For OpenAI piece, allow defining custom URL to use alternate AI server

[ikus060]

Describe your usecase.
I plan to make use of llama-cpp project to host a private instance train with our own data. llama-cpp provide a way to run AI model directly on CPU using various model.

Using: https://github.com/getumbrel/llama-gpt it's possible to how our own AI server to generate specific content.

Since most implementation provide a OpenAI compatible API, it's would be nice to allow user to define specific URL when creating OpenAI connection or to mimic the OpenAI piece to support self-hosted llamap-cpp server.

Describe alternatives you've considered
It might be possible to make direct call to HTTP REST API. I did not check if that was working properly.

But the fact, it's possible to run a self-hosted compatible OpenAI server might be highlight in the documentation or in the piece.