修复 workflow llm_call ToolContext 传播#1664
Merged
Merged
Conversation
Remove the lossy runtime LLM override mirror so llm_call derives its child request from the workflow ToolContext single source of truth. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
jason-aelf
added
crnd:lifecycle:managed
Collaborator
Author
🤖 PR #1664 测试覆盖 Review: approveTL;DR
详细说明
稳定性和测试质量方面，触达测试 diff 没有新增 📎 完整 codex 原始输出(存档备查)---
pr: 1664
role: tests
verdict: approve
---
## Verdict
Approve: the changed workflow ToolContext propagation behavior has focused tests with concrete payload assertions, and the diff does not add polling waits, skips, manual traits, or weakened assertions.
## Evidence
- `test/Aevatar.Integration.Tests/WorkflowAdditionalModulesCoverageTests.cs:2001` (`LlmCallModule_ShouldForwardTypedRuntimeContextOverrides`) exercises the new `LLMCallModule` dispatch path from runtime `AgentToolExecutionContext` to `ChatRequestEvent`, asserting all seven `LlmControl` fields plus child `ToolContext` sections (`Request`, `Caller`, `Channel`, sender binding, connected services, external metadata, and skill recovery) and verifying request metadata remains a single passthrough entry.
- `test/Aevatar.Integration.Tests/WorkflowAdditionalModulesCoverageTests.cs:2095` (`LlmCallModule_ShouldForwardControlOnlyRuntimeContextWithoutChildToolContext`) covers the control-only branch: control fields are promoted into `chatRequest.LlmControl`, while `chatRequest.ToolContext` remains null.
- `test/Aevatar.Workflow.Core.Tests/Execution/WorkflowExecutionRuntimeContextTests.cs:48` (`SetToolContext_ShouldStoreToolContextAsOnlyRuntimeFact`) and `:144` (`SetRequestMetadata_ShouldNotMutateExistingToolContext`) cover the runtime-context contract change: `ToolContext` is the stored fact, `LlmOverrides` is absent, and request metadata updates do not erase an existing tool context.
- `test/Aevatar.Workflow.Core.Tests/Execution/WorkflowExecutionContextAdapterTests.cs:31` keeps adapter coverage for exposing the same runtime context and verifies the new `ToolContext` value through the adapter.
- Static diff scan over touched tests found no added `[Skip]`, `[Trait("Category","Manual")]`, `Task.Delay(...)`, `WaitUntilAsync(...)`, `Assert.True(true)`, or not-throw-only tests; `tools/ci/test_polling_allowlist.txt` is unchanged.
⟦AI:AUTO-LOOP⟧
REVIEW_DONE:1664:tests:approve⟦AI:AUTO-LOOP⟧ |
Collaborator
Author
🤖 Architect review approve: ToolContext 单一事实源合规TL;DR
详细说明本次只看 PR #1664 / issue #1652。
📎 完整 codex 原始输出(存档备查)pr: 1664
|
Collaborator
Author
🤖 代码质量审查：通过TL;DR：从 readability / naming / simplicity 角度看，本 PR 可以通过。 详细说明
Raw artifact---
pr: 1664
role: quality
verdict: approve
---
## Verdict
Approve: the change is focused, readable, and keeps the ToolContext-to-LLM-control mapping simple without adding dead code or unnecessary abstraction.
## Evidence
- `src/workflow/Aevatar.Workflow.Core/Execution/WorkflowExecutionRuntimeContext.cs:24` keeps the runtime context surface small after removing the separate `LlmOverrides` mirror; `ApplyToolContext` at line 67 now stores the tool context directly with no extra state copy.
- `src/workflow/Aevatar.Workflow.Core/Modules/LLMCallModule.cs:336` uses the intent-revealing `ApplyWorkflowToolContext` entry point, and it is reachable from dispatch at line 655.
- `src/workflow/Aevatar.Workflow.Core/Modules/LLMCallModule.cs:352` through line 402 splits the mapping and presence checks into small, direct helpers; no helper introduces a new abstraction or unused private surface.
- `test/Aevatar.Integration.Tests/WorkflowAdditionalModulesCoverageTests.cs:2001` and line 2095 cover both full ToolContext forwarding and control-only forwarding without adding unrelated fixture complexity.
- `${HOST_REFACTOR_COMMENT_POLICY:-none}` resolved to `none`; the PR does not add new refactor self-doc comments in the changed code and deletes the obsolete `WorkflowLlmRuntimeOverrides` self-doc block with the removed type.
## What would change your verdict (only if comment or reject)
N/A.
⟦AI:AUTO-LOOP⟧
REVIEW_DONE:1664:quality:approve⟦AI:AUTO-LOOP⟧ |
Collaborator
Author
📊 当前状态 — review-gate 通过, 等 CI 完成(不需要人介入)
下一步自动会做:等待 🤖 controller status banner ⟦AI:AUTO-LOOP⟧ |
Codecov Report❌ Patch coverage is
@@ Coverage Diff @@
## auto-work-dev #1664 +/- ##
================================================
Coverage ? 83.05%
================================================
Files ? 1015
Lines ? 67502
Branches ? 8746
================================================
Hits ? 56062
Misses ? 7635
Partials ? 3805
Flags with carried forward coverage won't be shown. Click here to find out more.
🚀 New features to boost your workflow:
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
摘要
修复 #1652：
workflow llm_call子请求现在从 workflow runtimeToolContext单一事实源派生上下文。WorkflowLlmRuntimeOverrides/LlmOverridesruntime mirror。LLMCallModule从 runtimeToolContext派生 childLlmControl七个窄字段。ToolContext只在存在 caller/channel/sender binding/connected services/external metadata/skill recovery 等非控制面工具执行语义时写入；LlmControl-only 场景不写 childToolContext。Scope / 范围
6 files changed, 208 insertions, 97 deletions.
src/workflow/Aevatar.Workflow.Core/Execution/WorkflowExecutionRuntimeContext.cssrc/workflow/Aevatar.Workflow.Core/Modules/LLMCallModule.cstest/Aevatar.Integration.Tests/WorkflowAdditionalModulesCoverageTests.cstest/Aevatar.Integration.Tests/WorkflowGAgentCoverageTests.cstest/Aevatar.Workflow.Core.Tests/Execution/WorkflowExecutionContextAdapterTests.cstest/Aevatar.Workflow.Core.Tests/Execution/WorkflowExecutionRuntimeContextTests.cs验证
dotnet build aevatar.slnx --nologo— passeddotnet test test/Aevatar.Workflow.Core.Tests/Aevatar.Workflow.Core.Tests.csproj --filter "FullyQualifiedName~WorkflowExecutionRuntimeContext|FullyQualifiedName~WorkflowExecutionContextAdapter|FullyQualifiedName~LLMCallModule" --nologo— passed, 24 testsdotnet test test/Aevatar.Integration.Tests/Aevatar.Integration.Tests.csproj --filter "FullyQualifiedName~LlmCallModule_ShouldForwardTypedRuntimeContextOverrides|FullyQualifiedName~WorkflowRunGAgent" --nologo— passed, 36 testsdotnet test test/Aevatar.Integration.Tests/Aevatar.Integration.Tests.csproj --filter "FullyQualifiedName~LlmCallModule_ShouldForwardControlOnlyRuntimeContextWithoutChildToolContext" --nologo— passed, 1 testbash tools/ci/test_stability_guards.sh— passedbash tools/ci/architecture_guards.sh— passedConsensus-rnd
structuralIMPLEMENT_DONE:issue-1652:okVERIFY_DONE:issue-1652:pass🤖 Auto-loop / codex-refactor-loop issue-1652
⟦AI:AUTO-LOOP⟧