I have searched the existing issues and discussions.
Problem Description
some api(seems to be openai api V2) return the long COT (or thinking) content in the reasoning_content attribute, not in the content field, so the webui can not render the cot text.
I have tried gemini-2.0-flash-thinking-exp, which only shows the final anwser without the COT.
Desired Solution you'd like
support the rendering of thinking process.
Alternatives Considered
No response
Additional Context
No response
Originally created by @muziyongshixin on GitHub (Apr 1, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/12290
### Check Existing Issues
- [x] I have searched the existing issues and discussions.
### Problem Description
some api(seems to be openai api V2) return the long COT (or thinking) content in the `reasoning_content` attribute, not in the `content` field, so the webui can not render the cot text.
I have tried gemini-2.0-flash-thinking-exp, which only shows the final anwser without the COT.
### Desired Solution you'd like
support the rendering of thinking process.
### Alternatives Considered
_No response_
### Additional Context
_No response_
gemini 2.0 flash thinking experimental does not seem to return reasoning content from the API in general, since it's an experimental model. I suspect google didn't felt it is needed to provide the reasoning content for their experimental models via the API.
I also reference LiteLLM here, where this discussion was already held and some people over there said that there is no reasoning content from the experimental models via the API AT ALL
Also, for your original question, OpenWebUI already does render the reasoning content in the UI if it is provided in the reasoning_content attribute.
<!-- gh-comment-id:2769414006 -->
@Classic298 commented on GitHub (Apr 1, 2025):
gemini 2.0 flash thinking experimental does not seem to return reasoning content from the API in general, since it's an experimental model. I suspect google didn't felt it is needed to provide the reasoning content for their experimental models via the API.
I also reference LiteLLM here, where this discussion was already held and some people over there said that there is no reasoning content from the experimental models via the API **AT ALL**
Also, for your original question, OpenWebUI already does render the reasoning content in the UI if it is provided in the reasoning_content attribute.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Originally created by @muziyongshixin on GitHub (Apr 1, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/12290
Check Existing Issues
Problem Description
some api(seems to be openai api V2) return the long COT (or thinking) content in the
reasoning_contentattribute, not in thecontentfield, so the webui can not render the cot text.I have tried gemini-2.0-flash-thinking-exp, which only shows the final anwser without the COT.
Desired Solution you'd like
support the rendering of thinking process.
Alternatives Considered
No response
Additional Context
No response
@Classic298 commented on GitHub (Apr 1, 2025):
gemini 2.0 flash thinking experimental does not seem to return reasoning content from the API in general, since it's an experimental model. I suspect google didn't felt it is needed to provide the reasoning content for their experimental models via the API.
I also reference LiteLLM here, where this discussion was already held and some people over there said that there is no reasoning content from the experimental models via the API AT ALL
Also, for your original question, OpenWebUI already does render the reasoning content in the UI if it is provided in the reasoning_content attribute.
@tjbck commented on GitHub (Apr 1, 2025):
reasoning_content is already rendered and supported.