Expanding AI Api framework for offline / other models

Ideas
1

I already connected my Claude api to BS Studio and it was already useful to find some obscure bugs in my css.

It can burn through your tokens really fast tho if you don’t precisely define what the context is and what shall be returned.

So I wanted to add my own APIs as my organization hosts it’s own openwebui service that provides different models like gpt, qwen and so on. But when I try to add those models in BS Studio I can’t make this work yet.

I implemented those models already in openclaw and python, so I’m certain that they work – but there’s no socket yet in the custom tab in BS Studio. I tried adding my gpt-5.4 model using openai as provider in custom but it didn’t work. I don’t really understand what those provider presets are if you can provide your own base url anyway – are these only for completing the url parameters or is there more behind it? It would be great to make this work somehow as my organization has a pretty high token limit. Or maybe I am missing something.

Thank you in advance!

2

In the docs we show how custom models can be connected. According to the openwebui docs it supports the OpenAI responses format, you can try this when connecting your model.

3

Thanks for the answer!

It supports the OpenAI format, but apparently not entirely.

With our openwebui implementation only /api/(v1)/chat/completions is working, the new openai standard(?) /v1/responses is not.

I assume when selecting openai as the provider it tries to complete the url in that manner then?

Edit:

The connection does work with openrouter as the provider but returns a parameter error ‘reasoning.exclude’ does not exist.

4

If it doesn’t work you can try the anthropic endpoint.