wicm84266964 · wicm84266964 · Sep 10, 2026 · Sep 10, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -7,36 +7,102 @@
 Published 2026-09-10 09:14:36 UTC (2026-09-10 17:14:36 UTC+08:00): [v2.0.11](https://github.com/wicm84266964/Ant-Code/releases/tag/v2.0.11).
 Release commit: `474930e346a48f0ea919f0aaa1839498b067efc9`. PR #51 and main CI passed; local release verification passed 1,331 unit/integration tests and 38 browser tests, with no failures or skips. Additional process tests passed; cross-repository parity verified 27 exact paths and 13 reviewed differences.
 
-- Text-layer PDFs are answered from extracted text; scanned PDFs render the first few full pages for vision and cite those page numbers.
-- Full visual PDF reading is explicit: set start and end pages on the paperclip and resend. PNG pastes without MIME are classified by extension.
-- Joining a background subagent group can strengthen wakeup (`wakeParent` OR, `waitForGroup` all > any > none) but cannot disable an existing wakeup request.
-- Windows executable distributions include `@napi-rs/canvas` and the Windows x64 native binary beside the exe.
+This is a small Dashboard document and background-task release on the 2.0 TypeScript runtime.
+Composer PDFs with a text layer are answered from extracted text. Scanned PDFs render the first few full pages for vision and cite those pages. Full visual reading is explicit. Background subagent groups merge wakeup settings so a later disable cannot cancel an existing wakeup. Permission mode ids are unchanged.
 
-Validation, upgrade steps and limitations: [Chinese release note](docs/releases/2.0.11-pdf-scan-vision-wakeup_zh.md).
+### Added
+
+- Scanned paperclip PDFs render the first two full pages (text, vectors, and embedded images) for vision. Text-layer PDFs stay on extracted text.
+- PDF chips accept optional start and end pages. Empty fields use the default window; filling both pages is the explicit full visual read.
+- PNG pastes without a MIME type are classified by file extension.
+
+### Fixed
+
+- Joining a background subagent group merges `wakeParent` with OR and `waitForGroup` as all > any > none, so dispatch order cannot disable an existing wakeup.
+- Sparse PDF text detection no longer counts synthetic page headings as body text.
+
+### Upgrade
+
+```sh
+git pull
+npm ci
+npm run verify:install
+npm link
+ant-code --version
+```
+
+`ant-code --version` should print `2.0.11`. Restart a running Dashboard
+and hard-refresh the browser. Gateway config and `.lab-agent` sessions
+do not need to be recreated. Windows exe distributions now include
+`@napi-rs/canvas` and the Windows x64 native binary; do not copy only
+the exe.
 
 ## 2.0.10 - 2026-09-10
 
 Published 2026-09-10 03:38:31 UTC (2026-09-10 11:38:31 UTC+08:00): [v2.0.10](https://github.com/wicm84266964/Ant-Code/releases/tag/v2.0.10).
 Release commit: `e51b203038a44d72575c99d0bf87a9ab98a7c8eb`. PR #49 and main CI passed; local release verification passed 1,312 unit/integration tests and 37 browser tests, with no failures or skips. Additional process tests passed; cross-repository parity verified 23 exact paths and 6 reviewed differences.
 
-- Summarize older tool results in bounded model batches after the full request reaches its configured context budget; target a dynamic 10-20% reserve.
-- Retrieve pre-summary tool text through session-scoped `tool_result_read`, with existing transcript encryption, redaction, retention and deletion policies.
-- Distinguish tool summaries from conversation compaction in Dashboard and TUI; tool summaries no longer insert conversation boundaries.
-- Preserve original text on summary failure, cancellation or insufficient savings; protect recent results and error diagnostics.
+This is a small context-window release on the 2.0 TypeScript runtime.
+After the full request reaches its configured budget, older tool results are summarized in bounded model batches instead of a fixed character clip. Tool summaries are distinct from conversation compaction. Permission mode ids are unchanged.
 
-Validation, upgrade steps and limitations: [Chinese release note](docs/releases/2.0.10-tool-summary-budget_zh.md).
+### Added
+
+- Bounded model-batch summaries of older tool results, targeting a 10-20% context reserve.
+- `tool_result_read` retrieves pre-summary tool text through a session-scoped evidence reference.
+
+### Fixed
+
+- Tool summaries no longer insert conversation compaction boundaries in Dashboard and TUI.
+- Original tool text is kept when a summary fails, is cancelled, or does not save enough space.
+- Config replacement retries transient file sharing on Windows.
+
+### Upgrade
+
+```sh
+git pull
+npm ci
+npm run verify:install
+npm link
+ant-code --version
+```
+
+`ant-code --version` should print `2.0.10`. Restart a running Dashboard
+and hard-refresh the browser. Gateway config and `.lab-agent` sessions
+do not need to be recreated.
 
 ## 2.0.9 - 2026-09-10
 
 Published 2026-09-09 17:37:09 UTC (2026-09-10 01:37:09 UTC+08:00): [v2.0.9](https://github.com/wicm84266964/Ant-Code/releases/tag/v2.0.9).
 Release commit: `3f1a49b736c1f6ec1fbeba2d7854553382865f64`. PR and main CI passed; local release verification passed 1,304 unit/integration tests and 37 browser tests, with no failures or skips.
 
-- Preserve tool evidence until the complete request reaches its context budget; add continuous file excerpts and byte-budgeted search pagination.
-- Group Dashboard models by source URL and select one active credential per source while preserving each model's agent routes.
-- Require remarks for new same-source duplicate models and show readable remarks instead of internal configuration IDs.
-- Show configured idle gateways as connected, with real connection errors still reported.
+This is a small reliability and Dashboard model-source release on the 2.0 TypeScript runtime.
+Tool evidence stays until the full request reaches its context budget. Dashboard groups models by source URL and selects one active credential per source. Permission mode ids are unchanged.
+
+### Added
+
+- Continuous file excerpts with original line numbers, plus byte-budgeted search and directory pagination.
+- Dashboard model picker groups connections by source URL. New same-source duplicate models require a remark.
+- Per-source active credential selection, saved as `lab.sourceCredentialSelections`.
+
+### Fixed
+
+- Tool results are no longer stubbed before the complete request reaches its context budget.
+- Permission confirmation stays on screen and queues instead of being clipped by the composer.
+- Windows storage lock release retries transient sharing violations.
+
+### Upgrade
 
-Release candidate validation and limitations: [Chinese release note](docs/releases/2.0.9-context-source-credentials_zh.md).
+```sh
+git pull
+npm ci
+npm run verify:install
+npm link
+ant-code --version
+```
+
+`ant-code --version` should print `2.0.9`. Restart a running Dashboard
+and hard-refresh the browser. Gateway config and `.lab-agent` sessions
+do not need to be recreated.
 
 ## 2.0.8 - 2026-09-06
 

diff --git a/docs/releases/2.0.10-tool-summary-budget_zh.md b/docs/releases/2.0.10-tool-summary-budget_zh.md
@@ -2,7 +2,7 @@
 
 发布日期：2026-09-10
 
-这是面向长编程和数据分析任务的小版本更新。主会话达到配置的请求预算后，较旧工具结果改用模型批量摘要，替代固定字符裁剪；同时区分工具摘要和历史对话摘要，减少连续“聊天内容已压缩”提示造成的误解。本说明为发布候选记录，线上 CI、标签和 Release 状态以实际发布记录为准。
+这是 2.0 TypeScript 运行时上的小版本。主会话达到配置的请求预算后，较旧工具结果改用模型批量摘要，替代固定字符裁剪；同时区分工具摘要和历史对话摘要，减少连续“聊天内容已压缩”提示造成的误解。权限模式 id 不变。本说明处于公开发布候选阶段，不表示 CI、tag 或 GitHub Release 已经完成。
 
 ## 对用户工作流的直接变化
 
@@ -23,9 +23,17 @@
 
 ## 升级与产物位置
 
-需要 Node.js 22.18+。源码安装更新后执行 `npm ci`、`npm run verify:install`；使用全局链接的用户可执行 `npm link`。`ant-code --version` 应显示 `2.0.10`。完整重启 TUI/Dashboard，浏览器刷新以加载当前资源。
+需要 Node.js 22.18+。从源码更新：
 
-已有配置和会话继续使用；历史上已裁剪或摘要丢失的内容不会自动恢复。新建的工具证据归档随对应会话清理，无需手动搬迁。审计产物仍由发布检查写入既有审计目录。
+```sh
+git pull
+npm ci
+npm run verify:install
+npm link
+ant-code --version
+```
+
+`ant-code --version` 应显示 `2.0.10`。若以前 link 过旧安装，需要重新 `npm link`。正在跑的 Dashboard 需要重启并硬刷新浏览器。现有网关配置和会话不必重建。历史上已裁剪或摘要丢失的内容不会自动恢复。新建的工具证据归档随对应会话清理，无需手动搬迁。
 
 ## 验证、限制与待完成门禁
 

diff --git a/docs/releases/2.0.11-pdf-scan-vision-wakeup_zh.md b/docs/releases/2.0.11-pdf-scan-vision-wakeup_zh.md
@@ -2,7 +2,7 @@
 
 发布日期：2026-09-10
 
-这是面向文档阅读和后台子任务编排的小版本更新。回形针附上的 PDF 不再默认逐页全量视觉识别：有文字层时直接抽文本，扫描件先完整渲染前几页参与对话；全量视觉改为明确选择。同时修复同组后台子任务唤醒参数互相覆盖、主控不再接续的问题。本说明为发布候选记录，线上 CI、标签和 Release 状态以实际发布记录为准。
+这是 2.0 TypeScript 运行时上的小版本。回形针附上的 PDF 不再默认逐页全量视觉识别：有文字层时直接抽文本，扫描件先完整渲染前几页参与对话；全量视觉改为明确选择。同时修复同组后台子任务唤醒参数互相覆盖、主控不再接续的问题。权限模式 id 不变。本说明处于公开发布候选阶段，不表示 CI、tag 或 GitHub Release 已经完成。
 
 ## 对用户工作流的直接变化
 
@@ -24,9 +24,17 @@
 
 ## 升级与产物位置
 
-需要 Node.js 22.18+。源码安装更新后执行 `npm ci`、`npm run verify:install`；使用全局链接的用户可执行 `npm link`。`ant-code --version` 应显示 `2.0.11`。完整重启 TUI/Dashboard，浏览器刷新以加载当前资源。
+需要 Node.js 22.18+。从源码更新：
 
-已有配置和会话继续使用。审计产物仍由发布检查写入既有审计目录。
+```sh
+git pull
+npm ci
+npm run verify:install
+npm link
+ant-code --version
+```
+
+`ant-code --version` 应显示 `2.0.11`。若以前 link 过旧安装，需要重新 `npm link`。正在跑的 Dashboard 需要重启并硬刷新浏览器。现有网关配置和会话不必重建。Windows 可执行分发包会附带 `@napi-rs/canvas` 及 Windows x64 原生组件，不能只复制 exe。
 
 ## 验证、限制与待完成门禁
 

diff --git a/docs/releases/2.0.9-context-source-credentials_zh.md b/docs/releases/2.0.9-context-source-credentials_zh.md
@@ -2,7 +2,7 @@
 
 发布日期：2026-09-10
 
-这是 2.0 TypeScript 运行时的小版本改进，面向跨文件编程、多步数据分析和 Dashboard 多模型使用。减少工具结果过早移除造成的重复读取，将来源凭据与模型配套路由分开管理，并修复权限确认框可见性。本说明是公开发布候选说明，不预先宣称 CI、tag 或 GitHub Release 已完成。
+这是 2.0 TypeScript 运行时上的小版本。减少工具结果过早移除造成的重复读取，将来源凭据与模型配套路由分开管理，并修复权限确认框可见性。权限模式 id 不变。本说明处于公开发布候选阶段，不表示 CI、tag 或 GitHub Release 已经完成。
 
 ## 对用户工作流的直接变化
 
@@ -34,7 +34,7 @@ npm link
 ant-code --version
 ```
 
-版本输出应为 `2.0.9`。完整重启 TUI/Dashboard；浏览器强制刷新以加载更新的前端资源。已有配置与会话可继续使用，但旧会话已被压缩的正文不会自动恢复。首次使用多 Key 来源时检查设置页的生效凭据。无新增研究数据输出目录或需要搬迁的数据文件；审计产物仍由发布检查写入既有审计目录。
+`ant-code --version` 应显示 `2.0.9`。若以前 link 过旧安装，需要重新 `npm link`。正在跑的 Dashboard 需要重启并硬刷新浏览器。现有网关配置和会话不必重建。旧会话已被压缩的正文不会自动恢复。首次使用多 Key 来源时检查设置页的生效凭据。
 
 ## 验证、限制与待完成门禁