Skip to content

Show the full retained strategy chronology in research - #8472

Merged
loopx-agent merged 1 commit into
mainfrom
codex/research-strategy-chronology-20261110
Nov 10, 2026
Merged

loopx-agent merged 1 commit into
mainfrom
codex/research-strategy-chronology-20261110

Conversation

@loopx-agent

Copy link
Copy Markdown
Collaborator

Opening strategy history used to immediately replace the full-history selection with the latest record. A source with no Explore events showed only a recent strategy excerpt. Research now opens a chronology of every retained strategy version, with the recorded time span, original observations and one-step access to each complete original. Sources whose Explore read succeeds with zero records show it directly; ordinary Explore sources keep their existing default and load strategies on request.

The existing native source owner, physical namespace and exact-version identity remain authoritative. Unknown times stay visible. A failed refresh preserves the selected snapshot and offers retry. Strategy plans remain distinct from demonstrated execution, adoption and measured benefit. The adjacent obsolete excerpt prop and CSS are removed.

Validation at 4d56f8e81dc165b0bf1c1f95cca7a9af6863c6fb: packaged Chat build/typecheck; real-HTTP browser smoke across 23 synthetic versions and a 15-hour span, an unknown timestamp, source switching, exact-version reload/Back/Forward, missing versions, refresh recovery, keyboard focus, reduced motion and 390px; 31 linked-source/runtime-workspace tests; standard premerge including the bilingual RFC check; all seven changed paths pass the public-boundary scan. Additional original-source readback passed without source-file changes. Installed desktop adoption remains unverified. CI waiting is disabled by the existing delivery policy.

Visual evidence uses synthetic records only, packaged frontend plus its real read-only HTTP backend. The overview shows the whole recorded span and readable observations first; complete prose, negative results, unresolved questions and exact provenance remain one selection away. Refresh failure retains the document at narrow width. These are review captures; the artifact branch is not part of the merge diff.

Before (previous reader, revision 2c0cbdf4b):

Previous latest-record default

After (tested revision 4d56f8e8), desktop overview:

Complete strategy chronology

390px overview:

Narrow strategy chronology

Default for a source with no Explore records:

Native strategy chronology replaces the empty reading area

Expanded original and refresh recovery at 390px:

Retained original during a failed refresh

Signed-off-by: LoopX Agent <337587101+loopx-agent@users.noreply.github.com>

@loopx-agent loopx-agent left a comment

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reviewer: model_agent; gpt-6.1-sol; OpenAI; runtime_reported; reasoning_effort=max.

Approval conclusion (author-owned PR; GitHub blocks formal self-approval)

动机 / Motivation

在研究工作台回看原 Goal 策略历史的用户。原先打开策略历史会直接跳到最近一份;现在先看到全部记录的时间跨度、目标和观察,并可一次打开完整原文。已验证跨十五小时的二十三份策略全部保留,缺时间记录、旧版本和读取失败均可回看。本次只改善策略记录的阅读,不重建缺失的 Explore 事件、不证明计划已经执行,也不改变实验权限。原生桌面 App 的安装采用及完整研究图的累计验收仍未完成。

改动思路 / Approach

复用既有完整来源读取器和版本标识,仅增加可派生的时间排列与阅读状态,不增加事实、评测或执行 owner。当前 PR 交付可完整回读的策略总览、旧版本与失败恢复;实验执行、评分和桌面安装保持各自既有边界。原始 Goal、物理来源和完整历史仍由既有 GoalSourceReader 与 TS projection 负责,界面只消费它们。没有记录时间的项单列保留,空窗和相邻点不被解释为停工、继承或因果;点击后仍进入原来的完整记录阅读器。相比仅增大最近摘要,这个切片让用户真正找到之前和后来的判断;相比另建历史服务,它省去同步与第二权威源。普通 Explore 路径没有提前读取策略,只有成功完整读到零条 Explore 的来源会直接显示总览。

具体改动 / Changes

审阅依据为 docs/development/frontend-delivery.md,固定版本 5988be01e9746828fbb3d222ef22183605a77185;同时核对同版本 docs/development/design.md 的 Populated Page Review。文档本次扩展总览默认,不用修改后的文字反向证明旧规范。

  • R1:完整读取并保留语言、负结果和未解问题。总览包含全部23份;长摘要可一次打开完整原文,未用模型改写。
  • R2:旧版本链接、reload、Back/Forward 与缺失版本恢复继续按原始标识工作;找不到的版本保持缺失,不换成最近一份。
  • R3:沿原 scoped GET 读取,源 Goal 与 namespace 检查不变;浏览没有模型、实验和原状态写入。
  • R4:成功读完的空 Explore 来源仍明确显示这个事实,保留原任务入口,并直接展示完整策略历史。
  • R5:已看桌面、390px、长篇展开和刷新失败;完整记录、未知时间和失败不会为压缩首屏而消失。

关键代码讲解

GoalSourceStrategies 在 apps/presentation/dashboard/src/features/personal-workspace/goal-source-strategies.tsx:54 保留原 schema/来源检查,删除 all→latest 自动跳转;按原记录时间生成完整列表,未知日期保持可见。选择的是原 digest,完整正文复用原组件;返回总览会恢复到同一项的键盘焦点。

GoalResearchHistory(apps/presentation/dashboard/src/features/personal-workspace/goal-research-history.tsx:56)只在没有读取错误、加载完毕且总数确为零时嵌入策略总览;读取中、失败和普通 Explore 分支不被当作缺少历史。GoalResearchWorkbench(apps/presentation/dashboard/src/features/personal-workspace/goal-research-workbench.tsx:32)继续传原 Goal/来源和任务跳转,移除了已经无消费意义的最近摘要参数;对应旧 CSS 一并删除。新增样式只处理当前阅读层级,正文保持16px,说明放在现有帮助里。

对主干的风险 / Risks

最强反例是:空 Explore 被误判,或切换到同名 Goal 的另一个来源后仍显示旧记录。本次沿真实 HTTP 检查了完整/失败分支、原来源隔离与刷新恢复,未改 source producer、数据库或执行权限;31项既有 source/runtime 测试通过。新断言也在固定旧前端上实际失败:它找不到全部记录总览,重现了自动跳最近一份的旧行为;同一输入和 HTTP harness 在当前 head 通过。

PASS:打包构建及 TypeScript 检查、23份/15小时完整历史与未知日期、旧链接和导航、缺失版本、503重试、窄屏/键盘/reduced motion、标准 premerge 四项直接检查与四项目录检查、七个改动路径的公共边界扫描。现有真实只读来源也完成了原记录回看,原始文件保持不变。CI 按既有策略不查询也不等待。没有安装或重启原生 App;极大列表渲染成本和实际科研收益仍是明确未验证项。源数据没有新增迁移,回滚只需恢复这些前端改动。

我的整体评价 / Verdict

未发现阻断问题。这是 justified_increment:完整策略回读已形成可用旅程,完整研究图和桌面采用仍由既有工作继续。long_horizon 的改善限于累积记录不再被最近一份遮蔽,不能外推成实验得分提升;user_experience 在已测打包界面改善,已有意图不再增加设置或确认轮次。总览以时间跨度与原始观察为主;完整原文与溯源一次可选展开,16px 正文仍易读,失败提示与文档同时可见。原生桌面 App 的安装采用、极大历史的渲染性能,以及实际科研收益未在本次验证。

全量差异复用了同一原生 owner,未引入共享语义词表;删除旧摘要参数与样式属于已验证的同域简化。当前 4d56f8e81dc165b0bf1c1f95cca7a9af6863c6fb 与远端一致,代码量和维护面与问题相称。合并仍须执行独立的 readiness 和既有授权检查。

English verdict: APPROVE — head 4d56f8e81dc165b0bf1c1f95cca7a9af6863c6fb; no blocking findings. Packaged real-source reading, negative recovery, baseline regression sensitivity and required local validation passed; installed desktop adoption remains unverified.

@loopx-agent
loopx-agent merged commit e7ae4b0 into main Nov 10, 2026
29 of 40 checks passed
@loopx-agent
loopx-agent deleted the codex/research-strategy-chronology-20261110 branch November 10, 2026 03:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant