Avoid sending orphaned reasoning items to OpenAI Responses API

[siewcapital]

Summary\n- only include messages in when a later assistant message exists\n- skip trailing/orphaned reasoning items so we don't send invalid entries without the required follow-up item\n- add regression tests that cover both skipped trailing reasoning and preserved reasoning before assistant output\n\n## Testing\n- undefined

 ERR_PNPM_RECURSIVE_EXEC_FIRST_FAIL  Command "vitest" not found (fails in this environment: )\n\nCloses #11138

---

Continue Tasks: 🔄 7 running — View all

---

Summary by cubic

Stops sending orphaned reasoning items to the OpenAI Responses API by only including “thinking” entries when followed by an assistant message. Prevents invalid input and enforces correct item order. Fixes #11138.

• Bug Fixes
  • Skip trailing “thinking” messages without a following assistant.
  • Preserve reasoning when followed by assistant output.
  • Add unit tests covering both scenarios.

^{Written for commit d90c504. Summary will update on new commits.}

[github-actions]

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

---

_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}

[cubic-dev-ai]

No issues found across 2 files