<!-- 使用 release/v0.19.6-nightly.20260705.015ee4248 分支中 .github/release.yml 的配置生成的发布说明 -->
## 更新内容
* fix(triage): 通过批量检测、问题存在性检查和红旗模式加强 PR 门禁 by @pomelo-nwu 在 https://github.com/QwenLM/qwen-code/pull/5723
* feat(review): 将建议级别的发现路由到可更新的 PR 评论 by @pomelo-nwu 在 https://github.com/QwenLM/qwen-code/pull/5786
* fix(autofix): 在分支检出前无条件恢复已跟踪文件 (#6281) by @qwen-code-dev-bot 在 https://github.com/QwenLM/qwen-code/pull/6286
* feat(serve): 在守护进程状态 API 中添加 runtime.activity 字段 by @doudouOUC 在 https://github.com/QwenLM/qwen-code/pull/6270
* feat(web-shell): 基于 GET /daemon/status 添加守护进程状态页面 by @wenshao 在 https://github.com/QwenLM/qwen-code/pull/6272
* fix(vscode): 失去焦点时保持认证快捷输入框打开 by @barry166 在 https://github.com/QwenLM/qwen-code/pull/6274
* ci(audio): 明确 macOS 预构建产物后缀 by @barry166 在 https://github.com/QwenLM/qwen-code/pull/6275
* fix(cache): 在 Anthropic 提示词缓存命中时，在侧查询中保留 tools 前缀 by @kagura-agent 在 https://github.com/QwenLM/qwen-code/pull/6225
* feat(web-shell): 添加 MCP 提及和带图标的 @ 引用 by @BZ-D 在 https://github.com/QwenLM/qwen-code/pull/6279
* fix(core): 为 Stop-hook 续期提供全新的每轮工具调用预算；使上限可配置 by @tanzhenxin 在 https://github.com/QwenLM/qwen-code/pull/6238
* perf(glob): 在遍历期间剪枝被忽略的目录，而不仅限于后置过滤 by @minmax 在 https://github.com/QwenLM/qwen-code/pull/6123
* fix(web-shell): 使用主题色显示 @ 分组标题 by @BZ-D 在 https://github.com/QwenLM/qwen-code/pull/6294
* fix(acp): 显式传递会话级设置，避免在 this.settings 上竞态 by @doudouOUC 在 https://github.com/QwenLM/qwen-code/pull/6292
* feat(web-shell): 从侧边栏管理会话（归档、取消归档、删除） by @wenshao 在 https://github.com/QwenLM/qwen-code/pull/6293
* fix(core): 改进 debug txt 诊断信息 by @yiliang114 在 https://github.com/QwenLM/qwen-code/pull/6277
* fix(ci): 为 tier-1 快速路径要求维护者手动添加 `autofix/approved` 标签 by @yiliang114 在 https://github.com/QwenLM/qwen-code/pull/6276
* fix(auth): 防止 API 密钥更改后持续出现 401 错误 by @yiliang114 在 https://github.com/QwenLM/qwen-code/pull/6284
* feat(daemon): 添加会话导出端点 by @doudouOUC 在 https://github.com/QwenLM/qwen-code/pull/6297
* test(e2e): 使 Docker 沙箱可访问伪造的 OpenAI 服务 by @yiliang114 在 https://github.com/QwenLM/qwen-code/pull/6302
* fix(cli): 将长响应流式传输到滚动回溯中，以停止滚动到顶部锁定 by @MikeWang0316tw 在 https://github.com/QwenLM/qwen-code/pull/6170
* fix(qqbot): 流式空闲刷新，增加工具调用和过期回调保护 by @Eric-GoodBoy-Tech 在 https://github.com/QwenLM/qwen-code/pull/6204
* fix(openai): 保留无描述的工具 by @VectorPeak 在 https://github.com/QwenLM/qwen-code/pull/6243
* feat(acp): 在初始化响应中声明 vision-bridge 图像能力 by @yiliang114 在 https://github.com/QwenLM/qwen-code/pull/6269
* fix(core): 在前台子代理上强制执行代理并发上限 by @kagura-agent 在 https://github.com/QwenLM/qwen-code/pull/6300
* fix(ci): 停止对已关闭 PR 的审查机器人 by @yiliang114 在 https://github.com/QwenLM/qwen-code/pull/6304
* fix(core): 在 multiple_sentences 过滤器中跳过缩写 (#6077) by @qwen-code-dev-bot 在 https://github.com/QwenLM/qwen-code/pull/6193
* perf(cli): 按工作区缓存 LoadedSettings，并使用基于 stat 的失效机制 by @doudouOUC 在 https://github.com/QwenLM/qwen-code/pull/6310
* feat(web-shell): 支持紧凑版 echarts 完整数据块 by @zhangxy-zju 在 https://github.com/QwenLM/qwen-code/pull/6232
* fix(ci): 跳过过期的 PR 审查运行 by @yiliang114 在 https://github.com/QwenLM/qwen-code/pull/6313
* feat(acp): 批量会话加载重放 by @doudouOUC 在 https://github.com/QwenLM/qwen-code/pull/6309
* fix(core): 将请求超时 0 视为禁用，而非立即中止 by @Nas01010101 在 https://github.com/QwenLM/qwen-code/pull/6288
* fix(serve): 解决通过设置配置 API 密钥时预检中的虚假认证警告 by @doudouOUC 在 https://github.com/QwenLM/qwen-code/pull/6296
* fix(core): 在强制先读检查时将 @ 附加文件视为已读取 by @Nas01010101 在 https://github.com/QwenLM/qwen-code/pull/6295
* fix(cli): 保留部分远程输入 JSONL 记录 by @VectorPeak 在 https://github.com/QwenLM/qwen-code/pull/6317
* fix(core): 在非 DashScope 服务器上通过 chat_template_kwargs 禁用 qwen thinking by @beantownbytes 在 https://github.com/QwenLM/qwen-code/pull/6271
## 新贡献者
* @barry166 在 https://github.com/QwenLM/qwen-code/pull/6274 中首次贡献
* @minmax 在 https://github.com/QwenLM/qwen-code/pull/6123 中首次贡献
* @Nas01010101 在 https://github.com/QwenLM/qwen-code/pull/6288 中首次贡献
* @beantownbytes 在 https://github.com/QwenLM/qwen-code/pull/6271 中首次贡献
**完整更新日志**: https://github.com/QwenLM/qwen-code/compare/v0.19.6...v0.19.6-nightly.20260705.015ee4248
<!-- Release notes generated using configuration in .github/release.yml at release/v0.19.6-nightly.20260705.015ee4248 -->
## What's Changed
* fix(triage): strengthen PR gate with batch detection, problem existence check, and red flag patterns by @pomelo-nwu in https://github.com/QwenLM/qwen-code/pull/5723
* feat(review): route suggestion-level findings to an updatable PR comment by @pomelo-nwu in https://github.com/QwenLM/qwen-code/pull/5786
* fix(autofix): unconditionally restore tracked files before branch checkout (#6281) by @qwen-code-dev-bot in https://github.com/QwenLM/qwen-code/pull/6286
* feat(serve): Add runtime.activity fields to daemon status API by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/6270
* feat(web-shell): add a daemon status page backed by GET /daemon/status by @wenshao in https://github.com/QwenLM/qwen-code/pull/6272
* fix(vscode): keep auth quick inputs open on focus loss by @barry166 in https://github.com/QwenLM/qwen-code/pull/6274
* ci(audio): clarify macOS prebuild artifact suffix by @barry166 in https://github.com/QwenLM/qwen-code/pull/6275
* fix(cache): preserve tools prefix in side-query for Anthropic prompt-cache hits by @kagura-agent in https://github.com/QwenLM/qwen-code/pull/6225
* feat(web-shell): add MCP mentions and iconized @ references by @BZ-D in https://github.com/QwenLM/qwen-code/pull/6279
* fix(core): give Stop-hook continuations a fresh per-turn tool-call budget; make the cap configurable by @tanzhenxin in https://github.com/QwenLM/qwen-code/pull/6238
* perf(glob): prune ignored directories during traversal, not just post-filter by @minmax in https://github.com/QwenLM/qwen-code/pull/6123
* fix(web-shell): use theme color for @ group titles by @BZ-D in https://github.com/QwenLM/qwen-code/pull/6294
* fix(acp): pass per-session settings explicitly instead of racing on this.settings by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/6292
* feat(web-shell): manage sessions from the sidebar (archive, unarchive, delete) by @wenshao in https://github.com/QwenLM/qwen-code/pull/6293
* fix(core): improve debug txt diagnostics by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/6277
* fix(ci): require maintainer-applied `autofix/approved` label for tier-1 fast-path by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/6276
* fix(auth): prevent persistent 401 after API key change by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/6284
* feat(daemon): Add session export endpoint by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/6297
* test(e2e): make fake OpenAI reachable from Docker sandbox by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/6302
* fix(cli): stream long responses into scrollback to stop scroll-to-top lock by @MikeWang0316tw in https://github.com/QwenLM/qwen-code/pull/6170
* fix(qqbot): streaming idle-flush with tool-call and stale-callback protection by @Eric-GoodBoy-Tech in https://github.com/QwenLM/qwen-code/pull/6204
* fix(openai): preserve descriptionless tools by @VectorPeak in https://github.com/QwenLM/qwen-code/pull/6243
* feat(acp): advertise vision-bridge image capability in initialize response by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/6269
* fix(core): enforce agent concurrency cap on foreground sub-agents by @kagura-agent in https://github.com/QwenLM/qwen-code/pull/6300
* fix(ci): Stop review bots for closed PRs by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/6304
* fix(core): skip abbreviations in multiple_sentences filter (#6077) by @qwen-code-dev-bot in https://github.com/QwenLM/qwen-code/pull/6193
* perf(cli): cache LoadedSettings per workspace with stat-based invalidation by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/6310
* feat(web-shell): support compact echarts full data blocks by @zhangxy-zju in https://github.com/QwenLM/qwen-code/pull/6232
* fix(ci): skip stale PR review runs by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/6313
* feat(acp): Batch session load replay by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/6309
* fix(core): treat request timeout of 0 as disabled instead of aborting immediately by @Nas01010101 in https://github.com/QwenLM/qwen-code/pull/6288
* fix(serve): resolve false auth warning in preflight when API key is set via settings by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/6296
* fix(core): treat @-attached files as read for prior-read enforcement by @Nas01010101 in https://github.com/QwenLM/qwen-code/pull/6295
* fix(cli): preserve partial remote input JSONL records by @VectorPeak in https://github.com/QwenLM/qwen-code/pull/6317
* fix(core): disable qwen thinking via chat_template_kwargs on non-DashScope servers by @beantownbytes in https://github.com/QwenLM/qwen-code/pull/6271
## New Contributors
* @barry166 made their first contribution in https://github.com/QwenLM/qwen-code/pull/6274
* @minmax made their first contribution in https://github.com/QwenLM/qwen-code/pull/6123
* @Nas01010101 made their first contribution in https://github.com/QwenLM/qwen-code/pull/6288
* @beantownbytes made their first contribution in https://github.com/QwenLM/qwen-code/pull/6271
**Full Changelog**: https://github.com/QwenLM/qwen-code/compare/v0.19.6...v0.19.6-nightly.20260705.015ee4248