fix: handle reasoning_content in OpenRouter thinking model responses

[lireno]

Summary

When using OpenRouter with thinking models (Anthropic Claude, Google Gemini), the API returns a reasoning_content field alongside content in streaming deltas and response messages. The OpenAI completion handlers were only checking the content field, causing empty responses and ValueError: Invalid response from LLM call - None or empty.

Changes

• Streaming path (sync and async): Check reasoning_content as fallback when content is not a string in delta chunks
• Non-streaming path (sync and async): Check reasoning_content as fallback when message.content is not a string

Testing

All 121 existing tests pass (OpenAI completions and OpenAI-compatible providers).

Fixes #5537

🤖 Generated with Claude Code

Summary by CodeRabbit

• Bug Fixes
  • Improved OpenAI chat completion handling to support models that emit alternative response formats, with fallback mechanisms for both streaming and non-streaming requests. Enhanced compatibility and reliability across different model types.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: e3208b86-cd64-4879-9b99-8b58c61673a8

📥 Commits

Reviewing files that changed from the base of the PR and between c81b4fe and 295fa71.

📒 Files selected for processing (1)

• lib/crewai/src/crewai/llms/providers/openai/completion.py

---

📝 Walkthrough

Walkthrough

OpenAI completion handling now supports models that emit reasoning_content instead of or alongside standard content. Non-streaming and streaming paths in both sync and async variants apply consistent fallback logic: preferring content when available, otherwise using reasoning_content, and defaulting to an empty string if neither is present.

Changes

OpenAI completion reasoning_content fallback

Layer / File(s)	Summary
Non-streaming completion fallback lib/crewai/src/crewai/llms/providers/openai/completion.py	Synchronous and asynchronous non-streaming paths now derive final content from message.content when it is a string; if not, fall back to message.reasoning_content when that is a string; otherwise default to an empty string.
Streaming completion fallback lib/crewai/src/crewai/llms/providers/openai/completion.py	Synchronous and asynchronous streaming paths now extract per-chunk delta_text by preferring chunk_delta.content when it is a string, otherwise using chunk_delta.reasoning_content when that is a string, and only appending or emitting when delta_text is present.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Poem

🐰 Reasoning thoughts, now out in the open,
CrewAI's ears are listening,
Content or whispers—the model has spoken,
Each streaming chunk, a thought glistening!

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately describes the main change: adding support for reasoning_content in OpenAI completion handling, which directly addresses the OpenRouter thinking model response issue.
Linked Issues check	✅ Passed	The PR addresses the core coding requirement from issue #5537: robustly handling reasoning_content field when content is unavailable in OpenRouter thinking model responses for both streaming and non-streaming paths.
Out of Scope Changes check	✅ Passed	All changes are focused on OpenAI completion handlers to support reasoning_content fallback, which is directly scoped to fixing OpenRouter thinking model response handling without introducing unrelated modifications.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}