fix: keep question order and streamline contextual help
GoodBuddy Agents / Validate Agent source (push) Canceled after 0s
GoodBuddy Agents / Build darwin arm64 Agent (push) Canceled after 0s
GoodBuddy Agents / Build linux arm64 Agent (push) Canceled after 0s
GoodBuddy Agents / Build linux x64 Agent (push) Canceled after 0s
Cross-platform packages / Validate source (push) Canceled after 0s
Cross-platform packages / Authorize desktop packaging (push) Canceled after 0s
Cross-platform packages / linux arm64 (push) Canceled after 0s
Cross-platform packages / macos arm64 (push) Canceled after 0s
Cross-platform packages / windows arm64 (push) Canceled after 0s
Cross-platform packages / linux x64 (push) Canceled after 0s
Cross-platform packages / macos x64 (push) Canceled after 0s
Cross-platform packages / windows x64 (push) Canceled after 0s
Cross-platform packages / Publish GitHub and OSS release (push) Canceled after 0s

Keep interactive questions and answered cards at their original position in
conversation history, including remote reception, recovery, and reopening.
Preserve legacy history placement without a database migration.

Move supplementary guidance into shared accessible contextual-help popovers
across pages and settings. Add keyboard, modal, accessibility, and Electron
regressions, plus bounded opt-in live canvas-analysis validation.

Remove unused Agent, Runtime, and in-memory graph-search helpers, obsolete
mocks and browser copy; reuse the existing package-size formatter. Keep
production execution paths, protocol schemas, and released data migrations.

Validation: targeted cleanup tests, typecheck, lint, shared Linux Host
attach/read/re-attach, and added help/Electron checks passed. Final full-suite
verification passed 4,858 tests with 67 skipped and zero failures (412 files
passed, 9 skipped). Focused help checks passed 54 tests, including Electron
style comparisons and Chromium accessibility descriptions. Details and
validation boundaries are recorded in feature docs.

Release note: 修复交互式问答在回答后排到后续输出末尾的问题,重新打开会话仍保留提问位置。页面和设置中的补充说明改为标题或字段旁的按需帮助,支持鼠标与键盘操作。
This commit is contained in:
mesalogo
2026-09-20 01:26:36 +08:00
parent bf1d34e897
commit 9836a4cc9d
89 changed files with 1961 additions and 1111 deletions
@@ -359,6 +359,10 @@ Runtime 包。2026-09-19 完整桌面验证覆盖联合包安装、CN 项目创
未发布进程复用源码补齐 Desktop 重启后的多任务并行附加、实时待答快照及 Main 答案
持久化,详见[恢复与终态投影](../assistant-workbar/runtime-process-reuse-technical-design.md#63-恢复与终态投影);
不把该待答投影当作 durable transcript 或旧版本已经支持的行为。
Desktop Main 接收实时问题时保存其消息位置,回答后不移动到后续输出之后;实时待答
快照不伪造语义来源。位置去重、旧会话边界见
[Runtime 交互边界](../assistant-workbar/runtime-interactions.md#运行与失败),
当前源码的真实 Host 证据见[问答位置持久化](../assistant-workbar/progress.md#2026-09-19-问答位置持久化)。
- ACP 工具结果同时支持文本 `content` 和 `rawOutput.output`,移除 OpenCode 的
`<task><task_result>` 外层包装后渲染正文。恢复已有子代理时使用 ACP 所带的所属
`toolCallId` 继续路由,不等待新的父 Task metadata,也不重放模型请求。
@@ -851,6 +855,23 @@ goodbuddy-agent diagnostics --installation-id <installationId>
信任条件。上传到 Host 前重新序列化为确定性 LF JSON。签名目录、描述符、manifest 和
payload 仍按原始签名字节及 SHA-256 验证。
## 2026-09-20 冗余代码清理验证
- 删除没有生产调用的结构化命令 runner、两个进程身份辅助函数、占位 peer provider、
平台查询辅助函数和五个旧 Runtime 请求构造/摘要辅助函数。保留实际 CLI、进程检查器、
请求 schema、原生 peer 校验及包签名验证,不改变 Ask/Execute 或协议行为。
- 清理后的当前源码 Agent、Main attach transport 和 protocol client 在共享 Linux x64
Host 上通过固定 Host Key 连接,完成启动、状态/能力读取、只读工作区打开及真实文件读取,
断开后再次连接并重复读取。验证没有使用模拟 Agent 或旧安装包的 Agent 实现。
- 本轮测试进程剩余 0;专属 Host 目录、上传文件及本地隔离 profile 已删除。
文本和图片模型请求均为 **0 次**。这是未使用代码删除后的启动/连接验证,不是新增的
Runtime 模型调用、安装更新或其他架构验收。
- 临时证据为 `redundancy-cleanup-host-{build,run}.cjs` 与
`redundancy-cleanup-host-live.log`;定向回归 230 项通过,typecheck、lint 通过。
- 同期全量 `npm test` 为 4,844 通过、67 跳过、11 失败;失败来自并行修改中的五个
Renderer 帮助提示/焦点测试套件。更新后的五个套件连同更新设置大小展示回归共 64 项
定向复跑全部通过;未把这次复跑记为又一轮全量通过。
## 发布前完整回归(不能替代开发验证)
以下清单只用于复核开发期间已经在真实 Host 上通过的链路,不得作为 Agent 改动的首次