支持 OpenAI Responses API 及 Agent Node HITL 交互

[pcerypeng]

OpenAIModel 支持 OpenAI Responses API（use_responses_api 可选参数）
Agent Node 支持 HITL 多轮人机交互（LongRunningEvent 中断传播 + 恢复）
新增 is_tool_execution_error() 工具执行错误检测能力

[helloopenworld]

AI Code Review

我已经掌握了足够的信息。现在我来整理审查报告。

发现的问题

🚨 Critical

（无）

⚠️ Warning

• trpc_agent_sdk/models/_openai_model.py:1255-1256（_convert_api_params_to_responses 的 parameter_map）：Responses 模式下会把 top_logprobs 原样透传给 client.responses.create，但 Responses API 不支持顶层 top_logprobs（logprobs 需用 logprobs={"enabled":...} 结构）。当 use_responses_api=True 且请求配置开启 logprobs 时会触发上游 400 错误。建议在 Responses 路径中跳过 LOGPROBS/TOP_LOGPROBS 或映射为 Responses 的 logprobs 结构。

• trpc_agent_sdk/tools/_long_running_tool.py:20-36（is_tool_execution_error）：判定启发式过于宽松——只要响应含 error/error_message 键或 result/message 文本中出现 "json parse error" 等子串即判为错误。合法的 LongRunning/HITL 工具若以 {"status":"error",...} 表达业务级拒绝、或正常返回文本恰好包含这些 marker，会被错误降级为普通工具响应、跳过 HITL 挂起并触发模型重试。建议仅对明确的参数解析失败信号判定，或要求 status 与 marker 同时命中。

💡 Suggestion

• tests/trpc_agent_dsl/graph/test_agent_node_hitl.py:689 与 :737：received_call_ids: list[str] = [] 是类级可变属性，跨用例/重复执行会在同一进程内累积污染断言。建议在 fixture 或用例开头重置为实例属性，避免顺序依赖导致的脆弱测试。

总结

整体风险可控：HITL 多轮提升与 Responses API 适配的核心逻辑（interrupt 重放、子状态持久化、reasoning 往返）有测试覆盖且与 LangGraph 语义一致，未发现阻塞级问题。两处 Warning 分别是 Responses 路径下 logprobs 参数兼容性和工具错误判定启发式过宽，建议修复但非必须。

测试建议

• 补充 Responses 模式开启 response_logprobs/logprobs 的用例，验证不会向 /v1/responses 透传不支持的 top_logprobs。
• 补充 LongRunning 工具返回 {"status":"error"} 业务级失败响应的用例，确认 HITL 提升行为符合预期（不被误判为参数解析错误而降级）。

[github-actions]

CLA Assistant Lite bot:
Thank you for your submission, we really appreciate it. Like many open-source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution. You can sign the CLA by just posting a Pull Request Comment same as the below format.

---

I have read the CLA Document and I hereby sign the CLA

---

pcerypeng seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You can retrigger this bot by commenting recheck in this Pull Request}

[codecov]

Codecov Report

❌ Patch coverage is 90.07092% with 42 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (main@5ca3bf2). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
trpc_agent_sdk/models/_openai_model.py	88.40580%	32 Missing ⚠️
trpc_agent_sdk/agents/_llm_agent.py	79.16667%	5 Missing ⚠️
trpc_agent_sdk/dsl/graph/_node_action/_agent.py	96.20253%	3 Missing ⚠️
trpc_agent_sdk/tools/_long_running_tool.py	89.47368%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main        #226   +/-   ##
==========================================
  Coverage        ?   87.92616%           
==========================================
  Files           ?         479           
  Lines           ?       45230           
  Branches        ?           0           
==========================================
  Hits            ?       39769           
  Misses          ?        5461           
  Partials        ?           0           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.