iter93 cluster-093: type card action workflow resume + LLM selection payloads by loning · Pull Request #1005 · aevatarAI/aevatar

loning · 2026-05-24T19:39:46Z

Summary

chat_activity.proto:加 WorkflowResumeActionPayload + LlmSelectionActionPayload typed sub-messages
arguments map 降级 open-extension-only
Lark / Telegram composer + FeishuCardHumanInteractionPort:emit typed payload(adapter JSON compat 在边界)
NyxIdRelayTransport.BuildCardActionSubmission:known JSON → typed sub-message
ChannelConversationTurnRunner + InboundMessage.Extra:workflow/LLM handoff typed access;literal-key 路径 deprecated inbound fallback only
ChannelCardActionRouting:消费 typed WorkflowResumeActionPayload
TextUserLlmOptionsRenderer:LLM selection 走 typed payload
Tests:Lark card round-trip + negative

15 files: +547 / -81。

Phase 9 consensus

META_JUDGE_DONE:consensus:type-card-action-workflow-resume-and-llm-selection-payloads

⟦AI:AUTO-LOOP⟧

…control payloads (#1004) Phase 9 r1 consensus type-card-action-workflow-resume-and-llm-selection-payloads: - chat_activity.proto: add WorkflowResumeActionPayload + LlmSelectionActionPayload typed sub-messages on ActionElement / CardActionSubmission; arguments map demoted to open-extension-only. - FeishuCardHumanInteractionPort + LarkMessageComposer + TelegramMessageComposer: emit typed payload, keep adapter JSON compat at composer boundary. - NyxIdRelayTransport.BuildCardActionSubmission: map known callback JSON fields to typed sub-message. - ChannelConversationTurnRunner.ToInboundMessage + InboundMessage.Extra: workflow/LLM handoff via typed access; literal-key path becomes deprecated inbound fallback only. - ChannelCardActionRouting: consume typed WorkflowResumeActionPayload; literal-key rebuild path explicitly deprecated. - TextUserLlmOptionsRenderer: route LLM selection through typed payload. - Tests: Lark card round-trip (typed payload → relay → workflow resume); negative tests for missing/malformed typed payload. Closes #1004 ⟦AI:AUTO-LOOP⟧

loning · 2026-05-24T19:42:41Z

🤖 Phase 8 Reviewer — `quality` (round 1)

TL;DR

这是什么:quality 角度复核 PR iter93 cluster-093: type card action workflow resume + LLM selection payloads #1005 的 typed card action payloads 变更。
结论:reject，代码主体可读且聚焦，但缺少本轮 reviewer prompt 强制要求的 refactor self-doc。
需要 maintainer 做什么:补一个 // Refactor (iter93/cluster-093): Old/New 说明后再进下一轮。

详细说明

主要问题不是实现风格，而是流程质量门禁：本轮 reviewer-quality.md 明确要求 cluster 级重构必须有 // Refactor (iterN/cluster-XXX): 的 Old/New 自说明，并把“major refactor 缺失或不可读 self-doc”列为 reject 条件。

本 PR 在 chat_activity.proto 新增 WorkflowResumeActionPayload / LlmSelectionActionPayload，在 ChannelCardActionRouting 将 workflow resume 从 Extra map 权威读取切到 CardAction.WorkflowResume，并在 NyxIdRelayTransport 做 JSON 边界兼容到 typed payload 的映射。这些都是核心语义迁移，但 touched production files 里没有 // Refactor (iter93/cluster-093): 说明。建议在主 seam 附近补一句清晰说明：Old 是 workflow resume / LLM selection 控制语义放在 open arguments map；New 是仓库内用 typed payload，arguments 只保留 adapter/第三方扩展和旧 callback JSON 入站兼容。

📎 完整 codex 原始输出(存档备查)

---
pr: 1005
role: quality
verdict: reject
---

## Verdict
Reject from the quality checklist because the major typed card-action refactor is missing the mandated `// Refactor (iter93/cluster-093):` self-documentation block.

## Evidence
- `agents/Aevatar.GAgents.Channel.Abstractions/protos/chat_activity.proto:181` adds new repository-owned typed payload contracts (`WorkflowResumeActionPayload` / `LlmSelectionActionPayload`) and `ActionElement` / `CardActionSubmission` fields, but no touched production file contains the required `// Refactor (iter93/cluster-093): Old/New` explanation.
- `agents/Aevatar.GAgents.NyxidChat/ChannelCardActionRouting.cs:21` changes workflow resume routing from `Extra` map authority to typed `CardAction.WorkflowResume` plus deprecated fallback. This is exactly the kind of major semantic refactor the quality prompt asks to self-document for future readers.
- `agents/channels/Aevatar.GAgents.Channel.NyxIdRelay/NyxIdRelayTransport.cs:213` introduces boundary JSON-to-typed-payload mapping and strips known semantic keys from the open extension map, also without the mandated refactor self-doc comment.

## What would change your verdict
Add a concise `// Refactor (iter93/cluster-093):` comment near the main routing/boundary seam, with clear `Old:` and `New:` clauses that explain that workflow resume and LLM selection used to ride the open `arguments` map and now use typed payloads internally, while JSON compatibility remains only at adapter boundaries.

REVIEW_DONE:1005:quality-r1:reject
⟦AI:AUTO-LOOP⟧

REVIEW_DONE:1005:quality-r1:reject
⟦AI:AUTO-LOOP⟧

loning · 2026-05-24T19:42:56Z

🤖 Phase 8 Reviewer — `architect` (round 1)

TL;DR

这是什么: 架构 reviewer 对 PR iter93 cluster-093: type card action workflow resume + LLM selection payloads #1005 的 round 1 审查。
结论: comment-only，不阻塞；typed payload 方向符合强类型内核与 adapter JSON 边界规则。
需要 maintainer 做什么: 建议补上 refactor-loop 要求的 Old/New pattern 注释后继续。

详细说明

本轮改动把 workflow resume 与 LLM selection 这两类仓库内可控控制语义，从 ActionElement.arguments / CardActionSubmission.Arguments 这种开放 map 前移到 WorkflowResumeActionPayload / LlmSelectionActionPayload typed payload。Arguments 仍保留为第三方或平台扩展 map，旧 callback JSON 也只在 NyxId relay、Lark、Telegram 这类 adapter 边界兼容解析；这个方向符合「核心语义强类型」和「外部协议 JSON 只在边界转换」的架构要求。

我没有发现新的架构阻塞项：diff 中没有新增 Task.Delay(、GetAwaiter().GetResult()、TypeUrl.Contains(...)、SubscribeAsync<EventEnvelope>、raw HttpClient constructor、跳过测试或 generic actor query/reply。新增的 Dictionary<> 是方法内临时映射，不是中间层跨请求事实注册表；新增的 JsonSerializer 使用点在 adapter/test callback JSON 兼容路径，不是 actor state 或 committed payload 的内部持久化。

唯一 comment 是流程性/治理项：prompt 的 architect checklist 要求每个 refactored type/method 带 Refactor (iterN/cluster-XXX): Old pattern: ... New principle: ... 注释。当前新增/改造的 chat_activity.proto typed payload contract、ChannelCardActionRouting.TryBuildWorkflowResumeCommand、NyxIdRelayTransport.MapKnownPayloads、Lark/Telegram composer payload projection 方法都没有这类 Old/New pattern 注释。建议在这些边界补一条简短说明，说明旧模式是把控制语义塞入 Arguments map，新原则是仓库内控制语义走 typed payload，JSON/map 只留作 adapter/open-extension compat。

📎 完整 codex 原始输出(存档备查)

---
pr: 1005
role: architect
verdict: comment
---

## Verdict
comment-only: the typed card action payload architecture is directionally compliant, but the refactor-loop Old/New pattern comments are missing on the newly refactored typed-payload boundary methods/contracts.

## Evidence
- `agents/Aevatar.GAgents.Channel.Abstractions/protos/chat_activity.proto:181`: adds `WorkflowResumeActionPayload` / `LlmSelectionActionPayload` and `workflow_resume` / `llm_selection` fields, but no `Refactor (iter93/cluster-093): Old pattern: ... New principle: ...` comment is present near the refactored card action contract. This misses the reviewer checklist item: "Old/New pattern comment: each refactored type/method has `// Refactor (iterN/cluster-XXX): Old pattern: …  New principle: …`."
- `agents/Aevatar.GAgents.NyxidChat/ChannelCardActionRouting.cs:21`: `TryBuildWorkflowResumeCommand` now prefers `CardAction.WorkflowResume` and only falls back to deprecated `Extra` literal keys, which matches CLAUDE/AGENTS strong-typing intent, but the refactored routing method lacks the required Old/New pattern comment.
- `agents/channels/Aevatar.GAgents.Channel.NyxIdRelay/NyxIdRelayTransport.cs:213`: `MapKnownPayloads` maps legacy adapter JSON callback fields into typed payloads and removes known keys from `Arguments`, keeping JSON compatibility at the adapter boundary. This is architecturally acceptable under the serialization rule allowing JSON at Host/Adapter boundaries, but the new mapping boundary lacks the required Old/New pattern comment.
- `agents/platforms/Aevatar.GAgents.Platform.Lark/LarkMessageComposer.cs:333` and `agents/platforms/Aevatar.GAgents.Platform.Telegram/TelegramMessageComposer.cs:224`: composer methods project typed payloads back into platform callback JSON/compact callback data. This keeps platform compatibility at the adapter boundary, but the newly refactored methods also lack the required Old/New pattern comment.
- Anti-pattern scan over the PR diff found no new `Task.Delay(`, `GetAwaiter().GetResult()`, `TypeUrl.Contains(...)`, `SubscribeAsync<EventEnvelope>`, raw `HttpClient` constructor, skipped test, or generic actor query/reply pattern. The `Dictionary<>` additions are method-local projection/adapter maps, not service-level cross-request fact registries. The `JsonSerializer` additions are in adapter/test callback JSON compatibility paths, not actor state or committed payload storage.
- Proto evolution looks safe from this diff: new fields use previously unused field numbers (`ActionElement` 11/12 and `CardActionSubmission` 6/7); no field renumbering or field removals are visible in this PR.

## What would change your verdict
Add concise refactor-loop comments to the newly refactored type/method boundaries, for example around the proto typed payload contract and the adapter/routing methods:

`Refactor (iter93/cluster-093): Old pattern: workflow resume / LLM selection control fields were encoded in ActionElement.arguments / CardActionSubmission.Arguments. New principle: repository-owned control semantics use typed payloads; JSON maps remain adapter/open-extension compatibility only.`

No blocking CLAUDE/AGENTS architecture violation found.

REVIEW_DONE:1005:architect:comment
⟦AI:AUTO-LOOP⟧

REVIEW_DONE:1005:architect-r1:comment
⟦AI:AUTO-LOOP⟧

loning · 2026-05-24T19:43:35Z

🤖 Phase 8 Reviewer — `tests` (round 1)

TL;DR

这是什么: PR iter93 cluster-093: type card action workflow resume + LLM selection payloads #1005 的 tests 角色审查，范围只看测试覆盖与测试质量。
结论: reject；现有新增测试覆盖了 typed workflow resume 和 typed LLM select_service happy path，但漏了 typed apply_preset 与 Telegram adapter callback emission 分支。
需要 maintainer 做什么: 补 3 个聚焦行为测试后可重新请求 tests reviewer。

详细说明

LlmSelectionActionPayload 是把仓库内 LLM 选择语义从开放 Arguments map 迁到强类型字段的新路径。当前 ChannelConversationTurnRunner 对 typed payload 同时处理 select_service 和 apply_preset，但测试只证明了 typed select_service 能调用选择服务；没有 card action 测试证明 typed PresetId 会调用 ApplyPresetAsync。同样，NyxIdRelayTransport 新增了把 compact/action-id lp / llm_apply_preset 映射成 typed apply_preset 并填充 PresetId 的分支，但现有 relay 测试只覆盖 llm_select_service / service_id。

另外，TelegramMessageComposer 新增了从 ActionElement.LlmSelection 生成 callback JSON 兼容键的逻辑，这正是本 PR 的 adapter 边界 JSON compat 目标之一；目前没有测试断言 Telegram callback_data.v 中包含 llm_action/service_id/preset_id，因此该 adapter 分支没有被行为测试钉住。

验证情况: bash tools/ci/test_stability_guards.sh 通过；过滤运行相关测试项目通过 105 个测试、0 skipped。结论不是现有测试红，而是 net-new 逻辑分支缺覆盖。

📎 完整 codex 原始输出(存档备查)

---
pr: 1005
role: tests
verdict: reject
---

## Verdict
Reject from the tests angle: the added tests cover the typed workflow resume and typed LLM service-selection happy paths, but leave the new typed preset path and adapter callback emission branches unpinned.

## Evidence
- `agents/Aevatar.GAgents.NyxidChat/ChannelConversationTurnRunner.cs:677` adds typed `LlmSelectionActionPayload` routing for both `select_service` and `apply_preset`, but `test/Aevatar.GAgents.ChannelRuntime.Tests/ChannelConversationTurnRunnerTests.cs:680` only exercises typed `select_service`; the only preset-related hits are slash-command tests or the recording fake implementation, not a card-action typed `PresetId` path that proves `ApplyPresetAsync` receives the typed value.
- `agents/channels/Aevatar.GAgents.Channel.NyxIdRelay/NyxIdRelayTransport.cs:283` maps compact/action-id `"lp"` / `"llm_apply_preset"` into typed `apply_preset`, and `:305` maps `preset_id` / submitted value into `LlmSelectionActionPayload.PresetId`; `test/Aevatar.GAgents.ChannelRuntime.Tests/NyxIdRelayTransportTests.cs:269` covers only `"llm_select_service"` / `service_id`, so the new typed preset parsing branch has no regression test.
- `agents/platforms/Aevatar.GAgents.Platform.Telegram/TelegramMessageComposer.cs:224` now injects typed `LlmSelection` into Telegram callback arguments, but the touched test set has no `TelegramMessageComposer` behavior test asserting `callback_data.v.llm_action/service_id/preset_id`. This is a net-new adapter branch and should be pinned because the PR explicitly moves those values out of `ActionElement.Arguments`.

## What would change your verdict
Add focused behavior tests for:
- typed card action `apply_preset` in `ChannelConversationTurnRunnerTests`, asserting `RecordingUserLlmSelectionService.PresetId` receives the typed `LlmSelectionActionPayload.PresetId`;
- Nyx relay parsing for compact/apply-preset callback input, asserting `CardActionSubmission.LlmSelection.Action == "apply_preset"` and `PresetId` is populated from both `preset_id` or submitted value as intended;
- Telegram callback composition from an `ActionElement` with `LlmSelection`, asserting callback JSON carries the compatibility keys consumed by `NyxIdRelayTransport`.

Checked: `bash tools/ci/test_stability_guards.sh` passed; filtered `dotnet test test/Aevatar.GAgents.ChannelRuntime.Tests/Aevatar.GAgents.ChannelRuntime.Tests.csproj --nologo --filter "FullyQualifiedName~ChannelCardActionRoutingTests|FullyQualifiedName~ChannelConversationTurnRunnerTests|FullyQualifiedName~FeishuCardHumanInteractionPortRoundTripTests|FullyQualifiedName~NyxIdRelayTransportTests"` passed 105 tests, 0 skipped.

REVIEW_DONE:1005:tests-r1:reject
⟦AI:AUTO-LOOP⟧

REVIEW_DONE:1005:tests-r1:reject
⟦AI:AUTO-LOOP⟧

loning · 2026-05-24T19:44:41Z

🔧 Phase 8 r1 评审 — 2/3 reject + 1 comment → fix r2 派出

🤖 architect: comment(only self-doc)
🤖 tests: reject(typed apply_preset + NyxIdRelayTransport + Telegram callback emission 测试缺)
🤖 quality: reject(Refactor (iter93/cluster-093) 自说明缺)

🔧 fix r2 codex 派出。

⟦AI:AUTO-LOOP⟧

loning · 2026-05-24T19:48:02Z

📊 codex 进展 review-pr1005-architect-r1 (⏳ 进行中; 已跑 7 min)

+## Verdict
+comment-only: the typed card action payload architecture is directionally compliant, but the refactor-loop Old/New pattern comments are missing on the newly refactored typed-payload boundary methods/contracts.
+
+## Evidence
+- `agents/Aevatar.GAgents.Channel.Abstractions/protos/chat_activity.proto:181`: adds `WorkflowResumeActionPayload` / `LlmSelectionActionPayload` and `workflow_resume` / `llm_selection` fields, but no `Refactor (iter93/cluster-093): Old pattern: ... New principle: ...` comment is present near the refactored card action contract. This misses the reviewer checklist item: "Old/New pattern comment: each refactored type/method has `// Refactor (iterN/cluster-XXX): Old pattern: …  New principle: …`."
+- `agents/Aevatar.GAgents.NyxidChat/ChannelCardActionRouting.cs:21`: `TryBuildWorkflowResumeCommand` now prefers `CardAction.WorkflowResume` and only falls back to deprecated `Extra` literal keys, which matches CLAUDE/AGENTS strong-typing intent, but the refactored routing method lacks the required Old/New pattern comment.
+- `agents/channels/Aevatar.GAgents.Channel.NyxIdRelay/NyxIdRelayTransport.cs:213`: `MapKnownPayloads` maps legacy adapter JSON callback fields into typed payloads and removes known keys from `Arguments`, keeping JSON compatibility at the adapter boundary. This is architecturally acceptable under the serialization rule allowing JSON at Host/Adapter boundaries, but the new mapping boundary lacks the required Old/New pattern comment.
+- `agents/platforms/Aevatar.GAgents.Platform.Lark/LarkMessageComposer.cs:333` and `agents/platforms/Aevatar.GAgents.Platform.Telegram/TelegramMessageComposer.cs:224`: composer methods project typed payloads back into platform callback JSON/compact callback data. This keeps platform compatibility at the adapter boundary, but the newly refactored methods also lack the required Old/New pattern comment.
+- Anti-pattern scan over the PR diff found no new `Task.Delay(`, `GetAwaiter().GetResult()`, `TypeUrl.Contains(...)`, `SubscribeAsync<EventEnvelope>`, raw `HttpClient` constructor, skipped test, or generic actor query/reply pattern. The `Dictionary<>` additions are method-local projection/adapter maps, not service-level cross-request fact registries. The `JsonSerializer` additions are in adapter/test callback JSON compatibility paths, not actor state or committed payload storage.
+- Proto evolution looks safe from this diff: new fields use previously unused field numbers (`ActionElement` 11/12 and `CardActionSubmission` 6/7); no field renumbering or field removals are visible in this PR.
+
+## What would change your verdict
+Add concise refactor-loop comments to the newly refactored type/method boundaries, for example around the proto typed payload contract and the adapter/routing methods:
+
+`Refactor (iter93/cluster-093): Old pattern: workflow resume / LLM selection control fields were encoded in ActionElement.arguments / CardActionSubmission.Arguments. New principle: repository-owned control semantics use typed payloads; JSON maps remain adapter/open-extension compatibility only.`
+
+No blocking CLAUDE/AGENTS architecture violation found.
+
+REVIEW_DONE:1005:architect:comment
+⟦AI:AUTO-LOOP⟧

tokens used
83,651
Posted the architect round 1 review comment to PR #1005:

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 · 2026-05-24T19:48:03Z

📊 codex 进展 review-pr1005-quality-r1 (⏳ 进行中; 已跑 7 min)

@@ -0,0 +1,19 @@
+---
+pr: 1005
+role: quality
+verdict: reject
+---
+
+## Verdict
+Reject from the quality checklist because the major typed card-action refactor is missing the mandated `// Refactor (iter93/cluster-093):` self-documentation block.
+
+## Evidence
+- `agents/Aevatar.GAgents.Channel.Abstractions/protos/chat_activity.proto:181` adds new repository-owned typed payload contracts (`WorkflowResumeActionPayload` / `LlmSelectionActionPayload`) and `ActionElement` / `CardActionSubmission` fields, but no touched production file contains the required `// Refactor (iter93/cluster-093): Old/New` explanation.
+- `agents/Aevatar.GAgents.NyxidChat/ChannelCardActionRouting.cs:21` changes workflow resume routing from `Extra` map authority to typed `CardAction.WorkflowResume` plus deprecated fallback. This is exactly the kind of major semantic refactor the quality prompt asks to self-document for future readers.
+- `agents/channels/Aevatar.GAgents.Channel.NyxIdRelay/NyxIdRelayTransport.cs:213` introduces boundary JSON-to-typed-payload mapping and strips known semantic keys from the open extension map, also without the mandated refactor self-doc comment.
+
+## What would change your verdict
+Add a concise `// Refactor (iter93/cluster-093):` comment near the main routing/boundary seam, with clear `Old:` and `New:` clauses that explain that workflow resume and LLM selection used to ride the open `arguments` map and now use typed payloads internally, while JSON compatibility remains only at adapter boundaries.
+
+REVIEW_DONE:1005:quality-r1:reject
+⟦AI:AUTO-LOOP⟧

tokens used
131,493
Posted the Phase 8 quality review comment for PR #1005:

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 · 2026-05-24T19:48:04Z

📊 codex 进展 review-pr1005-tests-r1 (⏳ 进行中; 已跑 7 min)

+- `agents/Aevatar.GAgents.NyxidChat/ChannelConversationTurnRunner.cs:677` adds typed `LlmSelectionActionPayload` routing for both `select_service` and `apply_preset`, but `test/Aevatar.GAgents.ChannelRuntime.Tests/ChannelConversationTurnRunnerTests.cs:680` only exercises typed `select_service`; the only preset-related hits are slash-command tests or the recording fake implementation, not a card-action typed `PresetId` path that proves `ApplyPresetAsync` receives the typed value.
+- `agents/channels/Aevatar.GAgents.Channel.NyxIdRelay/NyxIdRelayTransport.cs:283` maps compact/action-id `"lp"` / `"llm_apply_preset"` into typed `apply_preset`, and `:305` maps `preset_id` / submitted value into `LlmSelectionActionPayload.PresetId`; `test/Aevatar.GAgents.ChannelRuntime.Tests/NyxIdRelayTransportTests.cs:269` covers only `"llm_select_service"` / `service_id`, so the new typed preset parsing branch has no regression test.
+- `agents/platforms/Aevatar.GAgents.Platform.Telegram/TelegramMessageComposer.cs:224` now injects typed `LlmSelection` into Telegram callback arguments, but the touched test set has no `TelegramMessageComposer` behavior test asserting `callback_data.v.llm_action/service_id/preset_id`. This is a net-new adapter branch and should be pinned because the PR explicitly moves those values out of `ActionElement.Arguments`.
+
+## What would change your verdict
+Add focused behavior tests for:
+- typed card action `apply_preset` in `ChannelConversationTurnRunnerTests`, asserting `RecordingUserLlmSelectionService.PresetId` receives the typed `LlmSelectionActionPayload.PresetId`;
+- Nyx relay parsing for compact/apply-preset callback input, asserting `CardActionSubmission.LlmSelection.Action == "apply_preset"` and `PresetId` is populated from both `preset_id` or submitted value as intended;
+- Telegram callback composition from an `ActionElement` with `LlmSelection`, asserting callback JSON carries the compatibility keys consumed by `NyxIdRelayTransport`.
+
+Checked: `bash tools/ci/test_stability_guards.sh` passed; filtered `dotnet test test/Aevatar.GAgents.ChannelRuntime.Tests/Aevatar.GAgents.ChannelRuntime.Tests.csproj --nologo --filter "FullyQualifiedName~ChannelCardActionRoutingTests|FullyQualifiedName~ChannelConversationTurnRunnerTests|FullyQualifiedName~FeishuCardHumanInteractionPortRoundTripTests|FullyQualifiedName~NyxIdRelayTransportTests"` passed 105 tests, 0 skipped.
+
+REVIEW_DONE:1005:tests-r1:reject
+⟦AI:AUTO-LOOP⟧

tokens used
99,290
已按 `reviewer-tests.md` 流程完成 PR #1005 tests-r1 审查，并直接发了 GitHub 评论：

`https://github.com/aevatarAI/aevatar/pull/1005#issuecomment-4529776287`

结论是 `reject`：现有测试通过，但 tests 角度认为 net-new typed `apply_preset` 路径、Nyx relay preset parsing、Telegram typed callback emission 仍缺行为覆盖。

验证已跑：
- `bash tools/ci/test_stability_guards.sh` 通过

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

…ply_preset + adapter callback emission tests - chat_activity.proto + ChannelCardActionRouting.TryBuildWorkflowResumeCommand + NyxIdRelayTransport.MapKnownPayloads + Lark/Telegram composer projection: add Refactor (iter93/cluster-093) Old/New self-doc blocks; old = workflow resume + LLM selection control in open arguments map; new = typed payload in repo-owned contract, arguments only for adapter/3p extension + inbound callback JSON compat. - ChannelConversationTurnRunnerTests: typed apply_preset behavior test (LlmSelection.PresetId → ApplyPresetAsync called with PresetId). - NyxIdRelayTransportTests: lp / llm_apply_preset action id maps to typed apply_preset with PresetId filled. - TelegramMessageComposerTests: ActionElement.LlmSelection projects into callback JSON v with llm_action/service_id/preset_id keys. ⟦AI:AUTO-LOOP⟧

loning · 2026-05-24T19:56:11Z

🔧 Phase 8 r2 fix applied

FIX_DONE:1005:round-2:applied-2:rejected-0:blocked-0

加 Refactor (iter93/cluster-093) Old/New 自说明到 proto + ChannelCardActionRouting + NyxIdRelayTransport + Lark/Telegram composer
补 3 net-new 行为测试:typed apply_preset(Runner)+ lp action id → typed apply_preset(NyxIdRelayTransport)+ Telegram LlmSelection → callback JSON v 含 llm_action/service_id/preset_id

👀 r2 reviewer 派出。

⟦AI:AUTO-LOOP⟧

loning · 2026-05-24T19:58:37Z

🤖 Phase 8 Reviewer — `quality` (round 2)

TL;DR

这是什么:quality 角度复核 PR iter93 cluster-093: type card action workflow resume + LLM selection payloads #1005 的 r2 修复。
结论:approve；r1 卡住的 refactor self-doc 已补齐，代码仍聚焦且可读。
需要 maintainer 做什么:无需 quality 侧继续修改，可交给后续 gate/其他 reviewer 决策。

详细说明

r2 修复覆盖了上一轮 quality reject 的核心问题：chat_activity.proto 的 typed payload 契约、ChannelCardActionRouting 的入站路由、NyxIdRelayTransport 的 JSON 边界映射，以及 Lark/Telegram composer 的 callback 投影处，都补了 Refactor (iter93/cluster-093) Old/New 说明。说明内容能让读者直接看懂旧模式是把 workflow resume / LLM selection 控制语义放进开放 arguments map，新模式是仓库内用 WorkflowResumeActionPayload / LlmSelectionActionPayload，map/JSON 只保留 adapter 与旧 callback 兼容边界。

我没有看到新的 quality 阻塞点：新增 helper 都有调用链，命名表达了业务意图，方法拆分没有引入单实现空抽象，r2 新增测试也聚焦在 typed apply_preset、relay compact apply-preset 解析、Telegram typed callback emission 这几个变更路径。已检查 git diff --check origin/auto-refact-dev...origin/iter93/cluster-093-card-action-typed，无 whitespace 问题；本轮 quality reviewer 没跑全量测试。

📎 完整 codex 原始输出(存档备查)

---
pr: 1005
role: quality
verdict: approve
---

## Verdict
Approve from the quality checklist: the r2 patch resolves the missing self-doc concern and the typed card-action changes remain readable, focused, and free of obvious dead code or over-engineering.

## Evidence
- `agents/Aevatar.GAgents.Channel.Abstractions/protos/chat_activity.proto:181` and `agents/Aevatar.GAgents.Channel.Abstractions/protos/chat_activity.proto:203` now include clear `Refactor (iter93/cluster-093)` Old/New comments before the typed payload contracts, explaining that repository-owned workflow resume and LLM selection semantics moved out of the open `arguments` map.
- `agents/Aevatar.GAgents.NyxidChat/ChannelCardActionRouting.cs:15` now documents the main routing seam, and the helper split (`HasWorkflowResumeIdentity`, `CopyWorkflowResumePayload`, `TryBuildDeprecatedWorkflowResumeValues`) keeps the method readable without adding unused abstraction.
- `agents/channels/Aevatar.GAgents.Channel.NyxIdRelay/NyxIdRelayTransport.cs:213`, `agents/platforms/Aevatar.GAgents.Platform.Lark/LarkMessageComposer.cs:355`, and `agents/platforms/Aevatar.GAgents.Platform.Telegram/TelegramMessageComposer.cs:228` document the adapter compatibility seam where typed payloads are projected to or recovered from platform callback JSON.
- The r2 test additions are behavior-focused and stay on the changed paths: typed `apply_preset` in `ChannelConversationTurnRunnerTests`, compact apply-preset relay parsing in `NyxIdRelayTransportTests`, and Telegram typed LLM callback projection in `TelegramMessageComposerTests`.
- `git diff --check origin/auto-refact-dev...origin/iter93/cluster-093-card-action-typed` reported no whitespace errors.

REVIEW_DONE:1005:quality-r2:approve
⟦AI:AUTO-LOOP⟧

REVIEW_DONE:1005:quality-r2:approve
⟦AI:AUTO-LOOP⟧

loning · 2026-05-24T19:58:41Z

🤖 Phase 8 Reviewer — `tests` (round 2)

TL;DR

这是什么: PR iter93 cluster-093: type card action workflow resume + LLM selection payloads #1005 tests 角色 r2 复审，范围只看测试覆盖与测试质量。
结论: approve；r1 要求的 typed apply_preset、Nyx relay preset parsing、Telegram callback emission 三项行为测试已补齐。
需要 maintainer 做什么: tests 角度无需继续修改，可进入其他 reviewer / controller 后续流程。

详细说明

r2 新增测试不是 bump-only：ChannelConversationTurnRunnerTests 现在用 typed LlmSelectionActionPayload(Action=apply_preset, PresetId=work-fast) 走完整 card action 路径，并断言 RecordingUserLlmSelectionService.PresetId、binding context、direct reply 与回复文案；NyxIdRelayTransportTests 覆盖 lp 和 llm_apply_preset compact callback 映射到 typed LlmSelection.Action == "apply_preset" / PresetId == "work-fast"；TelegramMessageComposerTests 解析 callback_data.v 并断言 typed LlmSelection 投影出 llm_action/service_id/preset_id 兼容键。

测试质量扫描未发现新增 Task.Delay、Thread.Sleep、[Skip]、manual trait、Assert.True(true) 或明显断言弱化。验证已跑：bash tools/ci/test_stability_guards.sh 通过；ChannelRuntime 相关过滤测试通过 108 个、0 skipped；Telegram composer 过滤测试通过 16 个、0 skipped。

📎 完整 codex 原始输出(存档备查)

---
pr: 1005
role: tests
verdict: approve
---

## Verdict
Approve from the tests angle: r2 adds behavior tests for the r1 gaps and does not introduce test-stability or assertion-quality regressions.

## Evidence
- `test/Aevatar.GAgents.ChannelRuntime.Tests/ChannelConversationTurnRunnerTests.cs:776` adds `RunInboundAsync_ShouldApplyTypedLlmPreset_WhenPayloadCarriesPresetId`, which sends a typed `LlmSelectionActionPayload` with `Action = apply_preset` and asserts `RecordingUserLlmSelectionService.PresetId == "work-fast"`, the binding context, direct-reply receipt, and user-facing reply.
- `test/Aevatar.GAgents.ChannelRuntime.Tests/NyxIdRelayTransportTests.cs:302` adds `Parse_ShouldMapCompactApplyPresetActionId_ToTypedLlmSelection` for both `lp` and `llm_apply_preset`, asserting the adapter callback becomes typed `LlmSelection.Action == "apply_preset"` and `PresetId == "work-fast"` while removing the known semantic key from `Arguments`.
- `test/Aevatar.GAgents.Platform.Telegram.Tests/TelegramMessageComposerTests.cs:129` adds `Compose_with_typed_llm_selection_projects_callback_arguments`, parsing Telegram `callback_data` and asserting `v.llm_action`, `v.service_id`, and `v.preset_id` are emitted from typed `ActionElement.LlmSelection`.
- Diff scan found no added `Task.Delay`, `Thread.Sleep`, `[Skip]`, manual-category trait, `Assert.True(true)`, or assertion-loosening pattern that would change the tests verdict. The added `NotBeNull` checks are followed by concrete field assertions.
- Checked: `bash tools/ci/test_stability_guards.sh` passed.
- Checked: `dotnet test test/Aevatar.GAgents.ChannelRuntime.Tests/Aevatar.GAgents.ChannelRuntime.Tests.csproj --nologo --filter "FullyQualifiedName~ChannelCardActionRoutingTests|FullyQualifiedName~ChannelConversationTurnRunnerTests|FullyQualifiedName~FeishuCardHumanInteractionPortRoundTripTests|FullyQualifiedName~NyxIdRelayTransportTests"` passed 108 tests, 0 skipped.
- Checked: `dotnet test test/Aevatar.GAgents.Platform.Telegram.Tests/Aevatar.GAgents.Platform.Telegram.Tests.csproj --nologo --filter FullyQualifiedName~TelegramMessageComposerTests` passed 16 tests, 0 skipped.

REVIEW_DONE:1005:tests-r2:approve
⟦AI:AUTO-LOOP⟧

REVIEW_DONE:1005:tests-r2:approve
⟦AI:AUTO-LOOP⟧

loning · 2026-05-24T19:59:55Z

🤖 Phase 8 Reviewer — `architect` (round 2)

TL;DR

这是什么: architect r2 复审 PR iter93 cluster-093: type card action workflow resume + LLM selection payloads #1005 的 typed card action payload 架构合规性。
结论: approve；r1 要求的 Old/New self-doc 已补齐，typed payload 主链没有新的 CLAUDE/AGENTS 架构违规。
下一步: controller 可以继续汇总其他 reviewer 结论。

详细说明

本轮重点复查 r1 architect 指出的 self-doc 缺口，以及 arguments 从控制语义降级为边界扩展 map 后有没有形成第二套权威路径。WorkflowResumeActionPayload / LlmSelectionActionPayload 已在 proto、NyxID relay adapter、Lark/Telegram composer、workflow resume routing 等关键 seam 补上 Refactor (iter93/cluster-093) Old/New 说明。

架构上认可当前形态：typed payload 是仓库内可控控制语义的主路径；Lark/Telegram/NyxID 的 JSON 仍只在 adapter/platform 边界投影；Extra / Arguments 的 literal-key 路径被标注为 deprecated inbound compatibility，不作为新的事实源。PR diff 的反模式扫描未发现新的 actor.HandleEventAsync(、SubscribeAsync<EventEnvelope>、生产 Task.Delay(、GetAwaiter().GetResult()、TypeUrl.Contains(...)、raw HttpClient、disabled tests 或中间层跨请求事实字典。

验证已跑过：dotnet test test/Aevatar.GAgents.ChannelRuntime.Tests/Aevatar.GAgents.ChannelRuntime.Tests.csproj --nologo --no-restore 861 passed；dotnet test test/Aevatar.GAgents.Platform.Telegram.Tests/Aevatar.GAgents.Platform.Telegram.Tests.csproj --nologo --no-restore 19 passed；bash tools/ci/architecture_guards.sh passed。

📎 完整 codex 原始输出(存档备查)

---
pr: 1005
role: architect
verdict: approve
---

## Verdict
approve: r2 addresses the r1 architecture self-doc gap and keeps the typed card-action payload refactor inside the intended adapter/proto/channel boundaries.

## Evidence
- `agents/Aevatar.GAgents.Channel.Abstractions/protos/chat_activity.proto:181`: the new `WorkflowResumeActionPayload` / `LlmSelectionActionPayload` contracts now carry the required `Refactor (iter93/cluster-093)` Old/New block, and the proto fields are additive (`ActionElement.workflow_resume = 11`, `ActionElement.llm_selection = 12`, `CardActionSubmission.workflow_resume = 6`, `CardActionSubmission.llm_selection = 7`) with no field-number reuse or renumbering.
- `agents/channels/Aevatar.GAgents.Channel.NyxIdRelay/NyxIdRelayTransport.cs:213`: relay callback JSON is mapped at the NyxID adapter boundary into typed payloads and known repository-owned keys are removed from `Arguments`; JSON stays confined to the external adapter boundary, matching the serialization rule that external JSON is allowed only at Host/Adapter conversion boundaries.
- `agents/Aevatar.GAgents.NyxidChat/ChannelCardActionRouting.cs:15` and `agents/Aevatar.GAgents.NyxidChat/ChannelConversationTurnRunner.cs:677`: workflow resume and LLM selection consume typed `CardActionSubmission` payloads first; literal-key lookup remains explicitly deprecated inbound compatibility, not a new authoritative path.
- `agents/platforms/Aevatar.GAgents.Platform.Lark/LarkMessageComposer.cs:355` and `agents/platforms/Aevatar.GAgents.Platform.Telegram/TelegramMessageComposer.cs:228`: composers project typed payloads back into platform callback JSON only at platform emission boundaries, which preserves current NyxID relay surface without requiring changes to the external `../NyxID` repository.
- `docs/canon/aevatar-channel-architecture.md:410`: the canonical channel document now states the typed card action payload rule and limits `Arguments` to third-party/platform extension and deprecated inbound compatibility.
- Anti-pattern scan over the PR diff found no new `actor.HandleEventAsync(`, `SubscribeAsync<EventEnvelope>`, production `Task.Delay(`, `GetAwaiter().GetResult()`, `TypeUrl.Contains(...)`, raw `HttpClient` constructors, disabled tests, or service-level middle-layer ID dictionaries. The added dictionaries are method-local/platform serialization maps rather than cross-request fact stores.
- Verification passed: `dotnet test test/Aevatar.GAgents.ChannelRuntime.Tests/Aevatar.GAgents.ChannelRuntime.Tests.csproj --nologo --no-restore` (861 passed), `dotnet test test/Aevatar.GAgents.Platform.Telegram.Tests/Aevatar.GAgents.Platform.Telegram.Tests.csproj --nologo --no-restore` (19 passed), and `bash tools/ci/architecture_guards.sh` (passed, including docs lint and channel card literal guard).

REVIEW_DONE:1005:architect-r2:approve

⟦AI:AUTO-LOOP⟧

REVIEW_DONE:1005:architect-r2:approve

⟦AI:AUTO-LOOP⟧

loning · 2026-05-24T20:00:28Z

✅ Phase 8 r2 集中评审

🤖 architect r2: approve
🤖 tests r2: approve
🤖 quality r2: approve

3/3 → auto-merge。

⟦AI:AUTO-LOOP⟧

loning · 2026-05-24T20:01:16Z

📊 codex 进展 review-pr1005-architect-r2 (⏳ 进行中; 已跑 5 min)

+## Evidence
+- `agents/Aevatar.GAgents.Channel.Abstractions/protos/chat_activity.proto:181`: the new `WorkflowResumeActionPayload` / `LlmSelectionActionPayload` contracts now carry the required `Refactor (iter93/cluster-093)` Old/New block, and the proto fields are additive (`ActionElement.workflow_resume = 11`, `ActionElement.llm_selection = 12`, `CardActionSubmission.workflow_resume = 6`, `CardActionSubmission.llm_selection = 7`) with no field-number reuse or renumbering.
+- `agents/channels/Aevatar.GAgents.Channel.NyxIdRelay/NyxIdRelayTransport.cs:213`: relay callback JSON is mapped at the NyxID adapter boundary into typed payloads and known repository-owned keys are removed from `Arguments`; JSON stays confined to the external adapter boundary, matching the serialization rule that external JSON is allowed only at Host/Adapter conversion boundaries.
+- `agents/Aevatar.GAgents.NyxidChat/ChannelCardActionRouting.cs:15` and `agents/Aevatar.GAgents.NyxidChat/ChannelConversationTurnRunner.cs:677`: workflow resume and LLM selection consume typed `CardActionSubmission` payloads first; literal-key lookup remains explicitly deprecated inbound compatibility, not a new authoritative path.
+- `agents/platforms/Aevatar.GAgents.Platform.Lark/LarkMessageComposer.cs:355` and `agents/platforms/Aevatar.GAgents.Platform.Telegram/TelegramMessageComposer.cs:228`: composers project typed payloads back into platform callback JSON only at platform emission boundaries, which preserves current NyxID relay surface without requiring changes to the external `../NyxID` repository.
+- `docs/canon/aevatar-channel-architecture.md:410`: the canonical channel document now states the typed card action payload rule and limits `Arguments` to third-party/platform extension and deprecated inbound compatibility.
+- Anti-pattern scan over the PR diff found no new `actor.HandleEventAsync(`, `SubscribeAsync<EventEnvelope>`, production `Task.Delay(`, `GetAwaiter().GetResult()`, `TypeUrl.Contains(...)`, raw `HttpClient` constructors, disabled tests, or service-level middle-layer ID dictionaries. The added dictionaries are method-local/platform serialization maps rather than cross-request fact stores.
+- Verification passed: `dotnet test test/Aevatar.GAgents.ChannelRuntime.Tests/Aevatar.GAgents.ChannelRuntime.Tests.csproj --nologo --no-restore` (861 passed), `dotnet test test/Aevatar.GAgents.Platform.Telegram.Tests/Aevatar.GAgents.Platform.Telegram.Tests.csproj --nologo --no-restore` (19 passed), and `bash tools/ci/architecture_guards.sh` (passed, including docs lint and channel card literal guard).
+
+REVIEW_DONE:1005:architect-r2:approve
+
+⟦AI:AUTO-LOOP⟧

tokens used
140,293
已完成 PR #1005 Phase 8 architect r2 review，并已直接评论到 GitHub：

https://github.com/aevatarAI/aevatar/pull/1005#issuecomment-4529810733

结论：`approve`

本地 artifact 已写入：
`.refactor-loop/runs/reviewer-architect-pr1005-r2.md`

验证已跑：

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 · 2026-05-24T20:01:18Z

📊 codex 进展 review-pr1005-quality-r2 (⏳ 进行中; 已跑 5 min)

+---
+pr: 1005
+role: quality
+verdict: approve
+---
+
+## Verdict
+Approve from the quality checklist: the r2 patch resolves the missing self-doc concern and the typed card-action changes remain readable, focused, and free of obvious dead code or over-engineering.
+
+## Evidence
+- `agents/Aevatar.GAgents.Channel.Abstractions/protos/chat_activity.proto:181` and `agents/Aevatar.GAgents.Channel.Abstractions/protos/chat_activity.proto:203` now include clear `Refactor (iter93/cluster-093)` Old/New comments before the typed payload contracts, explaining that repository-owned workflow resume and LLM selection semantics moved out of the open `arguments` map.
+- `agents/Aevatar.GAgents.NyxidChat/ChannelCardActionRouting.cs:15` now documents the main routing seam, and the helper split (`HasWorkflowResumeIdentity`, `CopyWorkflowResumePayload`, `TryBuildDeprecatedWorkflowResumeValues`) keeps the method readable without adding unused abstraction.
+- `agents/channels/Aevatar.GAgents.Channel.NyxIdRelay/NyxIdRelayTransport.cs:213`, `agents/platforms/Aevatar.GAgents.Platform.Lark/LarkMessageComposer.cs:355`, and `agents/platforms/Aevatar.GAgents.Platform.Telegram/TelegramMessageComposer.cs:228` document the adapter compatibility seam where typed payloads are projected to or recovered from platform callback JSON.
+- The r2 test additions are behavior-focused and stay on the changed paths: typed `apply_preset` in `ChannelConversationTurnRunnerTests`, compact apply-preset relay parsing in `NyxIdRelayTransportTests`, and Telegram typed LLM callback projection in `TelegramMessageComposerTests`.
+- `git diff --check origin/auto-refact-dev...origin/iter93/cluster-093-card-action-typed` reported no whitespace errors.
+
+REVIEW_DONE:1005:quality-r2:approve
+⟦AI:AUTO-LOOP⟧

tokens used
102,739
Posted the PR comment:

`POSTED:quality:1005:https://github.com/aevatarAI/aevatar/pull/1005#issuecomment-4529807997:## 🤖 Phase 8 Reviewer — quality (round 2)`

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 · 2026-05-24T20:01:20Z

📊 codex 进展 review-pr1005-tests-r2 (⏳ 进行中; 已跑 5 min)

+Approve from the tests angle: r2 adds behavior tests for the r1 gaps and does not introduce test-stability or assertion-quality regressions.
+
+## Evidence
+- `test/Aevatar.GAgents.ChannelRuntime.Tests/ChannelConversationTurnRunnerTests.cs:776` adds `RunInboundAsync_ShouldApplyTypedLlmPreset_WhenPayloadCarriesPresetId`, which sends a typed `LlmSelectionActionPayload` with `Action = apply_preset` and asserts `RecordingUserLlmSelectionService.PresetId == "work-fast"`, the binding context, direct-reply receipt, and user-facing reply.
+- `test/Aevatar.GAgents.ChannelRuntime.Tests/NyxIdRelayTransportTests.cs:302` adds `Parse_ShouldMapCompactApplyPresetActionId_ToTypedLlmSelection` for both `lp` and `llm_apply_preset`, asserting the adapter callback becomes typed `LlmSelection.Action == "apply_preset"` and `PresetId == "work-fast"` while removing the known semantic key from `Arguments`.
+- `test/Aevatar.GAgents.Platform.Telegram.Tests/TelegramMessageComposerTests.cs:129` adds `Compose_with_typed_llm_selection_projects_callback_arguments`, parsing Telegram `callback_data` and asserting `v.llm_action`, `v.service_id`, and `v.preset_id` are emitted from typed `ActionElement.LlmSelection`.
+- Diff scan found no added `Task.Delay`, `Thread.Sleep`, `[Skip]`, manual-category trait, `Assert.True(true)`, or assertion-loosening pattern that would change the tests verdict. The added `NotBeNull` checks are followed by concrete field assertions.
+- Checked: `bash tools/ci/test_stability_guards.sh` passed.
+- Checked: `dotnet test test/Aevatar.GAgents.ChannelRuntime.Tests/Aevatar.GAgents.ChannelRuntime.Tests.csproj --nologo --filter "FullyQualifiedName~ChannelCardActionRoutingTests|FullyQualifiedName~ChannelConversationTurnRunnerTests|FullyQualifiedName~FeishuCardHumanInteractionPortRoundTripTests|FullyQualifiedName~NyxIdRelayTransportTests"` passed 108 tests, 0 skipped.
+- Checked: `dotnet test test/Aevatar.GAgents.Platform.Telegram.Tests/Aevatar.GAgents.Platform.Telegram.Tests.csproj --nologo --filter FullyQualifiedName~TelegramMessageComposerTests` passed 16 tests, 0 skipped.
+
+REVIEW_DONE:1005:tests-r2:approve
+⟦AI:AUTO-LOOP⟧

tokens used
86,388
Posted PR #1005 tests r2 review:

https://github.com/aevatarAI/aevatar/pull/1005#issuecomment-4529808131

Verdict: `approve`

Verified:
- `bash tools/ci/test_stability_guards.sh` passed
- ChannelRuntime filtered tests passed: 108 passed, 0 skipped

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

…d-await dispatch port Coverage-quality CI on r2 still failed 39 tests in Aevatar.GAgents.ChannelRuntime.Tests because production callers (SkillRunnerCommandPort, UserAgentCatalogCommandPort, DeviceCommandFacade, ChannelRegistrationCommandFacade, ChannelBotRegistrationStartupService, TurnStreamingReplySink, NyxLark/NyxTelegram ProvisioningService) still used IActorDispatchPort admission-only contract but tests expected old handled-await semantics. r3 fix: - New OrleansActorHandledDispatchPort + HandledActorCommandTargetDispatcher providing handled-await semantics on Orleans path (parallel to Local r2 LocalActorHandledDispatchPort). - Production callers (SkillRunner / UserAgentCatalog / DeviceCommand / ChannelRegistration / TurnStreamingReplySink / NyxLark+NyxTelegram Provisioning / AgentRunReplyGenerationExecutor) migrated to IHandledActorDispatchPort for the dispatch paths that must await committed state. - DI registration (Orleans + Device + NyxIdRelay) wires the handled port. - NyxIdRelayTransport content parsing tests updated to match typed card action payload schema (from #1005 typed payload migration). - Test files updated to assert against new admission ACK shape OR handled completion as appropriate per call site. 23 files: r2 7 fail + r3 39 fail全过本地。r1 5 fail + Script integration 不回归。 ⟦AI:AUTO-LOOP⟧

…rkflowResume payload r3 fixed 39/40, only Aevatar.AI.Tests.NyxIdChatEndpointsCoverageTests.HandleRelayWebhookAsync_ShouldDispatchCardAction_ToConversationActor_ForWorkflowResumeSubmit still failed. Test asserted cardAction.Arguments["actor_id"] but #1005 typed card action migration moved actor_id/run_id/step_id/approved from Arguments map to cardAction.WorkflowResume.ActorId/RunId/StepId/Approved typed payload. 1 file, test assertions updated to read typed payload. ⟦AI:AUTO-LOOP⟧

loning · 2026-05-25T08:30:08Z

📊 codex 进展 hotfix-trunk-red-2026-05-25-r4 (⏳ 进行中; 已跑 7 min)

-        cardAction.Arguments.Should().ContainKey("step_id").WhoseValue.Should().Be("approval-1");
-        cardAction.Arguments.Should().ContainKey("approved").WhoseValue.Should().Be("False");
-        cardAction.FormFields.Should().ContainKey("user_input")
-            .WhoseValue.Should().Be("Need stronger hook");
+        cardAction!.WorkflowResume.Should().NotBeNull();
+        cardAction.WorkflowResume!.ActorId.Should().Be("workflow-actor-1");
+        cardAction.WorkflowResume.RunId.Should().Be("run-1");
+        cardAction.WorkflowResume.StepId.Should().Be("approval-1");
+        cardAction.WorkflowResume.Approved.Should().BeFalse();
+        cardAction.WorkflowResume.UserInput.Should().Be("Need stronger hook");
     }
 
     [Fact]

tokens used
98,055
Fixed [NyxIdChatEndpointsCoverageTests.cs](/Users/auric/aevatar-wt-hotfix-trunk-r4/test/Aevatar.AI.Tests/NyxIdChatEndpointsCoverageTests.cs:1689): the workflow resume card action assertions now read `WorkflowResume.ActorId/RunId/StepId/Approved/UserInput` instead of the deprecated `Arguments` / `FormFields` keys.

Verification passed:

```bash
dotnet test test/Aevatar.AI.Tests/Aevatar.AI.Tests.csproj --nologo --filter "FullyQualifiedName~Aevatar.AI.Tests.NyxIdChatEndpointsCoverageTests.HandleRelayWebhookAsync_ShouldDispatchCardAction_ToConversationActor_ForWorkflowResumeSubmit"

Result: 1 passed. Final scan found no remaining old actor_id/run_id/step_id/approved argument-map assertions in that file. No git operations performed.


> 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

…DispatchAdmission contract (#1015) * hotfix(trunk-r2): restore Script integration test compatibility with DispatchAdmission contract r1 hotfix 修了 5 个 fail 但破了 7 个 Script integration test。r2 修法: - 加 LocalActorHandledDispatchPort 提供旧"handled-await"语义给 Script 路径(integration test 期望 Script Bind/Replay 完成后才返回) - IActorDispatchPort 区分 admission-only(default) vs handled-await(opt-in) - ServiceCollectionExtensions: Local + Scripting + ConfiguredConnectorRegistry 走 handled-await - RuntimeScriptCommandService + RuntimeScriptProvisioningService 显式 await handled - HybridServiceUpgradeContinuityTests 调整断言适配新 ACK shape - architecture_guards.sh 新 guard 防新增 admission-only 路径触发 Script lifecycle await 漏洞 9 files: r1 5 test + r2 7 test 全 pass(本地),不回归。 ⟦AI:AUTO-LOOP⟧ * hotfix(trunk-r3): migrate ChannelRuntime production callers to handled-await dispatch port Coverage-quality CI on r2 still failed 39 tests in Aevatar.GAgents.ChannelRuntime.Tests because production callers (SkillRunnerCommandPort, UserAgentCatalogCommandPort, DeviceCommandFacade, ChannelRegistrationCommandFacade, ChannelBotRegistrationStartupService, TurnStreamingReplySink, NyxLark/NyxTelegram ProvisioningService) still used IActorDispatchPort admission-only contract but tests expected old handled-await semantics. r3 fix: - New OrleansActorHandledDispatchPort + HandledActorCommandTargetDispatcher providing handled-await semantics on Orleans path (parallel to Local r2 LocalActorHandledDispatchPort). - Production callers (SkillRunner / UserAgentCatalog / DeviceCommand / ChannelRegistration / TurnStreamingReplySink / NyxLark+NyxTelegram Provisioning / AgentRunReplyGenerationExecutor) migrated to IHandledActorDispatchPort for the dispatch paths that must await committed state. - DI registration (Orleans + Device + NyxIdRelay) wires the handled port. - NyxIdRelayTransport content parsing tests updated to match typed card action payload schema (from #1005 typed payload migration). - Test files updated to assert against new admission ACK shape OR handled completion as appropriate per call site. 23 files: r2 7 fail + r3 39 fail全过本地。r1 5 fail + Script integration 不回归。 ⟦AI:AUTO-LOOP⟧ * hotfix(trunk-r4): update NyxIdChat webhook test assertion to typed WorkflowResume payload r3 fixed 39/40, only Aevatar.AI.Tests.NyxIdChatEndpointsCoverageTests.HandleRelayWebhookAsync_ShouldDispatchCardAction_ToConversationActor_ForWorkflowResumeSubmit still failed. Test asserted cardAction.Arguments["actor_id"] but #1005 typed card action migration moved actor_id/run_id/step_id/approved from Arguments map to cardAction.WorkflowResume.ActorId/RunId/StepId/Approved typed payload. 1 file, test assertions updated to read typed payload. ⟦AI:AUTO-LOOP⟧ * hotfix(trunk-r5): fix ScopeDraftRunActor query test after dev merge dev_sync_daemon 自动 merge auto-refact-dev 进 PR1015 后(含 dev 22 commits + PR1014 skill update) 引入 1 新 fail: - Aevatar.GAgentService.Integration.Tests.ScopeDraftRunActorQueryIntegrationTests.DraftRunEndpoint_ShouldExposeCompletedActorSnapshotViaActorQuery 调整 test assertion 适配新 DispatchAdmission + ScopeDraftRun actor snapshot path。 1 test file。 ⟦AI:AUTO-LOOP⟧ * hotfix(trunk-r6): WorkflowRunGAgent await RoleGAgent creation via handled dispatch dev merge 引入 1 fail: - "给定 WorkflowGAgent 加载了 YAML，当触发时应创建 3 个 RoleGAgent 子 Agent" 根因:DispatchAdmission contract 后 WorkflowGAgent.HandleAsync 通过 admission-only port 派 RoleGAgent 创建,test 期望 RoleGAgent 实际可用 → 切 handled port + await committed。 1 file。 ⟦AI:AUTO-LOOP⟧

loning requested review from eanzhao, jason-aelf and louis4li as code owners May 24, 2026 19:39

loning added auto-loop Created by codex-refactor-loop skill 🚀 phase:pr-open 👀 phase:reviewing labels May 24, 2026

loning added 🔧 phase:fixing and removed 👀 phase:reviewing labels May 24, 2026

loning added 👀 phase:reviewing and removed 🔧 phase:fixing labels May 24, 2026

loning merged commit 04d3a53 into auto-refact-dev May 24, 2026
10 of 11 checks passed

loning mentioned this pull request May 24, 2026

iter93 cluster-093: type card action workflow resume + LLM selection control payloads #1004

Closed

loning mentioned this pull request May 25, 2026

持续重构集成 #690

Closed

Conversation

loning commented May 24, 2026

Summary

Phase 9 consensus

Uh oh!

loning commented May 24, 2026

🤖 Phase 8 Reviewer — quality (round 1)

TL;DR

详细说明

Uh oh!

loning commented May 24, 2026

🤖 Phase 8 Reviewer — architect (round 1)

TL;DR

详细说明

Uh oh!

loning commented May 24, 2026

🤖 Phase 8 Reviewer — tests (round 1)

TL;DR

详细说明

Uh oh!

loning commented May 24, 2026

🔧 Phase 8 r1 评审 — 2/3 reject + 1 comment → fix r2 派出

Uh oh!

loning commented May 24, 2026

📊 codex 进展 review-pr1005-architect-r1 (⏳ 进行中; 已跑 7 min)

Uh oh!

loning commented May 24, 2026

📊 codex 进展 review-pr1005-quality-r1 (⏳ 进行中; 已跑 7 min)

Uh oh!

loning commented May 24, 2026

📊 codex 进展 review-pr1005-tests-r1 (⏳ 进行中; 已跑 7 min)

Uh oh!

loning commented May 24, 2026

🔧 Phase 8 r2 fix applied

Uh oh!

loning commented May 24, 2026

🤖 Phase 8 Reviewer — quality (round 2)

TL;DR

详细说明

Uh oh!

loning commented May 24, 2026

🤖 Phase 8 Reviewer — tests (round 2)

TL;DR

详细说明

Uh oh!

loning commented May 24, 2026

🤖 Phase 8 Reviewer — architect (round 2)

TL;DR

详细说明

Uh oh!

loning commented May 24, 2026

✅ Phase 8 r2 集中评审

Uh oh!

Uh oh!

loning commented May 24, 2026

📊 codex 进展 review-pr1005-architect-r2 (⏳ 进行中; 已跑 5 min)

Uh oh!

loning commented May 24, 2026

📊 codex 进展 review-pr1005-quality-r2 (⏳ 进行中; 已跑 5 min)

Uh oh!

loning commented May 24, 2026

📊 codex 进展 review-pr1005-tests-r2 (⏳ 进行中; 已跑 5 min)

Uh oh!

loning commented May 25, 2026

📊 codex 进展 hotfix-trunk-red-2026-05-25-r4 (⏳ 进行中; 已跑 7 min)

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

🤖 Phase 8 Reviewer — `quality` (round 1)

🤖 Phase 8 Reviewer — `architect` (round 1)

🤖 Phase 8 Reviewer — `tests` (round 1)

🤖 Phase 8 Reviewer — `quality` (round 2)

🤖 Phase 8 Reviewer — `tests` (round 2)

🤖 Phase 8 Reviewer — `architect` (round 2)