<!-- 使用 release/v0.19.3-nightly.20260630.e00fe6a27 分支中 .github/release.yml 的配置生成发布说明 -->
## 更新内容
* docs(daemon): 为最近的 PR 刷新守护进程文档(第 2 波) by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5954
* feat(core): 添加可配置自动压缩阈值及 Stop hook 上下文使用 (#4025) by @ZijianZhang989 in https://github.com/QwenLM/qwen-code/pull/5868
* ci(review): 将 PR 审查超时时间从 90 分钟增加到 120 分钟 by @wenshao in https://github.com/QwenLM/qwen-code/pull/5959
* fix(core): 阻止重复的 shell 检查变体 by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/5944
* refactor(cli): 移除 serve 桥接重新导出垫片 by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5955
* feat(core,cli,sdk): 无需合成“继续”消息即可恢复中断的对话轮次 by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/5030
* ci: 允许更长的 PR 审查超时重试 by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/5961
* feat(desktop): 桌面应用支持语音输入 by @qqqys in https://github.com/QwenLM/qwen-code/pull/5856
* feat(browser-ext): 通过 daemon-direct 架构恢复 Chrome 扩展 by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/5777
* docs(telemetry): 全面更新文档以匹配当前实现 by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5960
* ci(autofix): 放宽问题候选者过滤器,以便代理能找到任务 by @qqqys in https://github.com/QwenLM/qwen-code/pull/5860
* feat(loop): 通过哨兵在触发时注入 .qwen/loop.md 任务文件 by @qqqys in https://github.com/QwenLM/qwen-code/pull/5890
* fix(cli): 在多选问题中按 Enter 键自动选择自定义输入 by @Alex-ai-future in https://github.com/QwenLM/qwen-code/pull/5791
* fix(core): 仅在启用自动记忆时触发记忆召回 by @Minerest in https://github.com/QwenLM/qwen-code/pull/5963
* fix(release): 使用相对 postinstall 补丁目录 by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/5973
* docs(qc-helper): 添加守护进程模式文档并修复系统设置路径 by @DragonnZhang in https://github.com/QwenLM/qwen-code/pull/5981
* fix(standalone): 通过 cli-entry 路由 serve 垫片 by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5977
* feat(channels): qwen 标签 — RFC + 第 0 阶段(多人频道驻留代理) by @qqqys in https://github.com/QwenLM/qwen-code/pull/5888
* fix(ui): 子代理显示输出 token 数而非累计 API 吞吐量 by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5972
* fix(cli): 在 serve 快速路径中避免 ACP 运行时预加载 by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5989
* fix(web-shell): 工具输出中优先显示原始文件差异 by @ytahdn in https://github.com/QwenLM/qwen-code/pull/5992
* [codex] 修复守护进程专用模型过滤 by @callmeYe in https://github.com/QwenLM/qwen-code/pull/5993
* feat(channels): 添加频道代理桥接抽象 by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5978
* fix(web-shell): 改进后续建议处理 by @ytahdn in https://github.com/QwenLM/qwen-code/pull/5996
* fix(ci): 覆盖发布集成回归测试 by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/5994
* feat(core): 添加 --insecure 标志以跳过自签名端点的 TLS 验证 (#3535) by @TianYuan1024 in https://github.com/QwenLM/qwen-code/pull/5962
* fix(core): 从 Stop hook 的 last_assistant_message 中过滤掉思考部分 by @DennisYu07 in https://github.com/QwenLM/qwen-code/pull/6009
* feat(cli): 添加 --safe-mode 标志以禁用所有自定义项用于故障排查 by @DennisYu07 in https://github.com/QwenLM/qwen-code/pull/4943
* fix(cli): 修复 VP 模式下思考查看器的截断、布局间隙和滚动卡顿 by @chiga0 in https://github.com/QwenLM/qwen-code/pull/6002
* fix(cli): 保护 serve 快速路径包闭包 by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5995
* feat(daemon): 支持 @extension 提及 by @callmeYe in https://github.com/QwenLM/qwen-code/pull/6008
* test(ci): 稳定 cron 交互式发布检查 by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/6016
## 新贡献者
* @Minerest 在 https://github.com/QwenLM/qwen-code/pull/5963 完成了首次贡献
* @TianYuan1024 在 https://github.com/QwenLM/qwen-code/pull/5962 完成了首次贡献
**完整更新日志**: https://github.com/QwenLM/qwen-code/compare/v0.19.3...v0.19.3-nightly.20260630.e00fe6a27
<!-- Release notes generated using configuration in .github/release.yml at release/v0.19.3-nightly.20260630.e00fe6a27 -->
## What's Changed
* docs(daemon): refresh daemon docs for recent PRs (wave 2) by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5954
* feat(core): add configurable auto-compact threshold and Stop hook context usage (#4025) by @ZijianZhang989 in https://github.com/QwenLM/qwen-code/pull/5868
* ci(review): increase PR review timeout from 90 to 120 minutes by @wenshao in https://github.com/QwenLM/qwen-code/pull/5959
* fix(core): halt repeated shell inspection variants by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/5944
* refactor(cli): Remove serve bridge re-export shims by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5955
* feat(core,cli,sdk): resume an interrupted turn without a synthetic "continue" message by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/5030
* ci: allow longer PR review timeout retries by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/5961
* feat(desktop): voice dictation in the desktop app by @qqqys in https://github.com/QwenLM/qwen-code/pull/5856
* feat(browser-ext): revive Chrome extension via daemon-direct architecture by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/5777
* docs(telemetry): comprehensive documentation update to match current implementation by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5960
* ci(autofix): loosen issue candidate filters so the agent finds work by @qqqys in https://github.com/QwenLM/qwen-code/pull/5860
* feat(loop): inject a .qwen/loop.md task file at fire time via sentinels by @qqqys in https://github.com/QwenLM/qwen-code/pull/5890
* fix(cli): auto-select custom input on Enter in multi-select questions by @Alex-ai-future in https://github.com/QwenLM/qwen-code/pull/5791
* fix(core): only spawn memory recall when auto-memory is enabled by @Minerest in https://github.com/QwenLM/qwen-code/pull/5963
* fix(release): use relative postinstall patch dir by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/5973
* docs(qc-helper): add daemon mode docs and fix system settings path by @DragonnZhang in https://github.com/QwenLM/qwen-code/pull/5981
* fix(standalone): Route serve shim through cli-entry by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5977
* feat(channels): qwen tag — RFC + Phase 0 (multiplayer channel-resident agent) by @qqqys in https://github.com/QwenLM/qwen-code/pull/5888
* fix(ui): display output tokens instead of cumulative API throughput for subagents by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5972
* fix(cli): Avoid ACP runtime preload on serve fast path by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5989
* fix(web-shell): prefer raw file diffs in tool output by @ytahdn in https://github.com/QwenLM/qwen-code/pull/5992
* [codex] fix daemon specialized model filtering by @callmeYe in https://github.com/QwenLM/qwen-code/pull/5993
* feat(channels): Add channel agent bridge abstraction by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5978
* fix(web-shell): improve follow-up suggestion handling by @ytahdn in https://github.com/QwenLM/qwen-code/pull/5996
* fix(ci): cover release integration regressions by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/5994
* feat(core): add --insecure flag to skip TLS verification for self-signed endpoints (#3535) by @TianYuan1024 in https://github.com/QwenLM/qwen-code/pull/5962
* fix(core): filter thought parts from Stop hook last_assistant_message by @DennisYu07 in https://github.com/QwenLM/qwen-code/pull/6009
* feat(cli): add --safe-mode flag to disable all customizations for troubleshooting by @DennisYu07 in https://github.com/QwenLM/qwen-code/pull/4943
* fix(cli): fix thought viewer truncation, layout gaps, and choppy scrolling in VP mode by @chiga0 in https://github.com/QwenLM/qwen-code/pull/6002
* fix(cli): Guard serve fast-path bundle closure by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/5995
* feat(daemon): support @extension mentions by @callmeYe in https://github.com/QwenLM/qwen-code/pull/6008
* test(ci): stabilize cron interactive release check by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/6016
## New Contributors
* @Minerest made their first contribution in https://github.com/QwenLM/qwen-code/pull/5963
* @TianYuan1024 made their first contribution in https://github.com/QwenLM/qwen-code/pull/5962
**Full Changelog**: https://github.com/QwenLM/qwen-code/compare/v0.19.3...v0.19.3-nightly.20260630.e00fe6a27