by starting a fresh session when the backend can no longer resume the old one. [#93547](https://github.com/openclaw/openclaw/pull/93547) Related [#87830](https://github.com/openclaw/openclaw/issues/87830). Thanks @amersheeny, @chouzz. - When Linux memory pressure kills a child command or session, systemd-managed OpenClaw gateways now stay running and keep channel connections alive while reporting the child failure. [#93585](https://github.com/openclaw/openclaw/pull/93585) Thanks @snowzlm. - Canceling an OpenClaw run during tool work now ends it promptly instead of starting another model turn or leaving the session locked. [#94412](https://github.com/openclaw/openclaw/pull/94412) Thanks @szsip239, @vincentkoc. - Scheduled OpenClaw jobs using cloud models now recover from silent, stuck model calls by default, helping prevent later cron work from backing up while local or self-hosted providers keep their existing timeout behavior. [#94445](https://github.com/openclaw/openclaw/pull/94445) Thanks @bek91. - Gateway readiness checks now turn unhealthy during a restart drain, preventing traffic managers from sending new work to a Gateway that is temporarily rejecting requests. [#94915](https://github.com/openclaw/openclaw/pull/94915) Related [#78136](https://github.com/openclaw/openclaw/issues/78136). Thanks @markoub, @maxschachere, @vincentkoc. - Mac users can keep LaunchAgent-managed gateways running through OpenClaw upgrades instead of seeing repeated crash-and-restart loops when older text-transform runtime code is still cached. [#95081](https://github.com/openclaw/openclaw/pull/95081) Related [#95057](https://github.com/openclaw/openclaw/issues/95057). Thanks @849261680, @yveslarose. - Codex-powered conversations in TUI, WebChat, and compatible streaming APIs now show replies as they are written, while replacing provisional text cleanly so the final answer does not include stale drafts. [#95404](https://github.com/openclaw/openclaw/pull/95404) Related [#95422](https://github.com/openclaw/openclaw/issues/95422). Thanks @agonza1, @vincentkoc. - After a gateway restart, users no longer see a misleading retry notice when OpenClaw is already resuming the interrupted reply or reporting the actual recovery failure, reducing unnecessary duplicate attempts. [#95431](https://github.com/openclaw/openclaw/pull/95431) Thanks @moeedahmed, @vincentkoc. - Long, tool-heavy agent sessions now retain prompt-cache savings as results accumulate, reducing avoidable delays and cost from resending rewritten history between turns. [#95624](https://github.com/openclaw/openclaw/pull/95624) Thanks @vincentkoc. - Gateway restarts no longer leave configured Codex, Copilot, or trusted plugin-based agents temporarily unavailable, and untrusted workspace plugins remain blocked from activating themselves. [#95652](https://github.com/openclaw/openclaw/pull/95652) Thanks @vincentkoc. - Long responses, busy tool streams, image-heavy requests, and memory recall now incur less CPU and filesystem overhead without requiring settings or workflow changes. [#95697](https://github.com/openclaw/openclaw/pull/95697) Thanks @vincentkoc. - Operators can again add or update scheduled announcements for known channels in no-config setups, while configured environments still reject disabled, stale, ownerless, or unknown destinations before delivery. [#95754](https://github.com/openclaw/openclaw/pull/95754) Thanks @vincentkoc. - macOS users are less likely to see a false port-conflict failure when stopping or updating a managed gateway, because OpenClaw briefly waits for normal shutdown to release the port while still reporting conflicts that persist. [#95886](https://github.com/openclaw/openclaw/pull/95886) Thanks @fuller-stack-dev. - Copilot-backed agents can now ask users a question and accept the answer through OpenClaw's normal chat reply flow, while compact tool-search and code-mode controls avoid loading the full tool catalog into the session. [#96005](https://github.com/openclaw/openclaw/pull/96005) Thanks @vincentkoc. - Gateway restarts on systemd or container setups no longer leave old Codex or Claude adapter processes behind, helping new ACPX sessions start without minutes-long cleanup stalls after repeated restarts. [#96032](https://github.com/openclaw/openclaw/pull/96032) Thanks @t2wei, @vincentkoc. - Copilot-backed sessions now show plan updates as work unfolds, and their native child tasks stay visible through completion or failure instead of disappearing from OpenClaw's task view. [#96062](https://github.com/openclaw/openclaw/pull/96062) Thanks @vincentkoc. - Connected agents such as OpenCode now start through OpenClaw even when their harness cannot select a requested model, while genuinely unsupported model choices still return the original error. [#96068](https://github.com/openclaw/openclaw/pull/96068) Related [#95869](https://github.com/openclaw/openclaw/issues/95869). Thanks @sabatech-dev, @vincentkoc. - Plugins using `heartbeat_prompt_contribution` now deliver their heartbeat-specific context to models when agents run through harness runtimes such as the Codex app-server, without affecting ordinary user turns or plugins that do not use the hook. [#96233](https://github.com/openclaw/openclaw/pull/96233) Thanks @azogheb, @vincentkoc. - Windows gateway cleanup and listener checks now handle UTF-16 WMIC command-line data consistently, reducing failed or conflicting identification of the running gateway process. [15c880a](https://github.com/openclaw/openclaw/commit/15c880aeff1f4e55964ad7204d14733a2d1362f7) Thanks @vincentkoc. - Long-context, tool-heavy agent sessions now keep prompt-cache reuse steadier across repeated turns without losing per-result size limits, while advanced operators can configure larger tool-result caps for large-context models without configuration rejection. [a60947f](https://github.com/openclaw/openclaw/commit/a60947fb3e92f45ea7eb2581da8877b10a8bebb2) Thanks @vincentkoc. - Long, tool-heavy agent sessions are less likely to bloat model requests as tool output accumulates, while repeated turns keep stable prompt-cache reuse. [2f33999](https://github.com/openclaw/openclaw/commit/2f3399989893e6af18be49fb810e58941d7a4a45) Thanks @vincentkoc. - Gateway restarts now use OpenClaw's durable state database for the handoff, while stale, malformed, wrong-process, or superseded requests are discarded before they can affect the restart. [0ad48da](https://github.com/openclaw/openclaw/commit/0ad48dad2c4747f255d5a156b94578f7d87386a1) - Gateway status, doctor, and restart diagnostics now retain recent restart details in OpenClaw's shared state database, while expired or malformed records are still discarded. [a39a3b7](https://github.com/openclaw/openclaw/commit/a39a3b74de05f06227ede904a73c1b4687679d3e) Thanks @vincentkoc. - Gateway restarts and managed-service updates now keep the correct continuation message, avoid reusing stale handoff state, and mark failed update handoffs consistently. [514b336](https://github.com/openclaw/openclaw/commit/514b3365b54c8b3493eaf8a94198b7c04ea34aec) Thanks @vincentkoc. - Gateway-launched agents no longer lose owner-only OpenClaw tools during tasks such as live cron checks, so authorized operations can use the intended tools with the correct request context. [c2ee9b0](https://github.com/openclaw/openclaw/commit/c2ee9b0be8aeeadedffc8c6aaa9f5f291283fea5) Thanks @vincentkoc. - Malformed gateway restart requests now fail clearly without scheduling a restart, preventing bad or accidental integration calls from unexpectedly restarting the gateway. [108d6d7](https://github.com/openclaw/openclaw/commit/108d6d7eca0000a736e28d198e77100f4d7774e5) Thanks @vincentkoc. - Stale node requests queued by the gateway now expire automatically, so old work is less likely to linger and affect later activity. [f6d432e](https://github.com/openclaw/openclaw/commit/f6d432e545e2e2be91d17badc48354a0135e5294) - Plugin workflows are less likely to stall or overload the gateway when an integration requests too much subagent session history, because each read is now capped at a safe limit. [b66b450](https://github.com/openclaw/openclaw/commit/b66b4504f87205dd8ba0393e763e23a4a6158a79) Thanks @vincentkoc. - Image descriptions now handle extremely large timeout settings consistently by capping them to a safe runtime limit instead of risking timer overflow. [88b21fc](https://github.com/openclaw/openclaw/commit/88b21fc30b4ac615b3d3870e483190bc832f9846) - Embedded agent sessions now wait reliably for another session to release its file lock, even with an extremely large timeout, instead of risking timer overflow. [4c736df](https://github.com/openclaw/openclaw/commit/4c736df975fed8e39f18db43bfaea9654b9ac0c7) Thanks @vincentkoc. - Queued commands with extremely large task timeout settings now time out reliably because OpenClaw caps the wait at the runtime's safe maximum. [1f6ae32](https://github.com/openclaw/openclaw/commit/1f6ae32cabb9d5ed308bb30715c287936ef483f1) Thanks @vincentkoc. - Fixes normalization-core exposing the wrong string-coercion entry point and ACP sessions showing an outdated fast-mode value, so integrations receive the intended API and users see the mode actually in effect. [93ad397](https://github.com/openclaw/openclaw/commit/93ad39772590a34be3821d83709a5050c186fdf9) Thanks @vincentkoc. - Fast auto runs now deliver final replies more consistently, with progress-reset handling limited to automatic mode so it does not interfere with responses or forwarded callbacks. [9e8ab08](https://github.com/openclaw/openclaw/commit/9e8ab083dd6b9df4dadf40ee523ff18ac1472bd3) Thanks @vincentkoc. - Agent sessions using OpenAI Responses now resume tool-based work without failing or losing progress when replayed history contains mismatched tool requests and results. [b4bc1f2](https://github.com/openclaw/openclaw/commit/b4bc1f20c9fca6d316561d42cbabdd793c67a6e7) Thanks @vincentkoc. - Completed plugin subagent and QA runs are no longer misreported as failures when gateways return alternate completion envelope shapes, making successful handoffs more reliable for plugin authors and operators. [d1b268f](https://github.com/openclaw/openclaw/commit/d1b268f7f7f7309dd5db99728019218b4d453e18) Thanks @vincentkoc. - Fixes completed subagent tasks sometimes ending without an update, so users receive the result or the parent agent's next step. [68a1e00](https://github.com/openclaw/openclaw/commit/68a1e00b73bd746f6fb577f4127fecb0ade9e228) Thanks @vincentkoc. #### Remote result and media delivery Additional [Gateway](https://docs.openclaw.ai/gateway/index) fixes return remote media and completed subagent results to the active conversation more reliably. - Generated images from a remote Codex app-server now arrive as attachments instead of showing `Media failed` or returning only text after successful generation. [#96212](https://github.com/openclaw/openclaw/pull/96212) Thanks @sjf-oa. - When a subagent finishes, its result now reaches the active parent run more reliably instead of appearing silent. [7fc4bbc](https://github.com/openclaw/openclaw/commit/7fc4bbc0bcbabc2aa99b1fd51e77099f2b26f4e1) Thanks @vincentkoc. #### Additional security and trust fixes Additional [secret-handling](https://docs.openclaw.ai/gateway/secrets) hardening keeps trusted package paths within the intended boundary. - Trusted OpenClaw package sources now reject lookalike sibling paths, so trusting `/artifactory/openclaw` no longer also admits paths such as `/artifactory/openclaw-malicious`. [12c34fc](https://github.com/openclaw/openclaw/commit/12c34fc3a95121f5a15f01c3f971a1bc5b0fe6f9) Thanks @vincentkoc. ### Clients and Interfaces #### Client sends and reconnects [WebChat](https://docs.openclaw.ai/web/webchat), the [Control UI](https://docs.openclaw.ai/web/control-ui), mobile clients, and the [terminal UI](https://docs.openclaw.ai/cli/tui) now recover completed, rejected, or interrupted sends and reconnects without leaving conversations looking stuck. - When a WebChat message fails before the agent starts, WebChat and Control UI now show the session as failed instead of leaving it looking like it is still running. [#84352](https://github.com/openclaw/openclaw/pull/84352) Thanks @jesse-merhi. - Fixes the Control UI session picker getting stuck behind hidden subagent sessions, so Load More reaches the next usable chat without showing a misleading total. [#89323](https://github.com/openclaw/openclaw/pull/89323) Related [#89249](https://github.com/openclaw/openclaw/issues/89249). Thanks @giodl73-repo, @originsecured-do. - When users reopen a Control UI conversation from History, their prompts now appear with the assistant's replies, preserving the question-and-answer context without blank gaps in long transcripts. [#93841](https://github.com/openclaw/openclaw/pull/93841) Related [#90241](https://github.com/openclaw/openclaw/issues/90241). Thanks @mushuiyu886, @pronzcw. - Control UI deployments behind a path prefix now keep manifest, favicon, and service-worker requests under that prefix, avoiding confusing root-level 403 errors after login. [#94204](https://github.com/openclaw/openclaw/pull/94204) Related [#94157](https://github.com/openclaw/openclaw/issues/94157). Thanks @hugenshen, @xrow. - Android users can now open Health log and Skill rows in Settings for readable details, making it easier to troubleshoot gateway activity, check skill setup and status, and understand how to pair with an existing setup code. [#95148](https://github.com/openclaw/openclaw/pull/95148) Thanks @tosko4. - Sent prompts no longer reappear in the Control UI composer after a send, so users can switch sessions or start their next message without risking a duplicate send or overwriting a new draft, while intentional re-entry still works. [#95503](https://github.com/openclaw/openclaw/pull/95503) Related [#89466](https://github.com/openclaw/openclaw/issues/89466). Thanks @vincentkoc, @zhangguiping-xydt, @zhong18804784882. - Android users now get a cleaner Overview where connection status, the configured agent, node health, approvals, recent sessions, and Chat and Talk actions are visible at a glance. [#95557](https://github.com/openclaw/openclaw/pull/95557) Thanks @joshavant, @solvely-colin. - Android users can now refresh and resolve gateway command approvals from the in-app Approvals screen, choosing Allow Once, Always, or Deny while connected. [#95593](https://github.com/openclaw/openclaw/pull/95593) Thanks @solvely-colin. - iOS users now avoid surprise notification prompts and get clear guidance when approval alerts are unavailable, with permission managed from one predictable Settings screen. [#95640](https://github.com/openclaw/openclaw/pull/95640) Thanks @joshavant. - Local TUI shutdowns now stay within safe timer limits even when `OPENCLAW_TUI_LOCAL_RUN_SHUTDOWN_GRACE_MS` is set extremely high. [c21dcfc](https://github.com/openclaw/openclaw/commit/c21dcfc7c272201484514bbc096bd51ab112bd47) - Canvas A2UI now serves only the current app assets after each build, so outdated compatibility images and leftover files are less likely to appear. [a89e65c](https://github.com/openclaw/openclaw/commit/a89e65c167f3280b047baff931f52e393875c892) Thanks @vincentkoc. - iOS push relay setup failures are easier to pinpoint because registration diagnostics show where setup stopped while keeping sensitive push credentials out of logs. [f2b8668](https://github.com/openclaw/openclaw/commit/f2b8668a549b50339a96e47020a12615734640a9) Thanks @joshavant. - iOS devices are now enrolled for push notifications only after users accept the hosted relay disclosure and allow notifications, preventing registration data from being published before consent. [8efed50](https://github.com/openclaw/openclaw/commit/8efed50c4ed33105cfed9f2f96532fc9fe2d7e6d) Thanks @joshavant. - Fixes chat, voice, TUI, and forwarded sends sometimes appearing stuck or disappearing after the gateway had already finished or rejected them, so affected clients now clear the pending state, restore retryable input, refresh history, or show a useful failure. [#91049](https://github.com/openclaw/openclaw/pull/91049) Related [#91048](https://github.com/openclaw/openclaw/issues/91048). Thanks @nxmxbbd. #### Additional client interface fixes Additional [Control UI](https://docs.openclaw.ai/web/control-ui), mobile, and desktop fixes improve display accuracy, accessibility, onboarding, and app behavior. - Restores the OpenAI/Codex usage quota in the expanded Control UI chat sidebar, so users can check their limits without leaving the conversation. [#94219](https://github.com/openclaw/openclaw/pull/94219) Related [#93041](https://github.com/openclaw/openclaw/issues/93041). Thanks @jazzroutine, @pick-cat. - iOS screens now use consistent OpenClaw accent and status colors across onboarding, settings, chat, approval prompts, voice permissions, widgets, and shared chat views. [#94627](https://github.com/openclaw/openclaw/pull/94627) Thanks @zats. - The Control UI can now create Early Morning jobs with the Silent preset in the main session and without notifications, instead of leaving the dialog open with no visible result. [#95459](https://github.com/openclaw/openclaw/pull/95459) Related [#95073](https://github.com/openclaw/openclaw/issues/95073). Thanks @vincentkoc, @vporton, @zoowh. - At the million-token boundary, Control UI badges and usage readouts now show "1M" instead of the confusing "1000k", while the underlying token counts remain unchanged. [#95485](https://github.com/openclaw/openclaw/pull/95485) Thanks @narahariraghava, @vincentkoc. - The Control UI Overview now counts and flags only enabled cron jobs that still need attention, while disabled jobs retain their past failure details without appearing as current problems. [#95723](https://github.com/openclaw/openclaw/pull/95723) Related [#95716](https://github.com/openclaw/openclaw/issues/95716). Thanks @voytas75, @zengwen-dt. - Control UI users now see shorter System, Light, and Dark theme tooltips, while screen readers announce less repetitive labels without losing the surrounding Color mode context. [#95837](https://github.com/openclaw/openclaw/pull/95837) Thanks @hannesrudolph, @sannidhyasah. - Raw configuration no longer appears missing in Settings after switching from the form view, because the JSON is brought back into view instead of retaining the previous scroll position. [#96145](https://github.com/openclaw/openclaw/pull/96145) Related [#94202](https://github.com/openclaw/openclaw/issues/94202). Thanks @sunlit-deng, @vporton. - New iOS users now reach OpenClaw's welcome and onboarding before iOS asks for local-network access, while existing users still get the request when opening gateway setup or otherwise needing LAN gateway discovery. [#96181](https://github.com/openclaw/openclaw/pull/96181) Thanks @joshavant. ### Plugins and Packaging #### Additional plugin and package fixes Additional [plugin management](https://docs.openclaw.ai/plugins/manage-plugins) fixes improve installation, verification, package resolution, and updates. - People installing or updating the official Yuanbao channel plugin through OpenClaw's trusted catalog now get version 2.15.0, with the expected integrity check and missing-plugin guidance aligned to that release. [#94470](https://github.com/openclaw/openclaw/pull/94470) Thanks @jase-283. - First-run onboarding can now install the bundled `gog` skill through Homebrew without failing on the removed third-party tap formula. [#95019](https://github.com/openclaw/openclaw/pull/95019) Related [#95017](https://github.com/openclaw/openclaw/issues/95017). Thanks @sedrak-hovhannisyan, @vincentkoc, @zengwen-dt. - Canvas, Discord, Slack, Voice Call, and WhatsApp users keep the same skill guidance with each installed or bundled plugin, while references to the former root `skills/...` paths need to move into the relevant plugin directory. [#95664](https://github.com/openclaw/openclaw/pull/95664) Thanks @vincentkoc. - ClawHub skill verification now accepts the same `@owner/<slug>` reference used for installs and updates, so users can check the intended publisher without switching to an ambiguous bare slug. [#95992](https://github.com/openclaw/openclaw/pull/95992) Thanks @patrick-erichsen. - OpenClaw's install-time package-manager warning now identifies npm, Yarn, Yarn Berry, and Corepack-style launchers correctly, avoiding misleading guidance when those tools run through alternate executable names. [11a2e03](https://github.com/openclaw/openclaw/commit/11a2e03bd4deda748336553710cb6426d448d952) Thanks @vincentkoc. - Package URL installs now handle oversized download timeouts without failing before available package data can be resolved. [c310f8c](https://github.com/openclaw/openclaw/commit/c310f8cfa4524453c7082bb5aab642c9decc6e99) - Plugin and CLI developers now see a clearer supported command-formatting API, while device pairing, node registration, and doctor guidance keep producing the same shell-safe commands. [23b4f33](https://github.com/openclaw/openclaw/commit/23b4f33195933ff4def4609d970f073293760683) - Windows ARM64 users now get matching ARM64 Node and MinGit downloads when running the PowerShell installer through an x64-emulated shell. [fac091b](https://github.com/openclaw/openclaw/commit/fac091b39de230d9ed90e8412123126cf676004f) Thanks @vincentkoc. - Default OpenClaw installs no longer spend time building optional llama.cpp support, avoiding native-build failures for users who did not enable it. [cc1b3a8](https://github.com/openclaw/openclaw/commit/cc1b3a8550dd9c29f581799934d140aaf5a84f1c) Thanks @vincentkoc. ### Docs and Admin Tools #### Setup and command reliability The [OpenClaw CLI](https://docs.openclaw.ai/cli/index), [tab completion](https://docs.openclaw.ai/cli/completion), and [`openclaw doctor`](https://docs.openclaw.ai/cli/doctor) now handle more setup and repair cases cleanly. - OpenClaw's zsh tab-completion menu now displays option descriptions containing `$` variables or backtick-wrapped examples literally instead of evaluating them as shell input and corrupting the menu. [#64490](https://github.com/openclaw/openclaw/pull/64490) Thanks @edenkangdw. - After upgrading from older sandbox storage, operators now get a clear `openclaw doctor` warning about leftover registry files and can use `openclaw doctor --fix` to migrate or clean them up. [#84326](https://github.com/openclaw/openclaw/pull/84326) Thanks @giodl73-repo. - Operators can now use `doctor --lint` to spot stale legacy Gateway services and preview cleanup, while intentional extra services remain informational and do not fail the default check. [#84340](https://github.com/openclaw/openclaw/pull/84340) Thanks @giodl73-repo. - macOS gateway operators now get a warning before reinstall, repair, or restart overwrites custom LaunchAgent wrapper behavior, while `openclaw status` distinguishes CLI-only missing-secret checks from the installed service. [#90537](https://github.com/openclaw/openclaw/pull/90537) Related [#90518](https://github.com/openclaw/openclaw/issues/90518). Thanks @turbotheturtle, @vincentkoc. - Long, multiline, or code-heavy prompts can now be sent to `openclaw agent` with `--message-file`, avoiding fragile shell quoting and reporting invalid files before dispatch. [#93351](https://github.com/openclaw/openclaw/pull/93351) Thanks @ooiuuii. - Fixes scheduled `doctor --fix --non-interactive` repairs restarting an already-running gateway after a temporary health-check failure, so unattended maintenance no longer interrupts the live service. [#94148](https://github.com/openclaw/openclaw/pull/94148) Related [#78217](https://github.com/openclaw/openclaw/issues/78217). Thanks @esqandil, @zhangguiping-xydt. - `openclaw configure` and bare `openclaw config` now stop with clear subcommand guidance when run from scripts or pipes, instead of opening a partial interactive wizard and exiting unclearly. [#94238](https://github.com/openclaw/openclaw/pull/94238) Related [#93953](https://github.com/openclaw/openclaw/issues/93953). Thanks @nianjiuzst, @ruomuxydt. - Multi-agent operators can now use `openclaw gateway usage-cost` to view costs for one configured agent or all agents while the existing default-agent command remains unchanged. [#94483](https://github.com/openclaw/openclaw/pull/94483) Thanks @ly-wang19. - Archived Workboard cards no longer clutter the default `openclaw workboard list` output, while `--include-archived` and JSON output still provide access when needed. [#94562](https://github.com/openclaw/openclaw/pull/94562) Related [#94555](https://github.com/openclaw/openclaw/issues/94555). Thanks @ecican, @vincentkoc, @zengwen-dt. - OpenClaw Doctor now gives accurate guidance for working isolated shell-prompt cron jobs instead of repeatedly suggesting a `--fix` command that cannot clear the warning. [#94784](https://github.com/openclaw/openclaw/pull/94784) Related [#94655](https://github.com/openclaw/openclaw/issues/94655). Thanks @altaywtf, @geekoagent, @zengwen-dt. - Fixes `openclaw doctor` showing a fix-required warning for healthy local GGUF memory setups after an intentionally skipped readiness check, while preserving the warning when the configured local model is actually missing. [#95393](https://github.com/openclaw/openclaw/pull/95393) Related [#92582](https://github.com/openclaw/openclaw/issues/92582). Thanks @mikasa0818, @neekolascmd, @vincentkoc. - On Windows, installer-created gateway tasks now run in the background without a console window that users could accidentally close and stop the gateway. [#95480](https://github.com/openclaw/openclaw/pull/95480) Related [#89231](https://github.com/openclaw/openclaw/issues/89231). Thanks @cameronweller, @mikasa0818, @vincentkoc. - Agent channel bindings now reject malformed account specs such as `matrix:work:extra` with a clear error instead of silently routing the agent to a different account. [#95572](https://github.com/openclaw/openclaw/pull/95572) Thanks @ly-wang19. - ClawHub skill updates now honor your configured install safety policy, and `openclaw skills update --all` updates only tracked ClawHub skills instead of unexpectedly installing other configured skills. [#95684](https://github.com/openclaw/openclaw/pull/95684) Thanks @vincentkoc. - Windows restart and gateway startup workflows are more reliable because OpenClaw now hands commands to the trusted system `cmd.exe` path instead of depending on process lookup. [7dd01d1](https://github.com/openclaw/openclaw/commit/7dd01d15c56da2ee50f55746ba725d708682fca9) Thanks @vincentkoc. - Windows gateway cleanup and listener checks are more reliable when PATH lookup is incomplete, so operators can identify the gateway process and free an occupied port without installed system tools being missed. [e9b694e](https://github.com/openclaw/openclaw/commit/e9b694ef9cd8b7528a76d85b8be2a830b296a5e1) Thanks @vincentkoc. - On Windows, OpenClaw startup and TUI Codex handoff now find `bun`, `codex`, and other runtime binaries through the trusted system locator even when another `where` command appears earlier on PATH. [72b9bc7](https://github.com/openclaw/openclaw/commit/72b9bc730370e3b1155fb231621c2bc7b2c87b56) Thanks @vincentkoc. - Windows port diagnostics now use the intended system tools even when PATH entries are missing or shadowed, so gateway and service port conflicts are less likely to be obscured by command-resolution failures. [c4facb2](https://github.com/openclaw/openclaw/commit/c4facb2bb372e99037f497e2640ca7bdc5cbc5f6) Thanks @vincentkoc. - Windows daemon recovery is less likely to miss process detection or cleanup when PATH is incomplete, unusual, or shadowed because scheduled-task fallback now finds PowerShell and taskkill in trusted system locations. [2a140e6](https://github.com/openclaw/openclaw/commit/2a140e6e6ae8c48edb5bd52d8d177915f0555d70) Thanks @vincentkoc. - OpenClaw now keeps config recovery markers, last-known-good snapshots, and suspicious-read history in its shared state through migration, without leaving a separate config-health log file behind. [6daabd2](https://github.com/openclaw/openclaw/commit/6daabd23f821c66154739de4b0f103e33343333c) Thanks @vincentkoc. - On Windows, Crabbox commands launched through Node package shims now receive provider flags, shell commands, and special shell characters as entered instead of losing or reinterpreting them. [54d24cd](https://github.com/openclaw/openclaw/commit/54d24cd956ff91f4fa8c4924f17c06798c1e0359) Thanks @vincentkoc. - Windows-targeted Crabbox workflows are less likely to fail or fall back to slower shell handling when launching Node tools through `.cmd` and `.bat` shims. [d48dcc6](https://github.com/openclaw/openclaw/commit/d48dcc664bc6e1106a61942a951745886f22d582) Thanks @vincentkoc. - Windows users can run `crabbox`, `git`, and other Node-backed tools through npm-installed command shims without Crabbox stopping before the tool opens. [77f4e45](https://github.com/openclaw/openclaw/commit/77f4e45c3518751b5f586eac193c4aee904f02d9) Thanks @vincentkoc. - `openclaw doctor` now checks profiles that omit tool policy settings without treating the valid omission as an error. [03ba09b](https://github.com/openclaw/openclaw/commit/03ba09bfa8676832d55bdc7724e79d9980fdd2d7) - `openclaw doctor` no longer shows misleading tool-section warnings when it cannot evaluate a custom preview profile. [420a0e6](https://github.com/openclaw/openclaw/commit/420a0e6fce4b2c5339e535e6b307f50df1c00bb2) Thanks @vincentkoc. - `openclaw doctor` now limits preview warnings to tool profiles it can evaluate, avoiding misleading configured-grant warnings for unknown profiles. [541f7ff](https://github.com/openclaw/openclaw/commit/541f7ffc6558c0e59a8afca066a9f00884d39b65) Thanks @vincentkoc. - Windows users can install OpenClaw from source without dependency setup being blocked by the installer forcing npm or pnpm scripts through `cmd.exe`. [1252378](https://github.com/openclaw/openclaw/commit/1252378018f899bfc110914bac7cba94b65b8930) Thanks @vincentkoc. #### Tools and scheduled work [Scheduled jobs](https://docs.openclaw.ai/cli/cron) and [tool workflows](https://docs.openclaw.ai/tools/index) now finish, validate, and report failures more consistently. - Isolated cron jobs using `deleteAfterRun` now remove their temporary session and transcript after finishing, including runs with delivery disabled, reducing stale files, accumulated context, and manual cleanup. [#84794](https://github.com/openclaw/openclaw/pull/84794) Related [#84707](https://github.com/openclaw/openclaw/issues/84707). Thanks @bottenbenny, @turbotheturtle. - Individual scheduled jobs can now use their own fallback models, run with fallbacks disabled, or return to normal fallback inheritance through the CLI instead of requiring operators to edit lower-level payload data. [#93369](https://github.com/openclaw/openclaw/pull/93369) Related [#90302](https://github.com/openclaw/openclaw/issues/90302). Thanks @849261680, @walliiee. - Cron history now reliably finds entries whose job IDs include extra surrounding spaces, and rejects nested or blank IDs before they can create log records that cannot be read back safely. [#93567](https://github.com/openclaw/openclaw/pull/93567) Thanks @alix-007, @vincentkoc. - Adding or removing a cron job no longer causes another recurring job that is already due to lose its pending run. [#94323](https://github.com/openclaw/openclaw/pull/94323) Thanks @yetval. - Word, PowerPoint, and Excel document reads and writes now use the intended `.docx`, `.pptx`, or `.xlsx` path instead of failing against a made-up extension. [#95805](https://github.com/openclaw/openclaw/pull/95805) Related [#93326](https://github.com/openclaw/openclaw/issues/93326). Thanks @bhnan, @lzyyzznl, @vincentkoc, @xzh-icenter. - Browser automation users keep the same reference-rich snapshots, including useful branches in compact results, with less avoidable processing during snapshot generation. [#96072](https://github.com/openclaw/openclaw/pull/96072) Thanks @vincentkoc. - Fixes timed-out commands and interrupted core updates on Windows sometimes leaving child processes running, so OpenClaw can stop the full process tree more reliably after cancellations, timeouts, or update cleanup. [a192b2e](https://github.com/openclaw/openclaw/commit/a192b2ea52b3166a7d190bf5f60f3feb030306bb) Thanks @vincentkoc. - Windows users are less likely to see agent-managed tool installs fail while unpacking ZIP downloads such as ripgrep, because OpenClaw now uses the built-in Windows extraction programs instead of relying on PATH lookup. [a5fde91](https://github.com/openclaw/openclaw/commit/a5fde9119c9c50685a392bff078a2f76a03d749d) - Windows setup and runtime checks now find required tools more reliably by using the trusted System32 resolver instead of depending on an unexpected PATH entry. [d3b4444](https://github.com/openclaw/openclaw/commit/d3b44442f6c8bedd765dc20a06316d3420b5d854) Thanks @vincentkoc. - SDK runs created with `timeoutMs: 0` now keep the requested zero timeout without an unwanted client-side watchdog. [2bdcc83](https://github.com/openclaw/openclaw/commit/2bdcc8314d3fce9ee2d0300759cd6a1d9bb45a7d) Thanks @vincentkoc. - Stalled OpenClaw commands now stop reliably even when callers supply extremely large execution or idle-output timeouts. [1425bb3](https://github.com/openclaw/openclaw/commit/1425bb3a03189813787194c23b8a38518166005a) Thanks @vincentkoc. - Commands given extremely large timeout settings now use a safe maximum instead of failing because the runtime cannot schedule the requested wait. [66b94ba](https://github.com/openclaw/openclaw/commit/66b94ba577b8836a4afa7cb59cd1a749bf2a8d68) - Provider-specific tool allow/deny settings now align more consistently with OpenClaw's doctor warnings, including configurations with provider aliases, model-specific keys, OpenRouter-style model IDs, or malformed policy entries. [8f2882f](https://github.com/openclaw/openclaw/commit/8f2882f94affbe5e89994ae175fdaf7304d7b392) #### Additional setup and CLI fixes Additional [`openclaw config`](https://docs.openclaw.ai/cli/config) and CLI fixes improve diagnostics, pairing, startup, and command behavior. - Larger OpenClaw configurations can initialize and generate UI hints more efficiently, while sensitive fields continue to be marked the same way. [#55018](https://github.com/openclaw/openclaw/pull/55018) Thanks @huangyandi-red, @vincentkoc, @xdhuangyandi. - Config changes that still need a manual gateway restart now show a clear restart-required notice with the original note preserved, instead of looking finished with a misleading `config-patch ok` message. [#83041](https://github.com/openclaw/openclaw/pull/83041) Related [#46797](https://github.com/openclaw/openclaw/issues/46797). Thanks @stache73, @xuruiray. - Help for `doctor`, `gateway`, `models`, `plugins`, `sessions`, and `tasks` now appears in tens of milliseconds, while commands such as `sessions --help` and `tasks --help` previously took about 1.6 to 1.8 seconds to begin responding. [#89628](https://github.com/openclaw/openclaw/pull/89628) Thanks @yyzquwu. - OpenTelemetry trace backends such as Langfuse now show the actual provider/model name instead of "unknown" for slash-qualified model IDs. [#89981](https://github.com/openclaw/openclaw/pull/89981) Thanks @mycarrysun, @vincentkoc. - Malformed or older device-pairing records no longer stop `openclaw devices list` from showing pending approval requests, while valid roles still appear normally. [#93504](https://github.com/openclaw/openclaw/pull/93504) Thanks @ly-wang19. - OpenClaw now rejects SSH targets with stray leading or trailing colons before they can produce invalid SSH configuration or tunnel startup failures for SSH-backed sandboxes and gateways. [#93887](https://github.com/openclaw/openclaw/pull/93887) Thanks @miorbnli. - Users whose non-interactive setup fails its local gateway health check now get runnable `openclaw onboard --install-daemon` or `openclaw onboard --skip-health` recovery commands instead of unsupported `setup` flags. [#93994](https://github.com/openclaw/openclaw/pull/93994) Related [#93947](https://github.com/openclaw/openclaw/issues/93947). Thanks @bk-z1, @nianjiuzst. - Gateway health and probe checks now accept the same custom `--port` used to start a local gateway, reject invalid ports early, and show the selected loopback target in JSON output. [#94687](https://github.com/openclaw/openclaw/pull/94687) Related [#79100](https://github.com/openclaw/openclaw/issues/79100). Thanks @bryantegomoh, @ozthedivine. - `gateway --force` now detects IPv4-only processes occupying the gateway port and still attempts cleanup when a port check is inconclusive, instead of mistakenly treating the port as free. [#94949](https://github.com/openclaw/openclaw/pull/94949) Related [#94426](https://github.com/openclaw/openclaw/issues/94426). Thanks @sunlit-deng, @vincentkoc, @wangwllu. - `openclaw config validate` now accepts command-based MCP server setups that explicitly use `transport: "stdio"`, avoiding false validation errors while still rejecting invalid remote-style stdio configurations. [#95102](https://github.com/openclaw/openclaw/pull/95102) Related [#95082](https://github.com/openclaw/openclaw/issues/95082). Thanks @ken-jo, @lzyyzznl. - CLI image edits can now return multiple variants in one command with `--count <n>`, instead of being limited to the provider's default single result. [#95300](https://github.com/openclaw/openclaw/pull/95300) Thanks @ly-wang19. - `openclaw sessions export-trajectory` now finds sessions that other session commands can already see when custom, `~`-based, or `{agentId}`-templated stores are configured, without requiring the store path again. [#95570](https://github.com/openclaw/openclaw/pull/95570) Related [#95568](https://github.com/openclaw/openclaw/issues/95568). Thanks @youngting520. - Fixes `infer inspect --name <id> --json` showing flags that the matching CLI commands did not accept, so developers and operators can reliably discover supported model, auth, and transcription options. [#95719](https://github.com/openclaw/openclaw/pull/95719) Thanks @ly-wang19, @vincentkoc. - People inspecting very large or out-of-order sessions can open usage details and still get the latest timestamped log entries without OpenClaw retaining the entire parsed log history in memory. [#96019](https://github.com/openclaw/openclaw/pull/96019) Thanks @vincentkoc. - Operators can now set up the auth monitor, systemd timer, and Termux widgets for their own OpenClaw host without first replacing maintainer-specific hostnames and filesystem paths. [af3e509](https://github.com/openclaw/openclaw/commit/af3e509ab823dac5f91b16915ee7067b369656a3) Thanks @vincentkoc. - Native Windows crabbox hydration now selects the required Windows daemon job automatically, avoiding failed or misrouted runs while leaving WSL2 and explicit job overrides unchanged. [d5d9a82](https://github.com/openclaw/openclaw/commit/d5d9a8256d6bc2ff8d699152923357bd61c606c1) Thanks @vincentkoc. #### User operation documentation The [Gateway configuration](https://docs.openclaw.ai/gateway/configuration) and related user guides now give clearer setup and operating instructions. - People setting a local agent avatar can avoid missing images by keeping workspace-relative files under 2 MB, while HTTP(S) and data URI avatars are not subject to that limit. [#78884](https://github.com/openclaw/openclaw/pull/78884) Related [#65312](https://github.com/openclaw/openclaw/issues/65312). Thanks @wangjieweb3-design, @nyx-nocturna. - OpenClaw's default agent instructions now ask agents to check for suitable free or open-source solutions before proposing a custom build, while still allowing custom work when it is the better fit. [#86608](https://github.com/openclaw/openclaw/pull/86608) Thanks @cablackmon. - Plugin authors can now use the documented `targetSessionKey` on `subagent_ended` events to match them with the corresponding spawn, instead of relying on `agentId` or `childSessionKey` fields that are not emitted. [#95191](https://github.com/openclaw/openclaw/pull/95191) Related [#95186](https://github.com/openclaw/openclaw/issues/95186). Thanks @ken-jo, @mahaohao-ch. - ClawHub skill links in OpenClaw docs and showcase cards now open the canonical owner-qualified pages, and install examples use copy-ready `openclaw skills install @owner/<slug>` references instead of older bare-slug routes. [#95972](https://github.com/openclaw/openclaw/pull/95972) Thanks @patrick-erichsen. #### Additional tool and automation fixes Additional [plugin SDK](https://docs.openclaw.ai/plugins/sdk-runtime) and automation fixes improve tool events, cron inputs, and Windows handoffs. - SDK applications now receive `tool.call.failed` when terminal tools fail or are blocked, instead of a misleading completion event, so existing failure handling can react correctly. [#95383](https://github.com/openclaw/openclaw/pull/95383) Thanks @ly-wang19. - Fixes cron add and update requests being rejected when recognized job fields arrive with harmless trailing spaces, so schedules can be saved without relaxing checks for ambiguous or unsafe input. [#95674](https://github.com/openclaw/openclaw/pull/95674) Related [#95407](https://github.com/openclaw/openclaw/issues/95407). Thanks @nassiel, @zw-xysk. - Codex subagent monitoring handles large sets of child agents and transcript files with less unnecessary scanning, while older transcript filename formats continue to resolve as before. [#96085](https://github.com/openclaw/openclaw/pull/96085) Thanks @vincentkoc. - Fixes native Windows crabbox hydration getting stuck or missing handoffs when the runner and daemon use different home directories, so both can find the same job state and stop files. [f354889](https://github.com/openclaw/openclaw/commit/f354889efa1c8bafca9304767afba2c270add549) Thanks @vincentkoc. ### Additional contributions Additional maintainer-facing contributions: [#95308](https://github.com/openclaw/openclaw/pull/95308), [#95465](https://github.com/openclaw/openclaw/pull/95465), [#95625](https://github.com/openclaw/openclaw/pull/95625), [#95649](https://github.com/openclaw/openclaw/pull/95649), [#95681](https://github.com/openclaw/openclaw/pull/95681), [#95857](https://github.com/openclaw/openclaw/pull/95857), [#95870](https://github.com/openclaw/openclaw/pull/95870), [#95872](https://github.com/openclaw/openclaw/pull/95872), [#95879](https://github.com/openclaw/openclaw/pull/95879), [#95890](https://github.com/openclaw/openclaw/pull/95890), [#95909](https://github.com/openclaw/openclaw/pull/95909), [#95922](https://github.com/openclaw/openclaw/pull/95922), [#95946](https://github.com/openclaw/openclaw/pull/95946), [#95967](https://github.com/openclaw/openclaw/pull/95967), [#95983](https://github.com/openclaw/openclaw/pull/95983), [#96258](https://github.com/openclaw/openclaw/pull/96258), [#95094](https://github.com/openclaw/openclaw/pull/95094), [#95466](https://github.com/openclaw/openclaw/pull/95466), [#95876](https://github.com/openclaw/openclaw/pull/95876), [#95880](https://github.com/openclaw/openclaw/pull/95880), [#95919](https://github.com/openclaw/openclaw/pull/95919), [#95928](https://github.com/openclaw/openclaw/pull/95928), [#95991](https://github.com/openclaw/openclaw/pull/95991), [#96235](https://github.com/openclaw/openclaw/pull/96235), [#94272](https://github.com/openclaw/openclaw/pull/94272), [#94622](https://github.com/openclaw/openclaw/pull/94622), [#95898](https://github.com/openclaw/openclaw/pull/95898), [#95901](https://github.com/openclaw/openclaw/pull/95901), [#95999](https://github.com/openclaw/openclaw/pull/95999), [#96055](https://github.com/openclaw/openclaw/pull/96055), [#96226](https://github.com/openclaw/openclaw/pull/96226), [#96271](https://github.com/openclaw/openclaw/pull/96271), [#97909](https://github.com/openclaw/openclaw/pull/97909), [#91502](https://github.com/openclaw/openclaw/pull/91502), [#91506](https://github.com/openclaw/openclaw/pull/91506), [#94700](https://github.com/openclaw/openclaw/pull/94700), [#95406](https://github.com/openclaw/openclaw/pull/95406), [#95858](https://github.com/openclaw/openclaw/pull/95858), [#95933](https://github.com/openclaw/openclaw/pull/95933), [#95944](https://github.com/openclaw/openclaw/pull/95944), [#95947](https://github.com/openclaw/openclaw/pull/95947), [#95952](https://github.com/openclaw/openclaw/pull/95952), [#95961](https://github.com/openclaw/openclaw/pull/95961), [#95971](https://github.com/openclaw/openclaw/pull/95971), [#95975](https://github.com/openclaw/openclaw/pull/95975), [#96003](https://github.com/openclaw/openclaw/pull/96003), [#96017](https://github.com/openclaw/openclaw/pull/96017), [#96030](https://github.com/openclaw/openclaw/pull/96030), [#96246](https://github.com/openclaw/openclaw/pull/96246), [#87121](https://github.com/openclaw/openclaw/pull/87121), [#90223](https://github.com/openclaw/openclaw/pull/90223), [#93378](https://github.com/openclaw/openclaw/pull/93378), [#95475](https://github.com/openclaw/openclaw/pull/95475), [#95499](https://github.com/openclaw/openclaw/pull/95499), [#95602](https://github.com/openclaw/openclaw/pull/95602), [#95653](https://github.com/openclaw/openclaw/pull/95653), [#76668](https://github.com/openclaw/openclaw/pull/76668), [#87861](https://github.com/openclaw/openclaw/pull/87861), [#95243](https://github.com/openclaw/openclaw/pull/95243), [#78715](https://github.com/openclaw/openclaw/pull/78715), [#93502](https://github.com/openclaw/openclaw/pull/93502), [#96044](https://github.com/openclaw/openclaw/pull/96044), [#96057](https://github.com/openclaw/openclaw/pull/96057), [#96061](https://github.com/openclaw/openclaw/pull/96061), [#91193](https://github.com/openclaw/openclaw/pull/91193), [#95706](https://github.com/openclaw/openclaw/pull/95706), [#96179](https://github.com/openclaw/openclaw/pull/96179), [#89912](https://github.com/openclaw/openclaw/pull/89912), [#90439](https://github.com/openclaw/openclaw/pull/90439), [#96182](https://github.com/openclaw/openclaw/pull/96182), [#96191](https://github.com/openclaw/openclaw/pull/96191), [#96193](https://github.com/openclaw/openclaw/pull/96193), [#96195](https://github.com/openclaw/openclaw/pull/96195), [#96204](https://github.com/openclaw/openclaw/pull/96204), [#96206](https://github.com/openclaw/openclaw/pull/96206), [#96213](https://github.com/openclaw/openclaw/pull/96213), [#96218](https://github.com/openclaw/openclaw/pull/96218), [#95893](https://github.com/openclaw/openclaw/pull/95893), [#95930](https://github.com/openclaw/openclaw/pull/95930). Thanks @aniruddhaadak80, @coder999999999, @davinci282828, @harjothkhara, @hugenshen, @jalehman, @jason-allen-oneal, @joshavant, @kklouzal, @lizuju, @mehrazmorshed, @mmyzwl, @patrick-erichsen, @romneyda, @rushindrasinha, @shuofengzhang, @tayoun, @vincentkoc, @wadydx, @wangmiao0668000666, @whiteyzy, @yachiyo1680, @zats. ### Release verification - npm package: https://www.npmjs.com/package/openclaw/v/2026.6.11 - registry tarball: https://registry.npmjs.org/openclaw/-/openclaw-2026.6.11.tgz - integrity: `sha512-T+P/g19IheeT1ckXMoPN61dYuE8vBF4MderI+kWkvpuFYxPkJxn8AXLpu9IXCnN9g36Acpm9+mMD/V+lsvOkyA==` - release SHA: `e085fa1a3ffd32d0ea6917e1e6fb4ecbffbb77d2` - full release CI report: https://github.com/openclaw/releases/blob/main/evidence/2026.6.11/release-evidence.md - release publish: https://github.com/openclaw/openclaw/actions/runs/28456822852 - npm preflight: https://github.com/openclaw/openclaw/actions/runs/28417979026 - full release validation: https://github.com/openclaw/openclaw/actions/runs/28415183795 - plugin npm publish: https://github.com/openclaw/openclaw/actions/runs/28457240090 - plugin ClawHub publish: dispatched separately, not awaited by this proof: https://github.com/openclaw/openclaw/actions/runs/28457242470 - plugin ClawHub bootstrap: not needed - OpenClaw npm publish: https://github.com/openclaw/openclaw/actions/runs/28458034111 - npm Telegram beta E2E: https://github.com/openclaw/openclaw/actions/runs/28419920660 - Windows Hub promotion: https://github.com/openclaw/openclaw/actions/runs/28458452299 from openclaw/[email protected]

### 亮点 - **更强的频道控制能力：** Slack 中继模式、原生 Mattermost `/oc_queue` 以及按 DM 模型覆盖，使频道操作更易于自动化和调优。(#94707, #95546, #95120) 感谢 @sjf-oa、@amknight、@xydigit-zt、@thomaszta 和 @gandalf-at-lerian。 - **更丰富的操作工作流：** `openclaw agent --message-file` 和 RAFT CLI 唤醒桥接提供了实用的文件驱动和远程唤醒路径。(#93351, #95497) 感谢 @ooiuuii 和 @vincentkoc。 - **更安全的插件分发：** 额外的官方插件被干净地外部化，安装的客户端可获得捆绑的插件图标元数据。(#95683, #95845) 感谢 @vincentkoc 和 @Patrick-Erichsen。 - **更强的移动端操作：** Android 设置详情面板提升了移动端的配置可见性和控制能力。(#95148) 感谢 @Tosko4。 - **更可靠的智能体交互轮次：** Codex 部分增量、工具架激活以及长上下文提示缓存稳定性减少了进度丢失和不一致运行。(#95404, #95652, #95624) 感谢 @agonza1 和 @vincentkoc。 ### 变更 - **网关与插件工具：** 频道身份钩子上下文和按智能体使用成本报告为集成和操作员提供了更精确的路由和计费。(#91903, #94483) 感谢 @lanzhi-lee、@vincentkoc 和 @ly-wang19。 - **提供商与模型覆盖：** 目录解析、推理控制、提供商模型解析以及加密推理支持现在能处理更多实时提供商变体。(#95283, #95710, #95268, #95744, #95686, #93956) 感谢 @ZengWen-DT、@vincentkoc、@Marvinthebored、@Darren2030、@daniel-alejandro-t、@parveshsaini、@geraint0923、@fuller-stack-dev 和 @jason-allen-oneal。 ### 修复 - **频道投递：** Telegram 进度渲染、Webhook 生命周期、反应指令、重复镜像写入、队列更新排空以及 WhatsApp 持久回复目标现在更加可靠。(#95532, #93002, #95183, #94506, #94977, #95069, #95577, #95007, #95914) 感谢 @amknight、@snowzlmbot、@zhangguiping-xydt、@shadow-enthusiast、@xialonglee、@travellingsoldier85、@obviyus、@hugenshen、@Cuttingwater、@heichaowo、@LiuwqGit、@freidrich-goldenflow、@mcaxtr 和 @vincentkoc。 - **WhatsApp 与消息身份：** 原生引用、Baileys 群组可靠性以及跨 JID 漂移的审批反应现在能保留预期的对话上下文。(#95483, #94338, #95935) 感谢 @mcaxtr、@xialonglee 和 @octopuslabs-fl。 - **网关与会话安全：** 卡住的释放声明、排空状态报告、远程探测超时、格式错误的配对访问列表以及未送达会话身份现在能在不静默丢失路由的情况下处理。(#95299, #94915, #89859, #92178, #95467) 感谢 @mikasa0818、@kriegerbangerz-ship-it、@markoub、@vincentkoc、@maxschachere、@mushuiyu886、@gozzbb2、@wangmiao0668000666、@ly-wang19、@EmilioNicolas、@yetval 和 @hellocli。 - **智能体与回退行为：** 中止的运行能干净停止，提供商响应体保持有界，Claude CLI 信用失败能继续通过回退，Codex 使用限制响应能正确分类。(#94412, #95218, #95508, #95420, #95418, #95417, #95400) 感谢 @szsip239、@vincentkoc、@Alix-007、@mikasa0818、@sallyom、@riazrahaman 和 @jason-allen-oneal。 - **提供商与模型边缘情况：** OpenRouter ID、Ollama 发现与嵌入、Gemini 新鲜度以及模型目录前缀现在能针对正确的运行时元数据进行解析。(#95268, #94811, #93956, #95682, #95744) 感谢 @Darren2030、@daniel-alejandro-t、@mushuiyu886、@jason-allen-oneal、@Sunjae-k、@parveshsaini、@vincentkoc 和 @shakkernerd。 - **配置与 UI 防护：** 非交互式配置失败时安全关闭，TLS 路径拒绝空值，内存工件已清理，UI 使用打了补丁的 DOMPurify 版本。(#94238, #94054, #95791, #95691) 感谢 @ruomuxydt、@NianJiuZst、@miorbnli、@vincentkoc、@SweetSophia 和 @YB0y。 - **定时任务与投递验证：** 无配置投递检查、线程感知去重以及待处理的周期性运行现在能保留其预期目标。(#95754, #95794, #94323) 感谢 @vincentkoc 和 @yetval。 ### 完整贡献记录 此审计记录涵盖了完整的 v2026.6.10..0a4d0daa8cc99e9b2c6bc78c2e67357272a9fd53 历史：308 个合并 PR。生成清单还提供了直接提交作为编辑输入；以上的分组说明优先考虑用户影响。 #### 拉取请求 - **PR #95406** 测试(质量保证)：使发布评分卡类别显式化。感谢 @RomneyDa。 - **PR #94700** 测试：将 HTTP API 脚本校验融入 QA 实验室。感谢 @RomneyDa。 - **PR #95499** 修复(测试)：单元快速流模拟。感谢 @RomneyDa。 - **PR #95308** 修复(CI)：在令牌铸造前过滤 ClawSweeper 评论调度。感谢 @vincentkoc。 - **PR #95532** 修复(Telegram)：将富消息换行符具体化为 `<br>`。关联 #95409。感谢 @amknight 和 @snowzlmbot。 - **PR #91786** 修复(插件)：协调托管 npm 根覆盖与托管对等固定。关联 #91772。感谢 @amknight 和 @mkdelta221。 - **PR #93002** 修复：在工具输出前清理 Telegram 进度草稿。关联 #90753。感谢 @zhangguiping-xydt 和 @shadow-enthusiast。 - **PR #95175** 修复：将移动端执行审批路由到审核者设备。感谢 @joshavant。 - **PR #94506** 修复(Telegram)：在频道重启时停止清除已注册的 webhook。关联 #90254。感谢 @xialonglee 和 @travellingsoldier85。 - **PR #95183** 修复(Telegram)：具体化流式进度占位符。关联 #95004。感谢 @snowzlmbot 和 @obviyus。 - **PR #95483** 修复(WhatsApp)：保留原生引用回复。感谢 @mcaxtr。 - **PR #94338** 修复(WhatsApp)：为群组消息可靠性添加缺失的 Baileys 重试/缓存钩子。关联 #7433。感谢 @xialonglee、@mcaxtr 和 @octopuslabs-fl。 - **PR #94707** 特性(Slack)：为入站消息添加中继模式。感谢 @sjf-oa。 - **PR #94977** 修复(Telegram)：遵守出站反应指令。关联 #71140。感谢 @hugenshen 和 @Cuttingwater。 - **PR #95069** 修复(Telegram)：当主回复已存在时跳过镜像写入 (#94930)。感谢 @heichaowo。 - **PR #95550** 在迁移期间保留继承的频道账户策略。感谢 @amknight。 - **PR #95390** 修复 #95378。感谢 @mikasa0818、@obviyus 和 @MaiDuy708。 - **PR #93143** 修复(iMessage)：保持拆分发合并的选择加入。感谢 @omarshahine。 - **PR #95008** 修复(Claude CLI)：在 `--print` 运行中也禁用原生后台 Bash 和 Monitor。感谢 @anagnorisis2peripeteia。 - **PR #94545** 修复：使用钩子注册表保留受信任的策略。感谢 @jesse-merhi。 - **PR #95007** 修复(Telegram)：将进度草稿行渲染为纯可读文本。关联 #95002。感谢 @snowzlmbot。 - **PR #95624** 修复(智能体)：保持长上下文工具结果提示缓存稳定。感谢 @vincentkoc。 - **PR #95625** 修复(CI)：平滑 PR 运行者注册突发。感谢 @vincentkoc。 - **PR #95572** 修复(智能体)：拒绝包含额外冒号段的绑定规范。感谢 @ly-wang19。 - **PR #95653** 测试(智能体)：增加大型提示缓存覆盖。感谢 @vincentkoc。 - **PR #95640** 合并 iOS 通知权限 UX。感谢 @joshavant。 - **PR #95546** 特性(Mattermost)：将 `/oc_queue` 注册为原生斜杠命令。感谢 @amknight。 - **PR #95019** 修复(技能)：将 gog brew 安装指向 homebrew-core gogcli (#95017)。感谢 @ZengWen-DT、@vincentkoc 和 @Sedrak-Hovhannisyan。 - **PR #93378** 测试(Telegram)：保持实时轮询租约受保护。关联 #93375。感谢 @mmyzwl 和 @Yachiyo1680。 - **PR #95084** 修复(Google Chat)：从出站文本中清除内部工具跟踪行 (#90684)。感谢 @jailbirt 和 @studentzhou-svg。 - **PR #95278** 避免在入队状态数据库打开时复制 process.env。关联 #94571。感谢 @kaka-srp。 - **PR #95577** 修复 #86957：立即排空工作线程缓冲的 Telegram 更新。感谢 @LiuwqGit 和 @freidrich-goldenflow。 - **PR #95128** 修复(压缩)：在切点估算器中计算用户消息图片块。感谢 @yetval。 - **PR #93887** 修复(SSH)：拒绝 parseSshTarget 中带有孤立前导或尾随冒号的主机名。感谢 @miorbnli。 - **PR #95191** 文档(插件)：记录 subagent_ended 钩子字段。关联 #95186。感谢 @MaHaoHao-ch 和 @ken-jo。 - **PR #95102** 修复(配置)：将 stdio 添加到 McpServerSchema 传输联合。关联 #95082。感谢 @lzyyzznl 和 @ken-jo。 - **PR #95465** 修复(SDK)：在打包根辅助函数中类型收窄 manifest.files。感谢 @wangmiao0668000666。 - **PR #95664** 重构(插件)：将所有者技能移入插件。感谢 @vincentkoc。 - **PR #95299** 修复 #95248：当声明被活动工作线程持有时，OpenClaw release_lane 无操作；卡住的 Telegram 入站事件阻塞智能体响应直到网关重启。感谢 @mikasa0818 和 @kriegerbangerz-ship-it。 - **PR #94687** 修复(网关)：接受健康检查和探测的端口。关联 #79100。感谢 @BryanTegomoh 和 @ozthedivine。 - **PR #95649** 修复(CI)：捆绑测试分片并适当调整运行者大小。感谢 @vincentkoc。 - **PR #95243** 修复(文档)：在 `docs:list` 中显示内联 read_when 提示。感谢 @hugenshen 和 @vincentkoc。 - **PR #95283** 修复(OpenAI Completions)：在 `/reasoning on` 下在答案之前密封原生推理。关联 #95280。感谢 @ZengWen-DT、@vincentkoc 和 @Marvinthebored。 - **PR #95497** 特性(RAFT)：添加 CLI 唤醒桥接频道。感谢 @vincentkoc。 - **PR #95459** 修复(Cron)：使用主会话系统事件进行静默快速创建预设。关联 #95073。感谢 @ZOOWH、@vincentkoc 和 @vporton。 - **PR #95503** 修复 #89466：[Bug]：控制 UI 聊天输入文本在发送后未清除。感谢 @zhangguiping-xydt、@vincentkoc 和 @zhong18804784882。 - **PR #95684** 修复(技能)：强化 ClawHub 更新策略。感谢 @vincentkoc。 - **PR #95683** 特性(插件)：将额外的官方插件外部化。感谢 @vincentkoc。 - **PR #95681** 修复(CI)：去抖规范主运行者准入。感谢 @vincentkoc。 - **PR #95652** 修复(智能体)：激活选定的工具架插件。感谢 @vincentkoc。 - **PR #95404** 修复(Codex)：将非最终答案的助手增量流式传输为部分内容。关联 #95422。感谢 @agonza1 和 @vincentkoc。 - **PR #58993** 修复(Google Chat)：支持 spaceType 字段用于 DM 与 Space 检测。感谢 @Starhappysh 和 @vincentkoc。 - **PR #94148** 修复(医生)：防止非交互式 `--fix` 自动重启网关。关联 #78217。感谢 @zhangguiping-xydt 和 @esqandil。 - **PR #89859** 修复(网关)：遵守远程状态探测超时。关联 #65355。感谢 @mushuiyu886 和 @hellocli。 - **PR #95466** 修复(CI)：增加易超时的进程组信号测试的超时。感谢 @jason-allen-oneal。 - **PR #95720** 修复(Matrix)：在强制重置中防止双重 bootstrapCrossSigning 重置。关联 #78396。感谢 @vincentkoc 和 @jteddy。 - **PR #95707** 修复(Synology Chat)：移除重复的本地投递超时。感谢 @vincentkoc。 - **PR #95706** 修复(WhatsApp)：移除死看门狗超时钳制。感谢 @vincentkoc。 - **PR #95719** 修复(CLI)：同步能力检查元数据标志与注册选项。感谢 @vincentkoc。 - **PR #95721** 修复(活跃记忆)：从资格门控中排除做梦叙述会话键。关联 #78500。感谢 @vincentkoc 和 @vishutdhar。 - **PR #95602** 测试：在网关会话工具中每月节省约 79 CI 小时。感谢 @zats 和 @vincentkoc。 - **PR #94412** 修复(智能体核心)：中止工具运行后停止循环。感谢 @szsip239 和 @vincentkoc。 - **PR #94915** 修复(网关)：在就绪状态中报告排空状态。关联 #78136。感谢 @markoub、@vincentkoc 和 @maxschachere。 - **PR #95691** 修复(UI)：将 dompurify 升级到已打补丁版本。感谢 @vincentkoc。 - **PR #95710** 修复(Vercel AI 网关)：解析动态模型选择。感谢 @vincentkoc。 - **PR #94072** 修复(智能体)：将消息工具源回复计作用户面向回复以用于工具错误警告。关联 #93875。感谢 @chenyangjun-xy、@vincentkoc 和 @hoyanhan。 - **PR #94784** 修复(医生)：停止为工作的孤立 shell 提示定时任务承诺 `--fix` (#94655)。感谢 @ZengWen-DT、@altaywtf 和 @geekoagent。 - **PR #93504** 修复(设备配对)：防止角色规范化处理非字符串条目。感谢 @ly-wang19。 - **PR #92178** 修复(网关)：规范化格式错误的配对访问列表。关联 #90654。感谢 @wangmiao0668000666、@ly-wang19 和 @EmilioNicolas。 - **PR #83041** 修复配置补丁重启所需通知。关联 #46797。感谢 @xuruiray 和 @Stache73。 - **PR #95754** 修复(Cron)：保留无配置投递验证。感谢 @vincentkoc。 - **PR #93351** 特性(CLI)：向 `openclaw agent` 添加 `--message-file`。感谢 @ooiuuii。 - **PR #95485** 修复(UI)：在紧凑令牌格式中将接近 1M 的值从 k 滚动到 M。感谢 @NarahariRaghava 和 @vincentkoc。 - **PR #94622** 修复(构建)：允许 tsdown 堆覆盖。感谢 @tayoun。 - **PR #76668** 元数据(ISSUE 模板)：添加专用文档错误报告表单。关联 #76664。感谢 @WadydX。 - **PR #91193** 修复(CLI)：在 action-reparse.ts 中记录 Commander rawArgs 内部 API 依赖。关联 #83893。感谢 @whiteyzy 和 @davinci282828。 - **PR #77339** 修复(自动回复)：重置时清除运行时模型缓存。关联 #77322。感谢 @mjamiv 和 @ZaynL。 - **PR #89628** 加速预计算命令帮助启动。感谢 @yyzquwu。 - **PR #95087** 重构：添加内存和 QMD 会话身份映射。感谢 @jalehman。 - **PR #89323** 修复(Web UI)：跳过隐藏的子智能体选择器页面。关联 #89249。感谢 @giodl73-repo 和 @originsecured-do。 - **PR #84794** 在运行后清理孤立的定时任务会话。关联 #84707。感谢 @TurboTheTurtle 和 @bottenbenny。 - **PR #95794** 修复(Cron)：在为失败目的地去重时比较线程 ID。感谢 @vincentkoc。 - **PR #95791** 修复(会话记忆)：在保存记忆前清理模型工件。感谢 @vincentkoc、@SweetSophia 和 @YB0y。 - **PR #95805** 修复(智能体)：规范化幻觉产生的 Office 文件扩展名。关联 #93326。感谢 @vincentkoc、@lzyyzznl、@xzh-icenter 和 @bhnan。 - **PR #89612** 修复：在迁移中包含持久化插件合约。关联 #89609。感谢 @zerone0x 和 @mugabuga。 - **PR #89981** 修复(诊断-OTel)：在跨度上保留完整模型 ID 而不是折叠为 "unknown"。感谢 @mycarrysun 和 @vincentkoc。 - **PR #90489** 修复(会话)：澄清跨智能体可见性指导。关联 #90443。感谢 @sahibzada-allahyar、@vincentkoc 和 @ramitrkar-hash。 - **PR #95383** 修复(SDK)：将失败/阻塞的工具事件归类为 tool.call.failed。感谢 @ly-wang19。 - **PR #95827** 修复(插件 SDK)：有界实时模型目录成功体。 - **PR #95792** 修复(引导)：在设置安装后刷新提供者插件注册表。关联 #95765。感谢 @snowzlmbot。 - **PR #93943** 修复(提供商使用)：遵守用于使用情况获取的代理环境变量。关联 #78714。感谢 @TurboTheTurtle 和 @TnzGit。 - **PR #89648** 修复(智能体)：恢复模型获取信息日志。关联 #89300。感谢 @xiaobao-k8s 和 @Enominera。 - **PR #93841** 修复(UI)：渲染持久化历史文本块。关联 #90241。感谢 @mushuiyu886 和 @pronzcw。 - **PR #93502** 文档：修复文档元数据拼写检查。感谢 @harjothkhara。 - **PR #86608** 文档：添加现有解决方案预检防护栏。感谢 @cablackmon。 - **PR #95322** 修复(回复)：在滚动时保留使用情况页脚。感谢 @litang9。 - **PR #91559** 修复(智能体)：按模型 ID 清理 Gemini 工具 schema。关联 #91542。感谢 @Pick-cat 和 @qiukui666。 - **PR #94701** 修复刷新会话活动后的嵌入式运行恢复。感谢 @imadal1n。 - **PR #95472** [AI] 修复(主会话)：为已完成的会话跳过当前世代中止控制器。感谢 @xydt-tanshanshan。 - **PR #95401** 修复(LM Studio)：规范化变体模型键。感谢 @MonkeyLeeT。 - **PR #94323** 修复(Cron)：阻止添加/删除操作丢弃一个到期的周期性作业的待处理运行。感谢 @yetval。 - **PR #94443** 修复(记忆维基)：重试瞬态源页面重写竞争而不是中止 wiki_status。关联 #92134。感谢 @ZengWen-DT 和 @cknzraposo。 - **PR #95120** 特性(频道)：添加 directUserId 支持用于每 DM 模型覆盖。关联 #53638。感谢 @xydigit-zt、@thomaszta 和 @gandalf-at-lerian。 - **PR #91049** 修复：处理终端聊天发送确认。关联 #91048。感谢 @nxmxbbd。 - **PR #95081** 修复：保持文本转换运行时导入哈希化。关联 #95057。感谢 @849261680 和 @YvesLaRose。 - **PR #93356** 修复(插件)：缓存插件设置注册表以终止 /models CPU 风暴。感谢 @obuchowski。 - **PR #85341** 重构：内部化 OpenClaw 智能体运行时。 - **PR #94233** 修复(模型回退)：合并认证决策日志。关联 #56979。感谢 @goutamadwant 和 @yanan1991。 - **PR #92399** 修复(LLM)：折叠累积的 openai-responses 消息快照而不是拼接 [AI 辅助]。关联 #91959。感谢 @amersheeny 和 @phoenixyy。 - **PR #94811** 修复(Ollama)：遵守记忆嵌入输出维度。感谢 @mushuiyu886。 - **PR #84708** 修复(智能体)：恢复消息工具镜像重放毒化。感谢 @anyech。 - **PR #95274** 修复(记忆)：保留 Windows QMD 命令路径。关联 #92302。感谢 @ly85206559 和 @Ardooken。 - **PR #95686** 修复(xAI)：为所有推理模型请求加密推理包含。感谢 @geraint0923 和 @fuller-stack-dev。 - **PR #82909** 修复(Telegram)：修复消息缓存回复类型。感谢 @lidge-jun。 - **PR #94942** 修复(Matrix)：修剪完成的假 indexeddb 事务以防止 OOM。关联 #90455。感谢 @xzh-icenter 和 @yar-sh。 - **PR #93994** 修复(设置)：将非交互式健康提示指向引导标志。关联 #93947。感谢 @BK-Z1 和 @NianJiuZst。 - **PR #93956** 修复(Ollama)：跳过远程/云端基础 URL 的自动发现。感谢 @jason-allen-oneal。 - **PR #93547** 修复(ACP)：通过结构化恢复所需代码恢复过时的持久会话 [AI 辅助]。关联 #87830。感谢 @amersheeny 和 @chouzz。 - **PR #94204** 修复(控制 UI)：为配置的基础路径重写清单 href。关联 #94157。感谢 @hugenshen 和 @xrow。 - **PR #95637** 修复(QQ 机器人)：识别带一个或两个破折号的 GFM 表格分隔符。感谢 @ly-wang19。 - **PR #95467** 修复(会话)：在非投递轮次中保持绑定的频道身份。感谢 @yetval。 - **PR #93919** 性能(插件)：在捆绑清单扫描中缓存存在性探测 [AI 辅助]。关联 #76209。感谢 @ml12580 和 @shenhonglong456-ai。 - **PR #94726** 修复(Google)：添加 gemini-3.5-flash 模型目录条目。关联 #94723。感谢 @ajwan8998 和 @anguslogan01。 - **PR #94238** 修复(配置)：当配置在没有交互式 TTY 的情况下运行时失败关闭 (#93953)。感谢 @ruomuxydt 和 @NianJiuZst。 - **PR #89949** 修复(媒体)：在任务开始时固定请求者投递路由。关联 #86034。感谢 @wangwllu 和 @tianxiaochannel-oss88。 - **PR #94470** 杂项：同步 yuanbao 插件目录到 2.15.0。感谢 @jase-283。 - **PR #93585** 修复(守护进程)：在子进程 OOM 后保持 systemd 网关运行。感谢 @snowzlm。 - **PR #94442** 修复(iMessage)：在持久化回声缓存中剥离前导回声损坏标记。感谢 @ly-wang19。 - **PR #93511** 修复(iMessage)：规范化前导 NUL 回声缓存前缀。感谢 @vincentkoc 和 @ly-wang19。 - **PR #87121** 测试(CLI)：添加横幅发射重置辅助函数。关联 #83903。感谢 @lizuju 和 @davinci282828。 - **PR #89172** 修复(飞书)：通过上传时长显示语音消息时长。关联 #53798。感谢 @areslp 和 @kinrocW。 - **PR #95118** 修复：有界智能体加载自己的入站工作空间上下文。关联 #92903。感谢 @849261680 和 @axjing。 - **PR #94752** 修复(回复)：澄清过时模型覆盖重置。关联 #94713。感谢 @snowzlmbot 和 @Gr4via。 - **PR #89716** 修复(提供商)：从非 Anthropic 提示中剥离缓存边界标记。感谢 @masatohoshino。 - **PR #95268** 修复(OpenRouter)：将短规范模型 ID 扩展为上游 API 别名（修复 #95198）。感谢 @Darren2030 和 @daniel-alejandro-t。 - **PR #90475** 修复(Telegram)：保持机器人回复锚定到当前消息。感谢 @moeedahmed。 - **PR #94219** 修复(控制 UI)：在侧边栏会话切换器中恢复提供商使用配额药丸（修复 #93041）。感谢 @Pick-cat 和 @jazzroutine。 - **PR #93369** 修复(Cron)：在 CLI 中暴露每作业回退。关联 #90302。感谢 @849261680 和 @Walliiee。 - **PR #87912** 处理工具结果截断中的 Codex toolResult 块。感谢 @AdrianIp0204。 - **PR #95744** 修复(模型目录)：按匹配长度剥离清单模型 ID 前缀。关联 #95743。感谢 @parveshsaini。 - **PR #95300** 修复(CLI)：在推理图片编辑上暴露 `--count`，与图片生成一致。感谢 @ly-wang19。 - **PR #94156** 修复：暴露 OpenAI 图片质量和审核 CLI 选项。感谢 @lastguru-net 和 @ly-wang19。 - **PR #94062** 修复(智能体)：将通用 "LLM request failed." 归类为瞬态超时。关联 #93931。感谢 @hugenshen 和 @hyphae-bot。 - **PR #93965** 修复(OpenCode Go)：当提供商结束响应时流式完成。关联 #93610。感谢 @zhangguiping-xydt 和 @ForceConstant。 - **PR #89017** 修复(WebChat)：会话在重连后持久化。关联 #87700。感谢 @zhangguiping-xydt 和 @asicoe。 - **PR #94627** 修复(iOS)：集中应用强调色。感谢 @zats。 - **PR #94054** 修复(网关 TLS)：拒绝空/空白 certPath 和 keyPath。感谢 @miorbnli。 - **PR #95674** 修复(Cron)：从识别的作业对象键中修剪尾随空白。关联 #95407。感谢 @zw-xysk 和 @Nassiel。 - **PR #95857** 修复 SQLite 用户版本防护栏白名单。感谢 @RomneyDa。 - **PR #95858** 特性(质量保证)：向场景暴露活跃记忆开关。感谢 @RomneyDa。 - **PR #84326** 医生：暴露沙盒注册表发现。感谢 @giodl73-repo。 - **PR #73079** 修复(MiniMax)：显式请求十六进制 TTS 输出。感谢 @efe-arv。 - **PR #94483** 特性(网关 CLI)：按智能体作用使用成本。感谢 @ly-wang19。 - **PR #91502** 特性(质量保证)：crabline 频道驱动。感谢 @RomneyDa。 - **PR #95872** 将 TUI PTY 测试移入 CI 节点分片。感谢 @RomneyDa。 - **PR #89800** 修复(智能体)：解析 webchat 当前会话状态。关联 #89773。感谢 @sweetcornna 和 @killo3967。 - **PR #95870** 修复(CI)：恢复 macOS 和 Windows QA 门控。感谢 @vincentkoc。 - **PR #94445** 修复(智能体)：保持定时任务云空闲看门狗启用。感谢 @bek91。 - **PR #91506** 修复(质量保证)：跨运行者预留共享 QA 套件标志。感谢 @RomneyDa。 - **PR #95876** CI：添加手动发布 QA 配置文件证据工作流。感谢 @RomneyDa。 - **PR #95879** 修复(CI)：使用工作流修订版进行证据检查。感谢 @vincentkoc。 - **PR #93113** 修复(记忆核心)：在状态中报告活跃做梦阶段。关联 #67868。感谢 @AgentArcLab 和 @mrossit。 - **PR #95837** 简化颜色模式按钮标签。感谢 @SannidhyaSah 和 @hannesrudolph。 - **PR #95797** 修复：对于固定模型会话，/status 过于冗长。感谢 @Solvely-Colin。 - **PR #90223** 测试：使 qqbot 符号链接媒体辅助测试在 Windows 上健壮。感谢 @aniruddhaadak80。 - **PR #95557** 改进：刷新 Android 概览控制界面。感谢 @Solvely-Colin 和 @joshavant。 - **PR #95593** 修复：将 Android 执行审批路由到应用内收件箱。感谢 @Solvely-Colin。 - **PR #95570** 修复(CLI)：一致地解析轨迹导出存储。关联 #95568。感谢 @youngting520。 - **PR #95880** CI：泛化 QA 配置文件证据工作流。感谢 @RomneyDa。 - **PR #95218** 修复(智能体)：有界提供商 JSON 响应读取。感谢 @Alix-007。 - **PR #95614** 修复(记忆维基)：在源重新导入时保留人类笔记块。感谢 @yetval。 - **PR #95148** 特性(Android)：添加设置详情面板。感谢 @Tosko4。 - **PR #95893** 添加 iOS 推送沙盒配置文件和继电器工具。感谢 @joshavant。 - **PR #95723** 修复(控制 UI)：从概览失败计数中排除禁用的定时任务。关联 #95716。感谢 @ZengWen-DT 和 @voytas75。 - **PR #78715** 修复插件文档中的微小语法问题（能力复数）。感谢 @mehrazmorshed。 - **PR #95890** 修复(CI)：恢复 QA 工作流门控。感谢 @vincentkoc。 - **PR #95661** 修复(Discord)：在围栏闭合行上保留闭合围栏空间。感谢 @ly-wang19。 - **PR #94272** CI：添加成熟度评分卡渲染器。感谢 @RomneyDa。 - **PR #95901** CI：添加成熟度评分卡渲染器。感谢 @RomneyDa。 - **PR #95914** 修复(WhatsApp)：保留持久回复目标。感谢 @mcaxtr。 - **PR #95697** 改进：减少热路径线性扫描和冗余 I/O。感谢 @vincentkoc。 - **PR #95916** 修复(记忆)：改进 node:sqlite 不可用指导。感谢 @vincentkoc 和 @rrrrrredy。 - **PR #78884** 文档：记录本地头像文件大小限制。关联 #65312。感谢 @wangjieweb3-design 和 @wraxle-geargrind。 - **PR #95909** 修复(CI)：使用可用的 Android SDK 平台。感谢 @vincentkoc。 - **PR #68389** 插件：当条目与发现的 ID 不匹配时澄清白名单警告。关联 #68352。感谢 @lyfuci 和 @JIRBOY。 - **PR #95898** CI：简化成熟度评分卡 QA 证据输入。感谢 @RomneyDa。 - **PR #95928** 修复(CI)：遵守可重用 QA 证据失败策略。感谢 @vincentkoc。 - **PR #95508** 修复 #95489：[Bug]：claude-cli 信用不足错误绕过模型回退链——错误文本作为最终响应投递。感谢 @mikasa0818、@sallyom 和 @riazrahaman。 - **PR #95420** 修复(智能体)：有界 OpenRouter 模型目录响应读取。感谢 @Alix-007 和 @sallyom。 - **PR #95103** 修复(网关)：有界定价目录流。感谢 @vincentkoc 和 @sallyom。 - **PR #95108** 修复(智能体)：有界 Anthropic 错误流。感谢 @vincentkoc 和 @sallyom。 - **PR #95418** 修复(智能体)：有界 OpenRouter 模型扫描目录成功体。感谢 @Alix-007。 - **PR #95922** 修复(CI)：在设置失败后最终确定 testbox 会话。感谢 @vincentkoc。 - **PR #95417** 修复(智能体)：有界 Google 提示缓存响应读取。感谢 @Alix-007。 - **PR #95935** 修复(WhatsApp)：在 JID 漂移中解析审批反应。感谢 @mcaxtr。 - **PR #84340** 医生：暴露额外的网关服务发现。感谢 @giodl73-repo。 - **PR #95946** 修复(CI)：在设置失败后最终确定 Windows Testbox。感谢 @vincentkoc。 - **PR #95947** 特性(质量保证)：添加 "all" 分类配置文件。感谢 @RomneyDa。 - **PR #91903** 特性(插件 SDK)：添加可扩展的频道身份钩子上下文。感谢 @lanzhi-lee 和 @vincentkoc。 - **PR #95930** 文档(Copilot)：刷新工具架对等说明。感谢 @vincentkoc。 - **PR #95952** 修复(CI)：要求 QA 实时证据工件。感谢 @vincentkoc。 - **PR #95666** 修复记忆维基桥接自导入循环。关联 #95657。感谢 @TurboTheTurtle、@vincentkoc 和 @Johannes0402。 - **PR #91724** 修复(智能体)：从限定模型 ID 推断运行时提供商。感谢 @yu-xin-c 和 @vincentkoc。 - **PR #95957** 改进：加速提供商工具调用流式传输。感谢 @vincentkoc。 - **PR #95541** 修复(CLI)：显示固定插件漂移的工作命令。感谢 @ooiuuii 和 @vincentkoc。 - **PR #95933** 文档：更新成熟度评分卡。感谢 @RomneyDa。 - **PR #78105** 修复(插件)：使空白白名单对新用户可操作。关联 #68780。感谢 @pahuchi-joe 和 @JIRBOY。 - **PR #95961** 修复(CI)：要求实时证明证据工件。感谢 @vincentkoc。 - **PR #95400** 修复(模型回退)：分类 Codex 使用限制负载。感谢 @jason-allen-oneal 和 @sallyom。 - **PR #95944** 性能(QA 实验室)：加速统一 QA 套件。感谢 @RomneyDa。 - **PR #95971** CI：在 QA 失败时标记 QA 配置文件证据失败。感谢 @RomneyDa。 - **PR #95845** 特性：添加捆绑插件图标清单 URL。感谢 @Patrick-Erichsen。 - **PR #95975** 修复(CI)：要求 iOS 外围设备证据工件。感谢 @vincentkoc。 - **PR #95972** 文档：更新 ClawHub 技能路由引用。感谢 @Patrick-Erichsen。 - **PR #90537** 警告生成的包装器覆盖和状态诊断。关联 #90518。感谢 @TurboTheTurtle 和 @vincentkoc。 - **PR #93567** 修复(Cron)：在写入时规范化运行日志 jobId 以匹配读取端验证。感谢 @Alix-007 和 @vincentkoc。 - **PR #95967** 性能(CI)：加宽主测试扇出并将 codeql 移出 blacksmith。感谢 @vincentkoc。 - **PR #95983** 修复(CI)：要求 OpenGrep SARIF 工件。感谢 @vincentkoc。 - **PR #64490** CLI：转义 zsh 补全描述。感谢 @EdenKangdw。 - **PR #95682** 修复 Gemini 天新鲜度时间范围处理。感谢 @Sunjae-k 和 @vincentkoc。 - **PR #93374** 修复(智能体)：为未知工具 ID 建议恢复。关联 #92273。感谢 @mushuiyu886、@vincentkoc 和 @poison。 - **PR #95991** 修复(CI)：要求发布 QA 证据工件。感谢 @vincentkoc。 - **PR #89886** 修复(上下文引擎)：通过委托桥接将 abortSignal 转发到运行时压缩。关联 #89868。感谢 @openperf、@vincentkoc 和 @lykeion-dev。 - **PR #95999** 修复(发布)：要求发布后证据工件。感谢 @vincentkoc。 - **PR #96003** 测试(质量保证)：强化全配置文件 QA 证据场景。感谢 @RomneyDa。 - **PR #95860** 修复(压缩)：当转录以超大工具结果结束时修剪前缀。关联 #78478。感谢 @yetval 和 @wzhgba。 - **PR #95934** 修复(Xiaomi)：将 mimo-v2.5 和 mimo-v2.5-pro 最大输出令牌修正为 128K。感谢 @idootop。 - **PR #95782** 修复(会话)：遵守配置的存储用于出站转录镜像。关联 #95781。感谢 @youngting520。 - **PR #96005** 特性(Copilot)：接线工具架对等辅助函数。感谢 @vincentkoc。 - **PR #95886** 修复：在网关关闭期间避免误报 macOS 更新失败。感谢 @fuller-stack-dev。 - **PR #91906** 性能：跳过子智能体实时流解析。感谢 @lanzhi-lee。 - **PR #87861** 修复(模型使用)：强制转换数字字符串成本并忽略非有限值。关联 #37878。感谢 @coder999999999、@vincentkoc 和 @shuofengzhang。 - **PR #96030** 修复(QA 实验室)：避免重复子证据文件。感谢 @RomneyDa。 - **PR #94369** 修复(记忆维基)：从陈旧报告中排除持久引用页面。感谢 @SunnyShu0925 和 @vincentkoc。 - **PR #94578** 修复长头部的近期会话恢复。关联 #94577。感谢 @rohitjavvadi 和 @vincentkoc。 - **PR #95919** CI：添加 Codex 成熟度评分卡智能体。感谢 @RomneyDa。 - **PR #92356** 修复(心跳)：在选择心跳回复时跳过推理负载。关联 #92260。感谢 @tangtaizong666、@vincentkoc 和 @jmpei。 - **PR #94328** 修复(智能体)：在压缩轮转期间保持压缩后用户重新发出保持尾提示。感谢 @yetval 和 @vincentkoc。 - **PR #95432** 修复(回复)：在多消息块流式传输中抑制每条消息的 finals。感谢 @yetval 和 @vincentkoc。 - **PR #95431** 修复(自动回复)：保持排空/重启中止回复路径静默。感谢 @moeedahmed 和 @vincentkoc。 - **PR #96017** 测试(质量保证)：门控成熟度文档基于通过的证据。感谢 @RomneyDa。 - **PR #95552** 特性(Mattermost)：持久化参与的线程以支持无需提及的后续。感谢 @amknight。 - **PR #95480** 修复 #89231：[Bug]：Windows 安装程序创建的定时任务启动 gateway.cmd 时显示控制台——应使用无窗口启动程序。感谢 @mikasa0818、@vincentkoc 和 @CameronWeller。 - **PR #96049** 修复(Copilot)：保留压缩元数据。感谢 @vincentkoc。 - **PR #96044** 文档：将顶级成熟度层级重命名为 Clawesome。感谢 @RomneyDa。 - **PR #92154** 门控私有 QQBot 群组命令。感谢 @sliverp。 - **PR #95484** 修复：在压缩摘要和后续转录中第一个保留用户之间的助手回复丢失。关联 #76729。感谢 @maweibin、@vincentkoc 和 @njuboy11。 - **PR #95094** CI：添加发布 QA 配置文件证据。感谢 @RomneyDa。 - **PR #96057** 文档：重新设计成熟度评分卡页面。感谢 @vincentkoc。 - **PR #96014** 性能(智能体)：索引移位工具结果。感谢 @vincentkoc。 - **PR #96019** 性能(使用)：有界会话日志保留。感谢 @vincentkoc。 - **PR #96013** 性能(Anthropic)：索引活跃流块。感谢 @vincentkoc。 - **PR #96061** 文档：将成熟度页面置于发布参考下。感谢 @vincentkoc。 - **PR #95589** 修复：npm 插件更新破坏运行中的网关导入。感谢 @ooiuuii 和 @vincentkoc。 - **PR #96068** 修复(ACPX)：消费 acpx 0.11.1 模型能力错误。关联 #95869。感谢 @vincentkoc 和 @SabaTech-dev。 - **PR #89912** 重构：添加转录更新身份合约。感谢 @jalehman。 - **PR #96055** 修复(维护)：使用变基 PR landing。感谢 @vincentkoc。 - **PR #96062** 特性(Copilot)：镜像原生计划和子智能体事件。感谢 @vincentkoc。 - **PR #96032** 修复(ACPX)：在任何 PPID 更改时检测包装器孤儿，而不仅仅是 init 重新父化。感谢 @t2wei 和 @vincentkoc。 - **PR #96124** 杂项(ACPX)：将捆绑客户端升级到 0.11.2。感谢 @vincentkoc。 - **PR #84352** 修复 WebChat 调度失败会话状态。感谢 @jesse-merhi。 - **PR #89518** 重构：迁移插件转录镜像。感谢 @jalehman。 - **PR #90439** 重构：添加嵌入式运行会话目标接缝。感谢 @jalehman。 - **PR #95699** 性能(网关)：丢弃冗余的每访问会话键案例扫描。感谢 @jzakirov 和 @jalehman。 - **PR #95992** 修复(技能)：接受所有者限定的 verify 引用。感谢 @Patrick-Erichsen。 - **PR #95987** 修复(插件)：移除 Simple Icons 颜色路径。感谢 @Patrick-Erichsen。 - **PR #96162** 重构：使用访问器支持的转录语料库用于记忆。感谢 @jalehman。 - **PR #96181** 将 iOS 本地网络权限推迟到引导阶段。感谢 @joshavant。 - **PR #95475** 测试(CLI)：在运行和更新套件中隔离服务环境。感谢 @kklouzal。 - **PR #95226** 修复(基础设施)：有界 ClawHub fetchJson 和错误响应体。感谢 @Alix-007。 - **PR #95240** 修复(Matrix)：在传输中有界非原始 JSON 响应体。感谢 @Alix-007。 - **PR #91742** 修复(记忆)：当 memory_search 超时时中止孤儿嵌入工作。关联 #91718。感谢 @dreamhunter2333 和 @NOVA-Openclaw。 - **PR #93394** 修复(记忆)：当 memory_search 超时时中止孤儿 qmd 搜索子进程。感谢 @Alix-007。 - **PR #89911** 重构：迁移捆绑转录目标查找。感谢 @jalehman。 - **PR #96191** 重构：通过访问器路由插件主机钩子状态。感谢 @jalehman。 - **PR #96179** 修复：通过会话访问器目标路由网关历史。感谢 @jalehman。 - **PR #96201** 重构：添加中止目标会话访问器。感谢 @jalehman。 - **PR #96193** 修复(记忆核心)：迁移做梦清理生命周期。感谢 @jalehman。 - **PR #96195** 修复：桥接 ACP 元数据到会话访问器。感谢 @jalehman。 - **PR #96182** 重构：迁移智能体会话访问器。感谢 @jalehman。 - **PR #96218** 重构：保护回复会话初始化。感谢 @jalehman。 - **PR #96213** 重构(网关)：添加别名变异访问器。感谢 @jalehman。 - **PR #96204** 重构：迁移命令会话持久化到访问器。感谢 @jalehman。 - **PR #96206** 重构：通过会话访问器路由实时模型新鲜读取。感谢 @jalehman。 - **PR #96220** 修复(WhatsApp)：在持久回复中引用当前后续消息。感谢 @mcaxtr。 - **PR #96226** 修复(macOS)：从聊天打包中移除 Textual。感谢 @vincentkoc。 - **PR #96212** 修复(Codex)：从远程应用服务器投递生成的图片。感谢 @sjf-oa。 - **PR #96235** 修复(crabbox)：需要 Xcode 用于 macOS 证明。感谢 @vincentkoc。 - **PR #96145** 修复：UI 故障：配置不可见。关联 #94202。感谢 @sunlit-deng 和 @vporton。 - **PR #96072** 性能(浏览器)：索引角色快照引用。感谢 @vincentkoc。 - **PR #96085** 性能(Codex)：索引推广转录 ID。感谢 @vincentkoc。 - **PR #96087** 性能(回复)：提升直接发送片段索引。感谢 @vincentkoc。 - **PR #94154** 修复(网关)：通过实时注册表解析插件注册的网关方法。关联 #94127。感谢 @Pick-cat、@vincentkoc 和 @BryceMurray。 - **PR #95393** 修复 #92582：[Bug]：医生错误地警告本地记忆嵌入未就绪。感谢 @mikasa0818、@vincentkoc 和 @neekolascmd。 - **PR #96246** 修复(QA)：接受 pnpm 分隔符用于 lab up。感谢 @vincentkoc。 - **PR #94949** 修复(端口)：通过 checkPortInUse 路由 isPortBusy 以捕获 IPv4-only 占用者。关联 #94426。感谢 @sunlit-deng、@vincentkoc 和 @wangwllu。 - **PR #94562** 修复(工作板)：默认在 CLI 列表中隐藏归档卡片。关联 #94555。感谢 @ZengWen-DT、@vincentkoc 和 @ecican。 - **PR #96243** 修复(Nextcloud Talk)：忽略签名的非消息 webhook 事件。关联 #81566。感谢 @arkyu2077、@vincentkoc 和 @rafaelmgbh。 - **PR #96140** 修复(执行)：在插件频道的审批后续中保留轮次来源路由目标。关联 #96103。感谢 @yetval、@vincentkoc 和 @lansenger-pm。 - **PR #96258** CI：将 CodeQL 质量扫描移至托管运行者。感谢 @vincentkoc。 - **PR #96271** 杂项(发布)：在 main 上关闭 2026.6.10。感谢 @vincentkoc。 - **PR #96233** 修复(智能体)：在工具架提示构建时运行 heartbeat_prompt_contribution。感谢 @azogheb 和 @vincentkoc。 - **PR #55018** 修复：避免 mapSensitivePaths schema 遍历中的 O(N²) 浅拷贝。感谢 @xdhuangyandi、@vincentkoc 和 @huangyandi-red。 - **PR #95831** 修复：压缩没有 API 密钥的 Codex OAuth OpenAI 会话。关联 #95693。感谢 @sallyom 和 @YUI-TIEN。 - **PR #96244** 修复(自动回复)：将频道介绍措辞与 chat_type 对齐。关联 #95645。感谢 @arkyu2077、@vincentkoc 和 @iloveleon19。 - **PR #97118** 修复(Telegram)：恢复停滞的入口 spool 声明。感谢 @vincentkoc。 - **PR #97543** 修复(Telegram)：恢复停滞的入口 spool 声明。感谢 @vincentkoc。 - **PR #97122** CI(Docker)：将发布发布到 Docker Hub。感谢 @vincentkoc。 ### 发布验证 - npm 包：https://www.npmjs.com/package/openclaw/v/2026.6.11-beta.2 - 注册表 tarball：https://registry.npmjs.org/openclaw/-/openclaw-2026.6.11-beta.2.tgz - 完整性：`sha512-aUA8lyvSetHvNMVsfbHvEveNhJhafvRl4OYZnxxwWSN4ocO8tNJ2uMUagxHMqZDdzwomjZ6VH3kSuolYjhU0wg==` - 发布 SHA：`4f9d5864f09f60a6eb9da6ad6317148bbf89e134` - 完整发布 CI 报告：https://github.com/openclaw/releases/blob/main/evidence/2026.6.11-beta.2/release-evidence.md - 发布发布：https://github.com/openclaw/openclaw/actions/runs/28338723855 - npm 预检：https://github.com/openclaw/openclaw/actions/runs/28336646700 - 完整发布验证：https://github.com/openclaw/openclaw/actions/runs/28333012052 - 插件 npm 发布：https://github.com/openclaw/openclaw/actions/runs/28338858442 - 插件 ClawHub 发布：单独调度，不由此证明等待：https://github.com/openclaw/openclaw/actions/runs/28338858960 - ClawHub 说明：Discord、Slack、语音通话和 WhatsApp 插件发布在 ClawHub 运行中失败 https://github.com/openclaw/openclaw/actions/runs/28338858960；核心 npm/GitHub 发布和插件 npm 包已上线。 - 插件 ClawHub 引导：不需要 - OpenClaw npm 发布：https://github.com/openclaw/openclaw/actions/runs/28338936121 - npm Telegram 测试版 E2E：未提供

### Highlights - **More capable channel control:** Slack relay mode, native Mattermost `/oc_queue`, and per-DM model overrides make channel operations easier to automate and tune. (#94707, #95546, #95120) Thanks @sjf-oa, @amknight, @xydigit-zt, @thomaszta, and @gandalf-at-lerian. - **Richer operator workflows:** `openclaw agent --message-file` and the RAFT CLI wake bridge add practical file-driven and remote wake-up paths. (#93351, #95497) Thanks @ooiuuii and @vincentkoc. - **Safer plugin distribution:** additional official plugins are externalized cleanly, with bundled plugin icon metadata available to installed clients. (#95683, #95845) Thanks @vincentkoc and @Patrick-Erichsen. - **Stronger mobile operations:** Android settings detail panels improve configuration visibility and control on mobile. (#95148) Thanks @Tosko4. - **More reliable agent turns:** Codex partial deltas, harness activation, and long-context prompt-cache stability reduce lost progress and inconsistent runs. (#95404, #95652, #95624) Thanks @agonza1 and @vincentkoc. ### Changes - **Gateway and plugin tooling:** channel identity hook context and per-agent usage-cost reporting give integrations and operators more precise routing and accounting. (#91903, #94483) Thanks @lanzhi-lee, @vincentkoc, and @ly-wang19. - **Provider and model coverage:** catalog parsing, reasoning controls, provider model resolution, and encrypted reasoning support now handle more live provider variants. (#95283, #95710, #95268, #95744, #95686, #93956) Thanks @ZengWen-DT, @vincentkoc, @Marvinthebored, @Darren2030, @daniel-alejandro-t, @parveshsaini, @geraint0923, @fuller-stack-dev, and @jason-allen-oneal. ### Fixes - **Channel delivery:** Telegram progress rendering, webhook lifecycle, reaction directives, duplicate mirror writes, queued update draining, and WhatsApp durable reply targets are now more reliable. (#95532, #93002, #95183, #94506, #94977, #95069, #95577, #95007, #95914) Thanks @amknight, @snowzlmbot, @zhangguiping-xydt, @shadow-enthusiast, @xialonglee, @travellingsoldier85, @obviyus, @hugenshen, @Cuttingwater, @heichaowo, @LiuwqGit, @freidrich-goldenflow, @mcaxtr, and @vincentkoc. - **WhatsApp and message identity:** native quotes, Baileys group reliability, and approval reactions across JID drift now preserve the intended conversation context. (#95483, #94338, #95935) Thanks @mcaxtr, @xialonglee, and @octopuslabs-fl. - **Gateway and session safety:** stuck release claims, draining-state reporting, remote probe timeouts, malformed paired access lists, and non-delivery session identity are handled without silent routing loss. (#95299, #94915, #89859, #92178, #95467) Thanks @mikasa0818, @kriegerbangerz-ship-it, @markoub, @vincentkoc, @maxschachere, @mushuiyu886, @gozzbb2, @wangmiao0668000666, @ly-wang19, @EmilioNicolas, @yetval, and @hellocli. - **Agent and fallback behavior:** aborted runs stop cleanly, provider response bodies stay bounded, Claude CLI credit failures continue through fallback, and Codex usage-limit responses classify correctly. (#94412, #95218, #95508, #95420, #95418, #95417, #95400) Thanks @szsip239, @vincentkoc, @Alix-007, @mikasa0818, @sallyom, @riazrahaman, and @jason-allen-oneal. - **Provider and model edge cases:** OpenRouter IDs, Ollama discovery and embeddings, Gemini freshness, and model-catalog prefixes now resolve against the right runtime metadata. (#95268, #94811, #93956, #95682, #95744) Thanks @Darren2030, @daniel-alejandro-t, @mushuiyu886, @jason-allen-oneal, @Sunjae-k, @parveshsaini, @vincentkoc, and @shakkernerd. - **Configuration and UI guardrails:** non-interactive configure fails closed, TLS paths reject empty values, memory artifacts are sanitized, and the UI uses the patched DOMPurify release. (#94238, #94054, #95791, #95691) Thanks @ruomuxydt, @NianJiuZst, @miorbnli, @vincentkoc, @SweetSophia, and @YB0y. - **Cron and delivery validation:** no-config delivery checks, thread-aware dedupe, and pending recurring runs retain their intended destinations. (#95754, #95794, #94323) Thanks @vincentkoc and @yetval. ### Complete contribution record This audited record covers the complete v2026.6.10..0a4d0daa8cc99e9b2c6bc78c2e67357272a9fd53 history: 308 merged PRs. The generation manifest also supplies direct commits as editorial input; the grouped notes above prioritize user impact. #### Pull requests - **PR #95406** test(qa): make release scorecard categories explicit. Thanks @RomneyDa. - **PR #94700** test: fold HTTP API script proof into QA Lab. Thanks @RomneyDa. - **PR #95499** fix(test): unit-fast flow mocks. Thanks @RomneyDa. - **PR #95308** fix(ci): filter ClawSweeper comment dispatches before token minting. Thanks @vincentkoc. - **PR #95532** fix(telegram): materialize rich message line breaks as <br>. Related #95409. Thanks @amknight and @snowzlmbot. - **PR #91786** fix(plugins): reconcile managed npm root overrides with managed peer pins. Related #91772. Thanks @amknight and @mkdelta221. - **PR #93002** Fix Telegram progress draft cleanup before tool output. Related #90753. Thanks @zhangguiping-xydt and @shadow-enthusiast. - **PR #95175** fix: route mobile exec approvals to reviewer device. Thanks @joshavant. - **PR #94506** fix(telegram): stop clearing registered webhook on channel restart. Related #90254. Thanks @xialonglee and @travellingsoldier85. - **PR #95183** fix(telegram): materialize streaming progress placeholders. Related #95004. Thanks @snowzlmbot and @obviyus. - **PR #95483** fix(whatsapp): preserve native quote replies. Thanks @mcaxtr. - **PR #94338** fix(whatsapp): wire missing Baileys retry/cache hooks for group message reliability. Related #7433. Thanks @xialonglee and @mcaxtr and @octopuslabs-fl. - **PR #94707** feat(slack): add relay mode for incoming messages. Thanks @sjf-oa. - **PR #94977** fix(telegram): honor outbound reaction directives. Related #71140. Thanks @hugenshen and @Cuttingwater. - **PR #95069** fix(telegram): skip mirror write when primary reply already exists (#94930). Thanks @heichaowo. - **PR #95550** Preserve inherited channel account policies during migration. Thanks @amknight. - **PR #95390** fix #95378: https://github.com/openclaw/openclaw/issues/95378. Thanks @mikasa0818 and @obviyus and @MaiDuy708. - **PR #93143** fix(imessage): keep split-send coalescing opt-in. Thanks @omarshahine. - **PR #95008** fix(claude-cli): also disable native background Bash and Monitor in --print runs. Thanks @anagnorisis2peripeteia. - **PR #94545** fix: keep trusted policies with hook registry. Thanks @jesse-merhi. - **PR #95007** fix(telegram): render progress draft rows as plain readable text. Related #95002. Thanks @snowzlmbot. - **PR #95624** fix(agents): keep long-context tool-result prompts cache-stable. Thanks @vincentkoc. - **PR #95625** fix(ci): smooth PR runner-registration bursts. Thanks @vincentkoc. - **PR #95572** fix(agents): reject bind specs with extra colon segments. Thanks @ly-wang19. - **PR #95653** test(agents): add large prompt cache coverage. Thanks @vincentkoc. - **PR #95640** Consolidate iOS notification permission UX. Thanks @joshavant. - **PR #95546** feat(mattermost): register /oc_queue as a native slash command. Thanks @amknight. - **PR #95019** fix(skills): point gog brew install at homebrew-core gogcli (#95017). Thanks @ZengWen-DT and @vincentkoc and @Sedrak-Hovhannisyan. - **PR #93378** test(telegram): keep live polling leases protected. Related #93375. Thanks @mmyzwl and @Yachiyo1680. - **PR #95084** fix(googlechat): sanitize internal tool-trace lines from outbound text (#90684). Thanks @jailbirt and @studentzhou-svg. - **PR #95278** Avoid copying process.env in ingress queue state DB opens. Related #94571. Thanks @kaka-srp. - **PR #95577** fix #86957: drain worker-spooled Telegram updates immediately. Thanks @LiuwqGit and @freidrich-goldenflow. - **PR #95128** fix(compaction): count user-message image blocks in cut-point estimator. Thanks @yetval. - **PR #93887** fix(ssh): reject hostnames with stray leading or trailing colons in parseSshTarget. Thanks @miorbnli. - **PR #95191** docs(plugins): document subagent_ended hook fields. Related #95186. Thanks @MaHaoHao-ch and @ken-jo. - **PR #95102** fix(config): add stdio to McpServerSchema transport union. Related #95082. Thanks @lzyyzznl and @ken-jo. - **PR #95465** fix(sdk): type-narrow manifest.files in pack staging root helper. Thanks @wangmiao0668000666. - **PR #95664** refactor(plugins): move owner skills into plugins. Thanks @vincentkoc. - **PR #95299** fix #95248: OpenClaw release_lane is a no-op when claim is held by a live worker; stuck Telegram inbound events block agent response until gateway restart. Thanks @mikasa0818 and @kriegerbangerz-ship-it. - **PR #94687** fix(gateway): accept port for health and probe. Related #79100. Thanks @BryanTegomoh and @ozthedivine. - **PR #95649** fix(ci): bundle test shards and right-size runners. Thanks @vincentkoc. - **PR #95243** fix(docs): show inline read_when hints in docs:list. Thanks @hugenshen and @vincentkoc. - **PR #95283** fix(openai-completions): seal native reasoning before the answer under /reasoning on. Related #95280. Thanks @ZengWen-DT and @vincentkoc and @Marvinthebored. - **PR #95497** feat(raft): add CLI wake bridge channel. Thanks @vincentkoc. - **PR #95459** fix(cron): use main-session systemEvent for silent quick-create preset. Related #95073. Thanks @ZOOWH and @vincentkoc and @vporton. - **PR #95503** fix #89466: [Bug]: Control UI chat input text not cleared after sending. Thanks @zhangguiping-xydt and @vincentkoc and @zhong18804784882. - **PR #95684** fix(skills): harden ClawHub update policy. Thanks @vincentkoc. - **PR #95683** feat(plugins): externalize additional official plugins. Thanks @vincentkoc. - **PR #95681** fix(ci): debounce canonical main runner admission. Thanks @vincentkoc. - **PR #95652** fix(agents): activate selected harness plugins. Thanks @vincentkoc. - **PR #95404** fix(codex): stream non-final-answer assistant deltas as partials. Related #95422. Thanks @agonza1 and @vincentkoc. - **PR #58993** fix(googlechat): support spaceType field for DM vs Space detection. Thanks @Starhappysh and @vincentkoc. - **PR #94148** fix(doctor): prevent non-interactive --fix from auto-restarting gateway. Related #78217. Thanks @zhangguiping-xydt and @esqandil. - **PR #89859** fix(gateway): honor remote status probe timeout. Related #65355. Thanks @mushuiyu886 and @hellocli. - **PR #95466** fix(ci): increase timeouts in flaky process-group signal test. Thanks @jason-allen-oneal. - **PR #95720** fix(matrix): prevent double bootstrapCrossSigning reset in forced reset. Related #78396. Thanks @vincentkoc and @jteddy. - **PR #95707** fix(synology-chat): remove duplicate local deliver timeout. Thanks @vincentkoc. - **PR #95706** fix(whatsapp): remove dead watchdog timeout clamp. Thanks @vincentkoc. - **PR #95719** fix(cli): sync capability inspect metadata flags with registered options. Thanks @vincentkoc. - **PR #95721** fix(active-memory): exclude dreaming-narrative session keys from eligibility gate. Related #78500. Thanks @vincentkoc and @vishutdhar. - **PR #95602** test: save ~79 CI hours/mo in gateway session utils. Thanks @zats and @vincentkoc. - **PR #94412** fix(agent-core): stop loop after aborted tool run. Thanks @szsip239 and @vincentkoc. - **PR #94915** fix(gateway): report draining state in readiness. Related #78136. Thanks @markoub and @vincentkoc and @maxschachere. - **PR #95691** fix(ui): bump dompurify to patched release. Thanks @vincentkoc. - **PR #95710** fix(vercel-ai-gateway): resolve dynamic model selections. Thanks @vincentkoc. - **PR #94072** fix(agents): count message-tool source reply as user-facing reply for tool error warnings. Related #93875. Thanks @chenyangjun-xy and @vincentkoc and @hoyanhan. - **PR #94784** fix(doctor): stop promising --fix for working isolated shell-prompt cron jobs (#94655). Thanks @ZengWen-DT and @altaywtf and @geekoagent. - **PR #93504** fix(device-pairing): guard role normalization against non-string entries. Thanks @ly-wang19. - **PR #92178** fix(gateway): normalize malformed paired access lists. Related #90654. Thanks @wangmiao0668000666 and @ly-wang19 and @EmilioNicolas. - **PR #83041** Fix config patch restart-required notices. Related #46797. Thanks @xuruiray and @Stache73. - **PR #95754** fix(cron): preserve no-config delivery validation. Thanks @vincentkoc. - **PR #93351** feat(cli): add --message-file to openclaw agent. Thanks @ooiuuii. - **PR #95485** fix(ui): roll values near 1M over from k to M in compact token format. Thanks @NarahariRaghava and @vincentkoc. - **PR #94622** fix(build): allow tsdown heap override. Thanks @tayoun. - **PR #76668** meta(issue-template): add dedicated docs bug report form. Related #76664. Thanks @WadydX. - **PR #91193** fix(cli): document Commander rawArgs internal API dependency in action-reparse.ts. Related #83893. Thanks @whiteyzy and @davinci282828. - **PR #77339** fix(auto-reply): clear runtime model cache on reset. Related #77322. Thanks @mjamiv and @ZaynL. - **PR #89628** Speed up precomputed command help startup. Thanks @yyzquwu. - **PR #95087** refactor: add memory and QMD session identity mapping. Thanks @jalehman. - **PR #89323** fix(web-ui): skip hidden subagent picker pages. Related #89249. Thanks @giodl73-repo and @originsecured-do. - **PR #84794** Clean up isolated cron sessions after runs. Related #84707. Thanks @TurboTheTurtle and @bottenbenny. - **PR #95794** fix(cron): compare thread IDs when deduping failure destinations. Thanks @vincentkoc. - **PR #95791** fix(session-memory): sanitize model artifacts before saving memory. Thanks @vincentkoc and @SweetSophia and @YB0y. - **PR #95805** fix(agents): normalize hallucinated Office file extensions. Related #93326. Thanks @vincentkoc and @lzyyzznl and @xzh-icenter and @bhnan. - **PR #89612** fix: include persisted plugin contracts for migrations. Related #89609. Thanks @zerone0x and @mugabuga. - **PR #89981** fix(diagnostics-otel): keep full model id on spans instead of collapsing to "unknown". Thanks @mycarrysun and @vincentkoc. - **PR #90489** fix(sessions): clarify cross-agent visibility guidance. Related #90443. Thanks @sahibzada-allahyar and @vincentkoc and @ramitrkar-hash. - **PR #95383** fix(sdk): classify failed/blocked tool events as tool.call.failed. Thanks @ly-wang19. - **PR #95827** fix(plugin-sdk): bound live model catalog success body. - **PR #95792** fix(onboard): refresh provider plugin registry after setup installs. Related #95765. Thanks @snowzlmbot. - **PR #93943** fix(provider-usage): honor proxy env for usage fetch. Related #78714. Thanks @TurboTheTurtle and @TnzGit. - **PR #89648** fix(agents): restore model-fetch info logs. Related #89300. Thanks @xiaobao-k8s and @Enominera. - **PR #93841** fix(ui): render persisted history text blocks. Related #90241. Thanks @mushuiyu886 and @pronzcw. - **PR #93502** docs: fix docs metadata spellcheck. Thanks @harjothkhara. - **PR #86608** docs: add existing-solutions preflight guardrail. Thanks @cablackmon. - **PR #95322** fix(reply): preserve usage footer across rollover. Thanks @litang9. - **PR #91559** fix(agents): clean Gemini tool schemas by model id. Related #91542. Thanks @Pick-cat and @qiukui666. - **PR #94701** Fix embedded-run recovery after refreshed session activity. Thanks @imadal1n. - **PR #95472** [AI] fix(main-session): skip current-gen abort controllers for completed sessions. Thanks @xydt-tanshanshan. - **PR #95401** fix(lmstudio): canonicalize variant model keys. Thanks @MonkeyLeeT. - **PR #94323** fix(cron): stop add/remove from dropping a due recurring job's pending run. Thanks @yetval. - **PR #94443** fix(memory-wiki): retry transient source-page rewrite race instead of aborting wiki_status. Related #92134. Thanks @ZengWen-DT and @cknzraposo. - **PR #95120** feat(channels): add directUserId support for per-DM model override. Related #53638. Thanks @xydigit-zt and @thomaszta and @gandalf-at-lerian. - **PR #91049** fix: handle terminal chat send acknowledgements. Related #91048. Thanks @nxmxbbd. - **PR #95081** fix: keep text transform runtime imports hashed. Related #95057. Thanks @849261680 and @YvesLaRose. - **PR #93356** fix(plugins): cache plugin setup registry to kill the /models CPU storm. Thanks @obuchowski. - **PR #85341** refactor: internalize OpenClaw agent runtime. - **PR #94233** fix(model-fallback): coalesce auth decision logs. Related #56979. Thanks @goutamadwant and @yanan1991. - **PR #92399** fix(llm): collapse cumulative openai-responses message snapshots instead of concatenating [AI-assisted]. Related #91959. Thanks @amersheeny and @phoenixyy. - **PR #94811** fix(ollama): honor memory embedding output dimensionality. Thanks @mushuiyu886. - **PR #84708** fix(agents): recover message-tool mirror replay poison. Thanks @anyech. - **PR #95274** fix(memory): preserve Windows QMD command paths. Related #92302. Thanks @ly85206559 and @Ardooken. - **PR #95686** fix(xai): request encrypted reasoning include for all reasoning models. Thanks @geraint0923 and @fuller-stack-dev. - **PR #82909** fix(telegram): repair message cache reply types. Thanks @lidge-jun. - **PR #94942** fix(matrix): prune finished fake-indexeddb transactions to prevent OOM. Related #90455. Thanks @xzh-icenter and @yar-sh. - **PR #93994** fix(setup): point non-interactive health hints at onboard flags. Related #93947. Thanks @BK-Z1 and @NianJiuZst. - **PR #93956** fix(ollama): skip auto-discovery for remote/cloud base URLs. Thanks @jason-allen-oneal. - **PR #93547** fix(acp): recover stale persistent sessions by structured resume-required code [AI-assisted]. Related #87830. Thanks @amersheeny and @chouzz. - **PR #94204** fix(control-ui): rewrite manifest hrefs for configured base path. Related #94157. Thanks @hugenshen and @xrow. - **PR #95637** fix(qqbot): recognize GFM table separators with one or two dashes. Thanks @ly-wang19. - **PR #95467** fix(sessions): keep bound channel identity across non-delivery turns. Thanks @yetval. - **PR #93919** perf(plugins): cache existence probes within bundle manifest scan [AI-assisted]. Related #76209. Thanks @ml12580 and @shenhonglong456-ai. - **PR #94726** fix(google): add gemini-3.5-flash model catalog entry. Related #94723. Thanks @ajwan8998 and @anguslogan01. - **PR #94238** fix(config): fail closed when configure runs without an interactive TTY (#93953). Thanks @ruomuxydt and @NianJiuZst. - **PR #89949** fix(media): pin requester delivery route when task starts. Related #86034. Thanks @wangwllu and @tianxiaochannel-oss88. - **PR #94470** chore: sync yuanbao plugin catalog to 2.15.0. Thanks @jase-283. - **PR #93585** fix(daemon): keep systemd gateway running after child OOM. Thanks @snowzlm. - **PR #94442** fix(imessage): strip leading echo corruption markers in the persisted echo cache. Thanks @ly-wang19. - **PR #93511** fix(imessage): normalize leading NUL echo-cache prefixes. Thanks @vincentkoc and @ly-wang19. - **PR #87121** test(cli): add banner emission reset helper. Related #83903. Thanks @lizuju and @davinci282828. - **PR #89172** fix(feishu): show voice message duration via upload duration. Related #53798. Thanks @areslp and @kinrocW. - **PR #95118** fix: bound agents load their own inbound workspace context. Related #92903. Thanks @849261680 and @axjing. - **PR #94752** fix(reply): clarify stale model override resets. Related #94713. Thanks @snowzlmbot and @Gr4via. - **PR #89716** fix(providers): strip cache-boundary marker from non-Anthropic prompts. Thanks @masatohoshino. - **PR #95268** fix(openrouter): expand short canonical model IDs to upstream API slugs (fixes #95198). Thanks @Darren2030 and @daniel-alejandro-t. - **PR #90475** fix(telegram): keep bot reply answers anchored to current message. Thanks @moeedahmed. - **PR #94219** fix(control-ui): restore provider usage quota pill in sidebar session switcher (fixes #93041). Thanks @Pick-cat and @jazzroutine. - **PR #93369** fix(cron): expose per-job fallbacks in CLI. Related #90302. Thanks @849261680 and @Walliiee. - **PR #87912** Handle Codex toolResult blocks in tool-result truncation. Thanks @AdrianIp0204. - **PR #95744** fix(model-catalog): strip manifest model-id prefixes by the matched length. Related #95743. Thanks @parveshsaini. - **PR #95300** fix(cli): expose --count on infer image edit, matching image generate. Thanks @ly-wang19. - **PR #94156** fix: expose OpenAI image quality and moderation CLI options. Thanks @lastguru-net and @ly-wang19. - **PR #94062** fix(agents): classify generic "LLM request failed." as transient time…. Related #93931. Thanks @hugenshen and @hyphae-bot. - **PR #93965** fix(opencode-go): streaming completes when provider ends responses. Related #93610. Thanks @zhangguiping-xydt and @ForceConstant. - **PR #89017** fix(webchat): sessions persist after reconnects. Related #87700. Thanks @zhangguiping-xydt and @asicoe. - **PR #94627** fix(ios): centralize app accent colors. Thanks @zats. - **PR #94054** fix(gateway.tls): reject empty/whitespace certPath and keyPath. Thanks @miorbnli. - **PR #95674** fix(cron): trim trailing whitespace from recognized job object keys. Related #95407. Thanks @zw-xysk and @Nassiel. - **PR #95857** Fix SQLite user version guardrail allowlist. Thanks @RomneyDa. - **PR #95858** feat(qa): expose active memory toggles to scenarios. Thanks @RomneyDa. - **PR #84326** Doctor: expose sandbox registry findings. Thanks @giodl73-repo. - **PR #73079** fix(minimax): request hex TTS output explicitly. Thanks @efe-arv. - **PR #94483** feat(gateway-cli): scope usage-cost by agent. Thanks @ly-wang19. - **PR #91502** feat(qa): crabline channel driver. Thanks @RomneyDa. - **PR #95872** Move TUI PTY tests into CI node shard. Thanks @RomneyDa. - **PR #89800** fix(agents): resolve webchat current session status. Related #89773. Thanks @sweetcornna and @killo3967. - **PR #95870** fix(ci): restore macOS and Windows QA gates. Thanks @vincentkoc. - **PR #94445** fix(agents): keep cron cloud idle watchdog enabled. Thanks @bek91. - **PR #91506** fix(qa): reserve shared QA suite flags across runners. Thanks @RomneyDa. - **PR #95876** ci: add manual release QA profile evidence workflow. Thanks @RomneyDa. - **PR #95879** fix(ci): use workflow revision for proof checks. Thanks @vincentkoc. - **PR #93113** fix(memory-core): report active dreaming phases in status. Related #67868. Thanks @AgentArcLab and @mrossit. - **PR #95837** Simplify color mode button labels. Thanks @SannidhyaSah and @hannesrudolph. - **PR #95797** fix: /status is too verbose for pinned model sessions. Thanks @Solvely-Colin. - **PR #90223** test: make qqbot symlinked media helper test robust on Windows. Thanks @aniruddhaadak80. - **PR #95557** improve: refresh Android overview control surface. Thanks @Solvely-Colin and @joshavant. - **PR #95593** fix: route Android exec approvals to in-app inbox. Thanks @Solvely-Colin. - **PR #95570** fix(cli): resolve trajectory export stores consistently. Related #95568. Thanks @youngting520. - **PR #95880** ci: generalize QA profile evidence workflow. Thanks @RomneyDa. - **PR #95218** fix(agents): bound provider JSON response reads. Thanks @Alix-007. - **PR #95614** fix(memory-wiki): preserve human notes block on source re-ingest. Thanks @yetval. - **PR #95148** feat(android): add settings detail panels. Thanks @Tosko4. - **PR #95893** Add iOS push sandbox profiles and relay tooling. Thanks @joshavant. - **PR #95723** fix(control-ui): exclude disabled cron jobs from Overview failed count. Related #95716. Thanks @ZengWen-DT and @voytas75. - **PR #78715** Fix minor grammar issue in plugin documentation (capabilities plural). Thanks @mehrazmorshed. - **PR #95890** fix(ci): restore QA workflow gates. Thanks @vincentkoc. - **PR #95661** fix(discord): reserve closing-fence space on fence-closing lines. Thanks @ly-wang19. - **PR #94272** ci: add maturity scorecard renderer. Thanks @RomneyDa. - **PR #95901** ci: add maturity scorecard renderer. Thanks @RomneyDa. - **PR #95914** fix(whatsapp): preserve durable reply target. Thanks @mcaxtr. - **PR #95697** improve: reduce hot-path linear scans and redundant I/O. Thanks @vincentkoc. - **PR #95916** fix(memory): improve node:sqlite unavailable guidance. Thanks @vincentkoc and @rrrrrredy. - **PR #78884** docs: document local avatar file size limit. Related #65312. Thanks @wangjieweb3-design and @wraxle-geargrind. - **PR #95909** fix(ci): use available Android SDK platform. Thanks @vincentkoc. - **PR #68389** plugins: clarify allowlist warning when entries don't match discovered ids. Related #68352. Thanks @lyfuci and @JIRBOY. - **PR #95898** ci: simplify maturity scorecard QA evidence inputs. Thanks @RomneyDa. - **PR #95928** fix(ci): honor reusable QA evidence failure policy. Thanks @vincentkoc. - **PR #95508** fix #95489: [Bug]: claude-cli out-of-credits error bypasses model fallback chain — error text delivered as final response. Thanks @mikasa0818 and @sallyom and @riazrahaman. - **PR #95420** fix(agents): bound OpenRouter model catalog response reads. Thanks @Alix-007 and @sallyom. - **PR #95103** fix(gateway): bound pricing catalog streams. Thanks @vincentkoc and @sallyom. - **PR #95108** fix(agents): bound Anthropic error streams. Thanks @vincentkoc and @sallyom. - **PR #95418** fix(agents): bound OpenRouter model-scan catalog success body. Thanks @Alix-007. - **PR #95922** fix(ci): finalize testbox sessions after setup failures. Thanks @vincentkoc. - **PR #95417** fix(agents): bound Google prompt cache response reads. Thanks @Alix-007. - **PR #95935** fix(whatsapp): resolve approval reactions across JID drift. Thanks @mcaxtr. - **PR #84340** Doctor: expose extra gateway service findings. Thanks @giodl73-repo. - **PR #95946** fix(ci): finalize Windows Testbox after setup failures. Thanks @vincentkoc. - **PR #95947** feat(qa): add "all" taxonomy profile. Thanks @RomneyDa. - **PR #91903** feat(plugin-sdk): add extensible channel identity hook context. Thanks @lanzhi-lee and @vincentkoc. - **PR #95930** docs(copilot): refresh harness parity notes. Thanks @vincentkoc. - **PR #95952** fix(ci): require QA live evidence artifacts. Thanks @vincentkoc. - **PR #95666** Fix memory-wiki bridge self-import loop. Related #95657. Thanks @TurboTheTurtle and @vincentkoc and @Johannes0402. - **PR #91724** fix(agents): infer runtime provider from qualified model ids. Thanks @yu-xin-c and @vincentkoc. - **PR #95957** improve: speed up provider tool-call streaming. Thanks @vincentkoc. - **PR #95541** fix(cli): show working commands for pinned plugin drift. Thanks @ooiuuii and @vincentkoc. - **PR #95933** docs: update maturity scorecard. Thanks @RomneyDa. - **PR #78105** fix(plugins): make empty-allowlist actionable for new users. Related #68780. Thanks @pahuchi-joe and @JIRBOY. - **PR #95961** fix(ci): require live proof evidence artifacts. Thanks @vincentkoc. - **PR #95400** fix(model-fallback): classify Codex usage-limit payloads. Thanks @jason-allen-oneal and @sallyom. - **PR #95944** perf(qa-lab): speed up unified QA suites. Thanks @RomneyDa. - **PR #95971** ci: fail QA profile evidence on QA failures. Thanks @RomneyDa. - **PR #95845** feat: add bundled plugin icon manifest URLs. Thanks @Patrick-Erichsen. - **PR #95975** fix(ci): require iOS Periphery evidence artifact. Thanks @vincentkoc. - **PR #95972** docs: update ClawHub skill route references. Thanks @Patrick-Erichsen. - **PR #90537** Warn on generated wrapper overwrites and status diagnostics. Related #90518. Thanks @TurboTheTurtle and @vincentkoc. - **PR #93567** fix(cron): normalize run-log jobId on write to match read-side validation. Thanks @Alix-007 and @vincentkoc. - **PR #95967** perf(ci): widen main test fanout and move codeql off blacksmith. Thanks @vincentkoc. - **PR #95983** fix(ci): require OpenGrep SARIF artifacts. Thanks @vincentkoc. - **PR #64490** CLI: escape zsh completion descriptions. Thanks @EdenKangdw. - **PR #95682** Fix Gemini day freshness time range handling. Thanks @Sunjae-k and @vincentkoc. - **PR #93374** fix(agents): suggest recovery for unknown tool ids. Related #92273. Thanks @mushuiyu886 and @vincentkoc and @poison. - **PR #95991** fix(ci): require release QA evidence artifacts. Thanks @vincentkoc. - **PR #89886** fix(context-engine): forward abortSignal through delegation bridge to runtime compaction. Related #89868. Thanks @openperf and @vincentkoc and @lykeion-dev. - **PR #95999** fix(release): require postpublish evidence artifact. Thanks @vincentkoc. - **PR #96003** test(qa): harden all-profile QA evidence scenarios. Thanks @RomneyDa. - **PR #95860** fix(compaction): trim prefix when transcript ends in an oversized tool result. Related #78478. Thanks @yetval and @wzhgba. - **PR #95934** fix(xiaomi): correct mimo-v2.5 and mimo-v2.5-pro max output tokens to 128K. Thanks @idootop. - **PR #95782** fix(sessions): honor configured store for outbound transcript mirrors. Related #95781. Thanks @youngting520. - **PR #96005** feat(copilot): wire harness parity helpers. Thanks @vincentkoc. - **PR #95886** fix: avoid false macOS update failures during gateway shutdown. Thanks @fuller-stack-dev. - **PR #91906** perf: skip subagent live stream parsing. Thanks @lanzhi-lee. - **PR #87861** fix(model-usage): coerce numeric-string costs and ignore non-finite values. Related #37878. Thanks @coder999999999 and @vincentkoc and @shuofengzhang. - **PR #96030** fix(qa-lab): avoid duplicate child evidence files. Thanks @RomneyDa. - **PR #94369** fix(memory-wiki): exclude durable reference pages from stale report. Thanks @SunnyShu0925 and @vincentkoc. - **PR #94578** Fix recent session resume with long headers. Related #94577. Thanks @rohitjavvadi and @vincentkoc. - **PR #95919** ci: add Codex maturity scorecard agent. Thanks @RomneyDa. - **PR #92356** fix(heartbeat): skip reasoning payloads when selecting heartbeat reply. Related #92260. Thanks @tangtaizong666 and @vincentkoc and @jmpei. - **PR #94328** fix(agents): keep post-compaction user re-issue of a kept-tail prompt during compaction rotation. Thanks @yetval and @vincentkoc. - **PR #95432** fix(reply): suppress per-message finals across multi-message block streaming. Thanks @yetval and @vincentkoc. - **PR #95431** fix(auto-reply): keep drain/restart-abort reply paths silent. Thanks @moeedahmed and @vincentkoc. - **PR #96017** test(qa): gate maturity docs on passing evidence. Thanks @RomneyDa. - **PR #95552** feat(mattermost): persist participated threads for mention-free follow-ups. Thanks @amknight. - **PR #95480** fix #89231: [Bug]: Windows installer-created scheduled task launches gateway.cmd with visible console — should use windowless launcher. Thanks @mikasa0818 and @vincentkoc and @CameronWeller. - **PR #96049** fix(copilot): preserve compaction metadata. Thanks @vincentkoc. - **PR #96044** docs: rename top maturity tier to Clawesome. Thanks @RomneyDa. - **PR #92154** Gate private QQBot group commands. Thanks @sliverp. - **PR #95484** fix: assistant reply lost between compaction summary and first kept user in successor transcript. Related #76729. Thanks @maweibin and @vincentkoc and @njuboy11. - **PR #95094** ci: add release QA profile evidence. Thanks @RomneyDa. - **PR #96057** docs: redesign maturity scorecard pages. Thanks @vincentkoc. - **PR #96014** perf(agents): index displaced tool results. Thanks @vincentkoc. - **PR #96019** perf(usage): bound session log retention. Thanks @vincentkoc. - **PR #96013** perf(anthropic): index active stream blocks. Thanks @vincentkoc. - **PR #96061** docs: place maturity pages under release reference. Thanks @vincentkoc. - **PR #95589** fix: npm plugin updates break running gateway imports. Thanks @ooiuuii and @vincentkoc. - **PR #96068** fix(acpx): consume acpx 0.11.1 model capability errors. Related #95869. Thanks @vincentkoc and @SabaTech-dev. - **PR #89912** refactor: add transcript update identity contract. Thanks @jalehman. - **PR #96055** fix(maint): use rebase PR landing. Thanks @vincentkoc. - **PR #96062** feat(copilot): mirror native plan and subagent events. Thanks @vincentkoc. - **PR #96032** fix(acpx): detect wrapper orphan on any PPID change, not just init reparenting. Thanks @t2wei and @vincentkoc. - **PR #96124** chore(acpx): bump bundled client to 0.11.2. Thanks @vincentkoc. - **PR #84352** Fix WebChat dispatch failure session status. Thanks @jesse-merhi. - **PR #89518** refactor: migrate plugin transcript mirrors. Thanks @jalehman. - **PR #90439** refactor: add embedded run session target seam. Thanks @jalehman. - **PR #95699** perf(gateway): drop redundant per-access session-key case scan. Thanks @jzakirov and @jalehman. - **PR #95992** fix(skills): accept owner-qualified verify refs. Thanks @Patrick-Erichsen. - **PR #95987** fix(plugins): remove Simple Icons color paths. Thanks @Patrick-Erichsen. - **PR #96162** refactor: use accessor-backed transcript corpus for memory. Thanks @jalehman. - **PR #96181** Defer iOS local network permission until onboarding. Thanks @joshavant. - **PR #95475** test(cli): isolate service env in run and update suites. Thanks @kklouzal. - **PR #95226** fix(infra): bound ClawHub fetchJson and error response bodies. Thanks @Alix-007. - **PR #95240** fix(matrix): bound non-raw JSON response body in transport. Thanks @Alix-007. - **PR #91742** fix(memory): abort orphaned embedding work when memory_search times out. Related #91718. Thanks @dreamhunter2333 and @NOVA-Openclaw. - **PR #93394** fix(memory): abort orphaned qmd search subprocess when memory_search times out. Thanks @Alix-007. - **PR #89911** refactor: migrate bundled transcript target lookups. Thanks @jalehman. - **PR #96191** refactor: route plugin host hook state through accessor. Thanks @jalehman. - **PR #96179** fix: route gateway history through session accessor target. Thanks @jalehman. - **PR #96201** refactor: add abort target session accessor. Thanks @jalehman. - **PR #96193** fix(memory-core): migrate dreaming cleanup lifecycle. Thanks @jalehman. - **PR #96195** fix: bridge ACP metadata to session accessors. Thanks @jalehman. - **PR #96182** refactor: migrate agent session accessors. Thanks @jalehman. - **PR #96218** refactor: guard reply session initialization. Thanks @jalehman. - **PR #96213** refactor(gateway): add alias mutation accessor. Thanks @jalehman. - **PR #96204** refactor: migrate command session persistence to accessor. Thanks @jalehman. - **PR #96206** refactor: route live model fresh reads through session accessor. Thanks @jalehman. - **PR #96220** fix(whatsapp): quote current follow-up in durable replies. Thanks @mcaxtr. - **PR #96226** fix(macos): drop Textual from chat packaging. Thanks @vincentkoc. - **PR #96212** fix(codex): deliver generated images from remote app-server. Thanks @sjf-oa. - **PR #96235** fix(crabbox): require Xcode for macOS proof. Thanks @vincentkoc. - **PR #96145** fix: UI glitch: config is not visible. Related #94202. Thanks @sunlit-deng and @vporton. - **PR #96072** perf(browser): index role snapshot references. Thanks @vincentkoc. - **PR #96085** perf(codex): index rollout transcript ids. Thanks @vincentkoc. - **PR #96087** perf(reply): hoist direct-send fragment index. Thanks @vincentkoc. - **PR #94154** fix(gateway): resolve plugin-registered gateway methods through live registry. Related #94127. Thanks @Pick-cat and @vincentkoc and @BryceMurray. - **PR #95393** fix #92582: Bug: doctor falsely warns local memory embeddings are not ready. Thanks @mikasa0818 and @vincentkoc and @neekolascmd. - **PR #96246** fix(qa): accept pnpm separator for lab up. Thanks @vincentkoc. - **PR #94949** fix(ports): route isPortBusy through checkPortInUse to catch IPv4-only occupants. Related #94426. Thanks @sunlit-deng and @vincentkoc and @wangwllu. - **PR #94562** fix(workboard): hide archived cards in CLI list by default. Related #94555. Thanks @ZengWen-DT and @vincentkoc and @ecican. - **PR #96243** fix(nextcloud-talk): ignore signed non-message webhook events. Related #81566. Thanks @arkyu2077 and @vincentkoc and @rafaelmgbh. - **PR #96140** fix(exec): preserve turn-source routing target in approval followups for plugin channels. Related #96103. Thanks @yetval and @vincentkoc and @lansenger-pm. - **PR #96258** ci: move CodeQL quality scans to hosted runners. Thanks @vincentkoc. - **PR #96271** chore(release): close out 2026.6.10 on main. Thanks @vincentkoc. - **PR #96233** fix(agents): run heartbeat_prompt_contribution on harness prompt builds. Thanks @azogheb and @vincentkoc. - **PR #55018** fix: avoid O(N²) shallow-copy in mapSensitivePaths schema traversal. Thanks @xdhuangyandi and @vincentkoc and @huangyandi-red. - **PR #95831** fix: compact Codex OAuth OpenAI sessions without API keys. Related #95693. Thanks @sallyom and @YUI-TIEN. - **PR #96244** fix(auto-reply): align channel intro wording with chat_type. Related #95645. Thanks @arkyu2077 and @vincentkoc and @iloveleon19. - **PR #97118** fix(telegram): recover stalled ingress spool claims. Thanks @vincentkoc. - **PR #97543** fix(telegram): recover stalled ingress spool claims. Thanks @vincentkoc. - **PR #97122** ci(docker): publish releases to Docker Hub. Thanks @vincentkoc. ### Release verification - npm package: https://www.npmjs.com/package/openclaw/v/2026.6.11-beta.2 - registry tarball: https://registry.npmjs.org/openclaw/-/openclaw-2026.6.11-beta.2.tgz - integrity: `sha512-aUA8lyvSetHvNMVsfbHvEveNhJhafvRl4OYZnxxwWSN4ocO8tNJ2uMUagxHMqZDdzwomjZ6VH3kSuolYjhU0wg==` - release SHA: `4f9d5864f09f60a6eb9da6ad6317148bbf89e134` - full release CI report: https://github.com/openclaw/releases/blob/main/evidence/2026.6.11-beta.2/release-evidence.md - release publish: https://github.com/openclaw/openclaw/actions/runs/28338723855 - npm preflight: https://github.com/openclaw/openclaw/actions/runs/28336646700 - full release validation: https://github.com/openclaw/openclaw/actions/runs/28333012052 - plugin npm publish: https://github.com/openclaw/openclaw/actions/runs/28338858442 - plugin ClawHub publish: dispatched separately, not awaited by this proof: https://github.com/openclaw/openclaw/actions/runs/28338858960 - ClawHub caveat: Discord, Slack, Voice Call, and WhatsApp plugin publishes failed in ClawHub run https://github.com/openclaw/openclaw/actions/runs/28338858960; core npm/GitHub release and plugin npm packages are live. - plugin ClawHub bootstrap: not needed - OpenClaw npm publish: https://github.com/openclaw/openclaw/actions/runs/28338936121 - npm Telegram beta E2E: not supplied

## 2026.6.11 ### 亮点 - **更强大的频道控制：** Slack 中继模式、原生 Mattermost `/oc_queue` 以及按私聊模型覆盖，让频道操作更易于自动化和调优。（#94707、#95546、#95120）感谢 @sjf-oa、@amknight、@xydigit-zt、@thomaszta 和 @gandalf-at-lerian。 - **更丰富的运维工作流：** `openclaw agent --message-file` 和 RAFT CLI 唤醒桥提供了实用的文件驱动和远程唤醒路径。（#93351、#95497）感谢 @ooiuuii 和 @vincentkoc。 - **更安全的插件分发：** 额外官方插件被干净地外部化，捆绑的插件图标元数据可供已安装的客户端使用。（#95683、#95845）感谢 @vincentkoc 和 @Patrick-Erichsen。 - **更强大的移动端操作：** Android 设置详情面板提升了移动端的配置可见性和控制能力。（#95148）感谢 @Tosko4。 - **更可靠的代理轮次：** Codex 部分增量、工具激活以及长上下文提示缓存稳定性，减少了进度丢失和不一致运行。（#95404、#95652、#95624）感谢 @agonza1 和 @vincentkoc。 ### 变更 - **网关与插件工具：** 频道身份钩子上下文和按代理的使用成本报告，为集成和运维人员提供了更精确的路由和计费能力。（#91903、#94483）感谢 @lanzhi-lee、@vincentkoc 和 @ly-wang19。 - **提供商与模型覆盖：** 目录解析、推理控制、提供商模型解析以及加密推理支持，现已处理更多实时提供商变体。（#95283、#95710、#95268、#95744、#95686、#93956）感谢 @ZengWen-DT、@vincentkoc、@Marvinthebored、@Darren2030、@daniel-alejandro-t、@parveshsaini、@geraint0923、@fuller-stack-dev 和 @jason-allen-oneal。 ### 修复 - **频道投递：** Telegram 进度渲染、Webhook 生命周期、反应指令、重复镜像写入、队列更新排空以及 WhatsApp 持久回复目标现已更可靠。（#95532、#93002、#95183、#94506、#94977、#95069、#95577、#95007、#95914）感谢 @amknight、@snowzlmbot、@zhangguiping-xydt、@shadow-enthusiast、@xialonglee、@travellingsoldier85、@obviyus、@hugenshen、@Cuttingwater、@heichaowo、@LiuwqGit、@freidrich-goldenflow、@mcaxtr 和 @vincentkoc。 - **WhatsApp 与消息身份：** 原生引用、Baileys 群组可靠性以及跨 JID 漂移的批准反应，现已保留预期的对话上下文。（#95483、#94338、#95935）感谢 @mcaxtr、@xialonglee 和 @octopuslabs-fl。 - **网关与会话安全：** 卡住的释放声明、排空状态报告、远程探测超时、格式错误的配对访问列表以及未送达的会话身份，现已处理而不导致静默路由丢失。（#95299、#94915、#89859、#92178、#95467）感谢 @mikasa0818、@kriegerbangerz-ship-it、@markoub、@vincentkoc、@maxschachere、@mushuiyu886、@gozzbb2、@wangmiao0668000666、@ly-wang19、@EmilioNicolas 和 @yetval。 - **代理与回退行为：** 中止的运行可干净停止、提供商响应体保持有界、Claude CLI 信用失败可通过回退继续、Codex 用量限制响应分类正确。（#94412、#95218、#95508、#95420、#95418、#95417、#95400）感谢 @szsip239、@vincentkoc、@Alix-007、@mikasa0818、@sallyom、@riazrahaman 和 @jason-allen-oneal。 - **提供商与模型边界情况：** OpenRouter ID、Ollama 发现与嵌入、Gemini 新鲜度以及模型目录前缀，现已针对正确的运行时元数据解析。（#95268、#94811、#93956、#95682、#95744）感谢 @Darren2030、@daniel-alejandro-t、@mushuiyu886、@jason-allen-oneal、@Sunjae-k、@parveshsaini、@vincentkoc 和 @shakkernerd。 - **配置与 UI 护栏：** 非交互式 configure 失败时关闭、TLS 路径拒绝空值、内存工件经过清理、UI 使用修补后的 DOMPurify 版本。（#94238、#94054、#95791、#95691）感谢 @ruomuxydt、@NianJiuZst、@miorbnli、@vincentkoc、@SweetSophia 和 @YB0y。 - **Cron 与投递验证：** 无配置投递检查、线程感知去重以及待处理的周期性运行，现已保留其预期目的地。（#95754、#95794、#94323）感谢 @vincentkoc 和 @yetval。 ### 完整贡献记录 此审计记录涵盖了完整的 v2026.6.10..ee421ef7da7edf152ab911d74ffcfd852ecf43e2 历史：共 305 个合并 PR。生成清单还提供了直接提交作为编辑输入；上述分组说明优先考虑了用户影响。 #### 拉取请求 - **PR #95406** test(qa): make release scorecard categories explicit. 感谢 @RomneyDa。 - **PR #94700** test: fold HTTP API script proof into QA Lab. 感谢 @RomneyDa。 - **PR #95499** fix(test): unit-fast flow mocks. 感谢 @RomneyDa。 - **PR #95308** fix(ci): filter ClawSweeper comment dispatches before token minting. 感谢 @vincentkoc。 - **PR #95532** fix(telegram): materialize rich message line breaks as <br>. 关联 #95409。感谢 @amknight 和 @snowzlmbot。 - **PR #91786** fix(plugins): reconcile managed npm root overrides with managed peer pins. 关联 #91772。感谢 @amknight 和 @mkdelta221。 - **PR #93002** Fix Telegram progress draft cleanup before tool output. 关联 #90753。感谢 @zhangguiping-xydt 和 @shadow-enthusiast。 - **PR #95175** fix: route mobile exec approvals to reviewer device. 感谢 @joshavant。 - **PR #94506** fix(telegram): stop clearing registered webhook on channel restart. 关联 #90254。感谢 @xialonglee 和 @travellingsoldier85。 - **PR #95183** fix(telegram): materialize streaming progress placeholders. 关联 #95004。感谢 @snowzlmbot 和 @obviyus。 - **PR #95483** fix(whatsapp): preserve native quote replies. 感谢 @mcaxtr。 - **PR #94338** fix(whatsapp): wire missing Baileys retry/cache hooks for group message reliability. 关联 #7433。感谢 @xialonglee、@mcaxtr 和 @octopuslabs-fl。 - **PR #94707** feat(slack): add relay mode for incoming messages. 感谢 @sjf-oa。 - **PR #94977** fix(telegram): honor outbound reaction directives. 关联 #71140。感谢 @hugenshen 和 @Cuttingwater。 - **PR #95069** fix(telegram): skip mirror write when primary reply already exists (#94930). 感谢 @heichaowo。 - **PR #95550** Preserve inherited channel account policies during migration. 感谢 @amknight。 - **PR #95390** fix #95378: https://github.com/openclaw/openclaw/issues/95378. 感谢 @mikasa0818、@obviyus 和 @MaiDuy708。 - **PR #93143** fix(imessage): keep split-send coalescing opt-in. 感谢 @omarshahine。 - **PR #95008** fix(claude-cli): also disable native background Bash and Monitor in --print runs. 感谢 @anagnorisis2peripeteia。 - **PR #94545** fix: keep trusted policies with hook registry. 感谢 @jesse-merhi。 - **PR #95007** fix(telegram): render progress draft rows as plain readable text. 关联 #95002。感谢 @snowzlmbot。 - **PR #95624** fix(agents): keep long-context tool-result prompts cache-stable. 感谢 @vincentkoc。 - **PR #95625** fix(ci): smooth PR runner-registration bursts. 感谢 @vincentkoc。 - **PR #95572** fix(agents): reject bind specs with extra colon segments. 感谢 @ly-wang19。 - **PR #95653** test(agents): add large prompt cache coverage. 感谢 @vincentkoc。 - **PR #95640** Consolidate iOS notification permission UX. 感谢 @joshavant。 - **PR #95546** feat(mattermost): register /oc_queue as a native slash command. 感谢 @amknight。 - **PR #95019** fix(skills): point gog brew install at homebrew-core gogcli (#95017). 感谢 @ZengWen-DT、@vincentkoc 和 @Sedrak-Hovhannisyan。 - **PR #93378** test(telegram): keep live polling leases protected. 关联 #93375。感谢 @mmyzwl 和 @Yachiyo1680。 - **PR #95084** fix(googlechat): sanitize internal tool-trace lines from outbound text (#90684). 感谢 @jailbirt 和 @studentzhou-svg。 - **PR #95278** Avoid copying process.env in ingress queue state DB opens. 关联 #94571。感谢 @kaka-srp。 - **PR #95577** fix #86957: drain worker-spooled Telegram updates immediately. 感谢 @LiuwqGit 和 @freidrich-goldenflow。 - **PR #95128** fix(compaction): count user-message image blocks in cut-point estimator. 感谢 @yetval。 - **PR #93887** fix(ssh): reject hostnames with stray leading or trailing colons in parseSshTarget. 感谢 @miorbnli。 - **PR #95191** docs(plugins): document subagent_ended hook fields. 关联 #95186。感谢 @MaHaoHao-ch 和 @ken-jo。 - **PR #95102** fix(config): add stdio to McpServerSchema transport union. 关联 #95082。感谢 @lzyyzznl 和 @ken-jo。 - **PR #95465** fix(sdk): type-narrow manifest.files in pack staging root helper. 感谢 @wangmiao0668000666。 - **PR #95664** refactor(plugins): move owner skills into plugins. 感谢 @vincentkoc。 - **PR #95299** fix #95248: OpenClaw release_lane is a no-op when claim is held by a live worker; stuck Telegram inbound events block agent response until gateway restart. 感谢 @mikasa0818 和 @kriegerbangerz-ship-it。 - **PR #94687** fix(gateway): accept port for health and probe. 关联 #79100。感谢 @BryanTegomoh 和 @ozthedivine。 - **PR #95649** fix(ci): bundle test shards and right-size runners. 感谢 @vincentkoc。 - **PR #95243** fix(docs): show inline read_when hints in docs:list. 感谢 @hugenshen 和 @vincentkoc。 - **PR #95283** fix(openai-completions): seal native reasoning before the answer under /reasoning on. 关联 #95280。感谢 @ZengWen-DT、@vincentkoc 和 @Marvinthebored。 - **PR #95497** feat(raft): add CLI wake bridge channel. 感谢 @vincentkoc。 - **PR #95459** fix(cron): use main-session systemEvent for silent quick-create preset. 关联 #95073。感谢 @ZOOWH、@vincentkoc 和 @vporton。 - **PR #95503** fix #89466: [Bug]: Control UI chat input text not cleared after sending. 感谢 @zhangguiping-xydt、@vincentkoc 和 @zhong18804784882。 - **PR #95684** fix(skills): harden ClawHub update policy. 感谢 @vincentkoc。 - **PR #95683** feat(plugins): externalize additional official plugins. 感谢 @vincentkoc。 - **PR #95681** fix(ci): debounce canonical main runner admission. 感谢 @vincentkoc。 - **PR #95652** fix(agents): activate selected harness plugins. 感谢 @vincentkoc。 - **PR #95404** fix(codex): stream non-final-answer assistant deltas as partials. 关联 #95422。感谢 @agonza1 和 @vincentkoc。 - **PR #58993** fix(googlechat): support spaceType field for DM vs Space detection. 感谢 @Starhappysh 和 @vincentkoc。 - **PR #94148** fix(doctor): prevent non-interactive --fix from auto-restarting gateway. 关联 #78217。感谢 @zhangguiping-xydt 和 @esqandil。 - **PR #89859** fix(gateway): honor remote status probe timeout. 关联 #65355。感谢 @mushuiyu886 和 @gozzbb2。 - **PR #95466** fix(ci): increase timeouts in flaky process-group signal test. 感谢 @jason-allen-oneal。 - **PR #95720** fix(matrix): prevent double bootstrapCrossSigning reset in forced reset. 关联 #78396。感谢 @vincentkoc 和 @jteddy。 - **PR #95707** fix(synology-chat): remove duplicate local deliver timeout. 感谢 @vincentkoc。 - **PR #95706** fix(whatsapp): remove dead watchdog timeout clamp. 感谢 @vincentkoc。 - **PR #95719** fix(cli): sync capability inspect metadata flags with registered options. 感谢 @vincentkoc。 - **PR #95721** fix(active-memory): exclude dreaming-narrative session keys from eligibility gate. 关联 #78500。感谢 @vincentkoc 和 @vishutdhar。 - **PR #95602** test: save ~79 CI hours/mo in gateway session utils. 感谢 @zats 和 @vincentkoc。 - **PR #94412** fix(agent-core): stop loop after aborted tool run. 感谢 @szsip239 和 @vincentkoc。 - **PR #94915** fix(gateway): report draining state in readiness. 关联 #78136。感谢 @markoub、@vincentkoc 和 @maxschachere。 - **PR #95691** fix(ui): bump dompurify to patched release. 感谢 @vincentkoc。 - **PR #95710** fix(vercel-ai-gateway): resolve dynamic model selections. 感谢 @vincentkoc。 - **PR #94072** fix(agents): count message-tool source reply as user-facing reply for tool error warnings. 关联 #93875。感谢 @chenyangjun-xy、@vincentkoc 和 @hoyanhan。 - **PR #94784** fix(doctor): stop promising --fix for working isolated shell-prompt cron jobs (#94655). 感谢 @ZengWen-DT、@altaywtf 和 @geekoagent。 - **PR #93504** fix(device-pairing): guard role normalization against non-string entries. 感谢 @ly-wang19。 - **PR #92178** fix(gateway): normalize malformed paired access lists. 关联 #90654。感谢 @wangmiao0668000666、@ly-wang19 和 @EmilioNicolas。 - **PR #83041** Fix config patch restart-required notices. 关联 #46797。感谢 @xuruiray 和 @Stache73。 - **PR #95754** fix(cron): preserve no-config delivery validation. 感谢 @vincentkoc。 - **PR #93351** feat(cli): add --message-file to openclaw agent. 感谢 @ooiuuii。 - **PR #95485** fix(ui): roll values near 1M over from k to M in compact token format. 感谢 @NarahariRaghava 和 @vincentkoc。 - **PR #94622** fix(build): allow tsdown heap override. 感谢 @tayoun。 - **PR #76668** meta(issue-template): add dedicated docs bug report form. 关联 #76664。感谢 @WadydX。 - **PR #91193** fix(cli): document Commander rawArgs internal API dependency in action-reparse.ts. 关联 #83893。感谢 @whiteyzy 和 @davinci282828。 - **PR #77339** fix(auto-reply): clear runtime model cache on reset. 关联 #77322。感谢 @mjamiv 和 @ZaynL。 - **PR #89628** Speed up precomputed command help startup. 感谢 @yyzquwu。 - **PR #95087** refactor: add memory and QMD session identity mapping. 感谢 @jalehman。 - **PR #89323** fix(web-ui): skip hidden subagent picker pages. 关联 #89249。感谢 @giodl73-repo 和 @originsecured-do。 - **PR #84794** Clean up isolated cron sessions after runs. 关联 #84707。感谢 @TurboTheTurtle 和 @bottenbenny。 - **PR #95794** fix(cron): compare thread IDs when deduping failure destinations. 感谢 @vincentkoc。 - **PR #95791** fix(session-memory): sanitize model artifacts before saving memory. 感谢 @vincentkoc、@SweetSophia 和 @YB0y。 - **PR #95805** fix(agents): normalize hallucinated Office file extensions. 关联 #93326。感谢 @vincentkoc、@lzyyzznl、@xzh-icenter 和 @bhnan。 - **PR #89612** fix: include persisted plugin contracts for migrations. 关联 #89609。感谢 @zerone0x 和 @mugabuga。 - **PR #89981** fix(diagnostics-otel): keep full model id on spans instead of collapsing to "unknown". 感谢 @mycarrysun 和 @vincentkoc。 - **PR #90489** fix(sessions): clarify cross-agent visibility guidance. 关联 #90443。感谢 @sahibzada-allahyar、@vincentkoc 和 @ramitrkar-hash。 - **PR #95383** fix(sdk): classify failed/blocked tool events as tool.call.failed. 感谢 @ly-wang19。 - **PR #95827** fix(plugin-sdk): bound live model catalog success body. - **PR #95792** fix(onboard): refresh provider plugin registry after setup installs. 关联 #95765。感谢 @snowzlmbot。 - **PR #93943** fix(provider-usage): honor proxy env for usage fetch. 关联 #78714。感谢 @TurboTheTurtle 和 @TnzGit。 - **PR #89648** fix(agents): restore model-fetch info logs. 关联 #89300。感谢 @xiaobao-k8s 和 @Enominera。 - **PR #93841** fix(ui): render persisted history text blocks. 关联 #90241。感谢 @mushuiyu886 和 @pronzcw。 - **PR #93502** docs: fix docs metadata spellcheck. 感谢 @harjothkhara。 - **PR #86608** docs: add existing-solutions preflight guardrail. 感谢 @cablackmon。 - **PR #95322** fix(reply): preserve usage footer across rollover. 感谢 @litang9。 - **PR #91559** fix(agents): clean Gemini tool schemas by model id. 关联 #91542。感谢 @Pick-cat 和 @qiukui666。 - **PR #94701** Fix embedded-run recovery after refreshed session activity. 感谢 @imadal1n。 - **PR #95472** [AI] fix(main-session): skip current-gen abort controllers for completed sessions. 感谢 @xydt-tanshanshan。 - **PR #95401** fix(lmstudio): canonicalize variant model keys. 感谢 @MonkeyLeeT。 - **PR #94323** fix(cron): stop add/remove from dropping a due recurring job's pending run. 感谢 @yetval。 - **PR #94443** fix(memory-wiki): retry transient source-page rewrite race instead of aborting wiki_status. 关联 #92134。感谢 @ZengWen-DT 和 @cknzraposo。 - **PR #95120** feat(channels): add directUserId support for per-DM model override. 关联 #53638。感谢 @xydigit-zt、@thomaszta 和 @gandalf-at-lerian。 - **PR #91049** fix: handle terminal chat send acknowledgements. 关联 #91048。感谢 @nxmxbbd。 - **PR #95081** fix: keep text transform runtime imports hashed. 关联 #95057。感谢 @849261680 和 @YvesLaRose。 - **PR #93356** fix(plugins): cache plugin setup registry to kill the /models CPU storm. 感谢 @obuchowski。 - **PR #85341** refactor: internalize OpenClaw agent runtime. - **PR #94233** fix(model-fallback): coalesce auth decision logs. 关联 #56979。感谢 @goutamadwant 和 @yanan1991。 - **PR #92399** fix(llm): collapse cumulative openai-responses message snapshots instead of concatenating [AI-assisted]. 关联 #91959。感谢 @amersheeny 和 @phoenixyy。 - **PR #94811** fix(ollama): honor memory embedding output dimensionality. 感谢 @mushuiyu886。 - **PR #84708** fix(agents): recover message-tool mirror replay poison. 感谢 @anyech。 - **PR #95274** fix(memory): preserve Windows QMD command paths. 关联 #92302。感谢 @ly85206559 和 @Ardooken。 - **PR #95686** fix(xai): request encrypted reasoning include for all reasoning models. 感谢 @geraint0923 和 @fuller-stack-dev。 - **PR #82909** fix(telegram): repair message cache reply types. 感谢 @lidge-jun。 - **PR #94942** fix(matrix): prune finished fake-indexeddb transactions to prevent OOM. 关联 #90455。感谢 @xzh-icenter 和 @yar-sh。 - **PR #93994** fix(setup): point non-interactive health hints at onboard flags. 关联 #93947。感谢 @zhouhe-xydt 和 @NianJiuZst。 - **PR #93956** fix(ollama): skip auto-discovery for remote/cloud base URLs. 感谢 @jason-allen-oneal。 - **PR #93547** fix(acp): recover stale persistent sessions by structured resume-required code [AI-assisted]. 关联 #87830。感谢 @amersheeny 和 @chouzz。 - **PR #94204** fix(control-ui): rewrite manifest hrefs for configured base path. 关联 #94157。感谢 @hugenshen 和 @xrow。 - **PR #95637** fix(qqbot): recognize GFM table separators with one or two dashes. 感谢 @ly-wang19。 - **PR #95467** fix(sessions): keep bound channel identity across non-delivery turns. 感谢 @yetval。 - **PR #93919** perf(plugins): cache existence probes within bundle manifest scan [AI-assisted]. 关联 #76209。感谢 @ml12580 和 @shenhonglong456-ai。 - **PR #94726** fix(google): add gemini-3.5-flash model catalog entry. 关联 #94723。感谢 @ajwan8998 和 @anguslogan01。 - **PR #94238** fix(config): fail closed when configure runs without an interactive TTY (#93953). 感谢 @ruomuxydt 和 @NianJiuZst。 - **PR #89949** fix(media): pin requester delivery route when task starts. 关联 #86034。感谢 @wangwllu 和 @tianxiaochannel-oss88。 - **PR #94470** chore: sync yuanbao plugin catalog to 2.15.0. 感谢 @jase-283。 - **PR #93585** fix(daemon): keep systemd gateway running after child OOM. 感谢 @snowzlm。 - **PR #94442** fix(imessage): strip leading echo corruption markers in the persisted echo cache. 感谢 @ly-wang19。 - **PR #93511** fix(imessage): normalize leading NUL echo-cache prefixes. 感谢 @vincentkoc 和 @ly-wang19。 - **PR #87121** test(cli): add banner emission reset helper. 关联 #83903。感谢 @lizuju 和 @davinci282828。 - **PR #89172** fix(feishu): show voice message duration via upload duration. 关联 #53798。感谢 @areslp 和 @kinrocW。 - **PR #95118** fix: bound agents load their own inbound workspace context. 关联 #92903。感谢 @849261680 和 @axjing。 - **PR #94752** fix(reply): clarify stale model override resets. 关联 #94713。感谢 @snowzlmbot 和 @Gr4via。 - **PR #89716** fix(providers): strip cache-boundary marker from non-Anthropic prompts. 感谢 @masatohoshino。 - **PR #95268** fix(openrouter): expand short canonical model IDs to upstream API slugs (fixes #95198). 感谢 @Darren2030 和 @daniel-alejandro-t。 - **PR #90475** fix(telegram): keep bot reply answers anchored to current message. 感谢 @moeedahmed。 - **PR #94219** fix(control-ui): restore provider usage quota pill in sidebar session switcher (fixes #93041). 感谢 @Pick-cat 和 @jazzroutine。 - **PR #93369** fix(cron): expose per-job fallbacks in CLI. 关联 #90302。感谢 @849261680 和 @Walliiee。 - **PR #87912** Handle Codex toolResult blocks in tool-result truncation. 感谢 @AdrianIp0204。 - **PR #95744** fix(model-catalog): strip manifest model-id prefixes by the matched length. 关联 #95743。感谢 @parveshsaini。 - **PR #95300** fix(cli): expose --count on infer image edit, matching image generate. 感谢 @ly-wang19。 - **PR #94156** fix: expose OpenAI image quality and moderation CLI options. 感谢 @lastguru-net 和 @ly-wang19。 - **PR #94062** fix(agents): classify generic "LLM request failed." as transient time…. 关联 #93931。感谢 @hugenshen 和 @hyphae-bot。 - **PR #93965** fix(opencode-go): streaming completes when provider ends responses. 关联 #93610。感谢 @zhangguiping-xydt 和 @ForceConstant。 - **PR #89017** fix(webchat): sessions persist after reconnects. 关联 #87700。感谢 @zhangguiping-xydt 和 @asicoe。 - **PR #94627** fix(ios): centralize app accent colors. 感谢 @zats。 - **PR #94054** fix(gateway.tls): reject empty/whitespace certPath and keyPath. 感谢 @miorbnli。 - **PR #95674** fix(cron): trim trailing whitespace from recognized job object keys. 关联 #95407。感谢 @zw-xysk 和 @Nassiel。 - **PR #95857** Fix SQLite user version guardrail allowlist. 感谢 @RomneyDa。 - **PR #95858** feat(qa): expose active memory toggles to scenarios. 感谢 @RomneyDa。 - **PR #84326** Doctor: expose sandbox registry findings. 感谢 @giodl73-repo。 - **PR #73079** fix(minimax): request hex TTS output explicitly. 感谢 @efe-arv。 - **PR #94483** feat(gateway-cli): scope usage-cost by agent. 感谢 @ly-wang19。 - **PR #91502** feat(qa): crabline channel driver. 感谢 @RomneyDa。 - **PR #95872** Move TUI PTY tests into CI node shard. 感谢 @RomneyDa。 - **PR #89800** fix(agents): resolve webchat current session status. 关联 #89773。感谢 @sweetcornna 和 @killo3967。 - **PR #95870** fix(ci): restore macOS and Windows QA gates. 感谢 @vincentkoc。 - **PR #94445** fix(agents): keep cron cloud idle watchdog enabled. 感谢 @bek91。 - **PR #91506** fix(qa): reserve shared QA suite flags across runners. 感谢 @RomneyDa。 - **PR #95876** ci: add manual release QA profile evidence workflow. 感谢 @RomneyDa。 - **PR #95879** fix(ci): use workflow revision for proof checks. 感谢 @vincentkoc。 - **PR #93113** fix(memory-core): report active dreaming phases in status. 关联 #67868。感谢 @AgentArcLab 和 @mrossit。 - **PR #95837** Simplify color mode button labels. 感谢 @SannidhyaSah 和 @hannesrudolph。 - **PR #95797** fix: /status is too verbose for pinned model sessions. 感谢 @Solvely-Colin。 - **PR #90223** test: make qqbot symlinked media helper test robust on Windows. 感谢 @aniruddhaadak80。 - **PR #95557** improve: refresh Android overview control surface. 感谢 @Solvely-Colin 和 @joshavant。 - **PR #95593** fix: route Android exec approvals to in-app inbox. 感谢 @Solvely-Colin。 - **PR #95570** fix(cli): resolve trajectory export stores consistently. 关联 #95568。感谢 @youngting520。 - **PR #95880** ci: generalize QA profile evidence workflow. 感谢 @RomneyDa。 - **PR #95218** fix(agents): bound provider JSON response reads. 感谢 @Alix-007。 - **PR #95614** fix(memory-wiki): preserve human notes block on source re-ingest. 感谢 @yetval。 - **PR #95148** feat(android): add settings detail panels. 感谢 @Tosko4。 - **PR #95893** Add iOS push sandbox profiles and relay tooling. 感谢 @joshavant。 - **PR #95723** fix(control-ui): exclude disabled cron jobs from Overview failed count. 关联 #95716。感谢 @ZengWen-DT 和 @voytas75。 - **PR #78715** Fix minor grammar issue in plugin documentation (capabilities plural). 感谢 @mehrazmorshed。 - **PR #95890** fix(ci): restore QA workflow gates. 感谢 @vincentkoc。 - **PR #95661** fix(discord): reserve closing-fence space on fence-closing lines. 感谢 @ly-wang19。 - **PR #94272** ci: add maturity scorecard renderer. 感谢 @RomneyDa。 - **PR #95901** ci: add maturity scorecard renderer. 感谢 @RomneyDa。 - **PR #95914** fix(whatsapp): preserve durable reply target. 感谢 @mcaxtr。 - **PR #95697** improve: reduce hot-path linear scans and redundant I/O. 感谢 @vincentkoc。 - **PR #95916** fix(memory): improve node:sqlite unavailable guidance. 感谢 @vincentkoc 和 @rrrrrredy。 - **PR #78884** docs: document local avatar file size limit. 关联 #65312。感谢 @wangjieweb3-design 和 @wraxle-geargrind。 - **PR #95909** fix(ci): use available Android SDK platform. 感谢 @vincentkoc。 - **PR #68389** plugins: clarify allowlist warning when entries don't match discovered ids. 关联 #68352。感谢 @lyfuci 和 @JIRBOY。 - **PR #95898** ci: simplify maturity scorecard QA evidence inputs. 感谢 @RomneyDa。 - **PR #95928** fix(ci): honor reusable QA evidence failure policy. 感谢 @vincentkoc。 - **PR #95508** fix #95489: [Bug]: claude-cli out-of-credits error bypasses model fallback chain — error text delivered as final response. 感谢 @mikasa0818、@sallyom 和 @riazrahaman。 - **PR #95420** fix(agents): bound OpenRouter model catalog response reads. 感谢 @Alix-007 和 @sallyom。 - **PR #95103** fix(gateway): bound pricing catalog streams. 感谢 @vincentkoc 和 @sallyom。 - **PR #95108** fix(agents): bound Anthropic error streams. 感谢 @vincentkoc 和 @sallyom。 - **PR #95418** fix(agents): bound OpenRouter model-scan catalog success body. 感谢 @Alix-007。 - **PR #95922** fix(ci): finalize testbox sessions after setup failures. 感谢 @vincentkoc。 - **PR #95417** fix(agents): bound Google prompt cache response reads. 感谢 @Alix-007。 - **PR #95935** fix(whatsapp): resolve approval reactions across JID drift. 感谢 @mcaxtr。 - **PR #84340** Doctor: expose extra gateway service findings. 感谢 @giodl73-repo。 - **PR #95946** fix(ci): finalize Windows Testbox after setup failures. 感谢 @vincentkoc。 - **PR #95947** feat(qa): add "all" taxonomy profile. 感谢 @RomneyDa。 - **PR #91903** feat(plugin-sdk): add extensible channel identity hook context. 感谢 @lanzhi-lee 和 @vincentkoc。 - **PR #95930** docs(copilot): refresh harness parity notes. 感谢 @vincentkoc。 - **PR #95952** fix(ci): require QA live evidence artifacts. 感谢 @vincentkoc。 - **PR #95666** Fix memory-wiki bridge self-import loop. 关联 #95657。感谢 @TurboTheTurtle、@vincentkoc 和 @Johannes0402。 - **PR #91724** fix(agents): infer runtime provider from qualified model ids. 感谢 @yu-xin-c 和 @vincentkoc。 - **PR #95957** improve: speed up provider tool-call streaming. 感谢 @vincentkoc。 - **PR #95541** fix(cli): show working commands for pinned plugin drift. 感谢 @ooiuuii 和 @vincentkoc。 - **PR #95933** docs: update maturity scorecard. 感谢 @RomneyDa。 - **PR #78105** fix(plugins): make empty-allowlist actionable for new users. 关联 #68780。感谢 @pahuchi-joe 和 @JIRBOY。 - **PR #95961** fix(ci): require live proof evidence artifacts. 感谢 @vincentkoc。 - **PR #95400** fix(model-fallback): classify Codex usage-limit payloads. 感谢 @jason-allen-oneal 和 @sallyom。 - **PR #95944** perf(qa-lab): speed up unified QA suites. 感谢 @RomneyDa。 - **PR #95971** ci: fail QA profile evidence on QA failures. 感谢 @RomneyDa。 - **PR #95845** feat: add bundled plugin icon manifest URLs. 感谢 @Patrick-Erichsen。 - **PR #95975** fix(ci): require iOS Periphery evidence artifact. 感谢 @vincentkoc。 - **PR #95972** docs: update ClawHub skill route references. 感谢 @Patrick-Erichsen。 - **PR #90537** Warn on generated wrapper overwrites and status diagnostics. 关联 #90518。感谢 @TurboTheTurtle 和 @vincentkoc。 - **PR #93567** fix(cron): normalize run-log jobId on write to match read-side validation. 感谢 @Alix-007 和 @vincentkoc。 - **PR #95967** perf(ci): widen main test fanout and move codeql off blacksmith. 感谢 @vincentkoc。 - **PR #95983** fix(ci): require OpenGrep SARIF artifacts. 感谢 @vincentkoc。 - **PR #64490** CLI: escape zsh completion descriptions. 感谢 @EdenKangdw。 - **PR #95682** Fix Gemini day freshness time range handling. 感谢 @Sunjae-k 和 @vincentkoc。 - **PR #93374** fix(agents): suggest recovery for unknown tool ids. 关联 #92273。感谢 @mushuiyu886、@vincentkoc 和 @poison。 - **PR #95991** fix(ci): require release QA evidence artifacts. 感谢 @vincentkoc。 - **PR #89886** fix(context-engine): forward abortSignal through delegation bridge to runtime compaction. 关联 #89868。感谢 @openperf、@vincentkoc 和 @lykeion-dev。 - **PR #95999** fix(release): require postpublish evidence artifact. 感谢 @vincentkoc。 - **PR #96003** test(qa): harden all-profile QA evidence scenarios. 感谢 @RomneyDa。 - **PR #95860** fix(compaction): trim prefix when transcript ends in an oversized tool result. 关联 #78478。感谢 @yetval 和 @wzhgba。 - **PR #95934** fix(xiaomi): correct mimo-v2.5 and mimo-v2.5-pro max output tokens to 128K. 感谢 @idootop。 - **PR #95782** fix(sessions): honor configured store for outbound transcript mirrors. 关联 #95781。感谢 @youngting520。 - **PR #96005** feat(copilot): wire harness parity helpers. 感谢 @vincentkoc。 - **PR #95886** fix: avoid false macOS update failures during gateway shutdown. 感谢 @fuller-stack-dev。 - **PR #91906** perf: skip subagent live stream parsing. 感谢 @lanzhi-lee。 - **PR #87861** fix(model-usage): coerce numeric-string costs and ignore non-finite values. 关联 #37878。感谢 @coder999999999、@vincentkoc 和 @shuofengzhang。 - **PR #96030** fix(qa-lab): avoid duplicate child evidence files. 感谢 @RomneyDa。 - **PR #94369** fix(memory-wiki): exclude durable reference pages from stale report. 感谢 @SunnyShu0925 和 @vincentkoc。 - **PR #94578** Fix recent session resume with long headers. 关联 #94577。感谢 @rohitjavvadi 和 @vincentkoc。 - **PR #95919** ci: add Codex maturity scorecard agent. 感谢 @RomneyDa。 - **PR #92356** fix(heartbeat): skip reasoning payloads when selecting heartbeat reply. 关联 #92260。感谢 @tangtaizong666、@vincentkoc 和 @jmpei。 - **PR #94328** fix(agents): keep post-compaction user re-issue of a kept-tail prompt during compaction rotation. 感谢 @yetval 和 @vincentkoc。 - **PR #95432** fix(reply): suppress per-message finals across multi-message block streaming. 感谢 @yetval 和 @vincentkoc。 - **PR #95431** fix(auto-reply): keep drain/restart-abort reply paths silent. 感谢 @moeedahmed 和 @vincentkoc。 - **PR #96017** test(qa): gate maturity docs on passing evidence. 感谢 @RomneyDa。 - **PR #95552** feat(mattermost): persist participated threads for mention-free follow-ups. 感谢 @amknight。 - **PR #95480** fix #89231: [Bug]: Windows installer-created scheduled task launches gateway.cmd with visible console — should use windowless launcher. 感谢 @mikasa0818、@vincentkoc 和 @CameronWeller。 - **PR #96049** fix(copilot): preserve compaction metadata. 感谢 @vincentkoc。 - **PR #96044** docs: rename top maturity tier to Clawesome. 感谢 @RomneyDa。 - **PR #92154** Gate private QQBot group commands. 感谢 @sliverp。 - **PR #95484** fix: assistant reply lost between compaction summary and first kept user in successor transcript. 关联 #76729。感谢 @maweibin、@vincentkoc 和 @njuboy11。 - **PR #95094** ci: add release QA profile evidence. 感谢 @RomneyDa。 - **PR #96057** docs: redesign maturity scorecard pages. 感谢 @vincentkoc。 - **PR #96014** perf(agents): index displaced tool results. 感谢 @vincentkoc。 - **PR #96019** perf(usage): bound session log retention. 感谢 @vincentkoc。 - **PR #96013** perf(anthropic): index active stream blocks. 感谢 @vincentkoc。 - **PR #96061** docs: place maturity pages under release reference. 感谢 @vincentkoc。 - **PR #95589** fix: npm plugin updates break running gateway imports. 感谢 @ooiuuii 和 @vincentkoc。 - **PR #96068** fix(acpx): consume acpx 0.11.1 model capability errors. 关联 #95869。感谢 @vincentkoc 和 @SabaTech-dev。 - **PR #89912** refactor: add transcript update identity contract. 感谢 @jalehman。 - **PR #96055** fix(maint): use rebase PR landing. 感谢 @vincentkoc。 - **PR #96062** feat(copilot): mirror native plan and subagent events. 感谢 @vincentkoc。 - **PR #96032** fix(acpx): detect wrapper orphan on any PPID change, not just init reparenting. 感谢 @t2wei 和 @vincentkoc。 - **PR #96124** chore(acpx): bump bundled client to 0.11.2. 感谢 @vincentkoc。 - **PR #84352** Fix WebChat dispatch failure session status. 感谢 @jesse-merhi。 - **PR #89518** refactor: migrate plugin transcript mirrors. 感谢 @jalehman。 - **PR #90439** refactor: add embedded run session target seam. 感谢 @jalehman。 - **PR #95699** perf(gateway): drop redundant per-access session-key case scan. 感谢 @jzakirov 和 @jalehman。 - **PR #95992** fix(skills): accept owner-qualified verify refs. 感谢 @Patrick-Erichsen。 - **PR #95987** fix(plugins): remove Simple Icons color paths. 感谢 @Patrick-Erichsen。 - **PR #96162** refactor: use accessor-backed transcript corpus for memory. 感谢 @jalehman。 - **PR #96181** Defer iOS local network permission until onboarding. 感谢 @joshavant。 - **PR #95475** test(cli): isolate service env in run and update suites. 感谢 @kklouzal。 - **PR #95226** fix(infra): bound ClawHub fetchJson and error response bodies. 感谢 @Alix-007。 - **PR #95240** fix(matrix): bound non-raw JSON response body in transport. 感谢 @Alix-007。 - **PR #91742** fix(memory): abort orphaned embedding work when memory_search times out. 关联 #91718。感谢 @dreamhunter2333 和 @NOVA-Openclaw。 - **PR #93394** fix(memory): abort orphaned qmd search subprocess when memory_search times out. 感谢 @Alix-007。 - **PR #89911** refactor: migrate bundled transcript target lookups. 感谢 @jalehman。 - **PR #96191** refactor: route plugin host hook state through accessor. 感谢 @jalehman。 - **PR #96179** fix: route gateway history through session accessor target. 感谢 @jalehman。 - **PR #96201** refactor: add abort target session accessor. 感谢 @jalehman。 - **PR #96193** fix(memory-core): migrate dreaming cleanup lifecycle. 感谢 @jalehman。 - **PR #96195** fix: bridge ACP metadata to session accessors. 感谢 @jalehman。 - **PR #96182** refactor: migrate agent session accessors. 感谢 @jalehman。 - **PR #96218** refactor: guard reply session initialization. 感谢 @jalehman。 - **PR #96213** refactor(gateway): add alias mutation accessor. 感谢 @jalehman。 - **PR #96204** refactor: migrate command session persistence to accessor. 感谢 @jalehman。 - **PR #96206** refactor: route live model fresh reads through session accessor. 感谢 @jalehman。 - **PR #96220** fix(whatsapp): quote current follow-up in durable replies. 感谢 @mcaxtr。 - **PR #96226** fix(macos): drop Textual from chat packaging. 感谢 @vincentkoc。 - **PR #96212** fix(codex): deliver generated images from remote app-server. 感谢 @sjf-oa。 - **PR #96235** fix(crabbox): require Xcode for macOS proof. 感谢 @vincentkoc。 - **PR #96145** fix: UI glitch: config is not visible. 关联 #94202。感谢 @sunlit-deng 和 @vporton。 - **PR #96072** perf(browser): index role snapshot references. 感谢 @vincentkoc。 - **PR #96085** perf(codex): index rollout transcript ids. 感谢 @vincentkoc。 - **PR #96087** perf(reply): hoist direct-send fragment index. 感谢 @vincentkoc。 - **PR #94154** fix(gateway): resolve plugin-registered gateway methods through live registry. 关联 #94127。感谢 @Pick-cat、@vincentkoc 和 @BryceMurray。 - **PR #95393** fix #92582: Bug: doctor falsely warns local memory embeddings are not ready. 感谢 @mikasa0818、@vincentkoc 和 @neekolascmd。 - **PR #96246** fix(qa): accept pnpm separator for lab up. 感谢 @vincentkoc。 - **PR #94949** fix(ports): route isPortBusy through checkPortInUse to catch IPv4-only occupants. 关联 #94426。感谢 @sunlit-deng、@vincentkoc 和 @wangwllu。 - **PR #94562** fix(workboard): hide archived cards in CLI list by default. 关联 #94555。感谢 @ZengWen-DT、@vincentkoc 和 @ecican。 - **PR #96243** fix(nextcloud-talk): ignore signed non-message webhook events. 关联 #81566。感谢 @arkyu2077、@vincentkoc 和 @rafaelmgbh。 - **PR #96140** fix(exec): preserve turn-source routing target in approval followups for plugin channels. 关联 #96103。感谢 @yetval、@vincentkoc 和 @lansenger-pm。 - **PR #96258** ci: move CodeQL quality scans to hosted runners. 感谢 @vincentkoc。 - **PR #96271** chore(release): close out 2026.6.10 on main. 感谢 @vincentkoc。 - **PR #96233** fix(agents): run heartbeat_prompt_contribution on harness prompt builds. 感谢 @azogheb 和 @vincentkoc。 - **PR #55018** fix: avoid O(N²) shallow-copy in mapSensitivePaths schema traversal. 感谢 @xdhuangyandi、@vincentkoc 和 @huangyandi-red。 - **PR #95831** fix: compact Codex OAuth OpenAI sessions without API keys. 关联 #95693。感谢 @sallyom 和 @YUI-TIEN。 - **PR #96244** fix(auto-reply): align channel intro wording with chat_type. 关联 #95645。感谢 @arkyu2077、@vincentkoc 和 @iloveleon19。 ### 发布验证 - npm 包：https://www.npmjs.com/package/openclaw/v/2026.6.11-beta.1 - 发布 SHA：`c862a644bff366599ead53df81f452aa913497ae` - 完整发布验证：https://github.com/openclaw/openclaw/actions/runs/28123941058 - npm 预检：https://github.com/openclaw/openclaw/actions/runs/28130211221 - npm 发布：https://github.com/openclaw/openclaw/actions/runs/28136271139 - ClawHub 后续：已有的四个代码插件族冲突和预先声明的 `tencent` ID 仍保留在子发布运行中；OpenClaw npm 和 GitHub 测试版发布不受影响。

## 2026.6.11 ### Highlights - **More capable channel control:** Slack relay mode, native Mattermost `/oc_queue`, and per-DM model overrides make channel operations easier to automate and tune. (#94707, #95546, #95120) Thanks @sjf-oa, @amknight, @xydigit-zt, @thomaszta, and @gandalf-at-lerian. - **Richer operator workflows:** `openclaw agent --message-file` and the RAFT CLI wake bridge add practical file-driven and remote wake-up paths. (#93351, #95497) Thanks @ooiuuii and @vincentkoc. - **Safer plugin distribution:** additional official plugins are externalized cleanly, with bundled plugin icon metadata available to installed clients. (#95683, #95845) Thanks @vincentkoc and @Patrick-Erichsen. - **Stronger mobile operations:** Android settings detail panels improve configuration visibility and control on mobile. (#95148) Thanks @Tosko4. - **More reliable agent turns:** Codex partial deltas, harness activation, and long-context prompt-cache stability reduce lost progress and inconsistent runs. (#95404, #95652, #95624) Thanks @agonza1 and @vincentkoc. ### Changes - **Gateway and plugin tooling:** channel identity hook context and per-agent usage-cost reporting give integrations and operators more precise routing and accounting. (#91903, #94483) Thanks @lanzhi-lee, @vincentkoc, and @ly-wang19. - **Provider and model coverage:** catalog parsing, reasoning controls, provider model resolution, and encrypted reasoning support now handle more live provider variants. (#95283, #95710, #95268, #95744, #95686, #93956) Thanks @ZengWen-DT, @vincentkoc, @Marvinthebored, @Darren2030, @daniel-alejandro-t, @parveshsaini, @geraint0923, @fuller-stack-dev, and @jason-allen-oneal. ### Fixes - **Channel delivery:** Telegram progress rendering, webhook lifecycle, reaction directives, duplicate mirror writes, queued update draining, and WhatsApp durable reply targets are now more reliable. (#95532, #93002, #95183, #94506, #94977, #95069, #95577, #95007, #95914) Thanks @amknight, @snowzlmbot, @zhangguiping-xydt, @shadow-enthusiast, @xialonglee, @travellingsoldier85, @obviyus, @hugenshen, @Cuttingwater, @heichaowo, @LiuwqGit, @freidrich-goldenflow, @mcaxtr, and @vincentkoc. - **WhatsApp and message identity:** native quotes, Baileys group reliability, and approval reactions across JID drift now preserve the intended conversation context. (#95483, #94338, #95935) Thanks @mcaxtr, @xialonglee, and @octopuslabs-fl. - **Gateway and session safety:** stuck release claims, draining-state reporting, remote probe timeouts, malformed paired access lists, and non-delivery session identity are handled without silent routing loss. (#95299, #94915, #89859, #92178, #95467) Thanks @mikasa0818, @kriegerbangerz-ship-it, @markoub, @vincentkoc, @maxschachere, @mushuiyu886, @gozzbb2, @wangmiao0668000666, @ly-wang19, @EmilioNicolas, and @yetval. - **Agent and fallback behavior:** aborted runs stop cleanly, provider response bodies stay bounded, Claude CLI credit failures continue through fallback, and Codex usage-limit responses classify correctly. (#94412, #95218, #95508, #95420, #95418, #95417, #95400) Thanks @szsip239, @vincentkoc, @Alix-007, @mikasa0818, @sallyom, @riazrahaman, and @jason-allen-oneal. - **Provider and model edge cases:** OpenRouter IDs, Ollama discovery and embeddings, Gemini freshness, and model-catalog prefixes now resolve against the right runtime metadata. (#95268, #94811, #93956, #95682, #95744) Thanks @Darren2030, @daniel-alejandro-t, @mushuiyu886, @jason-allen-oneal, @Sunjae-k, @parveshsaini, @vincentkoc, and @shakkernerd. - **Configuration and UI guardrails:** non-interactive configure fails closed, TLS paths reject empty values, memory artifacts are sanitized, and the UI uses the patched DOMPurify release. (#94238, #94054, #95791, #95691) Thanks @ruomuxydt, @NianJiuZst, @miorbnli, @vincentkoc, @SweetSophia, and @YB0y. - **Cron and delivery validation:** no-config delivery checks, thread-aware dedupe, and pending recurring runs retain their intended destinations. (#95754, #95794, #94323) Thanks @vincentkoc and @yetval. ### Complete contribution record This audited record covers the complete v2026.6.10..ee421ef7da7edf152ab911d74ffcfd852ecf43e2 history: 305 merged PRs. The generation manifest also supplies direct commits as editorial input; the grouped notes above prioritize user impact. #### Pull requests - **PR #95406** test(qa): make release scorecard categories explicit. Thanks @RomneyDa. - **PR #94700** test: fold HTTP API script proof into QA Lab. Thanks @RomneyDa. - **PR #95499** fix(test): unit-fast flow mocks. Thanks @RomneyDa. - **PR #95308** fix(ci): filter ClawSweeper comment dispatches before token minting. Thanks @vincentkoc. - **PR #95532** fix(telegram): materialize rich message line breaks as <br>. Related #95409. Thanks @amknight and @snowzlmbot. - **PR #91786** fix(plugins): reconcile managed npm root overrides with managed peer pins. Related #91772. Thanks @amknight and @mkdelta221. - **PR #93002** Fix Telegram progress draft cleanup before tool output. Related #90753. Thanks @zhangguiping-xydt and @shadow-enthusiast. - **PR #95175** fix: route mobile exec approvals to reviewer device. Thanks @joshavant. - **PR #94506** fix(telegram): stop clearing registered webhook on channel restart. Related #90254. Thanks @xialonglee and @travellingsoldier85. - **PR #95183** fix(telegram): materialize streaming progress placeholders. Related #95004. Thanks @snowzlmbot and @obviyus. - **PR #95483** fix(whatsapp): preserve native quote replies. Thanks @mcaxtr. - **PR #94338** fix(whatsapp): wire missing Baileys retry/cache hooks for group message reliability. Related #7433. Thanks @xialonglee and @mcaxtr and @octopuslabs-fl. - **PR #94707** feat(slack): add relay mode for incoming messages. Thanks @sjf-oa. - **PR #94977** fix(telegram): honor outbound reaction directives. Related #71140. Thanks @hugenshen and @Cuttingwater. - **PR #95069** fix(telegram): skip mirror write when primary reply already exists (#94930). Thanks @heichaowo. - **PR #95550** Preserve inherited channel account policies during migration. Thanks @amknight. - **PR #95390** fix #95378: https://github.com/openclaw/openclaw/issues/95378. Thanks @mikasa0818 and @obviyus and @MaiDuy708. - **PR #93143** fix(imessage): keep split-send coalescing opt-in. Thanks @omarshahine. - **PR #95008** fix(claude-cli): also disable native background Bash and Monitor in --print runs. Thanks @anagnorisis2peripeteia. - **PR #94545** fix: keep trusted policies with hook registry. Thanks @jesse-merhi. - **PR #95007** fix(telegram): render progress draft rows as plain readable text. Related #95002. Thanks @snowzlmbot. - **PR #95624** fix(agents): keep long-context tool-result prompts cache-stable. Thanks @vincentkoc. - **PR #95625** fix(ci): smooth PR runner-registration bursts. Thanks @vincentkoc. - **PR #95572** fix(agents): reject bind specs with extra colon segments. Thanks @ly-wang19. - **PR #95653** test(agents): add large prompt cache coverage. Thanks @vincentkoc. - **PR #95640** Consolidate iOS notification permission UX. Thanks @joshavant. - **PR #95546** feat(mattermost): register /oc_queue as a native slash command. Thanks @amknight. - **PR #95019** fix(skills): point gog brew install at homebrew-core gogcli (#95017). Thanks @ZengWen-DT and @vincentkoc and @Sedrak-Hovhannisyan. - **PR #93378** test(telegram): keep live polling leases protected. Related #93375. Thanks @mmyzwl and @Yachiyo1680. - **PR #95084** fix(googlechat): sanitize internal tool-trace lines from outbound text (#90684). Thanks @jailbirt and @studentzhou-svg. - **PR #95278** Avoid copying process.env in ingress queue state DB opens. Related #94571. Thanks @kaka-srp. - **PR #95577** fix #86957: drain worker-spooled Telegram updates immediately. Thanks @LiuwqGit and @freidrich-goldenflow. - **PR #95128** fix(compaction): count user-message image blocks in cut-point estimator. Thanks @yetval. - **PR #93887** fix(ssh): reject hostnames with stray leading or trailing colons in parseSshTarget. Thanks @miorbnli. - **PR #95191** docs(plugins): document subagent_ended hook fields. Related #95186. Thanks @MaHaoHao-ch and @ken-jo. - **PR #95102** fix(config): add stdio to McpServerSchema transport union. Related #95082. Thanks @lzyyzznl and @ken-jo. - **PR #95465** fix(sdk): type-narrow manifest.files in pack staging root helper. Thanks @wangmiao0668000666. - **PR #95664** refactor(plugins): move owner skills into plugins. Thanks @vincentkoc. - **PR #95299** fix #95248: OpenClaw release_lane is a no-op when claim is held by a live worker; stuck Telegram inbound events block agent response until gateway restart. Thanks @mikasa0818 and @kriegerbangerz-ship-it. - **PR #94687** fix(gateway): accept port for health and probe. Related #79100. Thanks @BryanTegomoh and @ozthedivine. - **PR #95649** fix(ci): bundle test shards and right-size runners. Thanks @vincentkoc. - **PR #95243** fix(docs): show inline read_when hints in docs:list. Thanks @hugenshen and @vincentkoc. - **PR #95283** fix(openai-completions): seal native reasoning before the answer under /reasoning on. Related #95280. Thanks @ZengWen-DT and @vincentkoc and @Marvinthebored. - **PR #95497** feat(raft): add CLI wake bridge channel. Thanks @vincentkoc. - **PR #95459** fix(cron): use main-session systemEvent for silent quick-create preset. Related #95073. Thanks @ZOOWH and @vincentkoc and @vporton. - **PR #95503** fix #89466: [Bug]: Control UI chat input text not cleared after sending. Thanks @zhangguiping-xydt and @vincentkoc and @zhong18804784882. - **PR #95684** fix(skills): harden ClawHub update policy. Thanks @vincentkoc. - **PR #95683** feat(plugins): externalize additional official plugins. Thanks @vincentkoc. - **PR #95681** fix(ci): debounce canonical main runner admission. Thanks @vincentkoc. - **PR #95652** fix(agents): activate selected harness plugins. Thanks @vincentkoc. - **PR #95404** fix(codex): stream non-final-answer assistant deltas as partials. Related #95422. Thanks @agonza1 and @vincentkoc. - **PR #58993** fix(googlechat): support spaceType field for DM vs Space detection. Thanks @Starhappysh and @vincentkoc. - **PR #94148** fix(doctor): prevent non-interactive --fix from auto-restarting gateway. Related #78217. Thanks @zhangguiping-xydt and @esqandil. - **PR #89859** fix(gateway): honor remote status probe timeout. Related #65355. Thanks @mushuiyu886 and @gozzbb2. - **PR #95466** fix(ci): increase timeouts in flaky process-group signal test. Thanks @jason-allen-oneal. - **PR #95720** fix(matrix): prevent double bootstrapCrossSigning reset in forced reset. Related #78396. Thanks @vincentkoc and @jteddy. - **PR #95707** fix(synology-chat): remove duplicate local deliver timeout. Thanks @vincentkoc. - **PR #95706** fix(whatsapp): remove dead watchdog timeout clamp. Thanks @vincentkoc. - **PR #95719** fix(cli): sync capability inspect metadata flags with registered options. Thanks @vincentkoc. - **PR #95721** fix(active-memory): exclude dreaming-narrative session keys from eligibility gate. Related #78500. Thanks @vincentkoc and @vishutdhar. - **PR #95602** test: save ~79 CI hours/mo in gateway session utils. Thanks @zats and @vincentkoc. - **PR #94412** fix(agent-core): stop loop after aborted tool run. Thanks @szsip239 and @vincentkoc. - **PR #94915** fix(gateway): report draining state in readiness. Related #78136. Thanks @markoub and @vincentkoc and @maxschachere. - **PR #95691** fix(ui): bump dompurify to patched release. Thanks @vincentkoc. - **PR #95710** fix(vercel-ai-gateway): resolve dynamic model selections. Thanks @vincentkoc. - **PR #94072** fix(agents): count message-tool source reply as user-facing reply for tool error warnings. Related #93875. Thanks @chenyangjun-xy and @vincentkoc and @hoyanhan. - **PR #94784** fix(doctor): stop promising --fix for working isolated shell-prompt cron jobs (#94655). Thanks @ZengWen-DT and @altaywtf and @geekoagent. - **PR #93504** fix(device-pairing): guard role normalization against non-string entries. Thanks @ly-wang19. - **PR #92178** fix(gateway): normalize malformed paired access lists. Related #90654. Thanks @wangmiao0668000666 and @ly-wang19 and @EmilioNicolas. - **PR #83041** Fix config patch restart-required notices. Related #46797. Thanks @xuruiray and @Stache73. - **PR #95754** fix(cron): preserve no-config delivery validation. Thanks @vincentkoc. - **PR #93351** feat(cli): add --message-file to openclaw agent. Thanks @ooiuuii. - **PR #95485** fix(ui): roll values near 1M over from k to M in compact token format. Thanks @NarahariRaghava and @vincentkoc. - **PR #94622** fix(build): allow tsdown heap override. Thanks @tayoun. - **PR #76668** meta(issue-template): add dedicated docs bug report form. Related #76664. Thanks @WadydX. - **PR #91193** fix(cli): document Commander rawArgs internal API dependency in action-reparse.ts. Related #83893. Thanks @whiteyzy and @davinci282828. - **PR #77339** fix(auto-reply): clear runtime model cache on reset. Related #77322. Thanks @mjamiv and @ZaynL. - **PR #89628** Speed up precomputed command help startup. Thanks @yyzquwu. - **PR #95087** refactor: add memory and QMD session identity mapping. Thanks @jalehman. - **PR #89323** fix(web-ui): skip hidden subagent picker pages. Related #89249. Thanks @giodl73-repo and @originsecured-do. - **PR #84794** Clean up isolated cron sessions after runs. Related #84707. Thanks @TurboTheTurtle and @bottenbenny. - **PR #95794** fix(cron): compare thread IDs when deduping failure destinations. Thanks @vincentkoc. - **PR #95791** fix(session-memory): sanitize model artifacts before saving memory. Thanks @vincentkoc and @SweetSophia and @YB0y. - **PR #95805** fix(agents): normalize hallucinated Office file extensions. Related #93326. Thanks @vincentkoc and @lzyyzznl and @xzh-icenter and @bhnan. - **PR #89612** fix: include persisted plugin contracts for migrations. Related #89609. Thanks @zerone0x and @mugabuga. - **PR #89981** fix(diagnostics-otel): keep full model id on spans instead of collapsing to "unknown". Thanks @mycarrysun and @vincentkoc. - **PR #90489** fix(sessions): clarify cross-agent visibility guidance. Related #90443. Thanks @sahibzada-allahyar and @vincentkoc and @ramitrkar-hash. - **PR #95383** fix(sdk): classify failed/blocked tool events as tool.call.failed. Thanks @ly-wang19. - **PR #95827** fix(plugin-sdk): bound live model catalog success body. - **PR #95792** fix(onboard): refresh provider plugin registry after setup installs. Related #95765. Thanks @snowzlmbot. - **PR #93943** fix(provider-usage): honor proxy env for usage fetch. Related #78714. Thanks @TurboTheTurtle and @TnzGit. - **PR #89648** fix(agents): restore model-fetch info logs. Related #89300. Thanks @xiaobao-k8s and @Enominera. - **PR #93841** fix(ui): render persisted history text blocks. Related #90241. Thanks @mushuiyu886 and @pronzcw. - **PR #93502** docs: fix docs metadata spellcheck. Thanks @harjothkhara. - **PR #86608** docs: add existing-solutions preflight guardrail. Thanks @cablackmon. - **PR #95322** fix(reply): preserve usage footer across rollover. Thanks @litang9. - **PR #91559** fix(agents): clean Gemini tool schemas by model id. Related #91542. Thanks @Pick-cat and @qiukui666. - **PR #94701** Fix embedded-run recovery after refreshed session activity. Thanks @imadal1n. - **PR #95472** [AI] fix(main-session): skip current-gen abort controllers for completed sessions. Thanks @xydt-tanshanshan. - **PR #95401** fix(lmstudio): canonicalize variant model keys. Thanks @MonkeyLeeT. - **PR #94323** fix(cron): stop add/remove from dropping a due recurring job's pending run. Thanks @yetval. - **PR #94443** fix(memory-wiki): retry transient source-page rewrite race instead of aborting wiki_status. Related #92134. Thanks @ZengWen-DT and @cknzraposo. - **PR #95120** feat(channels): add directUserId support for per-DM model override. Related #53638. Thanks @xydigit-zt and @thomaszta and @gandalf-at-lerian. - **PR #91049** fix: handle terminal chat send acknowledgements. Related #91048. Thanks @nxmxbbd. - **PR #95081** fix: keep text transform runtime imports hashed. Related #95057. Thanks @849261680 and @YvesLaRose. - **PR #93356** fix(plugins): cache plugin setup registry to kill the /models CPU storm. Thanks @obuchowski. - **PR #85341** refactor: internalize OpenClaw agent runtime. - **PR #94233** fix(model-fallback): coalesce auth decision logs. Related #56979. Thanks @goutamadwant and @yanan1991. - **PR #92399** fix(llm): collapse cumulative openai-responses message snapshots instead of concatenating [AI-assisted]. Related #91959. Thanks @amersheeny and @phoenixyy. - **PR #94811** fix(ollama): honor memory embedding output dimensionality. Thanks @mushuiyu886. - **PR #84708** fix(agents): recover message-tool mirror replay poison. Thanks @anyech. - **PR #95274** fix(memory): preserve Windows QMD command paths. Related #92302. Thanks @ly85206559 and @Ardooken. - **PR #95686** fix(xai): request encrypted reasoning include for all reasoning models. Thanks @geraint0923 and @fuller-stack-dev. - **PR #82909** fix(telegram): repair message cache reply types. Thanks @lidge-jun. - **PR #94942** fix(matrix): prune finished fake-indexeddb transactions to prevent OOM. Related #90455. Thanks @xzh-icenter and @yar-sh. - **PR #93994** fix(setup): point non-interactive health hints at onboard flags. Related #93947. Thanks @zhouhe-xydt and @NianJiuZst. - **PR #93956** fix(ollama): skip auto-discovery for remote/cloud base URLs. Thanks @jason-allen-oneal. - **PR #93547** fix(acp): recover stale persistent sessions by structured resume-required code [AI-assisted]. Related #87830. Thanks @amersheeny and @chouzz. - **PR #94204** fix(control-ui): rewrite manifest hrefs for configured base path. Related #94157. Thanks @hugenshen and @xrow. - **PR #95637** fix(qqbot): recognize GFM table separators with one or two dashes. Thanks @ly-wang19. - **PR #95467** fix(sessions): keep bound channel identity across non-delivery turns. Thanks @yetval. - **PR #93919** perf(plugins): cache existence probes within bundle manifest scan [AI-assisted]. Related #76209. Thanks @ml12580 and @shenhonglong456-ai. - **PR #94726** fix(google): add gemini-3.5-flash model catalog entry. Related #94723. Thanks @ajwan8998 and @anguslogan01. - **PR #94238** fix(config): fail closed when configure runs without an interactive TTY (#93953). Thanks @ruomuxydt and @NianJiuZst. - **PR #89949** fix(media): pin requester delivery route when task starts. Related #86034. Thanks @wangwllu and @tianxiaochannel-oss88. - **PR #94470** chore: sync yuanbao plugin catalog to 2.15.0. Thanks @jase-283. - **PR #93585** fix(daemon): keep systemd gateway running after child OOM. Thanks @snowzlm. - **PR #94442** fix(imessage): strip leading echo corruption markers in the persisted echo cache. Thanks @ly-wang19. - **PR #93511** fix(imessage): normalize leading NUL echo-cache prefixes. Thanks @vincentkoc and @ly-wang19. - **PR #87121** test(cli): add banner emission reset helper. Related #83903. Thanks @lizuju and @davinci282828. - **PR #89172** fix(feishu): show voice message duration via upload duration. Related #53798. Thanks @areslp and @kinrocW. - **PR #95118** fix: bound agents load their own inbound workspace context. Related #92903. Thanks @849261680 and @axjing. - **PR #94752** fix(reply): clarify stale model override resets. Related #94713. Thanks @snowzlmbot and @Gr4via. - **PR #89716** fix(providers): strip cache-boundary marker from non-Anthropic prompts. Thanks @masatohoshino. - **PR #95268** fix(openrouter): expand short canonical model IDs to upstream API slugs (fixes #95198). Thanks @Darren2030 and @daniel-alejandro-t. - **PR #90475** fix(telegram): keep bot reply answers anchored to current message. Thanks @moeedahmed. - **PR #94219** fix(control-ui): restore provider usage quota pill in sidebar session switcher (fixes #93041). Thanks @Pick-cat and @jazzroutine. - **PR #93369** fix(cron): expose per-job fallbacks in CLI. Related #90302. Thanks @849261680 and @Walliiee. - **PR #87912** Handle Codex toolResult blocks in tool-result truncation. Thanks @AdrianIp0204. - **PR #95744** fix(model-catalog): strip manifest model-id prefixes by the matched length. Related #95743. Thanks @parveshsaini. - **PR #95300** fix(cli): expose --count on infer image edit, matching image generate. Thanks @ly-wang19. - **PR #94156** fix: expose OpenAI image quality and moderation CLI options. Thanks @lastguru-net and @ly-wang19. - **PR #94062** fix(agents): classify generic "LLM request failed." as transient time…. Related #93931. Thanks @hugenshen and @hyphae-bot. - **PR #93965** fix(opencode-go): streaming completes when provider ends responses. Related #93610. Thanks @zhangguiping-xydt and @ForceConstant. - **PR #89017** fix(webchat): sessions persist after reconnects. Related #87700. Thanks @zhangguiping-xydt and @asicoe. - **PR #94627** fix(ios): centralize app accent colors. Thanks @zats. - **PR #94054** fix(gateway.tls): reject empty/whitespace certPath and keyPath. Thanks @miorbnli. - **PR #95674** fix(cron): trim trailing whitespace from recognized job object keys. Related #95407. Thanks @zw-xysk and @Nassiel. - **PR #95857** Fix SQLite user version guardrail allowlist. Thanks @RomneyDa. - **PR #95858** feat(qa): expose active memory toggles to scenarios. Thanks @RomneyDa. - **PR #84326** Doctor: expose sandbox registry findings. Thanks @giodl73-repo. - **PR #73079** fix(minimax): request hex TTS output explicitly. Thanks @efe-arv. - **PR #94483** feat(gateway-cli): scope usage-cost by agent. Thanks @ly-wang19. - **PR #91502** feat(qa): crabline channel driver. Thanks @RomneyDa. - **PR #95872** Move TUI PTY tests into CI node shard. Thanks @RomneyDa. - **PR #89800** fix(agents): resolve webchat current session status. Related #89773. Thanks @sweetcornna and @killo3967. - **PR #95870** fix(ci): restore macOS and Windows QA gates. Thanks @vincentkoc. - **PR #94445** fix(agents): keep cron cloud idle watchdog enabled. Thanks @bek91. - **PR #91506** fix(qa): reserve shared QA suite flags across runners. Thanks @RomneyDa. - **PR #95876** ci: add manual release QA profile evidence workflow. Thanks @RomneyDa. - **PR #95879** fix(ci): use workflow revision for proof checks. Thanks @vincentkoc. - **PR #93113** fix(memory-core): report active dreaming phases in status. Related #67868. Thanks @AgentArcLab and @mrossit. - **PR #95837** Simplify color mode button labels. Thanks @SannidhyaSah and @hannesrudolph. - **PR #95797** fix: /status is too verbose for pinned model sessions. Thanks @Solvely-Colin. - **PR #90223** test: make qqbot symlinked media helper test robust on Windows. Thanks @aniruddhaadak80. - **PR #95557** improve: refresh Android overview control surface. Thanks @Solvely-Colin and @joshavant. - **PR #95593** fix: route Android exec approvals to in-app inbox. Thanks @Solvely-Colin. - **PR #95570** fix(cli): resolve trajectory export stores consistently. Related #95568. Thanks @youngting520. - **PR #95880** ci: generalize QA profile evidence workflow. Thanks @RomneyDa. - **PR #95218** fix(agents): bound provider JSON response reads. Thanks @Alix-007. - **PR #95614** fix(memory-wiki): preserve human notes block on source re-ingest. Thanks @yetval. - **PR #95148** feat(android): add settings detail panels. Thanks @Tosko4. - **PR #95893** Add iOS push sandbox profiles and relay tooling. Thanks @joshavant. - **PR #95723** fix(control-ui): exclude disabled cron jobs from Overview failed count. Related #95716. Thanks @ZengWen-DT and @voytas75. - **PR #78715** Fix minor grammar issue in plugin documentation (capabilities plural). Thanks @mehrazmorshed. - **PR #95890** fix(ci): restore QA workflow gates. Thanks @vincentkoc. - **PR #95661** fix(discord): reserve closing-fence space on fence-closing lines. Thanks @ly-wang19. - **PR #94272** ci: add maturity scorecard renderer. Thanks @RomneyDa. - **PR #95901** ci: add maturity scorecard renderer. Thanks @RomneyDa. - **PR #95914** fix(whatsapp): preserve durable reply target. Thanks @mcaxtr. - **PR #95697** improve: reduce hot-path linear scans and redundant I/O. Thanks @vincentkoc. - **PR #95916** fix(memory): improve node:sqlite unavailable guidance. Thanks @vincentkoc and @rrrrrredy. - **PR #78884** docs: document local avatar file size limit. Related #65312. Thanks @wangjieweb3-design and @wraxle-geargrind. - **PR #95909** fix(ci): use available Android SDK platform. Thanks @vincentkoc. - **PR #68389** plugins: clarify allowlist warning when entries don't match discovered ids. Related #68352. Thanks @lyfuci and @JIRBOY. - **PR #95898** ci: simplify maturity scorecard QA evidence inputs. Thanks @RomneyDa. - **PR #95928** fix(ci): honor reusable QA evidence failure policy. Thanks @vincentkoc. - **PR #95508** fix #95489: [Bug]: claude-cli out-of-credits error bypasses model fallback chain — error text delivered as final response. Thanks @mikasa0818 and @sallyom and @riazrahaman. - **PR #95420** fix(agents): bound OpenRouter model catalog response reads. Thanks @Alix-007 and @sallyom. - **PR #95103** fix(gateway): bound pricing catalog streams. Thanks @vincentkoc and @sallyom. - **PR #95108** fix(agents): bound Anthropic error streams. Thanks @vincentkoc and @sallyom. - **PR #95418** fix(agents): bound OpenRouter model-scan catalog success body. Thanks @Alix-007. - **PR #95922** fix(ci): finalize testbox sessions after setup failures. Thanks @vincentkoc. - **PR #95417** fix(agents): bound Google prompt cache response reads. Thanks @Alix-007. - **PR #95935** fix(whatsapp): resolve approval reactions across JID drift. Thanks @mcaxtr. - **PR #84340** Doctor: expose extra gateway service findings. Thanks @giodl73-repo. - **PR #95946** fix(ci): finalize Windows Testbox after setup failures. Thanks @vincentkoc. - **PR #95947** feat(qa): add "all" taxonomy profile. Thanks @RomneyDa. - **PR #91903** feat(plugin-sdk): add extensible channel identity hook context. Thanks @lanzhi-lee and @vincentkoc. - **PR #95930** docs(copilot): refresh harness parity notes. Thanks @vincentkoc. - **PR #95952** fix(ci): require QA live evidence artifacts. Thanks @vincentkoc. - **PR #95666** Fix memory-wiki bridge self-import loop. Related #95657. Thanks @TurboTheTurtle and @vincentkoc and @Johannes0402. - **PR #91724** fix(agents): infer runtime provider from qualified model ids. Thanks @yu-xin-c and @vincentkoc. - **PR #95957** improve: speed up provider tool-call streaming. Thanks @vincentkoc. - **PR #95541** fix(cli): show working commands for pinned plugin drift. Thanks @ooiuuii and @vincentkoc. - **PR #95933** docs: update maturity scorecard. Thanks @RomneyDa. - **PR #78105** fix(plugins): make empty-allowlist actionable for new users. Related #68780. Thanks @pahuchi-joe and @JIRBOY. - **PR #95961** fix(ci): require live proof evidence artifacts. Thanks @vincentkoc. - **PR #95400** fix(model-fallback): classify Codex usage-limit payloads. Thanks @jason-allen-oneal and @sallyom. - **PR #95944** perf(qa-lab): speed up unified QA suites. Thanks @RomneyDa. - **PR #95971** ci: fail QA profile evidence on QA failures. Thanks @RomneyDa. - **PR #95845** feat: add bundled plugin icon manifest URLs. Thanks @Patrick-Erichsen. - **PR #95975** fix(ci): require iOS Periphery evidence artifact. Thanks @vincentkoc. - **PR #95972** docs: update ClawHub skill route references. Thanks @Patrick-Erichsen. - **PR #90537** Warn on generated wrapper overwrites and status diagnostics. Related #90518. Thanks @TurboTheTurtle and @vincentkoc. - **PR #93567** fix(cron): normalize run-log jobId on write to match read-side validation. Thanks @Alix-007 and @vincentkoc. - **PR #95967** perf(ci): widen main test fanout and move codeql off blacksmith. Thanks @vincentkoc. - **PR #95983** fix(ci): require OpenGrep SARIF artifacts. Thanks @vincentkoc. - **PR #64490** CLI: escape zsh completion descriptions. Thanks @EdenKangdw. - **PR #95682** Fix Gemini day freshness time range handling. Thanks @Sunjae-k and @vincentkoc. - **PR #93374** fix(agents): suggest recovery for unknown tool ids. Related #92273. Thanks @mushuiyu886 and @vincentkoc and @poison. - **PR #95991** fix(ci): require release QA evidence artifacts. Thanks @vincentkoc. - **PR #89886** fix(context-engine): forward abortSignal through delegation bridge to runtime compaction. Related #89868. Thanks @openperf and @vincentkoc and @lykeion-dev. - **PR #95999** fix(release): require postpublish evidence artifact. Thanks @vincentkoc. - **PR #96003** test(qa): harden all-profile QA evidence scenarios. Thanks @RomneyDa. - **PR #95860** fix(compaction): trim prefix when transcript ends in an oversized tool result. Related #78478. Thanks @yetval and @wzhgba. - **PR #95934** fix(xiaomi): correct mimo-v2.5 and mimo-v2.5-pro max output tokens to 128K. Thanks @idootop. - **PR #95782** fix(sessions): honor configured store for outbound transcript mirrors. Related #95781. Thanks @youngting520. - **PR #96005** feat(copilot): wire harness parity helpers. Thanks @vincentkoc. - **PR #95886** fix: avoid false macOS update failures during gateway shutdown. Thanks @fuller-stack-dev. - **PR #91906** perf: skip subagent live stream parsing. Thanks @lanzhi-lee. - **PR #87861** fix(model-usage): coerce numeric-string costs and ignore non-finite values. Related #37878. Thanks @coder999999999 and @vincentkoc and @shuofengzhang. - **PR #96030** fix(qa-lab): avoid duplicate child evidence files. Thanks @RomneyDa. - **PR #94369** fix(memory-wiki): exclude durable reference pages from stale report. Thanks @SunnyShu0925 and @vincentkoc. - **PR #94578** Fix recent session resume with long headers. Related #94577. Thanks @rohitjavvadi and @vincentkoc. - **PR #95919** ci: add Codex maturity scorecard agent. Thanks @RomneyDa. - **PR #92356** fix(heartbeat): skip reasoning payloads when selecting heartbeat reply. Related #92260. Thanks @tangtaizong666 and @vincentkoc and @jmpei. - **PR #94328** fix(agents): keep post-compaction user re-issue of a kept-tail prompt during compaction rotation. Thanks @yetval and @vincentkoc. - **PR #95432** fix(reply): suppress per-message finals across multi-message block streaming. Thanks @yetval and @vincentkoc. - **PR #95431** fix(auto-reply): keep drain/restart-abort reply paths silent. Thanks @moeedahmed and @vincentkoc. - **PR #96017** test(qa): gate maturity docs on passing evidence. Thanks @RomneyDa. - **PR #95552** feat(mattermost): persist participated threads for mention-free follow-ups. Thanks @amknight. - **PR #95480** fix #89231: [Bug]: Windows installer-created scheduled task launches gateway.cmd with visible console — should use windowless launcher. Thanks @mikasa0818 and @vincentkoc and @CameronWeller. - **PR #96049** fix(copilot): preserve compaction metadata. Thanks @vincentkoc. - **PR #96044** docs: rename top maturity tier to Clawesome. Thanks @RomneyDa. - **PR #92154** Gate private QQBot group commands. Thanks @sliverp. - **PR #95484** fix: assistant reply lost between compaction summary and first kept user in successor transcript. Related #76729. Thanks @maweibin and @vincentkoc and @njuboy11. - **PR #95094** ci: add release QA profile evidence. Thanks @RomneyDa. - **PR #96057** docs: redesign maturity scorecard pages. Thanks @vincentkoc. - **PR #96014** perf(agents): index displaced tool results. Thanks @vincentkoc. - **PR #96019** perf(usage): bound session log retention. Thanks @vincentkoc. - **PR #96013** perf(anthropic): index active stream blocks. Thanks @vincentkoc. - **PR #96061** docs: place maturity pages under release reference. Thanks @vincentkoc. - **PR #95589** fix: npm plugin updates break running gateway imports. Thanks @ooiuuii and @vincentkoc. - **PR #96068** fix(acpx): consume acpx 0.11.1 model capability errors. Related #95869. Thanks @vincentkoc and @SabaTech-dev. - **PR #89912** refactor: add transcript update identity contract. Thanks @jalehman. - **PR #96055** fix(maint): use rebase PR landing. Thanks @vincentkoc. - **PR #96062** feat(copilot): mirror native plan and subagent events. Thanks @vincentkoc. - **PR #96032** fix(acpx): detect wrapper orphan on any PPID change, not just init reparenting. Thanks @t2wei and @vincentkoc. - **PR #96124** chore(acpx): bump bundled client to 0.11.2. Thanks @vincentkoc. - **PR #84352** Fix WebChat dispatch failure session status. Thanks @jesse-merhi. - **PR #89518** refactor: migrate plugin transcript mirrors. Thanks @jalehman. - **PR #90439** refactor: add embedded run session target seam. Thanks @jalehman. - **PR #95699** perf(gateway): drop redundant per-access session-key case scan. Thanks @jzakirov and @jalehman. - **PR #95992** fix(skills): accept owner-qualified verify refs. Thanks @Patrick-Erichsen. - **PR #95987** fix(plugins): remove Simple Icons color paths. Thanks @Patrick-Erichsen. - **PR #96162** refactor: use accessor-backed transcript corpus for memory. Thanks @jalehman. - **PR #96181** Defer iOS local network permission until onboarding. Thanks @joshavant. - **PR #95475** test(cli): isolate service env in run and update suites. Thanks @kklouzal. - **PR #95226** fix(infra): bound ClawHub fetchJson and error response bodies. Thanks @Alix-007. - **PR #95240** fix(matrix): bound non-raw JSON response body in transport. Thanks @Alix-007. - **PR #91742** fix(memory): abort orphaned embedding work when memory_search times out. Related #91718. Thanks @dreamhunter2333 and @NOVA-Openclaw. - **PR #93394** fix(memory): abort orphaned qmd search subprocess when memory_search times out. Thanks @Alix-007. - **PR #89911** refactor: migrate bundled transcript target lookups. Thanks @jalehman. - **PR #96191** refactor: route plugin host hook state through accessor. Thanks @jalehman. - **PR #96179** fix: route gateway history through session accessor target. Thanks @jalehman. - **PR #96201** refactor: add abort target session accessor. Thanks @jalehman. - **PR #96193** fix(memory-core): migrate dreaming cleanup lifecycle. Thanks @jalehman. - **PR #96195** fix: bridge ACP metadata to session accessors. Thanks @jalehman. - **PR #96182** refactor: migrate agent session accessors. Thanks @jalehman. - **PR #96218** refactor: guard reply session initialization. Thanks @jalehman. - **PR #96213** refactor(gateway): add alias mutation accessor. Thanks @jalehman. - **PR #96204** refactor: migrate command session persistence to accessor. Thanks @jalehman. - **PR #96206** refactor: route live model fresh reads through session accessor. Thanks @jalehman. - **PR #96220** fix(whatsapp): quote current follow-up in durable replies. Thanks @mcaxtr. - **PR #96226** fix(macos): drop Textual from chat packaging. Thanks @vincentkoc. - **PR #96212** fix(codex): deliver generated images from remote app-server. Thanks @sjf-oa. - **PR #96235** fix(crabbox): require Xcode for macOS proof. Thanks @vincentkoc. - **PR #96145** fix: UI glitch: config is not visible. Related #94202. Thanks @sunlit-deng and @vporton. - **PR #96072** perf(browser): index role snapshot references. Thanks @vincentkoc. - **PR #96085** perf(codex): index rollout transcript ids. Thanks @vincentkoc. - **PR #96087** perf(reply): hoist direct-send fragment index. Thanks @vincentkoc. - **PR #94154** fix(gateway): resolve plugin-registered gateway methods through live registry. Related #94127. Thanks @Pick-cat and @vincentkoc and @BryceMurray. - **PR #95393** fix #92582: Bug: doctor falsely warns local memory embeddings are not ready. Thanks @mikasa0818 and @vincentkoc and @neekolascmd. - **PR #96246** fix(qa): accept pnpm separator for lab up. Thanks @vincentkoc. - **PR #94949** fix(ports): route isPortBusy through checkPortInUse to catch IPv4-only occupants. Related #94426. Thanks @sunlit-deng and @vincentkoc and @wangwllu. - **PR #94562** fix(workboard): hide archived cards in CLI list by default. Related #94555. Thanks @ZengWen-DT and @vincentkoc and @ecican. - **PR #96243** fix(nextcloud-talk): ignore signed non-message webhook events. Related #81566. Thanks @arkyu2077 and @vincentkoc and @rafaelmgbh. - **PR #96140** fix(exec): preserve turn-source routing target in approval followups for plugin channels. Related #96103. Thanks @yetval and @vincentkoc and @lansenger-pm. - **PR #96258** ci: move CodeQL quality scans to hosted runners. Thanks @vincentkoc. - **PR #96271** chore(release): close out 2026.6.10 on main. Thanks @vincentkoc. - **PR #96233** fix(agents): run heartbeat_prompt_contribution on harness prompt builds. Thanks @azogheb and @vincentkoc. - **PR #55018** fix: avoid O(N²) shallow-copy in mapSensitivePaths schema traversal. Thanks @xdhuangyandi and @vincentkoc and @huangyandi-red. - **PR #95831** fix: compact Codex OAuth OpenAI sessions without API keys. Related #95693. Thanks @sallyom and @YUI-TIEN. - **PR #96244** fix(auto-reply): align channel intro wording with chat_type. Related #95645. Thanks @arkyu2077 and @vincentkoc and @iloveleon19. ### Release verification - npm package: https://www.npmjs.com/package/openclaw/v/2026.6.11-beta.1 - release SHA: `c862a644bff366599ead53df81f452aa913497ae` - full release validation: https://github.com/openclaw/openclaw/actions/runs/28123941058 - npm preflight: https://github.com/openclaw/openclaw/actions/runs/28130211221 - npm publish: https://github.com/openclaw/openclaw/actions/runs/28136271139 - ClawHub follow-up: four existing code-plugin family conflicts and the pre-claimed `tencent` id remain in the child publish runs; OpenClaw npm and GitHub beta publication are unaffected.

## What's Changed * fix(core): strip thoughts from scrubbed history turns and resolve thought leakage by @amelidev in https://github.com/google-gemini/gemini-cli/pull/27971 * Refactor: exclude transient CI configuration files from workspace context by @DavidAPierce in https://github.com/google-gemini/gemini-cli/pull/28216 * feat(caretaker-triage): add triage worker core foundational modules by @chadd28 in https://github.com/google-gemini/gemini-cli/pull/28163 * feat(caretaker-egress): implement octokit github action handler for egress service by @chadd28 in https://github.com/google-gemini/gemini-cli/pull/28303 * chore(release): bump version to 0.52.0-nightly.20260707.g27a3da3e8 by @gemini-cli-robot in https://github.com/google-gemini/gemini-cli/pull/28323 * Changelog for v0.51.0-preview.0 by @gemini-cli-robot in https://github.com/google-gemini/gemini-cli/pull/28320 * Changelog for v0.50.0 by @gemini-cli-robot in https://github.com/google-gemini/gemini-cli/pull/28322 * fix(core-tools): bypass LLM correction for JSON and IPYNB files in write_file and replace by @amelidev in https://github.com/google-gemini/gemini-cli/pull/28223 * fix(core): use unambiguous previous intent label in fallback summary by @amelidev in https://github.com/google-gemini/gemini-cli/pull/28343 * feat(caretaker-triage): implement main worker execution loop and egress action publisher by @chadd28 in https://github.com/google-gemini/gemini-cli/pull/28306 **Full Changelog**: https://github.com/google-gemini/gemini-cli/compare/v0.51.0-nightly.20260707.g15a9429b6...v0.52.0-nightly.20260710.ga4c91ce19

## 变更内容 * 修复/验证：发布时 npm ci 忽略脚本 由 @rmedranollamas 提交于 https://github.com/google-gemini/gemini-cli/pull/28116 * 修复(CI)：防止发布验证中的工作区二进制文件遮蔽问题 由 @galdawave 提交于 https://github.com/google-gemini/gemini-cli/pull/28132 * 功能：工具注册表发现 由 @ved015 提交于 https://github.com/google-gemini/gemini-cli/pull/28113 * 修复(CI)：防止错误的 NPM 发布并避免任务崩溃 由 @galdawave 提交于 https://github.com/google-gemini/gemini-cli/pull/28147 **完整变更日志**：https://github.com/google-gemini/gemini-cli/compare/v0.49.0...v0.50.0

## What's Changed * fix/verify release npm ci ignore scripts by @rmedranollamas in https://github.com/google-gemini/gemini-cli/pull/28116 * fix(ci): prevent workspace binary shadowing in release verification by @galdawave in https://github.com/google-gemini/gemini-cli/pull/28132 * Feat/tool registry discovery by @ved015 in https://github.com/google-gemini/gemini-cli/pull/28113 * fix(ci): prevent bad NPM releases and promote job crashes by @galdawave in https://github.com/google-gemini/gemini-cli/pull/28147 **Full Changelog**: https://github.com/google-gemini/gemini-cli/compare/v0.49.0...v0.50.0

## 变更内容 * 为 v0.50.0-preview.1 添加变更日志，由 @gemini-cli-robot 在 https://github.com/google-gemini/gemini-cli/pull/28150 提交 * 修复 no_proxy 测试，由 @jerrylin3321 在 https://github.com/google-gemini/gemini-cli/pull/28131 提交 * chore(release): 将版本提升至 0.51.0-nightly.20260625.g3fbf93e26，由 @gemini-cli-robot 在 https://github.com/google-gemini/gemini-cli/pull/28151 提交 * 更新 Vertex 基础 URL，由 @DavidAPierce 在 https://github.com/google-gemini/gemini-cli/pull/28145 提交 * fix(security): 强制敏感路径黑名单和 vscode hitl 不区分大小写，由 @luisfelipe-alt 在 https://github.com/google-gemini/gemini-cli/pull/27966 提交 * fix(core-tools): 解决 at-参考文件的防御性路径解析并修复 macOS 测试，由 @luisfelipe-alt 在 https://github.com/google-gemini/gemini-cli/pull/28053 提交 * feat(caretaker): 实现 Cloud Run Webhook 接收服务，由 @chadd28 在 https://github.com/google-gemini/gemini-cli/pull/28015 提交 * fix(core): 解决内存导入处理器中的符号链接目录逃逸问题，由 @luisfelipe-alt 在 https://github.com/google-gemini/gemini-cli/pull/28233 提交 * feat(caretaker): 出口 Cloud Run 服务骨架，由 @chadd28 在 https://github.com/google-gemini/gemini-cli/pull/28167 提交 * fix(sandbox): 在 macOS 沙盒中使 ~/.gitconfig 只读，由 @ompatel-aiml 在 https://github.com/google-gemini/gemini-cli/pull/28221 提交 * fix(core): 为现代模型保留字符串字面量中的转义序列，由 @luisfelipe-alt 在 https://github.com/google-gemini/gemini-cli/pull/28299 提交 * fix(core): 从清洗过的历史对话中剥离思考内容并解决思考泄露问题，由 @amelidev 在 https://github.com/google-gemini/gemini-cli/pull/27971 提交 ## 新贡献者 * @jerrylin3321 在 https://github.com/google-gemini/gemini-cli/pull/28131 提交了首次贡献 **完整变更日志**: https://github.com/google-gemini/gemini-cli/compare/v0.50.0-preview.1...v0.51.0-preview.0

## What's Changed * Changelog for v0.50.0-preview.1 by @gemini-cli-robot in https://github.com/google-gemini/gemini-cli/pull/28150 * Fix no_proxy test by @jerrylin3321 in https://github.com/google-gemini/gemini-cli/pull/28131 * chore(release): bump version to 0.51.0-nightly.20260625.g3fbf93e26 by @gemini-cli-robot in https://github.com/google-gemini/gemini-cli/pull/28151 * Vertex base url update by @DavidAPierce in https://github.com/google-gemini/gemini-cli/pull/28145 * fix(security): enforce case-insensitive sensitive path blocklist and vscode hitl by @luisfelipe-alt in https://github.com/google-gemini/gemini-cli/pull/27966 * fix(core-tools): resolve defensive path resolution for at-reference files and fix macOS tests by @luisfelipe-alt in https://github.com/google-gemini/gemini-cli/pull/28053 * feat(caretaker): implement Cloud Run webhook ingestion service by @chadd28 in https://github.com/google-gemini/gemini-cli/pull/28015 * fix(core): resolve symbolic link directory escape in memory import processor by @luisfelipe-alt in https://github.com/google-gemini/gemini-cli/pull/28233 * feat(caretaker): egress cloud run service skeleton by @chadd28 in https://github.com/google-gemini/gemini-cli/pull/28167 * fix(sandbox): make ~/.gitconfig read-only in the macOS sandbox by @ompatel-aiml in https://github.com/google-gemini/gemini-cli/pull/28221 * fix(core): preserve escape sequences in string literals for modern models by @luisfelipe-alt in https://github.com/google-gemini/gemini-cli/pull/28299 * fix(core): strip thoughts from scrubbed history turns and resolve thought leakage by @amelidev in https://github.com/google-gemini/gemini-cli/pull/27971 ## New Contributors * @jerrylin3321 made their first contribution in https://github.com/google-gemini/gemini-cli/pull/28131 **Full Changelog**: https://github.com/google-gemini/gemini-cli/compare/v0.50.0-preview.1...v0.51.0-preview.0

## 变更内容 * 修复(沙盒)：在 macOS 沙盒中将 ~/.gitconfig 设为只读，作者：@ompatel-aiml 于 https://github.com/google-gemini/gemini-cli/pull/28221 * 修复(核心)：为现代模型保留字符串字面量中的转义序列，作者：@luisfelipe-alt 于 https://github.com/google-gemini/gemini-cli/pull/28299 **完整变更日志**：https://github.com/google-gemini/gemini-cli/compare/v0.51.0-nightly.20260706.gf7af4e518...v0.51.0-nightly.20260707.g15a9429b6

## What's Changed * fix(sandbox): make ~/.gitconfig read-only in the macOS sandbox by @ompatel-aiml in https://github.com/google-gemini/gemini-cli/pull/28221 * fix(core): preserve escape sequences in string literals for modern models by @luisfelipe-alt in https://github.com/google-gemini/gemini-cli/pull/28299 **Full Changelog**: https://github.com/google-gemini/gemini-cli/compare/v0.51.0-nightly.20260706.gf7af4e518...v0.51.0-nightly.20260707.g15a9429b6

**完整变更日志**: https://github.com/google-gemini/gemini-cli/compare/v0.51.0-nightly.20260705.gf7af4e518...v0.51.0-nightly.20260706.gf7af4e518

**Full Changelog**: https://github.com/google-gemini/gemini-cli/compare/v0.51.0-nightly.20260705.gf7af4e518...v0.51.0-nightly.20260706.gf7af4e518

Release 0.145.0-alpha.4

Release 0.145.0-alpha.3

Release 0.145.0-alpha.3

发布 0.145.0-alpha.2

Release 0.145.0-alpha.2

## Bug 修复 - 修复了当 GitHub 返回压缩或重新排序的发布元数据时独立安装失败的问题。(#31913) - 确保 macOS 安装包在 `codex` 可执行文件旁同时暴露代码模式主机。(#31913) - 通过回退到嵌入式运行时，在伴随主机二进制文件不可用时保持代码模式正常工作。(#31913) ## 变更日志 完整变更日志：https://github.com/openai/codex/compare/rust-v0.144.0...rust-v0.144.1 - #31913 [0.144] 回移植安装程序和代码模式可靠性修复 @bolinfest

## Bug Fixes - Fixed standalone installs failing when GitHub returns compact or reordered release metadata. (#31913) - Ensured macOS package installs expose the code-mode host alongside the `codex` executable. (#31913) - Kept code mode working when the companion host binary is unavailable by falling back to the embedded runtime. (#31913) ## Changelog Full Changelog: https://github.com/openai/codex/compare/rust-v0.144.0...rust-v0.144.1 - #31913 [0.144] Backport installer and code-mode reliability fixes @bolinfest

发布 0.145.0-alpha.1

Release 0.145.0-alpha.1