Skip to content

Conversation

@mostlygeek
Copy link
Contributor

@mostlygeek mostlygeek commented Aug 3, 2025

When llama-server sends reasoning content in choices[0].delta.reasoning_content add that content within <think>...</think> tags.

Additionally fixes issue with Qwen3 thinking models that do not send a tag by default.

Ref: Addresses feedback from @ngxson in #14997 (comment) by using Message.content instead of introducing a new reasoningContent?: string.

When llama-server sends reasoning content in
choices[0].delta.reasoning_content add that content within
<think></think> tags.

Additionally fixes issue with QWen3 thinking models that do not send a
<think> tag by default.
@mostlygeek
Copy link
Contributor Author

Tested with Qwen3-30B-A3B-Thinking-2507-UD-Q4_K_XL.gguf. With PR, Thinking content from model is correctly handled in Message.content:

image

@mostlygeek
Copy link
Contributor Author

@ngxson if you have time can you review this? If this is acceptable I can update the index.html.gz archive. If not, I’ll close the PR.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant