iter56 cluster-917: ChatRequestEvent.Telegram typed + 删 hot-path Headers/Metadata bag by loning · Pull Request #922 · aevatarAI/aevatar

loning · 2026-05-23T18:46:02Z

摘要

iter56 cluster-917 — Phase 9 r3 unanimous(3/3 after reflector r1 narrow):

核心:ChatRequestEvent.Telegram typed submessage + 删 hot-path Headers/Metadata bag

加 ChatRequestEvent.Telegram typed submessage:chatId / operation / wait / poll / login 等 control fields
TelegramBridgeGAgent 不再用 `Headers` 当 primary source for control reads
RoleGAgent / ChatRuntime / StreamingProxy 用 typed LLM/tool contexts(不 merge Headers/Metadata)
删除 hot-path Headers/Metadata bag reader
Workflow producers 填 typed Telegram/tool context fields

边界(per reflector r1 narrowing constraint)

不动 `docs/canon/*`
不加新 bridge inbox event / envelope kind / actor type / Projection phase
不加 generic `ControlMetadata` / `Any` / 第二个 bag
复用现有 typed contexts(per consensus)

影响范围

21 files changed,LOC +808 / -354

Coverage

TelegramBridgeGAgent line-rate 89.7%(超 80% threshold)
Root report line-rate 2.91% 是 coverlet 包了 full dependency closure,不代表本 PR coverage

验证

bash tools/ci/test_stability_guards.sh PASS
bash tools/ci/architecture_guards.sh PASS
dotnet test aevatar.slnx --nologo PASS

Closes #917

🤖 Auto-loop / codex-refactor-loop iter56

⟦AI:AUTO-LOOP⟧

loning · 2026-05-23T18:47:10Z

📊 状态卡片 — reviewer 派出

维度	值
阶段	派出 codex(role=`reviewer`)
codex log	`review-pr922-omnibus-r1.log`
工作目录	`/Users/auric/aevatar`
timeout	5400s(~90 min 上限)
上下文	omnibus r1 派出(typed Telegram + delete Headers bag,21 files +808/-354)
下一步自动会做	1. 三 reviewer 完成 verdict marker 2. controller 计算 consensus 3. unanimous → auto-merge / reject → fix r<N+1>
是否需要人介入	❌ 否(自动推进)

🤖 controller status banner

⟦AI:AUTO-LOOP⟧

loning · 2026-05-23T18:51:58Z

📊 状态卡片 — fix 派出

维度	值
阶段	派出 codex(role=`fix`)
codex log	`fix-pr922-r2.log`
工作目录	`/Users/auric/aevatar-wt-iter56-cluster-917-workflow-llm-control-metadata`
timeout	5400s(~90 min 上限)
上下文	r1 reject(proto3 默认值变更:poll_timeout/settle_polls absent→0 而非 old 8/1 regression)→ fix r2 改 optional + has 判 absent
下一步自动会做	1. fix r 完成 marker `FIX_DONE:...` 2. controller commit + push 3. 派 reviewer r<N+1>
是否需要人介入	❌ 否(自动推进)

🤖 controller status banner

⟦AI:AUTO-LOOP⟧

loning · 2026-05-23T18:56:09Z

📊 状态卡片 — reviewer 派出

维度	值
阶段	派出 codex(role=`reviewer`)
codex log	`review-pr922-omnibus-r3.log`
工作目录	`/Users/auric/aevatar`
timeout	5400s(~90 min 上限)
上下文	omnibus r3 派出 verify r2 fix(proto optional + Has presence)
下一步自动会做	1. 三 reviewer 完成 verdict marker 2. controller 计算 consensus 3. unanimous → auto-merge / reject → fix r<N+1>
是否需要人介入	❌ 否(自动推进)

🤖 controller status banner

⟦AI:AUTO-LOOP⟧

loning · 2026-05-23T18:56:33Z

📊 codex 进展 review-pr922-omnibus-r1 (⏳ 进行中; 已跑 9 min)

`TelegramBridgeRequest` is typed, and the hot-path `TelegramBridgeGAgent` no longer reads Telegram control values from `Headers`/`Metadata`. Tests also cover typed Telegram flow and legacy header non-routing, and local guards passed:

- `test_stability_guards.sh`: PASS
- `architecture_guards.sh`: PASS
- `TelegramBridgeGAgent` coverage: `line-rate="0.897"` in coverage XML

Reject reason: `poll_timeout_seconds` and `settle_polls_after_match` are non-optional proto3 scalar fields, so absent values arrive as `0`. In `TelegramBridgeGAgent`, that changes default wait-reply behavior: absent `poll_timeout_seconds` now clamps to `1` instead of defaulting to `8`, and absent `settle_polls_after_match` becomes `0` instead of defaulting to `1`. This is a behavior regression caused by the typed migration.

Reject verbatim demand: Make default-sensitive Telegram numeric control fields presence-aware, at minimum `poll_timeout_seconds` and `settle_polls_after_match`, and add regression coverage proving absent typed fields keep the old defaults while explicit `0` still has the intended explicit behavior.

REVIEW_DONE:922:omnibus:reject
⟦AI:AUTO-LOOP⟧
tokens used
88,316
Found one blocking issue.

`TelegramBridgeRequest` is typed, and the hot-path `TelegramBridgeGAgent` no longer reads Telegram control values from `Headers`/`Metadata`. Tests also cover typed Telegram flow and legacy header non-routing, and local guards passed:

- `test_stability_guards.sh`: PASS
- `architecture_guards.sh`: PASS
- `TelegramBridgeGAgent` coverage: `line-rate="0.897"` in coverage XML

Reject reason: `poll_timeout_seconds` and `settle_polls_after_match` are non-optional proto3 scalar fields, so absent values arrive as `0`. In `TelegramBridgeGAgent`, that changes default wait-reply behavior: absent `poll_timeout_seconds` now clamps to `1` instead of defaulting to `8`, and absent `settle_polls_after_match` becomes `0` instead of defaulting to `1`. This is a behavior regression caused by the typed migration.

Reject verbatim demand: Make default-sensitive Telegram numeric control fields presence-aware, at minimum `poll_timeout_seconds` and `settle_polls_after_match`, and add regression coverage proving absent typed fields keep the old defaults while explicit `0` still has the intended explicit behavior.

自动更新每 10 分钟;edit-in-place 不堆评论;codex 完成后此 comment 自动删除(per Auric "完成后删掉就好了否则太占空间")。
🤖 controller progress reporter

loning · 2026-05-23T18:59:37Z

📊 状态卡片 — fix 派出

维度	值
阶段	派出 codex(role=`fix`)
codex log	`fix-pr922-r4.log`
工作目录	`/Users/auric/aevatar-wt-iter56-cluster-917-workflow-llm-control-metadata`
timeout	5400s(~90 min 上限)
上下文	r3 reject 2/4 fields 仍 fallback → fix r4 完整 4/4
下一步自动会做	1. fix r 完成 marker `FIX_DONE:...` 2. controller commit + push 3. 派 reviewer r<N+1>
是否需要人介入	❌ 否(自动推进)

🤖 controller status banner

⟦AI:AUTO-LOOP⟧

loning · 2026-05-23T19:02:05Z

📊 状态卡片 — reviewer 派出

维度	值
阶段	派出 codex(role=`reviewer`)
codex log	`review-pr922-omnibus-r5.log`
工作目录	`/Users/auric/aevatar`
timeout	5400s(~90 min 上限)
上下文	omnibus r5 派出 verify 4/4 fields explicit 0 preservation
下一步自动会做	1. 三 reviewer 完成 verdict marker 2. controller 计算 consensus 3. unanimous → auto-merge / reject → fix r<N+1>
是否需要人介入	❌ 否(自动推进)

🤖 controller status banner

⟦AI:AUTO-LOOP⟧

loning · 2026-05-23T19:06:53Z

📊 codex 进展 fix-pr922-r4 (⏳ 进行中; 已跑 7 min)

 
         await agent.HandleEventAsync(Envelope(request), CancellationToken.None);
 
         var command = publisher.Sent.Should().ContainSingle().Subject.evt
             .Should().BeOfType<TelegramWaitForReplyCommand>().Subject;
+        command.WaitTimeoutMs.Should().Be(0);
         command.PollTimeoutSeconds.Should().Be(0);
         command.SettlePollsAfterMatch.Should().Be(0);
+        command.ConnectorParameters["timeout_ms"].Should().Be("0");
     }
 
     [Fact]

tokens used
84,173
Implemented explicit zero preservation for all 4 Telegram controls.

Changed:
- `wait_timeout_ms`: presence-based, explicit `0` preserved.
- `timeout_ms`: presence-based, explicit `0` forwarded to connector parameters.
- Regression test expanded to cover `wait_timeout_ms`, `poll_timeout_seconds`, `settle_polls_after_match`, and `timeout_ms`.

Verified:
- `dotnet test test/Aevatar.Workflow.Host.Api.Tests/Aevatar.Workflow.Host.Api.Tests.csproj --nologo --filter TelegramBridgeGAgentTests` passed: 25 passed, 0 skipped.
- `bash tools/ci/test_stability_guards.sh` passed.

自动更新每 10 分钟;edit-in-place 不堆评论;codex 完成后此 comment 自动删除(per Auric "完成后删掉就好了否则太占空间")。
🤖 controller progress reporter

loning · 2026-05-23T19:07:23Z

📊 状态卡片 — fix 派出

维度	值
阶段	派出 codex(role=`fix`)
codex log	`fix-pr922-r5.log`
工作目录	`/Users/auric/aevatar-wt-iter56-cluster-917-workflow-llm-control-metadata`
timeout	5400s(~90 min 上限)
上下文	r5 reject 3 blockers(build break + LLMCallModule producer + regression coverage)→ fix r5
下一步自动会做	1. fix r 完成 marker `FIX_DONE:...` 2. controller commit + push 3. 派 reviewer r<N+1>
是否需要人介入	❌ 否(自动推进)

🤖 controller status banner

⟦AI:AUTO-LOOP⟧

loning · 2026-05-23T19:09:11Z

📊 codex 进展 review-pr922-omnibus-r3 (⏳ 进行中; 已跑 13 min)

- Presence checks exist for all 4.
- Absent `poll_timeout_seconds` / `settle_polls_after_match` regressions cover legacy `8/1`.
- Explicit `0` is preserved only for `poll_timeout_seconds` and `settle_polls_after_match`.
- `test_stability_guards.sh`: PASS.
- `architecture_guards.sh`: PASS.
- `dotnet test ... --filter Telegram --nologo`: PASS, 31/31.
- `dotnet test ... --filter ChatBridge --nologo`: no matching tests found in this project.

REVIEW_DONE:922:omnibus:reject
⟦AI:AUTO-LOOP⟧
tokens used
68,004
Reject.

Finding: r2 does not preserve explicit `0` for all 4 optional numeric fields. [TelegramBridgeGAgent.cs](/Users/auric/aevatar-wt-iter56-cluster-917-workflow-llm-control-metadata/src/workflow/extensions/Aevatar.Workflow.Extensions.Bridge/TelegramBridgeGAgent.cs:224) still maps explicit `wait_timeout_ms = 0` to `DefaultWaitReplyTimeoutMs`, and [TelegramBridgeGAgent.cs](/Users/auric/aevatar-wt-iter56-cluster-917-workflow-llm-control-metadata/src/workflow/extensions/Aevatar.Workflow.Extensions.Bridge/TelegramBridgeGAgent.cs:290) still maps explicit `timeout_ms = 0` to null / LLM fallback. The new regression test only covers explicit zero for `poll_timeout_seconds` and `settle_polls_after_match`, not all 4 fields.

Verified:
- Proto fields 12, 13, 14, 20 are `optional`.
- Presence checks exist for all 4.
- Absent `poll_timeout_seconds` / `settle_polls_after_match` regressions cover legacy `8/1`.
- Explicit `0` is preserved only for `poll_timeout_seconds` and `settle_polls_after_match`.
- `test_stability_guards.sh`: PASS.
- `architecture_guards.sh`: PASS.
- `dotnet test ... --filter Telegram --nologo`: PASS, 31/31.
- `dotnet test ... --filter ChatBridge --nologo`: no matching tests found in this project.

自动更新每 10 分钟;edit-in-place 不堆评论;codex 完成后此 comment 自动删除(per Auric "完成后删掉就好了否则太占空间")。
🤖 controller progress reporter

loning · 2026-05-23T19:09:12Z

📊 codex 进展 review-pr922-omnibus-r5 (⏳ 进行中; 已跑 7 min)

2. **BLOCKER**: 4/4 explicit-zero preservation is not complete across the producer path. Consumer-side `TelegramBridgeGAgent` is now presence-based for all 4 fields, but [LLMCallModule.cs](</tmp/aevatar-pr922/src/workflow/Aevatar.Workflow.Core/Modules/LLMCallModule.cs:465>) still parses `wait_timeout_ms` with `TryParsePositiveInt32`, and [LLMCallModule.cs](</tmp/aevatar-pr922/src/workflow/Aevatar.Workflow.Core/Modules/LLMCallModule.cs:506>) does the same for `telegram.timeout_ms` / `timeout_ms`. So workflow parameters containing `"0"` do not set proto presence for 2/4 fields.

3. **BLOCKER**: Regression coverage is incomplete. [TelegramBridgeGAgentTests.cs](</tmp/aevatar-pr922/test/Aevatar.Workflow.Host.Api.Tests/TelegramBridgeGAgentTests.cs:286>) covers manually constructed typed `TelegramBridgeRequest` with explicit zero, but it does not cover workflow parameter promotion from string `"0"` into typed proto presence, which is the path still broken above.

Verification:
- `bash /tmp/aevatar-pr922/tools/ci/test_stability_guards.sh`: PASS
- `bash /tmp/aevatar-pr922/tools/ci/architecture_guards.sh`: PASS
- `dotnet test --filter Telegram --nologo`: FAIL, compile errors CS0117 in `StreamingProxyGAgent.cs`

REVIEW_DONE:922:omnibus:reject
⟦AI:AUTO-LOOP⟧
tokens used
172,920
Findings:

1. **BLOCKER**: PR head does not compile under requested verification. `dotnet test --filter Telegram --nologo` exits `1` with CS0117 because [StreamingProxyGAgent.cs](</tmp/aevatar-pr922/agents/Aevatar.GAgents.StreamingProxy/StreamingProxyGAgent.cs:45>) still references `StreamingProxyRoomChatCommandEnvelopeFactory.AccessTokenHeader/PreferredRouteHeader/DefaultModelHeader`, but those constants were removed from the factory.

2. **BLOCKER**: 4/4 explicit-zero preservation is not complete across the producer path. Consumer-side `TelegramBridgeGAgent` is now presence-based for all 4 fields, but [LLMCallModule.cs](</tmp/aevatar-pr922/src/workflow/Aevatar.Workflow.Core/Modules/LLMCallModule.cs:465>) still parses `wait_timeout_ms` with `TryParsePositiveInt32`, and [LLMCallModule.cs](</tmp/aevatar-pr922/src/workflow/Aevatar.Workflow.Core/Modules/LLMCallModule.cs:506>) does the same for `telegram.timeout_ms` / `timeout_ms`. So workflow parameters containing `"0"` do not set proto presence for 2/4 fields.

3. **BLOCKER**: Regression coverage is incomplete. [TelegramBridgeGAgentTests.cs](</tmp/aevatar-pr922/test/Aevatar.Workflow.Host.Api.Tests/TelegramBridgeGAgentTests.cs:286>) covers manually constructed typed `TelegramBridgeRequest` with explicit zero, but it does not cover workflow parameter promotion from string `"0"` into typed proto presence, which is the path still broken above.

Verification:
- `bash /tmp/aevatar-pr922/tools/ci/test_stability_guards.sh`: PASS
- `bash /tmp/aevatar-pr922/tools/ci/architecture_guards.sh`: PASS
- `dotnet test --filter Telegram --nologo`: FAIL, compile errors CS0117 in `StreamingProxyGAgent.cs`

自动更新每 10 分钟;edit-in-place 不堆评论;codex 完成后此 comment 自动删除(per Auric "完成后删掉就好了否则太占空间")。
🤖 controller progress reporter

loning · 2026-05-23T19:11:44Z

📊 状态卡片 — reviewer 派出

维度	值
阶段	派出 codex(role=`reviewer`)
codex log	`review-pr922-omnibus-r7.log`
工作目录	`/Users/auric/aevatar`
timeout	5400s(~90 min 上限)
上下文	omnibus r7 派出 verify 3 blockers fixed(build + LLMCallModule + regression)
下一步自动会做	1. 三 reviewer 完成 verdict marker 2. controller 计算 consensus 3. unanimous → auto-merge / reject → fix r<N+1>
是否需要人介入	❌ 否(自动推进)

🤖 controller status banner

⟦AI:AUTO-LOOP⟧

loning · 2026-05-23T19:19:26Z

📊 codex 进展 fix-pr922-r5 (⏳ 进行中; 已跑 12 min)

+            CancellationToken.None);
+
+        var absentRequest = ctx.Published.Select(x => x.evt).OfType<ChatRequestEvent>().Single();
+        absentRequest.Telegram.HasWaitTimeoutMs.Should().BeFalse();
+        absentRequest.Telegram.HasTimeoutMs.Should().BeFalse();
+    }
+
+    [Fact]
     public async Task LLMCallModule_TextMessageEndAndChatResponse_ShouldCompleteMatchingPendingStep()
     {
         var module = new LLMCallModule();

tokens used
108,152
Implemented the PR #922 r5 fixes.

Changed:
- `StreamingProxyGAgent` now reads access token, preferred route, and default model from typed `ChatRequestEvent.ToolContext`, removing stale constant references.
- `LLMCallModule` now accepts `"0"` for Telegram `wait_timeout_ms`, `telegram.timeout_ms`, and shared timeout promotion, preserving explicit proto presence.
- Added regression coverage for workflow parameter string `"0"` -> typed Telegram proto presence, and absent timeout -> no presence.

Verified:
- `dotnet build aevatar.slnx --nologo` passed
- `dotnet test aevatar.slnx --nologo --filter "Telegram|StreamingProxy"` passed
- `bash tools/ci/test_stability_guards.sh` passed

自动更新每 10 分钟;edit-in-place 不堆评论;codex 完成后此 comment 自动删除(per Auric "完成后删掉就好了否则太占空间")。
🤖 controller progress reporter

loning · 2026-05-23T19:20:54Z

📊 codex 进展 review-pr922-omnibus-r7 (⏳ 进行中; 已跑 9 min)


codex
Findings: none. The 3 r5 blockers are fixed at reviewed head `9a3d43a8`.

Verified:
- `dotnet build aevatar.slnx --nologo` PASS
- `dotnet test aevatar.slnx --nologo --filter 'Telegram|StreamingProxy'` PASS
- `bash tools/ci/test_stability_guards.sh` PASS
- `bash tools/ci/architecture_guards.sh` PASS

Note: local `origin/pr-922` still pointed at `67686f04`, but the requested worktree branch is at `9a3d43a8`; I reviewed that head.

REVIEW_DONE:922:omnibus:approve
⟦AI:AUTO-LOOP⟧
tokens used
145,950
Findings: none. The 3 r5 blockers are fixed at reviewed head `9a3d43a8`.

Verified:
- `dotnet build aevatar.slnx --nologo` PASS
- `dotnet test aevatar.slnx --nologo --filter 'Telegram|StreamingProxy'` PASS
- `bash tools/ci/test_stability_guards.sh` PASS
- `bash tools/ci/architecture_guards.sh` PASS

Note: local `origin/pr-922` still pointed at `67686f04`, but the requested worktree branch is at `9a3d43a8`; I reviewed that head.

自动更新每 10 分钟;edit-in-place 不堆评论;codex 完成后此 comment 自动删除(per Auric "完成后删掉就好了否则太占空间")。
🤖 controller progress reporter

codecov · 2026-05-23T19:28:37Z

Codecov Report

❌ Patch coverage is 96.01329% with 12 lines in your changes missing coverage. Please review.
✅ Project coverage is 83.52%. Comparing base (a3bca79) to head (5553b3c).
⚠️ Report is 2 commits behind head on auto-refact-dev.

Files with missing lines	Patch %	Lines
...low/Aevatar.Workflow.Core/Modules/LLMCallModule.cs	96.42%	1 Missing and 4 partials ⚠️
...low/Aevatar.Workflow.Core/Modules/ReflectModule.cs	87.50%	1 Missing and 1 partial ⚠️
...s/ToolProviders/AgentToolExecutionContextMapper.cs	98.75%	0 Missing and 1 partial ⚠️
src/Aevatar.AI.Core/AIGAgentBase.cs	80.00%	0 Missing and 1 partial ⚠️
src/Aevatar.AI.Core/Chat/ChatRuntime.cs	96.77%	0 Missing and 1 partial ⚠️
src/Aevatar.AI.Core/RoleGAgent.cs	83.33%	0 Missing and 1 partial ⚠️
....Core/Execution/WorkflowExecutionRuntimeContext.cs	87.50%	1 Missing ⚠️

@@                 Coverage Diff                 @@
##           auto-refact-dev     #922      +/-   ##
===================================================
+ Coverage            83.21%   83.52%   +0.30%     
===================================================
  Files                  991      991              
  Lines                63471    63678     +207     
  Branches              8211     8232      +21     
===================================================
+ Hits                 52820    53184     +364     
+ Misses                7107     6933     -174     
- Partials              3544     3561      +17

Flag	Coverage Δ
ci	`83.52% <96.01%> (+0.30%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...ication/Runs/WorkflowChatRequestEnvelopeFactory.cs	`84.37% <ø> (-0.25%)`	⬇️
...ion/WorkflowRequestMetadataRuntimeContextAccess.cs	`92.59% <100.00%> (+1.28%)`	⬆️
...ow/Aevatar.Workflow.Core/Modules/EvaluateModule.cs	`65.02% <100.00%> (+0.98%)`	⬆️
...orkflow/Aevatar.Workflow.Core/WorkflowRunGAgent.cs	`83.07% <100.00%> (+0.60%)`	⬆️
...s/ToolProviders/AgentToolExecutionContextMapper.cs	`98.93% <98.75%> (-0.14%)`	⬇️
src/Aevatar.AI.Core/AIGAgentBase.cs	`82.55% <80.00%> (-2.03%)`	⬇️
src/Aevatar.AI.Core/Chat/ChatRuntime.cs	`89.45% <96.77%> (+0.27%)`	⬆️
src/Aevatar.AI.Core/RoleGAgent.cs	`82.47% <83.33%> (+0.57%)`	⬆️
....Core/Execution/WorkflowExecutionRuntimeContext.cs	`87.12% <87.50%> (-1.45%)`	⬇️
...low/Aevatar.Workflow.Core/Modules/ReflectModule.cs	`77.87% <87.50%> (+0.51%)`	⬆️
... and 1 more

... and 8 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

loning · 2026-05-23T19:29:36Z

📊 状态卡片 — fix 派出

维度	值
阶段	派出 codex(role=`fix`)
codex log	`fix-pr922-r6-codecov.log`
工作目录	`/Users/auric/aevatar-wt-iter56-cluster-917-workflow-llm-control-metadata`
timeout	5400s(~90 min 上限)
上下文	codecov/patch fail → fix r6 补 coverage
下一步自动会做	1. fix r 完成 marker `FIX_DONE:...` 2. controller commit + push 3. 派 reviewer r<N+1>
是否需要人介入	❌ 否(自动推进)

🤖 controller status banner

⟦AI:AUTO-LOOP⟧

…ers/Metadata bag #917 Phase 9 r3 consensus(3/3 after reflector narrow): - 加 ChatRequestEvent.Telegram typed submessage(chatId/operation/wait/poll/login 等) - TelegramBridgeGAgent 不再用 Headers 当 control field source - RoleGAgent / ChatRuntime / StreamingProxy 用 typed LLM/tool contexts(不 merge Headers/Metadata) - 删除 hot-path Headers/Metadata bag reader - Workflow producers 填 typed Telegram/tool context fields 边界(per reflector r1): - 不动 docs/canon/* - 不加新 bridge inbox event / envelope kind / actor type / Projection phase - 不加 generic ControlMetadata/Any/第二个 bag LOC 21 files +808/-354 TelegramBridgeGAgent line-rate 89.7% test_stability_guards.sh + architecture_guards.sh PASS dotnet test aevatar.slnx PASS ⟦AI:AUTO-LOOP⟧

…+ Has presence check omnibus r1 reject: - proto3 non-optional scalar default 0 → absent typed value 变成 0 不是 old default(8/1) - TelegramBridgeGAgent regression:absent poll_timeout_seconds → 1(clamp),应为 8;absent settle_polls_after_match → 0,应为 1 修复: - TelegramBridgeRequest 4 个 default-sensitive numeric field 改 optional: wait_timeout_ms, poll_timeout_seconds, settle_polls_after_match, timeout_ms - TelegramBridgeGAgent 用 Has* presence check 判 absent vs explicit 0 - absent → 走 old default(8/1) - explicit 0 → 走 0(用户明示) - 加 regression tests:absent → 默认,explicit 0 → 0 local PASS:test_stability + architecture + Telegram 25/25 + Telegram filter 31/31 ⟦AI:AUTO-LOOP⟧

omnibus r3 reject:r2 只覆盖 2/4 fields,wait_timeout_ms + timeout_ms 仍 fallback 修复: - wait_timeout_ms / timeout_ms 同样 presence-based,explicit 0 preserve - regression test 扩 4/4 fields(wait/poll/settle/timeout) local PASS:Telegram 25/25 + test_stability_guards.sh ⟦AI:AUTO-LOOP⟧

…ression omnibus r5 reject 3 blockers: 1. StreamingProxyGAgent build break(引用被删 constants) 2. LLMCallModule producer 用 TryParsePositiveInt32 不 accept "0" 3. regression test 没覆盖 workflow string "0" → typed proto presence 修复: - StreamingProxyGAgent 读 typed ChatRequestEvent.ToolContext(per consensus,删 stale constants) - LLMCallModule 接受 "0":wait_timeout_ms / telegram.timeout_ms / shared timeout promotion - "0" preserve typed proto presence(explicit 0) - 加 regression:workflow param "0" → typed presence,absent → no presence(走 default) local PASS: - dotnet build aevatar.slnx PASS - dotnet test --filter Telegram|StreamingProxy PASS - test_stability_guards.sh PASS ⟦AI:AUTO-LOOP⟧

补 test 覆盖: - LLMCallModule.TryApplyTelegramParameter typed promotion + aliases + optional zero/invalid parse - TelegramBridgeGAgent typed send payload + connector params + EmitChatResponse + wait-reply controls + clamps + run/step command id local:line 88.7%/branch 72.7% PASS normalized patch:806/915 = 88.09%(>80% threshold) ⟦AI:AUTO-LOOP⟧

loning · 2026-05-23T19:44:56Z

📊 状态卡片 — fix 派出

维度	值
阶段	派出 codex(role=`fix`)
codex log	`fix-pr922-r6-codecov.log`
工作目录	`/Users/auric/aevatar`
timeout	5400s(~90 min 上限)
上下文	r6 fix done(+216 coverage,patch 88%)+ rebased → CI re-run
下一步自动会做	1. fix r 完成 marker `FIX_DONE:...` 2. controller commit + push 3. 派 reviewer r<N+1>
是否需要人介入	❌ 否(自动推进)

🤖 controller status banner

⟦AI:AUTO-LOOP⟧

loning · 2026-05-23T20:10:48Z

📊 状态卡片 — fix 派出

维度	值
阶段	派出 codex(role=`fix`)
codex log	`fix-pr922-r7-deep-codecov.log`
工作目录	`/Users/auric/aevatar-wt-iter56-cluster-917-workflow-llm-control-metadata`
timeout	5400s(~90 min 上限)
上下文	r6 仍 codecov 56% (vs local 88%) → r7 deep investigation Codecov per-file detail
下一步自动会做	1. fix r 完成 marker `FIX_DONE:...` 2. controller commit + push 3. 派 reviewer r<N+1>
是否需要人介入	❌ 否(自动推进)

🤖 controller status banner

⟦AI:AUTO-LOOP⟧

Add focused coverage for typed tool-context payload serialization, ChatRuntime routing merge, StreamingProxy typed lifecycle fields, and LLM Telegram parameter variants. ⟦AI:AUTO-LOOP⟧

loning · 2026-05-23T20:25:54Z

📊 状态卡片 — fix 派出

维度	值
阶段	派出 codex(role=`fix`)
codex log	`fix-pr922-r7-deep-codecov.log`
工作目录	`/Users/auric/aevatar`
timeout	5400s(~90 min 上限)
上下文	r7 fix done(applied-12 coverage:typed tool context + ChatRuntime + StreamingProxy + LLMCallModule;local 83.25%)→ CI re-run
下一步自动会做	1. fix r 完成 marker `FIX_DONE:...` 2. controller commit + push 3. 派 reviewer r<N+1>
是否需要人介入	❌ 否(自动推进)

🤖 controller status banner

⟦AI:AUTO-LOOP⟧

loning requested review from eanzhao, jason-aelf and louis4li as code owners May 23, 2026 18:46

loning added auto-loop Created by codex-refactor-loop skill 🚀 phase:pr-open 👀 phase:reviewing 🤖 human:auto-推进 labels May 23, 2026

loning added 🔧 phase:fixing and removed 👀 phase:reviewing labels May 23, 2026

loning added 👀 phase:reviewing and removed 🔧 phase:fixing labels May 23, 2026

loning added 🔧 phase:fixing and removed 👀 phase:reviewing labels May 23, 2026

loning added 👀 phase:reviewing and removed 🔧 phase:fixing labels May 23, 2026

loning added 🔧 phase:fixing and removed 👀 phase:reviewing labels May 23, 2026

loning added 👀 phase:reviewing and removed 🔧 phase:fixing labels May 23, 2026

loning added 🔧 phase:fixing and removed 👀 phase:reviewing labels May 23, 2026

loning added 5 commits May 24, 2026 03:44

loning force-pushed the refactor/iter56-cluster-917-workflow-llm-control-metadata branch from 9a3d43a to f8045a7 Compare May 23, 2026 19:44

loning added ⚙️ phase:ci-running and removed 🔧 phase:fixing labels May 23, 2026

loning added 🔧 phase:fixing and removed ⚙️ phase:ci-running labels May 23, 2026

test: cover PR 922 codecov gaps

5553b3c

Add focused coverage for typed tool-context payload serialization, ChatRuntime routing merge, StreamingProxy typed lifecycle fields, and LLM Telegram parameter variants. ⟦AI:AUTO-LOOP⟧

loning force-pushed the refactor/iter56-cluster-917-workflow-llm-control-metadata branch from eb3b808 to 5553b3c Compare May 23, 2026 20:24

loning added ⚙️ phase:ci-running and removed 🔧 phase:fixing labels May 23, 2026

loning merged commit b4070cc into auto-refact-dev May 23, 2026
12 checks passed

loning added 🎉 phase:merged and removed 🚀 phase:pr-open labels May 23, 2026

loning mentioned this pull request May 23, 2026

[refactor-design] cluster-001-workflow-llm-control-metadata #917

Closed

Conversation

loning commented May 23, 2026

摘要

边界(per reflector r1 narrowing constraint)

影响范围

Coverage

验证

Uh oh!

loning commented May 23, 2026

📊 状态卡片 — reviewer 派出

Uh oh!

loning commented May 23, 2026

📊 状态卡片 — fix 派出

Uh oh!

loning commented May 23, 2026

📊 状态卡片 — reviewer 派出

Uh oh!

loning commented May 23, 2026

📊 codex 进展 review-pr922-omnibus-r1 (⏳ 进行中; 已跑 9 min)

Uh oh!

loning commented May 23, 2026

📊 状态卡片 — fix 派出

Uh oh!

loning commented May 23, 2026

📊 状态卡片 — reviewer 派出

Uh oh!

loning commented May 23, 2026

📊 codex 进展 fix-pr922-r4 (⏳ 进行中; 已跑 7 min)

Uh oh!

loning commented May 23, 2026

📊 状态卡片 — fix 派出

Uh oh!

loning commented May 23, 2026

📊 codex 进展 review-pr922-omnibus-r3 (⏳ 进行中; 已跑 13 min)

Uh oh!

loning commented May 23, 2026

📊 codex 进展 review-pr922-omnibus-r5 (⏳ 进行中; 已跑 7 min)

Uh oh!

loning commented May 23, 2026

📊 状态卡片 — reviewer 派出

Uh oh!

loning commented May 23, 2026

📊 codex 进展 fix-pr922-r5 (⏳ 进行中; 已跑 12 min)

Uh oh!

loning commented May 23, 2026

📊 codex 进展 review-pr922-omnibus-r7 (⏳ 进行中; 已跑 9 min)

Uh oh!

codecov Bot commented May 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

loning commented May 23, 2026

📊 状态卡片 — fix 派出

Uh oh!

loning commented May 23, 2026

📊 状态卡片 — fix 派出

Uh oh!

loning commented May 23, 2026

📊 状态卡片 — fix 派出

Uh oh!

loning commented May 23, 2026

📊 状态卡片 — fix 派出

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

codecov Bot commented May 23, 2026 •

edited

Loading