Hide thoughts

[SMInsmed]

Does gptel-include-reasoning turn reasoning on/off or does it just impact what's displayed in the chat buffer? It looks like it does both, but I'm not quite sure. When I inspect fsm in gptel-request with gptel-include-reasoning off, :thinkingConfig is not present in :generationConfig. When it's on :thinkingConfig is true.

Is the only way to include reasoning without showing thoughts in the chat buffer is to redirect them to a different buffer?

Would it be possible to fold the reasoning blocks like chatgpt and toggle them open when you want to view them?

[karthink]

Does gptel-include-reasoning turn reasoning on/off or does it just impact what's displayed in the chat buffer? It looks like it does both, but I'm not quite sure.

It depends on the API. Some (like Gemini) allow you to specify if you want reasoning output in the response. Others (like Deepseek) always include reasoning.

If the API allows it gptel-include-reasoning turns it on/off.

Is the only way to include reasoning without showing thoughts in the chat buffer
is to redirect them to a different buffer?

If you are using gptel's default chat callbacks, yes.

If you are writing your own gptel-request calls you can do whatever you want with the reasoning output.

Would it be possible to fold the reasoning blocks like chatgpt and toggle them
open when you want to view them?

From your description I'm assuming you're using gptel in markdown-mode.

In Org mode reasoning content is folded into blocks automatically.

markdown-mode and text-mode do not have built-in facilities for folding regions of text. See #845 for a proposal to add this to Markdown.

[SMInsmed]

Ah, guess I'll switch to org mode then. Thank you for info. And appreciate all the work done on gptel, it's been awesome having llm access in emacs