Skip to content

fix(frontend): reconcile correction evidence observations - #5669

Merged
huangruiteng merged 6 commits into
mainfrom
codex/l1-observation-consistency-20261005
Oct 6, 2026
Merged

fix(frontend): reconcile correction evidence observations#5669
huangruiteng merged 6 commits into
mainfrom
codex/l1-observation-consistency-20261005

Conversation

@loopx-agent

@loopx-agent loopx-agent commented Oct 5, 2026 •
edited
Loading

Copy link
Copy Markdown
Collaborator

A correction recheck could withdraw requester adoption in the trace while surrounding details retained earlier success, or leave the old report visible after core evidence failed. It also erased a valid source/Markdown comparison after a successful unchanged-version check. The App/Chat evidence panel now shares one checked observation and preserves reading choices by session, execution identity and exact source/output bindings. Downstream loss retains valid correction evidence; changed, missing or ambiguous selected versions withdraw the comparison, and core loss clears the evidence view until explicit recovery.

This reuses the existing delegation projection and local TypeScript comparison owner. It replaces array-index choice ownership with exact ref/hash bindings, preserving choices across reordering and unrelated observation changes. It adds no capability, protocol vocabulary, persisted receipt, execution or messaging authority. CLI and Lark contracts are unchanged. The original L1 acceptance remains open for a real objection, independent exact-version verification and requester adoption.

Validation on c226c945168e60cbbde495ad549df242e62de280, integrating main f97189ebf0cbefb747a428760ceeec0b531f6ae6:

  • The new packaged regression fails on the preceding 6505b1a5b49c351cff8f6fe6c3efdd528f61ff8e after an unchanged successful check and passes with the repair. It retains the selected Markdown output, source button and raw reading mode through successful recheck and downstream loss. Existing core-loss, recovery, keyboard return, mobile and pagination checks remain covered.
  • Official source bundle build/verification, the full packaged workspace entry (44 scenarios plus the correction-reader regression), comparison/typecheck, semantic advisory/full vocabulary validation and diff/privacy checks passed. The scripted steward journey still reports its semantic readiness/correction/recovery/return gaps. All six branch commits carry DCO sign-off; exact-head CI and independent review remain pending.
  • On the final source, actual disposable SQLite/Chat HTTP and native CLI reads match. Unchanged checks, consumer/core loss and exact-byte restoration pass; the typed choice rule retains or withdraws the corresponding real version bindings. Canonical Todos are unchanged, all four deterministic worker invocation counts remain one, and no live model was called. Full manual production-backed desktop/mobile keyboard evidence was collected on 6505; the final selection repair is qualified by packaged UI tests and real backend packet integration separately. The 27 focused backend/bundle tests passed on 6505; that backend is unchanged by the later reader repair.
  • The current first viewport, layout and essential actions are unchanged. Desktop/390px packaged screenshots were inspected. The related refactor removes duplicate observation authority and index-based choice ownership; no broader refactor is needed. Later main updates compose without conflicts, but are not part of this tested head.

This candidate is proposed and packaged, not installed or released. Full L1/G1 semantics and ordinary native-App qualification remain open. Leave merge to the maintainer.

中文：一次核验统一更新整份证据，同版本保留来源、Markdown 产物选择与原文阅读模式；核心证据失效撤回旧成功。完整打包回归与真实 SQLite/HTTP/CLI 版本读回、失效恢复通过。最终候选尚未安装，独立准确版本验收与真实采用仍未完成。

@mergify

mergify Bot commented Oct 5, 2026

Copy link
Copy Markdown

This pull request has merge conflicts with main and cannot be merged
until they are resolved. Please rebase or merge the base branch, @loopx-agent.

Choose the remote for the base repository, not an out-of-date fork.
For a fork clone, first inspect git remote -v; upstream must point
to https://github.com/loopx-project/loopx.git. If it is absent, add it
with git remote add upstream https://github.com/loopx-project/loopx.git.
Then run:

git fetch upstream
git rebase upstream/main
# Resolve each conflict, git add the resolved files, then git rebase --continue.
git push --force-with-lease origin HEAD

For a same-repository clone whose origin points to
https://github.com/loopx-project/loopx.git, use origin instead of
upstream for fetch/rebase. If you prefer merging the base, use
git merge <base-remote>/main and push normally.

Keep the DCO Signed-off-by trailer on every commit when you rebase.
https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

@mergify mergify Bot added the needs-rebase Mergify: the pull request has merge conflicts with its base branch label Oct 5, 2026
loopx-agent
loopx-agent commented Oct 5, 2026
View reviewed changes

@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 | model=gpt-6.1-sol | provider=OpenAI | runtime_reported | effort=xhigh

Request changes conclusion (author-owned PR; GitHub blocks formal self-review)

动机

在 Goal 对话中反复核验多成员纠偏报告的操作者，需要同时阅读版本对照和当前采用证据。旧版关联核验失败后，详情仍可能保留采用成功；新版把失效同步到整个证据视图，但成功核验同一版本也会清空已选的正文对照。打包页面的下游失联、核心失效和显式恢复通过；同输入对照证明 base 保留 Markdown 选择，head 清除选择，且没有启动模型。本次只核验按需证据展示；不认证真实异议、独立语义验收、真实请求方采用、安装升级或业务结论。

改动思路

复用既有 delegation readback；原产物、复核回应和修订先按 operation/request/Agent/Todo/ref/hash 核验，后续消费者再按采用回执核验。下游失联只撤掉采用，核心失效清除整份报告、basis 和关系；unavailable 是本地只读观察，没有改持久回执。删除 child 独立 error 状态是有价值的相关收敛，但 parent 新对象必须与仍有效的用户选择相容。

具体改动

完整 base b391abd635efe56fcdf575fee96a78295a3487ec → head dd4ce39e3a8f2ea95a9b857e1dca731485c193a7，8 文件 +62/-20。规范先读 docs/architecture/rfcs/live-team-workspace-v0.md，revision b391abd635efe56fcdf575fee96a78295a3487ec。V1 Truth：实现失效统一撤回；V3 Experience：not_met，仍有效的选择被清空；V6 Rollback：核心失效和显式恢复通过；First live outcome: one understandable correction：原 L1 owner 的真实语义验收仍 deferred，不能用 read-only 修复关闭它。

关键代码讲解

  • trace（apps/presentation/dashboard/src/features/personal-workspace/goal-team-episode.tsx:24）：Read exact original, response, revised and consumer observations; propagate checked revision or null.
  • onObservation（apps/presentation/dashboard/src/features/personal-workspace/goal-team-evidence.tsx:69）：Share the checked result with report, comparison and lineage; update on-demand timestamp.
  • GoalTeamComparison effect（apps/presentation/dashboard/src/features/personal-workspace/goal-team-comparison.tsx:16）：An unchanged sibling resets source/output selection when the result object changes, causing the new regression.
  • GoalTeamLineage（apps/presentation/dashboard/src/features/personal-workspace/goal-team-lineage.tsx:24）：Render every adoption from the current parent result, without writing receipts.

其余改动覆盖整个补丁：comparison 开发 smoke 改用现有 Vite SSR 模式并保留 tsc；locale 检查验证 capability-specific options；两个 packaged 脚本补失效/恢复断言；RFC 替换当前 checkpoint。无生成资产、API 或新配置字段。

对主干的风险

[P2] 保留成功复核后的有效正文对照。 从 Goal Chat 打开团队证据，选修订依据并将对照产物设为 report.md，再核验完全相同的原产物/回应/修订/采用：base 的 comparisonPresent=1、outputIndex=1 保持；head 变成0/null。调用链是新增 callback 更换 result 对象，然后 unchanged comparison effect 把来源与 outputIndex 重置。没有错误、没有新模型调用，也没有版本变化，现有全套回归仍能绿。

最小修复：在共享观察更新时，按 session/operation/request 及所选来源和输出的 ref/hash 重新核验现有对照；仅在这些绑定改变或失效时清空。保留仍有效的来源、输出选择和阅读上下文，继续撤回真正失效的报告与采用。

完整 Chat build/typecheck、43 packaged scenarios 加 evidence-return、comparison SSR smoke、8个实际 File/SQLite delegation/Turn 检查、semantic advisory/full、diff/private-boundary 全部通过。核心撤回/下游失联/恢复、键盘、390px、分页和无新增执行有现成真实组件检查。额外同一脚本在 base/head 验证了上述新体验回归；初次脚本选择器超时是评审 harness 错误，改用实际可访问名称后两侧完整结果保留，不当作产品故障。

本机 authority 的19条注册关系来自10个 Goal、16个 Agent；有 profile 的关系据领域描述生成合成报告并走同一打包 UI，5条缺 profile 的关系保留资料缺失，不推断领域或授权。19组完成的配对结果均为 base 保留选择、head 清除选择；每组输入字节指纹相同，页面错误及模型调用均为0。它验证的是读取/选择行为，不是真实业务、真实模型、账户权限或 L1 采用。私有身份与领域文本只留本地，不进入这份公开 review。

我的整体评价

统一撤回陈旧成功对后续决策是正向的，也降低了双份状态维护成本；当前选择丢失对反复核验效率是负向的，会增加每个阅读周期的重复选择。修复绑定失效规则后才可把整个用户流程判断为正向。没有量化 token/吞吐/长跑提升结论，业务语义和 requester adoption 仍开着。相关收敛已应用 child error 合并；comparison 身份失效规则应作为当前同边界修复，不需要新抽象或新任务。结论 REQUEST_CHANGES；这不是要求本小补丁实现整个协作路线图。

English verdict: REQUEST_CHANGES - 5669@dd4ce39e3a8f2ea95a9b857e1dca731485c193a7; successful unchanged linked-work verification clears a valid source/Markdown comparison selection. Identical packaged-path oracle passes at base and fails at head; withdrawal/recovery and required affected checks pass. Preserve exact-bound choices while still withholding stale core/adoption evidence.

@mergify mergify Bot removed the needs-rebase Mergify: the pull request has merge conflicts with its base branch label Oct 5, 2026
@loopx-agent

loopx-agent commented Oct 5, 2026

Copy link
Copy Markdown
Collaborator Author

Addressed the comparison-retention finding on c226c945168e60cbbde495ad549df242e62de280 (behavior commit e1968fadf). The regression reproduces on 6505: a successful unchanged-version check removes the selected Markdown comparison. The reader now retains the exact source/output refs and hashes, session and execution identity rather than resetting on object replacement or relying on an array index. Raw reading mode is retained; changed identity, selected hash, missing/duplicate bindings or unavailable selected source clear the comparison.

The packaged same-version/consumer-loss regression and typed identity/reordering/invalidation cases pass. Existing core-loss/recovery and navigation checks are retained. Full packaged validation and production version-readback evidence are summarized in the updated PR body. This is a fix response, not an approval of the new head; please review that exact head. No native installation, real semantic-verifier acceptance or full L1 completion is claimed.

loopx-agent
loopx-agent commented Oct 5, 2026
View reviewed changes

@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 xhigh.

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

动机

Goal 对话的主人阅读纠偏产物时，需要整份证据反映同一次核验，同时保留当前仍有效的阅读选择。
原先，纠偏路径已撤回采用，旁边仍可能显示旧成功；核心失效后旧报告也可能留在页面。前一版本共享新观察时，又清空了成功核验前选好的 Markdown 对照。
当前版本在下游失联时保留有效核心、撤回旧采用；核心失效清空证据；同一版本成功核验保留来源、Markdown 产物与阅读模式。
它不修改 API、持久采用回执、执行权限或模型验收。
真实异议解决、准确版本的独立验收与请求方采用，仍是原有 L1/G1 的未完成工作；本次合成测试不关闭这些验收。

改动思路

评审先读不可变 base f97189ebf0cbefb747a428760ceeec0b531f6ae6 的 docs/architecture/rfcs/live-team-workspace-v0.md，以 V1 的真实证据、V3 的选择保留及 V6 的失效恢复为当前标准，并核对 local-delegation 的 Inspect accepted evidence and return 及现有设计/前端交付规则。做空修复会保留矛盾成功；只修子路径仍会让周边报告过期。当前方案复用现有父观察，并在现有 comparison owner 中按精确版本保留显式意图，避免第二个证据或权限 owner。

完整检索 base/head 的 episode/evidence/comparison/lineage、typed readback、现有 smoke 和入口。unavailable 沿用既有 TS 类型；ComparisonSelection 是局部阅读意图类型，复用已有 dependency 和 readback，不是新协议、持久回执或能力。数组索引仅保留为 select 控件投影。

具体改动

最终整个 PR 为 10 文件 +146/-32，生产 TS 为 +66/-22。GoalTeamEpisode.trace（goal-team-episode.tsx:25）按 operation/request/agent/Todo/ref/hash 核验原始、回应、修订及可选采用，77 行传播新观察，87 行核心失败传播 null。GoalTeamEvidence（goal-team-evidence.tsx:68）统一报告、关系及验收标识。GoalTeamComparison（goal-team-comparison.tsx:14）用准确版本选择替代旧索引 owner。currentComparisonSelection（team-artifact-comparison.ts:14）匹配相同 session/执行身份、唯一当前来源及唯一输出版本，拒绝变化、缺失、歧义和失效，允许仅重排及无关观察变化。

全部 companion 也已核对：package/比较 smoke 复用现有 Vite SSR 与 TS 源，增加版本选择的语义反例；reader-return 保留核心/采用撤回、恢复和导航断言，并加入先选 Markdown/原文再成功核验的回归；team-evidence 延伸整份旧成功撤回和恢复；workspace 契约保留英中 capability-specific Explore Harness 语义；RFC 更新已合并检查点和当前只读选择规则。没有生成资产、私有状态、新模块、权限或后台效果。

上次评审到当前版本单独核对：旧 dd4/6505 的 reset 回归没有靠提交变旧被认定解决。我在 6505 和不可变 base 的官方打包 UI 中独立复现：相同版本成功核验，base 保留 report.md，而6505变为0/null。c226 增加五文件有界修复；当前官方 bundle 和现有打包 reader 已独立重跑，保留 report.md、来源 pressed 状态和原文 PRE，连同下游失联后的选择保留均通过。

对主干的风险

最强反方向风险是保留选择时误保留失效证据。当前 reader 的核心失效仍撤回整份报告与路径，准确恢复须显式重读；可选下游失联保留新核验核心、详情撤回旧采用，并显式恢复。当前比较语义矩阵覆盖 session/执行身份变化、hash/ref变化、缺失/重复绑定、顺序变化及无关观察，失效选择被拒绝，不换成新字节。整个任务中未启动模型或额外成员工作；普通 off Chat、执行/反馈/CLI/Lark 及 canonical receipt 规则未改变。

验证区分版本：c226 官方 Chat bundle/typecheck、比较矩阵、完整 correction-reader 返回/恢复回归通过，包括 Markdown/原文选择、390px、键盘、分页和无执行。6505 的 44 场景完整打包套件及 17 项实际隔离 Chat/CLI 与 bundle 测试通过，仅在逐项核对五文件 delta 与最终整个 diff 后复用未变化分支/后端；没有用旧批准代替当前证据。当前 diff advisory 后重跑 full semantic vocabulary 检查，保留其 unresolved producer/cross-runtime 分析边界，diff/private-boundary 扫描通过。未查询或等待 CI。

真实打包 React 行为是本次被验证边界；HTTP 关系输入是同一组既有合成 fixture，不伪造被审查的选择保留结果。真实 canonical Chat/CLI 入口另有隔离测试，作者的最终集成 SQLite/UI 声明本轮未独立复跑，已保留为剩余资格范围。已查看实际6505整体视口及当前 reader 导航检查；当前布局/首屏未改变。候选未安装或发布；真实模型 L1/G1 仍未验收。最初目录/占用端口/脚本路径操作失败保留，正确源、隔离端口、原断言的成功读回分别记录。

我的整体评价

APPROVE：当前精确 head c226c945168e60cbbde495ad549df242e62de280 没有发现阻塞项。它交付一致的当前观察，同时修复成功检查丢失阅读意图，V1/V3/V6 的受影响有界路径均得到当前证据。长期执行规则保持，用户阅读与恢复改善。相关维护性改造已应用：删除重复子错误 owner，按版本绑定替代索引 owner；没有必要扩成通用框架或引入新能力。完整真实纠偏、独立准确版本验收、安装和请求方采用仍由既有 owner 继续资格化。批准不代表 GitHub aggregate APPROVED 或合并授权；此生产界面变更由维护者在 closeout/readiness 后合并。

English verdict: APPROVE - c226c94; current packaged validation preserves exact source, Markdown output and raw mode after successful rechecks while withdrawing invalid core/consumer success. The preceding6505 regression was independently reproduced against the immutable base, then retested on this exact head. Current bundle/typecheck, identity matrix and reader recovery pass; unchanged broad-suite/backend evidence is revision-qualified. Live-model and installed-App qualification remain separate.

@huangruiteng
huangruiteng merged commit d0333d8 into main Oct 6, 2026
33 of 37 checks passed
@huangruiteng
huangruiteng deleted the codex/l1-observation-consistency-20261005 branch October 6, 2026 05:09
@loopx-agent loopx-agent mentioned this pull request Oct 6, 2026
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@huangruiteng huangruiteng Awaiting requested review from huangruiteng huangruiteng is a code owner
@maxliux5 maxliux5 Awaiting requested review from maxliux5 maxliux5 is a code owner

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@loopx-agent @huangruiteng