agentjido / req_llm / 86b1026876abd71c2db4ea498a871b70403058f2
49%

DEFAULT BRANCH: main
Ran
Jobs 0
Files 0
Run time –
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
86b1026876abd71c2db4ea498a871b70403058f2

push

github

web-flow 
Add first-class reasoning_details support for OpenRouter (#267)

* Add first-class reasoning_details support for OpenRouter

Add reasoning_details field to ReqLLM.Message to preserve provider-specific
reasoning metadata across multi-turn conversations with reasoning models
(Gemini 3, DeepSeek R1, etc.).

Changes:
- Add reasoning_details field to Message struct with documentation
- Extract reasoning_details from OpenRouter responses
- Encode reasoning_details in subsequent requests for multi-turn continuity
- Update context.messages to preserve reasoning_details across turns
- Add comprehensive tests for extraction, encoding, and round-trip preservation

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

* Format code

Amp-Thread-ID: https://ampcode.com/threads/T-019b2a19-aaab-759d-b128-69ba82a4e62e
Co-authored-by: Amp <amp@ampcode.com>

* Fix reasoning_details context message ordering bug

The original code updated the first message in context.messages, but
Context.merge_response appends the assistant message at the end. This
caused incorrect ordering and duplicated messages when context was
non-empty.

Now correctly updates the last message in context.messages to match
how merge_response appends the assistant reply.

Added test to verify reasoning_details is attached to last context
message without duplication.

---------

Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Mike Hostetler <84222+mikehostetler@users.noreply.github.com>
Co-authored-by: Amp <amp@ampcode.com>
Source Files on build 86b1026876abd71c2db4ea498a871b70403058f2
Detailed source file information is not available for this build.