[loopx-agent]

Goal And Delivered Outcome

A reader checking the conclusion of a saved case study had to expand its scope and then open a second disclosure to see any citations. Now one expansion shows the cited originals first, with the observation window and version in the same disclosure. The conclusion and study outline still lead the page.

Intended base: main.

Author Declaration

• Written by: model_agent, OpenAI Codex.
• Implemented against: Research Workspace, which keeps sources and scope one expansion away from a saved finding.

Criterion	Disposition	Validation
Expose the conclusion citations and scope together	Implemented	Paired desktop and narrow captures after one expansion
Keep version and source boundaries intact	Preserved	Both saved versions opened their own cited range; the wheel's HTTP owner rejected a wrong range and analysis key
Keep the conclusion readable before opening details	Preserved	First viewport and eight-section populated study checked at both widths

Scope And Continuation

The change is limited to the case reading component. Section citations still use their own disclosure and the same source reader. No model runs, source registrations, or extra requests occur when a reader expands the conclusion. Separate source error improvements in #8384 can integrate independently.

Validation

• Tested revision: 6b08201e3613e69ef27bbe83e637ebcd0f35e401.
• Run state: finished.
• Input classes: synthetic, authorized_private_read_only.

Check kind	Result	Evidence
static	passed	TypeScript, dashboard CSS checks, wheel asset manifest, and local premerge with verified change-quality receipt
unit	passed	Existing research analysis read smoke
real_entrypoint	passed	Exact candidate wheel, isolated production Chat HTTP owner, synthetic saved studies in two versions at 1200px and 390px; switching versions retained their exact source text
regression_parity	passed	Invalid range and a mismatched version key returned HTTP 409. After one injected HTTP 503, closing the panel returned focus to the original citation and reopening it read the same range. Source records stayed unchanged and the fixture observed zero mutations.

The packaged candidate was validated in isolation. The installed app was used to reproduce the existing journey; it has not been updated to this commit.

Frontend / Visual Evidence

The before and after captures use the same public-safe synthetic study at both widths: two saved versions, eight sections, and five conclusion citations, including an unknown outcome. Baseline: 71190b6b4ef4e146f3bf4a10d31ce61ac42fd021 compiled Chat bundle. Candidate: 6b08201e3613e69ef27bbe83e637ebcd0f35e401 wheel. The collapsed first viewport keeps the conclusion and outline; after a single expansion the candidate shows the evidence before the scope metadata.

Viewport	Candidate first viewport
Desktop
390px

After one expansion	Before	Candidate
Desktop
390px

Type of Change

• Bug fix

LoopX Area

• Dashboard

Boundary Checklist

• Screenshots and description contain public-safe evidence.
• The commit includes a DCO sign-off.

[loopx-agent]

Reviewer: model_agent · gpt-6.1-sol · OpenAI · runtime_reported · max

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

动机 / Motivation

用户读到已保存的 Case 结论，想打开对应原文核查依据。

旧界面先展开“范围与来源”，只能看到范围说明，还要再展开“依据与原文”才出现引用按钮。新版点一次“结论依据与范围”就能看到引用，观察窗口和版本也在同一处。

同一份八节、五条结论引用的合成解读在桌面与 390px 下，旧版展开一次显示零条，新版显示五条；版本切换后仍读取各自的原文。

此 PR 改变证据入口的位置，不重新分析案例，也不改变来源、权限或原文读取。 候选只在隔离包验证，尚未更新本机；独立评审与正式采用仍待确认。

改动思路 / Approach

结论与章节可以共享引用按钮，但文档来源、版本和失败恢复继续由既有阅读器负责。删除重复展开就能完成改动，无需新增数据或路由。 本次只改已保存案例的呈现；仍由原引用接口按当前 Goal、解读版本和行号返回原文。

对照 Research Workspace（docs/development/design.md#research-workspace--研究工作台，基线 71190b6b4ef4e146f3bf4a10d31ce61ac42fd021），检查 case-evidence-one-expansion：首屏先读结论、一次展开看依据与范围。只改标题不能消除原来的第二次展开；默认铺开全部引用又会挤掉正文。因此保留关闭状态，把原来的引用列表移入同一折叠。

具体改动 / Changes

评审提交为 6b08201e3613e69ef27bbe83e637ebcd0f35e401；远端 diff 与本地验证内容一致。

关键代码讲解

• ResearchCaseStudy 仍把结论、历史来源提醒和原文语言放在相同位置。展开标题后先列原文，再显示作者分析声明、观察窗口和版本；不额外发请求。
• citationLinks 是结论和各节共享的引用列表。每个按钮沿用原来的来源 ID、起止行和触发按钮，仍由现有面板读取。它不从文件名推断证据类型。
• .case-citation-links 给引文保留原来的点击高度和长文本换行。章节自己的“依据与原文”入口仍保留；只有结论下的双层展开被移除。

对主干的风险 / Risks

最需防的是切换版本后拿到另一份解读的引文，或者来源短暂失败后丢掉阅读位置。隔离 wheel 接真实 HTTP owner 时，两版读取各自的原文；错版本键和越界行号均拒绝。故意让同一个引文读失败一次，按 Escape 退回按钮，再打开读到了同一版本的正确行。观测到零状态写入。常规构建、静态检查、原阅读 smoke 和本地预合并通过；按当前策略未查看远端 CI。

我的整体评价 / Verdict

两种宽度的首屏都保留结论和正文目录。展开后原文入口排在版本元数据前；窄屏里版本信息可继续向下阅读，没有挤压正文。 视觉检查使用隔离包和合成资料；正式桌面宿主安装后仍需读回。 本改动保留已保存结论和原文之间的出处关联，删掉了实际多余的一步，适合独立集成。此评论为作者复核，仍需独立评审。

English verdict: APPROVE the bounded saved-case reading change at the reviewed commit; independent approval and installed readback remain open.