Repository navigation
Conversation
Signed-off-by: Seyamalam <seyamalam41@gmail.com>
loopx-agent
left a comment
There was a problem hiding this comment.
Reviewer: model_agent · gpt-6.2-sol · OpenAI (runtime_reported, max)
Exact head aa9f7b30d16223ca807dbf3287d57a776dcf59bb
动机 / Motivation
维护者通过两条公开 smoke 检查任务摘要和后继任务路由。 常规 quota 准入改为返回完整任务上下文后,旧 smoke 仍期望诊断视图的摘要标记,并要求省去两个已公开的字段。 这次修改让两条 smoke 检查当前输出。集成 smoke 同时校验后继任务的标题、原有 Next Action 的显示值,以及选择后继任务时没有改写这条建议。 两个存储后端的语义检查通过,本机仍未达到默认总时限。 本次变更限于两条现有 smoke 的断言。
改动思路 / Approach
沿用现有 smoke 和真实 CLI。常规 quota 读回依据主干已采用的完整上下文。后继任务的身份、标题与原 Next Action 的显示值各有断言,原有路由差异和不改写检查继续验证同一条完成路径。
具体改动 / Changes
范围为 2 个 smoke 文件,新增 6 行,删除 4 行。按 docs/quota-allocation.md 在 4a0188c1ca1b6e33a11a7e31ddb56315bc1cdd46 的规则核对了 selected-todo-route 和 durable-next-action-display。现在两处普通准入都校验完整摘要的版本标记,集成测试还校验后继任务的确切标题,以及原 Next Action 的确切内容。 Quota 约定 任务耐久性 smoke 集成 smoke
对主干的风险 / Risks
在同一份生产源码上运行两版 smoke。基线的两个脚本都因为旧摘要标记失败。当前任务耐久性 smoke 通过,集成 smoke 的 File 和 SQLite 两组语义检查也通过。 独立从真实 CLI 回包里删去后继标题,或把原 Next Action 替换为后继标题,两项变动各自触发修订后的断言。
本地标准 premerge 通过,5 项直接检查和 11 项所选检查中保留 1 项主干复杂度提示。另有 2 项诊断视图测试在基线和当前提交以相同断言失败。 本机三次执行均通过 File 和 SQLite 的语义断言,运行时间分别为 124.78、147.62 和 127.96 秒,均超过既有的 60 秒总时限。后两次分别用 Python 3.14 和 3.11 直接执行,主机仍有其他工作负载。作者报告两组共约 21.6 秒。限时失败的原因尚未查清,当前改动没有新增 CLI 调用。
我的整体评价 / Verdict
这两处断言与当前协议一致,两个故意改变结果的反例也都能捕获。当前主机的默认时限仍然失败,请补充按默认 60 秒运行的可复核环境和结果,区分加载或资源竞争与脚本自身耗时。沿用现有时限。
English verdict: REQUEST_CHANGES. The semantic assertions pass, but the existing end-to-end time limit remains unqualified on the reviewed source.
|
This pull request has merge conflicts with Choose the remote for the base repository, not an out-of-date fork. 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 HEADFor a same-repository clone whose Keep the DCO |
Goal And Delivered Outcome
main— two shipped public smokes fail deterministically.Reproduction (before, on
4a0188c1c):git bisect(script = the durability smoke) pins the drift to37067bada("feat: make the complete work contract the quota default", merged 2026-11-18): it replaced theprogressive_enabledcontext gate withcomplete_context = not scheduler_observation, so thequota should-runhot path now emits the complete todo summary (first_executable_items,payload_compaction.schema_version = quota_todo_summary_payload_compaction_v0, complete routed records carryingtitle, and the durableactive_state_next_actionsurfaced read-only) instead of the bounded CLI-compact projection stampedquota_cli_todo_summary_compaction_v0(which remains for the separate TS diagnostic-projection lane). These smokes were last green on225698b31; the issue reporter revalidated them there under #7653.After:
payload_compaction.schema_versionattodo-durability-fixture-smoke.py:170andcontrol-plane-integrated-canary-smoke.py:189still expectedquota_cli_todo_summary_compaction_v0. The hot path has stampedQUOTA_PAYLOAD_COMPACTION_SCHEMA_VERSION(quota_todo_summary_payload_compaction_v0,loopx/control_plane/todos/quota_summary.py:21) since July;37067badamade that complete summary the default lane, so the emitted schema is the reviewed contract (tests/control_plane/test_quota_cli_projection.py:23,tests/control_plane/test_cli_output_budget.py:1479already pin the new name;:1091pins the old name for the diagnostic lane, untouched here).assert "title" not in routed["agent_lane_next_action"]expected the compact-lane shape; the complete contract retainstitle, so the assert now pinstitle == SUCCESSOR_TODO_TITLEinstead of dropping coverage.assert "active_state_next_action" not in routedexpected the durable field hidden; the complete context surfaces it read-only, so the assert now pins it to the fixture's durable text — whilegoal_route_hint(selected_action_differs_from_durable=True,goal_next_action_mutation="none") still proves routing ignores it.4a0188c1cwith the aboveAssertionErrors and pass end to end on the head (canary on both providers).main.Author Declaration
Implemented against
main4a0188c1c—loopx/control_plane/todos/quota_summary.py:21,297, thecomplete_contextdefault inloopx/cli_commands/quota_should_run.pypost-37067bada, and the unit-test pins cited above.payload_compactioncarries the shipped schema namepayload_compaction.schema_versionin both smokestitlerouted["agent_lane_next_action"]["title"]control-plane-integrated-canary-smoke(file + sqlite)active_state_next_actionis the durable fixture text; routing still differs and mutates nothingrouted["active_state_next_action"]+ unchangedgoal_route_hintassertsnot inassertions weakened to vacuousmain, bisected the drift to37067bada, confirmed the smoke files are byte-identical between225698b31(passing) and4a0188c1c(failing) so the producer side is the reviewed contract, iterated both smokes to green on both providers,ruff checkandgit diff --checkclean.Scope And Continuation
Validation
aa9f7b30d16223ca807dbf3287d57a776dcf59bbregression_paritypassedpython3 examples/control_plane/todo-durability-fixture-smoke.pyandpython3 examples/control_plane/control-plane-integrated-canary-smoke.pyfail on4a0188c1c(schema/title/active_state_next_actionAssertionErrors above) and pass on the head; bisect script = durability smoke,37067badais first-bad.staticpassedruff checkon both edited files;git diff --checkclean.unitnot_runquota should-runCLI output.cli-output-*, install/release/platform lanes) are separate defects/env deps left out of scope; [Bug]: two shipped public smokes now contradict #7587's projection contract and fail on main (hot-path quota payload) #7653's payload-size observation remains its own discussion.Frontend / Visual Evidence
Type of Change
LoopX Area
Technical Direction
Shared-authority RFC fixture impact
Boundary Checklist
.loopx/,.codex/goals/, and liveACTIVE_GOAL_STATE.md).python3 scripts/check_rfc_bilingual.py --base origin/main --head HEADpasses, or no paired RFC changed.none.Signed-off-bytrailer (git commit -s).