Hey HExLEsEO,

Indeed, getting the various language model WebUIs out there to work properly can be quite a pain. When I originally built AutoComplete, which feels like not that long ago, following the instructions on the oobabooga WebUI repository worked fine with GPT-4chan.

Nowadays, it should still work with Wurst 7.42 and oobabooga's new OpenAI-compatible API, but I don't know for sure. The settings would be as follows:

Setting	Value	Notes
Custom model	`gpt-4chan`
Custom model type	Legacy
OpenAI chat endpoint	`http://127.0.0.1:5000/v1/chat/completions`	Replace `5000` with your WebUI port if it's different.
OpenAI legacy endpoint	`http://127.0.0.1:5000/v1/completions`	Replace `5000` with your WebUI port if it's different.
Environment variable `WURST_OPENAI_KEY`	`sk-123`	Can't be blank, but the value is otherwise ignored.

Regarding gpt4all, it seems like it doesn't support the OpenAI format for API calls, and its own API is different from the old oobabooga API too, so I'm afraid that one just won't work. I know this sucks, but there are so many different WebUIs and hosting providers out there, with so many competing API formats, that I just can't support all of them.

I hope that makes sense.