Skip to content

Distinguish prepared research input from analyzer judgments - #8317

Merged
loopx-agent merged 1 commit into
mainfrom
codex/research-judgment-input-state-20261108
Nov 7, 2026
Merged

loopx-agent merged 1 commit into
mainfrom
codex/research-judgment-input-state-20261108

Conversation

@loopx-agent

Copy link
Copy Markdown
Collaborator

A Goal with prepared analysis input now shows that no analyzer judgments are available yet. The input count and sampling remain in scope help. Saved judgments still show their coverage, and original measurements stay visible. The analysis panel uses a compact input label and preserves Escape after a successful read retry.

First viewport

Packaged Chat at base e58ba30 and head f06fccb reads the same synthetic source through the read-only HTTP owner. Desktop is 1280 × 900 and the narrow view is 390 × 844.

Previous desktop Proposed desktop Proposed narrow
Prepared input count beside research results Judgment status and original measurements Judgment status in the narrow view

Validation

  • The full packaged research browser scenario covers prepared input, saved judgments, original measurements, a failed read and successful retry, keyboard dismissal, citations and scope isolation.
  • Candidate assets also read two live linked sources in a separate browser session. Both showed one prepared input and no analyzer judgments. The original records stayed visible in their own scope, with zero write requests.
  • Chat packaging, the research analysis read smoke and standard local premerge passed. The navigation assertion now renders the workbench through its public entry point.

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

打开已有 Goal 研究页的读者,需要判断页面能否提供分析结论。 同一份只读样本已有一个输入片段,旧版把计数放在研究结果旁,看起来像已有判读。原生测量已经可读,分析结果此时仍为空。 新版直接写明「尚无分析器判读」,在帮助里保留输入片段数和抽样范围。 此提交处理研究页的判读状态、分析面板的输入标签和读取回归。 科学结论仍需来自授权分析和可复核证据,已安装桌面的读回在合并后验证。

改动思路 / Approach

同一个分析包已经提供输入范围和可选判读。直接使用现有结果字段即可区分状态,另存一份完成标志会带来同步和失败恢复成本。 两个既有 TypeScript 读取组件调整展示,继续从已有 HTTP 来源取证和重试。 只改帮助无法消除首屏歧义。直接从同一份分析包读取可选结果,可以分别说明已登记输入、已保存判读和已采集的原生结果。打开研究页仍只读取。分析仍通过原有明确入口发起。

具体改动 / Changes

依据 docs/development/design.md 的 Research Workspace,规格版本为 e58ba30。research-judgment-availability 在正常入口区分判读状态和原生记录,已用同一只读来源对照首屏。research-read-authority 保留原来选中 Goal 的来源验证和按需分析,已用打包浏览器和两个真实来源检查。 Research Workspace

关键代码讲解

  • ResearchProgress 依据原分析包的可选结果渲染。待分析时标题旁显示明确状态,片段数和抽样规则移入帮助。已保存判读继续显示覆盖范围和未校准标记,原生测量由自己的分支显示。 ResearchProgress
  • ResearchAnalysisPanel 用「输入片段」给尚未分析的范围命名,减少重试后提示按钮位置的变化。已保存结果保留原范围标签,重试、关闭和执行权限仍由既有控制器处理。 ResearchAnalysisPanel
  • 既有浏览器用例断言原生测量与判读状态独立可读、已有判读仍标明覆盖,读取失败后重试和一次 Escape 关闭有效。读取 smoke 改走公开的工作台入口。 浏览器回归 工作台读取 smoke

对主干的风险 / Risks

主要风险是把待分析误画成零收益,跨来源显示旧判读,或者文案撑宽后改变关闭行为。真实来源错传命名空间返回 400,响应没有分析包。较长的候选标签曾在重试后让范围提示消耗第一次 Escape,短标签和新增断言已在完整打包用例中复验。打包构建、读取 smoke、完整研究浏览用例和标准本地预合并均通过。浏览两个真实来源时记录到零写请求。

语义与 CI 对齐

沿用已有 pack、result 和 source namespace。状态文案仅参与用户界面展示。按此 Goal 配置,评审证据使用本地验证。

我的整体评价 / Verdict

长期使用仍沿原来的真实 Goal、来源和恢复路径。打开研究页时更容易判断有没有现成的分析结论,原生证据仍可直接读。同一合成数据的前后整屏显示主标题与原生测量均可读。宽度缩到 390 px 时,判读状态留在标题旁,数值仍完整显示,页面没有横向溢出。 首屏图来自合成只读来源。两个本机真实来源已用候选打包资产读回,安装后的桌面仍待验证。 当前切片的改动与可复现的问题相称,适合合入同一提交。 前后首屏与验证

English verdict: APPROVE f06fccb. No blocking findings. Packaged Research, scope isolation, retry and standard local premerge passed.

@loopx-agent
loopx-agent merged commit 8a4d4f9 into main Nov 7, 2026
8 of 9 checks passed
@loopx-agent
loopx-agent deleted the codex/research-judgment-input-state-20261108 branch November 7, 2026 20:39
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