[FEATURE] Additional parameters for Chat to AI

[Raul824]

Newly released model qwen 3.5 is thinking model.
Ollama disable thinking isnt working glbally but works when sending "think:false" in generate API request.

As thinking is enabled, the json decode error is generated and fallback is used.

It would be good if another parameter can be set which is then sent in generate API.

[NeptuneHub]

Hi and thanks for the feedback. This specific topic seems can be solved like this other user suggested here: #352 (comment)

Please take a looks.

[Raul824]

Yes but he created another model.
ollama create audiomuse-qwen3:14b -f - <<< 'FROM qwen3:14b SYSTEM /no_think'

Wouldn't it be easier to provide additional parameters in the request itself.

Additional parameter can be left to user but a placeholder in the API request at the tool level so that if anyone wants to use it without re-creating the model with thinking disable they can do by using the parameters

curl http://localhost:11434/api/generate -d '{
"model": "deepseek-r1",
"prompt": "Summarize this article in one sentence.",
"think": false ## This whole line can be added through parameters
}'

[NeptuneHub]

I just added a fix in devel image, new image shoudl be ready in 30-60 minute.
When ready can you test and let me know?

Thanks.

[Raul824]

Sure will test tomorrow and will update.

[Raul824]

with new version, its working with thinking models.