<!-- qwen-release-notes:v1 -->
## 亮点
_详见下方的完整变更列表。_
## 破坏性变更
- 移除了频道中可配置的消息前缀过滤功能，符合条件的消息现在将遵循正常的发送者、群组、提及和配对策略，不再依赖前缀。 ([#11571](https://github.com/QwenLM/qwen-code/pull/11571)) by @qqqys
- fix(core)!: 以秒为单位读取命令钩子超时时间 ([#11615](https://github.com/QwenLM/qwen-code/pull/11615)) by @qqqys
## 完整变更列表
### 新功能
- 获批准的 Web Shell 目标提案将在其所属回合结束后自动启动，无需再手动复制 `/goal set` 命令。 ([#11360](https://github.com/QwenLM/qwen-code/pull/11360)) by @qqqys
- feat(sdk): 添加对等端点，使 Qwen Code 外部的程序能够加入跨会话消息传递 ([#11560](https://github.com/QwenLM/qwen-code/pull/11560)) by @qqqys
- 通过 model.goalMaxTurns 和 model.goalMaxActiveMinutes 添加可选的目标限制，并在恢复时重新计算活跃目标时间。 ([#11457](https://github.com/QwenLM/qwen-code/pull/11457)) by @qqqys
- 当选择“每次运行新建会话”时，Web Shell 的计划任务现在可以路由到配置的模型和会话组。 ([#11396](https://github.com/QwenLM/qwen-code/pull/11396)) by @qqqys
- Web Shell 的浏览器通知现在会标识会话、显示提示词和回复摘要，点击时可打开目标会话。 ([#11447](https://github.com/QwenLM/qwen-code/pull/11447)) by @doudouOUC
- 提升 Web Shell 编辑器的紧凑性，使编辑器、对话和侧边栏中的上下文使用情况更加清晰。 ([#11509](https://github.com/QwenLM/qwen-code/pull/11509)) by @wenshao
- 为独立版 Qwen Live 体验添加视觉输入、主动协助和本地记忆功能。 ([#11369](https://github.com/QwenLM/qwen-code/pull/11369)) by @wangxiongts
- 对小型静态文档导航变更的自动审查现在使用范围更窄、结果受限的专注型审查器。 ([#11456](https://github.com/QwenLM/qwen-code/pull/11456)) by @wenshao
- feat(subagents): 添加 Codex 执行器和原生代理内置函数 ([#11474](https://github.com/QwenLM/qwen-code/pull/11474)) by @wenshao
- feat(core): 添加仅代码模式工具执行功能 ([#10607](https://github.com/QwenLM/qwen-code/pull/10607)) by @DragonnZhang
- 将工作流编写参考文档移至内置的 workflow-authoring 技能中，减少了默认工具描述的开销。 ([#11520](https://github.com/QwenLM/qwen-code/pull/11520)) by @qqqys
- 默认隐藏 Web Shell 源面板的“添加源”按钮；可通过 URL 参数 `addSource=1` 重新启用。 ([#11639](https://github.com/QwenLM/qwen-code/pull/11639)) by @callmeYe
- CLI 恢复、Web Shell 和会话预览中的频道回合现在将精确显示模型接收到的提示词。 ([#11646](https://github.com/QwenLM/qwen-code/pull/11646)) by @qqqys
- 新增支持桌面端和移动端宽度的 Web 预览面板，具备刷新、外部打开、保存 HTML 工件功能，并支持按会话记忆 URL 和宽度。 ([#11276](https://github.com/QwenLM/qwen-code/pull/11276)) by @wenshao
- 在 Web Shell 中新增最新用户消息的内联编辑与重发功能，生成新回复时保留已附加的图片和文件。 ([#11650](https://github.com/QwenLM/qwen-code/pull/11650)) by @ytahdn
- feat(core): 向每个钩子输入添加 permission_mode、agent_id 和 prompt_id ([#11618](https://github.com/QwenLM/qwen-code/pull/11618)) by @qqqys
- 在 Web Shell 中为可恢复的中断会话添加“继续执行”操作，允许用户在不重复消息的情况下恢复原始对话。 ([#11545](https://github.com/QwenLM/qwen-code/pull/11545)) by @wenshao
- feat(skills): 扩展技能以其扩展名命名 ([#10841](https://github.com/QwenLM/qwen-code/pull/10841)) by @nerdalytics
- feat(vscode): 添加显式的会话源切换开关 ([#11584](https://github.com/QwenLM/qwen-code/pull/11584)) by @yiliang114
- feat(core): 在无法使用 Ctrl+Y 时自动重试瞬态网络错误（EOF） ([#10347](https://github.com/QwenLM/qwen-code/pull/10347)) by @qwen-code-dev-bot
- feat(workflows): 向 agent() 选项添加 effort 和 disallowedTools ([#11691](https://github.com/QwenLM/qwen-code/pull/11691)) by @qqqys
- feat(goal): 分批重新运行停滞的检查点，每次停滞后批次规模缩小 ([#11690](https://github.com/QwenLM/qwen-code/pull/11690)) by @qqqys
- feat(web-shell): 添加模型角色和上下文窗口配置 ([#11342](https://github.com/QwenLM/qwen-code/pull/11342)) by @wenshao
- feat(cua): 添加应用绑定操作和紧凑的原生观测 ([#11705](https://github.com/QwenLM/qwen-code/pull/11705)) by @LaZzyMan
- feat: 代理看板 — 跨独立启动的代理共享工作 ([#9402](https://github.com/QwenLM/qwen-code/pull/9402)) by @yiliang114
- feat(review): 在计划中记录默认截止日期，确保每次运行都有边界限制 ([#11686](https://github.com/QwenLM/qwen-code/pull/11686)) by @wenshao
- feat(web-shell): 使会话工作流依赖关系可导航，并精简其界面元素 ([#10938](https://github.com/QwenLM/qwen-code/pull/10938)) by @yiliang114
- feat(web-shell): 改进上下文概览并添加手动压缩功能 ([#11700](https://github.com/QwenLM/qwen-code/pull/11700)) by @wenshao
- feat(core): 使 web_search 预算可配置，并限制提取器的回退机制 ([#11692](https://github.com/QwenLM/qwen-code/pull/11692)) by @qqqys
- feat(web-shell): 在侧边栏悬停详情中恢复工作区 Git 选择器 ([#11818](https://github.com/QwenLM/qwen-code/pull/11818)) by @ytahdn
- feat(web-shell): 在编辑器上下文悬停提示中添加工具压缩功能 ([#11782](https://github.com/QwenLM/qwen-code/pull/11782)) by @wenshao
### 错误修复
- 通过放宽伪造的时间线事件容差（从 75 秒扩大到 300 秒），修复了 CI 回放循环测试的不稳定问题，且未更改工作流或产品行为。([#11581](https://github.com/QwenLM/qwen-code/pull/11581)) by @wenshao
- 修复了自动修复的门控逻辑，使回归问题计入刹车机制，并限制测试弱化行为，防止修复轮次免费引入新问题。([#10188](https://github.com/QwenLM/qwen-code/pull/10188)) by @wenshao
- 为 `UserPromptSubmit` 钩子保留原始提交的提示词，使自动回忆功能适用于 Web Shell 提交及已选择的 ACP 或守护进程客户端。([#11455](https://github.com/QwenLM/qwen-code/pull/11455)) by @doudouOUC
- 当 `qwen serve` 启动时未指定显式通道参数，恢复使用工作区配置的启动通道，同时显式通道选择仍优先。([#11472](https://github.com/QwenLM/qwen-code/pull/11472)) by @qqqys
- 修复 `qwen serve` 启动流程，确保在 ACP 预热后加载持久化的工作区范围 MCP 配置。([#11145](https://github.com/QwenLM/qwen-code/pull/11145)) by @kabishou11
- 修复 VS Code 差异处理逻辑，使差异视图可通过原始路径或工作区解析路径可靠关闭。([#10494](https://github.com/QwenLM/qwen-code/pull/10494)) by @CanReader
- 终端退出时立即释放 Web 终端 PTY 资源，而非等待后续释放。([#11572](https://github.com/QwenLM/qwen-code/pull/11572)) by @yiliang114
- 为 E2E 构建产物上传增加一次有界重试，以减少临时性 CI 失败。([#11375](https://github.com/QwenLM/qwen-code/pull/11375)) by @qwen-code-dev-bot
- 在显式关闭或终止代理会话时保留子错误详情，提升诊断日志的完整性。([#11190](https://github.com/QwenLM/qwen-code/pull/11190)) by @now-ing
- Web Shell 现可在页面重载时保留已取消的响应，并显示请求在取消前运行的时长。([#11608](https://github.com/QwenLM/qwen-code/pull/11608)) by @ytahdn
- 将集群更新器安装路径固定为 `/usr/local` 前缀，使运行器执行新安装版本，而非继续使用 PATH 中的旧版本。([#11593](https://github.com/QwenLM/qwen-code/pull/11593)) by @yiliang114
- fix(core): 从被拒绝的 Responses 加密推理中恢复 ([#11596](https://github.com/QwenLM/qwen-code/pull/11596)) by @DragonnZhang
- 通过移除 Markdown 代码块围栏，并在遇到某些格式错误或不一致的验证器答案时重试一次修正性模型调用，提升检查点验证能力。([#11578](https://github.com/QwenLM/qwen-code/pull/11578)) by @qqqys
- 仅在附件或工作区文件加载失败后，在源预览中显示“重试”操作，并在重试成功后清除该操作。([#11629](https://github.com/QwenLM/qwen-code/pull/11629)) by @callmeYe
- 将成功的纯文本钩子标准输出路由至 `additionalContext`，适用于 `SessionStart`、`UserPromptSubmit` 和 `UserPromptExpansion` 事件，其他事件仍保持系统消息行为。([#11612](https://github.com/QwenLM/qwen-code/pull/11612)) by @qqqys
- 修复 OpenTUI 渲染器在转录、编辑器、页脚、对话框、键盘及启动审批等界面的显示间隙，使其与 ink 保持一致。([#11630](https://github.com/QwenLM/qwen-code/pull/11630)) by @chiga0
- 重新附加的图像快照现标记为“可能过时”，以免将早期会话图像误认为当前 UI 状态。([#11602](https://github.com/QwenLM/qwen-code/pull/11602)) by @yiliang114
- 本地化处理了编辑已发送消息时 VS Code 回滚预检失败的提示，将原始传输错误替换为翻译后的友好消息。([#11628](https://github.com/QwenLM/qwen-code/pull/11628)) by @yiliang114
- fix(core): 保留 Responses 重试元数据及助手阶段 ([#11652](https://github.com/QwenLM/qwen-code/pull/11652)) by @DragonnZhang
- fix(cli): 确保展开的确认对话框保持在视口内 ([#11656](https://github.com/QwenLM/qwen-code/pull/11656)) by @qwen-code-dev-bot
- fix(core): 对已建立流中的临时服务器错误进行重试 ([#11634](https://github.com/QwenLM/qwen-code/pull/11634)) by @DragonnZhang
- fix(vscode): 在 Remote-SSH 下保留 IDE 工作区环境变量 (#11556) ([#11624](https://github.com/QwenLM/qwen-code/pull/11624)) by @yiliang114
- fix(vscode): 关闭权限差异视图时恢复编辑权限 ([#11171](https://github.com/QwenLM/qwen-code/pull/11171)) by @yiliang114
- 在目标证据检查点失败时持久化一行诊断信息，并在目标停止前显示，提升对检查点失败原因的可见性。([#11576](https://github.com/QwenLM/qwen-code/pull/11576)) by @qqqys
- 当提交因预检取消或中止被拒绝时，抑制 Web Shell 中不必要的错误提示，避免造成困惑的失败消息。([#11676](https://github.com/QwenLM/qwen-code/pull/11676)) by @yiliang114
- fix(core): 在 Responses 调试日志中脱敏请求体前缀 ([#11668](https://github.com/QwenLM/qwen-code/pull/11668)) by @yiliang114
- 在 Web Shell 中将受阻的目标卡片正确标记为“目标受阻”，而非更宽泛的“目标已中止”，提升状态清晰度。([#11174](https://github.com/QwenLM/qwen-code/pull/11174)) by @yu-xin-c
- 更新 CLI serve 测试，在结束前等待“即发即忘”处理器完成，在不改变生产行为的前提下提升测试可靠性。([#11417](https://github.com/QwenLM/qwen-code/pull/11417)) by @qwen-code-dev-bot
- 在构建请求时剥离或忽略非标准的 `thoughtSignature` 载荷，防止 Responses 推理回放数据发送至不兼容的提供商。([#11567](https://github.com/QwenLM/qwen-code/pull/11567)) by @yiliang114
- 通过保留解析状态并防止挂起的标签页在断开连接时清除已存储的授权，修复本地文件桥接的信任及旁观者问题。([#11169](https://github.com/QwenLM/qwen-code/pull/11169)) by @wenshao
- 将 DashScope 对话缓存断点移至重新附加的图片之前，使缓存前缀在轮次间不再发生偏移。 ([#11640](https://github.com/QwenLM/qwen-code/pull/11640)) by @yiliang114
- fix(core): 在 Stop 钩子上报告真实的 stop_hook_active 状态 ([#11613](https://github.com/QwenLM/qwen-code/pull/11613)) by @qqqys
- fix(core): 将网络搜索来源引用为纯 URL ([#11677](https://github.com/QwenLM/qwen-code/pull/11677)) by @qqqys
- ACP 子进程现在根据实际主机内存派生 V8 old-space 目标，而非盲目信任无限制的 cgroup 哨兵值。 ([#11653](https://github.com/QwenLM/qwen-code/pull/11653)) by @doudouOUC
- 防止过期的并发领导者分配将同一任务分派给多个队友。 ([#10237](https://github.com/QwenLM/qwen-code/pull/10237)) by @AaronZ345
- fix(telemetry): 基于 logPrompts 限制 request_text/response_text 的上报 ([#11670](https://github.com/QwenLM/qwen-code/pull/11670)) by @yiliang114
- fix(vscode): 统一会话历史并移除源切换功能 ([#11713](https://github.com/QwenLM/qwen-code/pull/11713)) by @yiliang114
- fix(acp): 在资源链接提示中保留原始文件名 ([#11721](https://github.com/QwenLM/qwen-code/pull/11721)) by @zhangxy-zju
- fix(vscode): 优雅关闭 ACP CLI 而非直接强制终止 ([#11642](https://github.com/QwenLM/qwen-code/pull/11642)) by @yiliang114
- fix(core): 在用量统计遥测数据接收端脱敏错误文本 ([#11649](https://github.com/QwenLM/qwen-code/pull/11649)) by @yiliang114
- fix(core): 在 Windows 上回收存活的钩子进程树 ([#11623](https://github.com/QwenLM/qwen-code/pull/11623)) by @yiliang114
- fix(core): 在夏令时回退期间使用纪元算术正确计算 cron 下次触发时间 ([#11723](https://github.com/QwenLM/qwen-code/pull/11723)) by @yiliang114
- fix(goal): 从评审回复中移除回溯栅栏解包逻辑 ([#11693](https://github.com/QwenLM/qwen-code/pull/11693)) by @yiliang114
- fix(core): 在 Fireworks 请求中移除镜像推理字段 ([#11701](https://github.com/QwenLM/qwen-code/pull/11701)) by @bluefateludi
- fix(core): 在委托边界和用户轮次触发活跃待办提醒 ([#10963](https://github.com/QwenLM/qwen-code/pull/10963)) by @yiliang114
- fix(core): 防护内部 git 调用免受配置的辅助程序干扰 ([#11669](https://github.com/QwenLM/qwen-code/pull/11669)) by @yiliang114
- fix(lsp): 在查询前同步目标文档 ([#11443](https://github.com/QwenLM/qwen-code/pull/11443)) by @shenyankm
- fix(core): 在捆绑的 ConPTY 后端上运行 Web 终端 PTY ([#11643](https://github.com/QwenLM/qwen-code/pull/11643)) by @yiliang114
- fix(core): 对无状态码的上游错误进行重试，而非直接结束轮次 ([#11291](https://github.com/QwenLM/qwen-code/pull/11291)) by @wenshao
- fix(dashscope): 仅对 qwen 系列模型发送请求元数据 ([#11606](https://github.com/QwenLM/qwen-code/pull/11606)) by @shoemoney
- fix(web-shell): 防止终端查询冻结并清理协议拒绝逻辑 ([#11748](https://github.com/QwenLM/qwen-code/pull/11748)) by @yiliang114
- fix(cli): 打破 Ink useBoxMetrics 提交阶段的 setState 循环 (React #185) ([#11565](https://github.com/QwenLM/qwen-code/pull/11565)) by @zaalipro
- fix(dingtalk): 处理已回复的富文本媒体 ([#11759](https://github.com/QwenLM/qwen-code/pull/11759)) by @qqqys
- fix(dingtalk): 保留回调文本 ([#11757](https://github.com/QwenLM/qwen-code/pull/11757)) by @qqqys
- fix(cli): 防止在短终端中对话框被截断 ([#9040](https://github.com/QwenLM/qwen-code/pull/9040)) by @shenyankm
- fix(web-shell): 在空闲时保留守护进程拒绝的轮次中间消息 ([#11289](https://github.com/QwenLM/qwen-code/pull/11289)) by @wenshao
- fix(core): 从 stdio MCP 子进程中剥离 AppImage 的 PYTHONHOME/PYTHONPATH ([#11763](https://github.com/QwenLM/qwen-code/pull/11763)) by @yiliang114
- fix(cua): 解决 OSWorld 运行中的工具失败与效率问题 ([#11683](https://github.com/QwenLM/qwen-code/pull/11683)) by @tanzhenxin
- fix(cua): 限制观测数量并减少模型往返调用 ([#11735](https://github.com/QwenLM/qwen-code/pull/11735)) by @DragonnZhang
- fix(web-shell): 在两种构建配置中固定 es2021 最低版本要求，并强化终端测试 ([#11758](https://github.com/QwenLM/qwen-code/pull/11758)) by @yiliang114
- fix(core): 由生产者自身的预算决定 Shell 输出大小 ([#11727](https://github.com/QwenLM/qwen-code/pull/11727)) by @doudouOUC
- fix(review): 记录截止时间的暂停上限并改进捕获本地错误 ([#11779](https://github.com/QwenLM/qwen-code/pull/11779)) by @wenshao
- fix(ci): 将构建堆内存上限从 3072 MB 提升至 4096 MB ([#11781](https://github.com/QwenLM/qwen-code/pull/11781)) by @doudouOUC
- fix(cli): 当进程因未捕获异常退出时，回收正在运行的监控器 ([#11742](https://github.com/QwenLM/qwen-code/pull/11742)) by @yiliang114
- fix(web-shell): 将源码下载移至头部 ([#11641](https://github.com/QwenLM/qwen-code/pull/11641)) by @callmeYe
- fix(skills): 恢复会话时重新应用 Skill 的副作用 ([#11280](https://github.com/QwenLM/qwen-code/pull/11280)) by @TianYuan1024
- fix(web-shell): 使驾驶舱视觉捕获结果确定化 ([#11786](https://github.com/QwenLM/qwen-code/pull/11786)) by @yiliang114
- fix(dingtalk): 移除后台代理的标题 ([#11800](https://github.com/QwenLM/qwen-code/pull/11800)) by @qqqys
- fix(cli): 将写入侧 PTY 的 EIO 错误视为无害的清理竞态条件 ([#11788](https://github.com/QwenLM/qwen-code/pull/11788)) by @yiliang114
- fix(goal): 防止完整声明检查点重放消耗停滞配额，并整理批处理后续操作 ([#11796](https://github.com/QwenLM/qwen-code/pull/11796)) by @qqqys
- fix(cli): 针对活动目标目录解析 ACP 核心设置 ([#11647](https://github.com/QwenLM/qwen-code/pull/11647)) by @yiliang114
- 修复(sdk)：支持通过 HTTP 创建独立会话 ([#11812](https://github.com/QwenLM/qwen-code/pull/11812)) by @doudouOUC
- 修复(core)：在工具钩子匹配器中接受 Claude Code 工具名称 ([#11826](https://github.com/QwenLM/qwen-code/pull/11826)) by @qqqys
- 修复(cli)：在混合基础时间戳下保持历史项 ID 严格递增 ([#11743](https://github.com/QwenLM/qwen-code/pull/11743)) by @yiliang114
- 修复(ci)：将自动修复推理开销传递给 Kimi K3 ([#11843](https://github.com/QwenLM/qwen-code/pull/11843)) by @yiliang114
- 修复(ci)：恢复 Windows 测试基线 ([#11787](https://github.com/QwenLM/qwen-code/pull/11787)) by @yiliang114
- 修复(cua)：恢复 0.20.7 版本的 macOS 应用观察与输入 ([#11841](https://github.com/QwenLM/qwen-code/pull/11841)) by @LaZzyMan
- 修复(cli)：仅在因未获批准而拒绝工具调用时建议 -y ([#11828](https://github.com/QwenLM/qwen-code/pull/11828)) by @qqqys
- 修复(web-shell)：稳定目标提案审批布局 ([#11839](https://github.com/QwenLM/qwen-code/pull/11839)) by @qqqys
- 修复(dws)：对齐群聊与单聊的访问控制 ([#11836](https://github.com/QwenLM/qwen-code/pull/11836)) by @qqqys
### 性能优化
- 通过移除对话历史的冗余深拷贝，提升会话恢复性能。 ([#11595](https://github.com/QwenLM/qwen-code/pull/11595)) by @DragonnZhang
- 性能优化(web-shell)：按需加载元数据并复用能力检查 ([#11644](https://github.com/QwenLM/qwen-code/pull/11644)) by @ytahdn
### 文档
- 文档(daemon)：发布 REST API 参考文档 ([#11592](https://github.com/QwenLM/qwen-code/pull/11592)) by @yiliang114
- 文档(vscode)：记录跨源删除的后果 ([#11730](https://github.com/QwenLM/qwen-code/pull/11730)) by @yiliang114
- 文档(daemon)：分别描述提示词限流和队列已满的响应 ([#11733](https://github.com/QwenLM/qwen-code/pull/11733)) by @yiliang114
- 文档：翻译权限卡片与浏览器通知设计 ([#11775](https://github.com/QwenLM/qwen-code/pull/11775)) by @wenshao
### 内部变更
- 移除过时的钉钉后台响应聚合设置，同时保留即时 Agent 响应、传输、附件、拆分和交互卡片。 ([#11570](https://github.com/QwenLM/qwen-code/pull/11570)) by @qqqys
- 增加审查测试，固定多行条目的折叠不变量，并更新相关注释，不改变可执行产品的行为。 ([#11533](https://github.com/QwenLM/qwen-code/pull/11533)) by @wenshao
- 重构 DashScope 网页搜索客户端，将其置于后端接口之后，且不改变面向用户的行为。 ([#11490](https://github.com/QwenLM/qwen-code/pull/11490)) by @qqqys
- 加强集成测试，以验证 /compress 命令确实能减少 token 消耗。 ([#11259](https://github.com/QwenLM/qwen-code/pull/11259)) by @yu-xin-c
- 移除未使用的可信钩子管理器，并使 unregisterSkillHooks 真正移除由技能根目录注册的会话钩子。 ([#11621](https://github.com/QwenLM/qwen-code/pull/11621)) by @qqqys
- 测试(integration)：将未设置 QWEN_HOME 的情况与真实 $HOME 隔离 (#10900) ([#10920](https://github.com/QwenLM/qwen-code/pull/10920)) by @qwen-code-dev-bot
- 测试(core)：针对尾部输出固定延迟的退出时 PTY 释放 ([#11585](https://github.com/QwenLM/qwen-code/pull/11585)) by @yiliang114
- 在设置模式和钩子执行总线中添加缺失的钩子事件，使编辑器不再将其标记为未知键，并确保所有钩子事件一致分发。 ([#11620](https://github.com/QwenLM/qwen-code/pull/11620)) by @qqqys
- 将共享的 Web Shell 侧边栏测试设置整合到单一测试框架中，以防止不同测试套件间行为不一致。 ([#9797](https://github.com/QwenLM/qwen-code/pull/9797)) by @bluefateludi
- 维护(chore)：强制 Playwright 版本一致性，并关闭延迟的 #11101 后续工作 ([#11371](https://github.com/QwenLM/qwen-code/pull/11371)) by @yiliang114
- 维护(pnpm)：以 package-lock 为条件控制 pnpm-lock，并声明提升的导入 ([#11625](https://github.com/QwenLM/qwen-code/pull/11625)) by @yiliang114
- 测试(lockfile)：固定 Playwright 一致性检查的第一个清单及两个嵌套方向 ([#11716](https://github.com/QwenLM/qwen-code/pull/11716)) by @yiliang114
- 重构(core)：将单一匹配器规则应用于每个钩子事件 ([#11619](https://github.com/QwenLM/qwen-code/pull/11619)) by @qqqys
- 测试(web-shell)：为 200 条记录的历史视口冒烟测试用例分配独立预算 ([#11740](https://github.com/QwenLM/qwen-code/pull/11740)) by @yiliang114
- 测试(core)：在尾部增长会话模拟中计入额外的 Windows fstat ([#11750](https://github.com/QwenLM/qwen-code/pull/11750)) by @Bug-killer-prog
- 维护(release)：将 node repl 升级至 0.1.4 ([#11771](https://github.com/QwenLM/qwen-code/pull/11771)) by @LaZzyMan
- 维护(pnpm)：限定构建审批范围，固定 ACP zod 依赖，添加锁文件新鲜度检查通道 ([#11797](https://github.com/QwenLM/qwen-code/pull/11797)) by @yiliang114
- CI(lint)：阻止新的 CLI 导入从 core 包根目录引入 ([#11798](https://github.com/QwenLM/qwen-code/pull/11798)) by @yiliang114
## 新贡献者
- @kabishou11 在 [#11145](https://github.com/QwenLM/qwen-code/pull/11145) 中首次贡献
- @CanReader 在 [#10494](https://github.com/QwenLM/qwen-code/pull/10494) 中首次贡献
- @wangxiongts 在 [#11369](https://github.com/QwenLM/qwen-code/pull/11369) 中首次贡献
- @Bug-killer-prog 在 [#11750](https://github.com/QwenLM/qwen-code/pull/11750) 中首次贡献
- @zaalipro 在 [#11565](https://github.com/QwenLM/qwen-code/pull/11565) 中首次贡献
**完整更新日志**: https://github.com/QwenLM/qwen-code/compare/v0.23.3...v0.23.4
<!-- qwen-release-notes:v1 -->
## Highlights
_See the complete change list below._
## Breaking Changes
- Removes configurable message-prefix filtering from channels, so eligible messages now follow normal sender, group, mention, and pairing policies without a prefix. ([#11571](https://github.com/QwenLM/qwen-code/pull/11571)) by @qqqys
- fix(core)!: read command hook timeout in seconds ([#11615](https://github.com/QwenLM/qwen-code/pull/11615)) by @qqqys
## Complete Change List
### Features
- Lets approved Web Shell Goal proposals start automatically after their owning turn finishes, removing the need to copy a manual /goal set command. ([#11360](https://github.com/QwenLM/qwen-code/pull/11360)) by @qqqys
- feat(sdk): add a peer endpoint so a program outside Qwen Code can join cross-session messaging ([#11560](https://github.com/QwenLM/qwen-code/pull/11560)) by @qqqys
- Adds optional Goal limits via model.goalMaxTurns and model.goalMaxActiveMinutes, and rebases active Goal time on restore. ([#11457](https://github.com/QwenLM/qwen-code/pull/11457)) by @qqqys
- Web Shell scheduled runs can now route to a configured model and session group when New session each run is selected. ([#11396](https://github.com/QwenLM/qwen-code/pull/11396)) by @qqqys
- Browser notifications in Web Shell now identify the session, show prompt and reply excerpts, and open the target session when clicked. ([#11447](https://github.com/QwenLM/qwen-code/pull/11447)) by @doudouOUC
- Improves Web Shell composer compactness and makes context usage clearer across composer, conversation, and sidebar. ([#11509](https://github.com/QwenLM/qwen-code/pull/11509)) by @wenshao
- Adds visual input, proactive assistance, and local memory to the standalone Qwen Live experience. ([#11369](https://github.com/QwenLM/qwen-code/pull/11369)) by @wangxiongts
- Automatic reviews of small static documentation navigation changes now use a focused reviewer with narrower scope and capped outcomes. ([#11456](https://github.com/QwenLM/qwen-code/pull/11456)) by @wenshao
- feat(subagents): add Codex executor and native agent builtins ([#11474](https://github.com/QwenLM/qwen-code/pull/11474)) by @wenshao
- feat(core): add CodeModeOnly tool execution ([#10607](https://github.com/QwenLM/qwen-code/pull/10607)) by @DragonnZhang
- Moved Workflow authoring reference into the bundled workflow-authoring skill, reducing default tool description overhead. ([#11520](https://github.com/QwenLM/qwen-code/pull/11520)) by @qqqys
- Hid the Web Shell sources panel add-source button by default; it can be re-enabled with the addSource=1 URL parameter. ([#11639](https://github.com/QwenLM/qwen-code/pull/11639)) by @callmeYe
- Channel turns in CLI resume, Web Shell, and session previews now display exactly the prompt the model received. ([#11646](https://github.com/QwenLM/qwen-code/pull/11646)) by @qqqys
- Added Web preview panels with desktop and mobile widths, refresh, external opening, saved HTML artifacts, and per-session URL and width memory. ([#11276](https://github.com/QwenLM/qwen-code/pull/11276)) by @wenshao
- Added inline editing and resending of the latest user message in Web Shell, retaining attached images and files while generating a new reply. ([#11650](https://github.com/QwenLM/qwen-code/pull/11650)) by @ytahdn
- feat(core): add permission_mode, agent_id and prompt_id to every hook input ([#11618](https://github.com/QwenLM/qwen-code/pull/11618)) by @qqqys
- Adds a Continue execution action in Web Shell for recoverable interrupted sessions, allowing users to resume the original conversation without duplicating messages. ([#11545](https://github.com/QwenLM/qwen-code/pull/11545)) by @wenshao
- feat(skills): extension skills are named by their extension ([#10841](https://github.com/QwenLM/qwen-code/pull/10841)) by @nerdalytics
- feat(vscode): add an explicit session source switch ([#11584](https://github.com/QwenLM/qwen-code/pull/11584)) by @yiliang114
- feat(core): auto-retry transient network errors (EOF) where Ctrl+Y is unavailable ([#10347](https://github.com/QwenLM/qwen-code/pull/10347)) by @qwen-code-dev-bot
- feat(workflows): add effort and disallowedTools to agent() options ([#11691](https://github.com/QwenLM/qwen-code/pull/11691)) by @qqqys
- feat(goal): rerun a stalled checkpoint in batches, smaller after each stall ([#11690](https://github.com/QwenLM/qwen-code/pull/11690)) by @qqqys
- feat(web-shell): add model role and context window configuration ([#11342](https://github.com/QwenLM/qwen-code/pull/11342)) by @wenshao
- feat(cua): add app-bound actions and compact native observations ([#11705](https://github.com/QwenLM/qwen-code/pull/11705)) by @LaZzyMan
- feat: agent board — share work across independently started agents ([#9402](https://github.com/QwenLM/qwen-code/pull/9402)) by @yiliang114
- feat(review): record a default deadline in the plan so every run has a wall ([#11686](https://github.com/QwenLM/qwen-code/pull/11686)) by @wenshao
- feat(web-shell): make Session Workflow dependencies navigable and quiet its chrome ([#10938](https://github.com/QwenLM/qwen-code/pull/10938)) by @yiliang114
- feat(web-shell): Improve context overview and add manual compression ([#11700](https://github.com/QwenLM/qwen-code/pull/11700)) by @wenshao
- feat(core): make the web_search budget configurable and bound the extractor fallback ([#11692](https://github.com/QwenLM/qwen-code/pull/11692)) by @qqqys
- feat(web-shell): restore the workspace Git picker in the sidebar hover details ([#11818](https://github.com/QwenLM/qwen-code/pull/11818)) by @ytahdn
- feat(web-shell): Add manual compression to composer context hover ([#11782](https://github.com/QwenLM/qwen-code/pull/11782)) by @wenshao
### Bug Fixes
- Deflakes a replayed-loop CI test by widening the fabricated timeline-event margin from 75 to 300 seconds without changing workflow or product behavior. ([#11581](https://github.com/QwenLM/qwen-code/pull/11581)) by @wenshao
- Fixes autofix gating so regressions count against the brake and test weakening is gated, preventing a fix round from shipping new problems for free. ([#10188](https://github.com/QwenLM/qwen-code/pull/10188)) by @wenshao
- Preserves the originally submitted prompt for UserPromptSubmit hooks so Auto Recall works for Web Shell submissions and opted-in ACP or daemon clients. ([#11455](https://github.com/QwenLM/qwen-code/pull/11455)) by @doudouOUC
- Restores workspace-configured startup channels when qwen serve starts without an explicit channel argument, while explicit channel selection still takes precedence. ([#11472](https://github.com/QwenLM/qwen-code/pull/11472)) by @qqqys
- Fixes qwen serve startup so persisted workspace-scope MCP configuration is loaded after ACP preheat. ([#11145](https://github.com/QwenLM/qwen-code/pull/11145)) by @kabishou11
- Fixes VS Code diff handling so diffs can be closed reliably using either the original or workspace-resolved path. ([#10494](https://github.com/QwenLM/qwen-code/pull/10494)) by @CanReader
- Frees web terminal PTY resources when the terminal exits instead of waiting for later release. ([#11572](https://github.com/QwenLM/qwen-code/pull/11572)) by @yiliang114
- Adds one bounded retry to E2E build artifact uploads to reduce transient CI failures. ([#11375](https://github.com/QwenLM/qwen-code/pull/11375)) by @qwen-code-dev-bot
- Preserves child error details when agent sessions are explicitly closed or killed, improving diagnostic logging. ([#11190](https://github.com/QwenLM/qwen-code/pull/11190)) by @now-ing
- Web Shell now preserves cancelled responses across reloads and shows how long the request ran before cancellation. ([#11608](https://github.com/QwenLM/qwen-code/pull/11608)) by @ytahdn
- Pins the fleet updater install to the /usr/local prefix so runners execute the newly installed version instead of continuing to use an older install on PATH. ([#11593](https://github.com/QwenLM/qwen-code/pull/11593)) by @yiliang114
- fix(core): recover from rejected Responses encrypted reasoning ([#11596](https://github.com/QwenLM/qwen-code/pull/11596)) by @DragonnZhang
- Improves checkpoint verification by stripping markdown fences and retrying with one corrective model call for certain malformed or inconsistent verifier answers. ([#11578](https://github.com/QwenLM/qwen-code/pull/11578)) by @qqqys
- Shows the Try again action in source previews only after attachment or workspace-file load failures, and clears it after a successful retry. ([#11629](https://github.com/QwenLM/qwen-code/pull/11629)) by @callmeYe
- Routes successful plain-text hook stdout to additionalContext for SessionStart, UserPromptSubmit, and UserPromptExpansion events while other events keep system-message behavior. ([#11612](https://github.com/QwenLM/qwen-code/pull/11612)) by @qqqys
- Fixed OpenTUI renderer display gaps across transcript, composer, footer, dialogs, keyboard, and startup approval to match ink. ([#11630](https://github.com/QwenLM/qwen-code/pull/11630)) by @chiga0
- Reattached image snapshots are now labeled as potentially outdated so earlier session images are not treated as current UI state. ([#11602](https://github.com/QwenLM/qwen-code/pull/11602)) by @yiliang114
- Localized VS Code rewind preflight failures when editing a sent message, replacing raw transport errors with a translated message. ([#11628](https://github.com/QwenLM/qwen-code/pull/11628)) by @yiliang114
- fix(core): preserve Responses retry metadata and assistant phases ([#11652](https://github.com/QwenLM/qwen-code/pull/11652)) by @DragonnZhang
- fix(cli): keep the expanded confirmation dialog inside the viewport ([#11656](https://github.com/QwenLM/qwen-code/pull/11656)) by @qwen-code-dev-bot
- fix(core): retry transient server errors in established streams ([#11634](https://github.com/QwenLM/qwen-code/pull/11634)) by @DragonnZhang
- fix(vscode): preserve IDE workspace env var on Remote-SSH (#11556) ([#11624](https://github.com/QwenLM/qwen-code/pull/11624)) by @yiliang114
- fix(vscode): give the edit back when a permission diff is closed ([#11171](https://github.com/QwenLM/qwen-code/pull/11171)) by @yiliang114
- Persists a one-line diagnostic when a Goal evidence checkpoint fails and displays it before the Goal stops, improving visibility into why checkpoints fail. ([#11576](https://github.com/QwenLM/qwen-code/pull/11576)) by @qqqys
- Suppresses unnecessary error toasts in Web Shell when a submission is rejected due to preflight cancellation or abort, avoiding confusing failure messages. ([#11676](https://github.com/QwenLM/qwen-code/pull/11676)) by @yiliang114
- fix(core): redact request-body prefix in Responses debug logs ([#11668](https://github.com/QwenLM/qwen-code/pull/11668)) by @yiliang114
- Correctly labels blocked Goal cards as Goal blocked in Web Shell instead of the broader Goal aborted label, improving status clarity. ([#11174](https://github.com/QwenLM/qwen-code/pull/11174)) by @yu-xin-c
- Updates a CLI serve test to wait for the fire-and-forget handler to finish before completing, improving test reliability without changing production behavior. ([#11417](https://github.com/QwenLM/qwen-code/pull/11417)) by @qwen-code-dev-bot
- Prevents Responses reasoning replay data from being sent to incompatible providers by stripping or ignoring foreign thoughtSignature payloads during request building. ([#11567](https://github.com/QwenLM/qwen-code/pull/11567)) by @yiliang114
- Fixes local-files bridge trust and bystander issues by reserving a resolving state and preventing parked tabs from wiping stored grants on disconnect. ([#11169](https://github.com/QwenLM/qwen-code/pull/11169)) by @wenshao
- Moves the DashScope conversation cache breakpoint before reattached images so the cached prefix no longer shifts between turns. ([#11640](https://github.com/QwenLM/qwen-code/pull/11640)) by @yiliang114
- fix(core): report real stop_hook_active on Stop hooks ([#11613](https://github.com/QwenLM/qwen-code/pull/11613)) by @qqqys
- fix(core): cite web search sources as bare URLs ([#11677](https://github.com/QwenLM/qwen-code/pull/11677)) by @qqqys
- ACP child spawns now derive V8 old-space targets from actual host memory instead of trusting an unlimited cgroup sentinel value. ([#11653](https://github.com/QwenLM/qwen-code/pull/11653)) by @doudouOUC
- Prevents stale concurrent leader assignments from dispatching the same task to multiple teammates. ([#10237](https://github.com/QwenLM/qwen-code/pull/10237)) by @AaronZ345
- fix(telemetry): gate request_text/response_text on logPrompts ([#11670](https://github.com/QwenLM/qwen-code/pull/11670)) by @yiliang114
- fix(vscode): unify session history and remove source switch ([#11713](https://github.com/QwenLM/qwen-code/pull/11713)) by @yiliang114
- fix(acp): preserve original filenames in resource-link prompts ([#11721](https://github.com/QwenLM/qwen-code/pull/11721)) by @zhangxy-zju
- fix(vscode): shut the ACP CLI down gracefully instead of killing it ([#11642](https://github.com/QwenLM/qwen-code/pull/11642)) by @yiliang114
- fix(core): redact error text in usage-statistics telemetry sink ([#11649](https://github.com/QwenLM/qwen-code/pull/11649)) by @yiliang114
- fix(core): reap surviving hook process trees on Windows ([#11623](https://github.com/QwenLM/qwen-code/pull/11623)) by @yiliang114
- fix(core): compute cron next-fire time with epoch arithmetic across DST fall-back ([#11723](https://github.com/QwenLM/qwen-code/pull/11723)) by @yiliang114
- fix(goal): remove backtracking fence unwrap from judge replies ([#11693](https://github.com/QwenLM/qwen-code/pull/11693)) by @yiliang114
- fix(core): keep the mirrored reasoning field off Fireworks requests ([#11701](https://github.com/QwenLM/qwen-code/pull/11701)) by @bluefateludi
- fix(core): fire active-todo reminders at delegation boundaries and user turns ([#10963](https://github.com/QwenLM/qwen-code/pull/10963)) by @yiliang114
- fix(core): guard internal git calls against configured helper programs ([#11669](https://github.com/QwenLM/qwen-code/pull/11669)) by @yiliang114
- fix(lsp): synchronize target documents before queries ([#11443](https://github.com/QwenLM/qwen-code/pull/11443)) by @shenyankm
- fix(core): run web terminal PTYs on the bundled ConPTY backend ([#11643](https://github.com/QwenLM/qwen-code/pull/11643)) by @yiliang114
- fix(core): retry status-less upstream errors instead of ending the turn ([#11291](https://github.com/QwenLM/qwen-code/pull/11291)) by @wenshao
- fix(dashscope): send request metadata only for qwen-family models ([#11606](https://github.com/QwenLM/qwen-code/pull/11606)) by @shoemoney
- fix(web-shell): prevent terminal query freezes and clean up protocol rejection ([#11748](https://github.com/QwenLM/qwen-code/pull/11748)) by @yiliang114
- fix(cli): break Ink useBoxMetrics commit-phase setState loop (React #185) ([#11565](https://github.com/QwenLM/qwen-code/pull/11565)) by @zaalipro
- fix(dingtalk): handle replied rich-text media ([#11759](https://github.com/QwenLM/qwen-code/pull/11759)) by @qqqys
- fix(dingtalk): preserve callback text ([#11757](https://github.com/QwenLM/qwen-code/pull/11757)) by @qqqys
- fix(cli): prevent dialog clipping in short terminals ([#9040](https://github.com/QwenLM/qwen-code/pull/9040)) by @shenyankm
- fix(web-shell): keep mid-turn messages the daemon rejects at idle ([#11289](https://github.com/QwenLM/qwen-code/pull/11289)) by @wenshao
- fix(core): strip AppImage PYTHONHOME/PYTHONPATH from stdio MCP children ([#11763](https://github.com/QwenLM/qwen-code/pull/11763)) by @yiliang114
- fix(cua): address tool failures and efficiency in OSWorld runs ([#11683](https://github.com/QwenLM/qwen-code/pull/11683)) by @tanzhenxin
- fix(cua): bound observations and reduce model round trips ([#11735](https://github.com/QwenLM/qwen-code/pull/11735)) by @DragonnZhang
- fix(web-shell): pin es2021 floor across both build configs and harden terminal tests ([#11758](https://github.com/QwenLM/qwen-code/pull/11758)) by @yiliang114
- fix(core): let the producer's own budget decide shell output size ([#11727](https://github.com/QwenLM/qwen-code/pull/11727)) by @doudouOUC
- fix(review): document the deadline's pause ceilings and improve capture-local errors ([#11779](https://github.com/QwenLM/qwen-code/pull/11779)) by @wenshao
- fix(ci): raise the build heap cap from 3072 to 4096 MB ([#11781](https://github.com/QwenLM/qwen-code/pull/11781)) by @doudouOUC
- fix(cli): reap running monitors when the process dies on an uncaught exception ([#11742](https://github.com/QwenLM/qwen-code/pull/11742)) by @yiliang114
- fix(web-shell): move source download to the header ([#11641](https://github.com/QwenLM/qwen-code/pull/11641)) by @callmeYe
- fix(skills): re-apply a Skill's side effects when a session is resumed ([#11280](https://github.com/QwenLM/qwen-code/pull/11280)) by @TianYuan1024
- fix(web-shell): make the cockpit visual capture deterministic ([#11786](https://github.com/QwenLM/qwen-code/pull/11786)) by @yiliang114
- fix(dingtalk): remove background agent headings ([#11800](https://github.com/QwenLM/qwen-code/pull/11800)) by @qqqys
- fix(cli): treat write-side PTY EIO as a benign teardown race ([#11788](https://github.com/QwenLM/qwen-code/pull/11788)) by @yiliang114
- fix(goal): keep a full-claim checkpoint replay from spending a stall, and tidy batching follow-ups ([#11796](https://github.com/QwenLM/qwen-code/pull/11796)) by @qqqys
- fix(cli): resolve ACP core settings against the active target dir ([#11647](https://github.com/QwenLM/qwen-code/pull/11647)) by @yiliang114
- fix(sdk): support standalone session creation over HTTP ([#11812](https://github.com/QwenLM/qwen-code/pull/11812)) by @doudouOUC
- fix(core): accept Claude Code tool names in tool hook matchers ([#11826](https://github.com/QwenLM/qwen-code/pull/11826)) by @qqqys
- fix(cli): keep history item ids strictly increasing across mixed base timestamps ([#11743](https://github.com/QwenLM/qwen-code/pull/11743)) by @yiliang114
- fix(ci): pass autofix reasoning effort to Kimi K3 ([#11843](https://github.com/QwenLM/qwen-code/pull/11843)) by @yiliang114
- fix(ci): restore Windows test baseline ([#11787](https://github.com/QwenLM/qwen-code/pull/11787)) by @yiliang114
- fix(cua): recover macOS App observation and input for 0.20.7 ([#11841](https://github.com/QwenLM/qwen-code/pull/11841)) by @LaZzyMan
- fix(cli): only suggest -y for tool calls denied for lack of approval ([#11828](https://github.com/QwenLM/qwen-code/pull/11828)) by @qqqys
- fix(web-shell): stabilize goal proposal approval layout ([#11839](https://github.com/QwenLM/qwen-code/pull/11839)) by @qqqys
- fix(dws): align group and direct chat access controls ([#11836](https://github.com/QwenLM/qwen-code/pull/11836)) by @qqqys
### Performance
- Improves session recovery performance by removing a redundant deep copy of conversation history. ([#11595](https://github.com/QwenLM/qwen-code/pull/11595)) by @DragonnZhang
- perf(web-shell): load metadata on demand and reuse capability checks ([#11644](https://github.com/QwenLM/qwen-code/pull/11644)) by @ytahdn
### Documentation
- docs(daemon): publish the REST API reference ([#11592](https://github.com/QwenLM/qwen-code/pull/11592)) by @yiliang114
- docs(vscode): record the cross-source delete consequence ([#11730](https://github.com/QwenLM/qwen-code/pull/11730)) by @yiliang114
- docs(daemon): describe the prompt rate-limit and queue-full responses separately ([#11733](https://github.com/QwenLM/qwen-code/pull/11733)) by @yiliang114
- docs: translate permission card and browser notification designs ([#11775](https://github.com/QwenLM/qwen-code/pull/11775)) by @wenshao
### Internal Changes
- Removes the obsolete DingTalk background response aggregation setting while keeping immediate Agent responses, transports, attachments, splitting, and interactive cards. ([#11570](https://github.com/QwenLM/qwen-code/pull/11570)) by @qqqys
- Adds review tests that pin the fold invariant for multi-line entries and updates related comments without changing executable product behavior. ([#11533](https://github.com/QwenLM/qwen-code/pull/11533)) by @wenshao
- Refactors the DashScope web search client behind a backend interface with no user-facing behavior change. ([#11490](https://github.com/QwenLM/qwen-code/pull/11490)) by @qqqys
- Strengthens integration tests to verify /compress actually reduces token usage. ([#11259](https://github.com/QwenLM/qwen-code/pull/11259)) by @yu-xin-c
- Removes the unused trusted hooks manager and makes unregisterSkillHooks actually remove session hooks registered by a skill's root directory. ([#11621](https://github.com/QwenLM/qwen-code/pull/11621)) by @qqqys
- test(integration): isolate the QWEN_HOME-unset case from the real $HOME (#10900) ([#10920](https://github.com/QwenLM/qwen-code/pull/10920)) by @qwen-code-dev-bot
- test(core): pin the deferred exit-time PTY release against trailing output ([#11585](https://github.com/QwenLM/qwen-code/pull/11585)) by @yiliang114
- Adds missing hook events to the settings schema and hook execution bus so editors no longer flag them as unknown keys and all hook events are dispatched consistently. ([#11620](https://github.com/QwenLM/qwen-code/pull/11620)) by @qqqys
- Consolidates shared Web Shell sidebar test setup into a single harness to prevent inconsistent behavior across suites. ([#9797](https://github.com/QwenLM/qwen-code/pull/9797)) by @bluefateludi
- chore: enforce Playwright version parity and close the deferred #11101 follow-ups ([#11371](https://github.com/QwenLM/qwen-code/pull/11371)) by @yiliang114
- chore(pnpm): gate pnpm-lock on package-lock and declare hoisted imports ([#11625](https://github.com/QwenLM/qwen-code/pull/11625)) by @yiliang114
- test(lockfile): pin the Playwright parity check's first manifest and both nested directions ([#11716](https://github.com/QwenLM/qwen-code/pull/11716)) by @yiliang114
- refactor(core): apply one matcher rule to every hook event ([#11619](https://github.com/QwenLM/qwen-code/pull/11619)) by @qqqys
- test(web-shell): give the 200-record history-viewport smoke case its own budget ([#11740](https://github.com/QwenLM/qwen-code/pull/11740)) by @yiliang114
- test(core): account for the extra Windows fstat in the tail-growth session mocks ([#11750](https://github.com/QwenLM/qwen-code/pull/11750)) by @Bug-killer-prog
- chore(release): bump node repl to 0.1.4 ([#11771](https://github.com/QwenLM/qwen-code/pull/11771)) by @LaZzyMan
- chore(pnpm): scope build approvals, pin the ACP zod peer, add a lockfile freshness lane ([#11797](https://github.com/QwenLM/qwen-code/pull/11797)) by @yiliang114
- ci(lint): stop new cli imports from the core package root ([#11798](https://github.com/QwenLM/qwen-code/pull/11798)) by @yiliang114
## New Contributors
- @kabishou11 made their first contribution in [#11145](https://github.com/QwenLM/qwen-code/pull/11145)
- @CanReader made their first contribution in [#10494](https://github.com/QwenLM/qwen-code/pull/10494)
- @wangxiongts made their first contribution in [#11369](https://github.com/QwenLM/qwen-code/pull/11369)
- @Bug-killer-prog made their first contribution in [#11750](https://github.com/QwenLM/qwen-code/pull/11750)
- @zaalipro made their first contribution in [#11565](https://github.com/QwenLM/qwen-code/pull/11565)
**Full Changelog**: https://github.com/QwenLM/qwen-code/compare/v0.23.3...v0.23.4