`enable_thinking` parameter sent to OpenAI-compatible APIs that don't support it

[Harsche]

When using an OpenAI-compatible provider (Groq in my case), the AI commit message generation fails with HTTP 400 because SourceGit is sending enable_thinking in the request body, which Groq does not support.

[dan2dev]

Same issue here.

Asking AI to generate commit message...


[ERROR]
HTTP 404 (invalid_request_error: )
Parameter: model

This is not a chat model and thus not supported in the v1/chat/completions endpoint. Did you mean to use v1/completions?

[dougcunha]

Same here

[love-linger]

Fixed. Instead of directly disabling thinking mode (the way to disable thinking mode varies across different services, and some services do not allow unsupported parameters to be passed), the current approach is to return the reasoning_content to the AI service as-is if it is included in the response.

[love-linger]

You can download the latest CI build from GitHub Action and test it.