Skip to content

refine(research): consolidate contextual explanations - #7270

Merged
loopx-agent merged 1 commit into
mainfrom
codex/research-contextual-reading-20261025
Oct 24, 2026
Merged

loopx-agent merged 1 commit into
mainfrom
codex/research-contextual-reading-20261025

Conversation

@loopx-agent

@loopx-agent loopx-agent commented Oct 24, 2026 •

Copy link
Copy Markdown
Collaborator

研究页已采用按需说明,但结果标题仍并排显示“指标说明”和“尚未分析”,历史分析/测量提醒也各自游离在标题之外。现在每个所属标题只保留一个按当前状态命名的帮助入口;历史警示使用原有警告图标,完整口径和未核验原因仍可悬停、聚焦或点按读取。

Consolidate result state and interpretation in the existing heading help. Attach historical measurement/source warnings to their owning heading instead of adding standalone glyphs. Keep original values, unknowns, historical identity, source citations, model/cost consequences and recovery actions.

Validation

  • Tested 8a7b4d6ffe42472abd11d904522be4fe443bf8d5, based on bf21a2cf1f5783c34acf4c2796c6cb0205192731; the accepted design contract already requires one contextual entry per fact.
  • TypeScript noEmit, production Chat build and full packaged research-history passed. Existing report/subset/source return, Chinese/English, narrow reading, stale results and original evidence remain covered.
  • Same synthetic fixture through the real isolated HTTP/SQLite owner: frozen base renders two result-help controls and fails the one-entry expectation; head renders one. Source bytes are unchanged and mutation attempts remain zero.
  • Five desktop/narrow/expanded captures manually inspected, including metadata; all are synthetic. One intermediate test targeted the native-results section instead of the parent results heading; the corrected selector and independent before/after capture verify the actual count.
  • Exact-scope change-quality receipt recorded. Repository-native premerge and exact-head review remain separate merge gates; remote CI is not consulted under the configured policy.
  • This PR adds no model call, new state contract, settings or Agent instruction. Installed App readback is recorded below; captures do not claim researcher comprehension or scientific benefit.

Visual evidence

1280×900 and 390×844, same fixture. The final historical-source label refinement does not affect these current-source capture states; the final complete scenario covers the historical path.

Desktop before Desktop after
Before desktop After desktop
Narrow before Narrow after
Before narrow After narrow
Expanded contextual explanation

Complete explanation on focus

Installed readback

After merge, the native macOS App running source 6edc94ccdb4c46a7f461003e8eb9077fd1868d7b was read back: one result-help entry, complete explanation, original citation and Escape return were confirmed. Browsing started no new analysis. Public screenshots above remain synthetic; researcher comprehension and scientific benefit are not established by this check.

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 · max

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

动机

在真实 Goal 的研究页查阅测量、判读和来源的用户。 此前结果标题并排放置指标说明与尚未分析两个图标,历史提醒又游离在标题之外;现在状态和说明合为所属标题的一处入口,避免来回分辨。 桌面与窄屏先读原测量和未知值,再从同一个按状态命名的图标查看完整解释;历史警示仍有警告图标和完整原因。 不改原始数据、模型执行、计分、权限、分析范围和原文,也不把图标减少当作研究收益证明。 这次新补丁的已安装 App 回读尚待合入后验证;真实研究者理解和净效率收益未测量。

这次核对原任务要求的整页主次与恢复路径,而非把“文字已收进悬浮”当作完成。重复的小图标仍会让用户分辨说明归属;删除警示则会损失真实状态。当前选择保留读数与未知,以一个入口说明同一区域。无需新增配置、复制来源或重复确认,聚焦/点按属于自愿查证,不启动分析。

改动思路

复用 ContextHelp 与现有 typed applicability/result,把同一标题的帮助合并即可;不需要新组件、配置、存储或文字分类器。 当前 PR 仅收敛研究结果和原生测量标题的说明归属,保留完整定义、原值与执行恢复边界。

结果帮助的名字由现有 applicability 和保存结果是否存在派生;历史来源没有结果时不会叫作已有历史分析。原生测量独立使用其来源状态,在标题处保留历史名称与警告图标。原文、指标定义和不确定性没有删去;执行模型的按钮、单独用量后果、原分析/引用错误和重试仍各由既有owner负责。

具体改动

完整范围是一个既有生产展示文件、原完整浏览器场景和五张合成对照图,文本14增5减。关键代码为 apps/presentation/dashboard/src/features/personal-workspace/goal-research-progress.tsx:124 的 NativeResults 和同文件239行的 ResearchProgress:前者合并测量说明与历史原因,后者把当前结果状态和解释放在同一个标题帮助中。没有新增状态存储、通用组件、协议或全局CSS隐藏;原始数值、单位、来源范围、判读、报告与引用路径保持原值。

先验规范 docs/development/design.md 固定在 bf21a2cf1f5783c34acf4c2796c6cb0205192731。已有 explanations-on-demand 明确一个事实使用一个所属图标,populated-page-review 要求默认/展开及窄屏混合状态观察。两项均落实到源符号、完整场景和同输入截图;没有为当前实现重写规范。基线与head均通过隔离的实际HTTP/SQLite owner读取同一合成来源。

对主干的风险

主要风险是少一个图标却丢失历史或未知含义。当前完整 packaged research-history 检查仍覆盖原值、unknown、历史来源、完整解释、原文Escape、报告返回、中英文和390px;来源不变且写入尝试为零。TypeScript与最终Chat构建通过,标准premerge的4项直接及13项选中检查通过;CQR cqr_161b1443d66c9b0202ed 有效。5张图片逐张核对像素及元数据,premerge原始manual_review_required照实保留,没有将人工门称为自动全绿。未读取或等待远端CI。

旧包在独立“一处所属帮助”断言下实际得到2而失败,head得到1并通过。最初测试选错为原生测量子区而得到0的设置失败另行保留,并非产品回归证据;修正后完整场景与独立基线/head捕获共同覆盖该不变量。无后台写入、模型执行或新权限,非synthesis/错误读取和原引用恢复仍沿原实现。

我的整体评价

APPROVE,交付判断 justified_increment,long_horizon preserved。user_experience improved 仅指实际观察到的说明归属、重复入口和恢复体验,不宣称研究者理解、模型效果或token净收益已证明。

复用 ContextHelp 与现有 typed applicability/result,把同一标题的帮助合并即可;不需要新组件、配置、存储或文字分类器。 当前 PR 仅收敛研究结果和原生测量标题的说明归属,保留完整定义、原值与执行恢复边界。 已观察相同合成记录的桌面、390px默认与展开页面:原生测量和未知值不变,结果标题的双图标合为一个;完整解释在聚焦后可读,Escape回到原入口。 未把合成截图当作已安装 App、新分析质量、研究者理解或量化效率的验证。

这次相邻的有界整理符合简洁风格,保持用户真正的工作和证据可读;没有为局部展示变化引入框架或新数据owner。正式安装回读由同一产品任务继续,不能由截图、合并或CQR替代。结论只对应 8a7b4d6ffe42472abd11d904522be4fe443bf8d5;head改变须重新核验。

English verdict: APPROVE — 8a7b4d6ffe42472abd11d904522be4fe443bf8d5. Existing heading help now combines state and interpretation, retaining historical warnings, complete definitions, original values and recovery. Full packaged reading, type/build and native checks passed; five synthetic desktop/narrow/expanded images were inspected. No new model call, persisted state or authority. Installed adoption and comprehension remain separate qualifications.

@loopx-agent
loopx-agent merged commit 6edc94c into main Oct 24, 2026
24 of 30 checks passed
@loopx-agent
loopx-agent deleted the codex/research-contextual-reading-20261025 branch October 24, 2026 16:26
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