[Question]: The response api for openai

[mansiibm]

Hi Danny, can you please correct me if I am wrong-

Recently you added the response api - for openai . There is a capability of- web search also added. I went through the documentation of the openai chat completions vs response , and understood that the these are basically 2 different ways of generating the response.

If suppose I am using open ai's gpt-4.1, and I enable the web-search without activating the response api, the model is able to perform web search ,

As evident the responses api is toggled off. In the openai's documentation - it is mentioned that using responses api the model can use the web search action -

But the response api was off for the above case...? How was it able to perform web search? Where exactly is it using the response api or is this api only for o1-pro and o3-pro model?

Source- https://platform.openai.com/docs/models/gpt-4.1

[oliver-31415]

I believe that this has been fixed in a new pr, #8186. Can you share the librechat.yaml config file as well?

[mansiibm]

Hi, I believe the response api has been set for the openai models- o3-pro and o1-pro only. So if in LC you want to use these models then you need to enable the response api and then you can interact with the models otherwise it will throw an error for the v1/chatcompletions api not supported. Also in the commit description it is mentioned that -

No extra yaml config is required, only add the models- o1-pro and o3-pro to the openai models list and done,

Hope this helps :) !

[hardware-is-not-software]

With o3-pro from Azure the user needs to manually toggle responses API active to get it working. Does the auto-select only work for OpenAi hosted o3-pro?