feature request: Ollama LLM HTTP authentication

[xorinzor]

It would be great if there's a way to provide some form of authentication to the Ollama LLM method.
Options could be an Authentication header, Token, or even (though not ideal) basic-auth.

This would allow someone to self-host an LLM and expose it to the internet securely, while keeping goose working no matter where they are using it.

[salman1993]

This makes sense. I think we'll hold off on this one till Ollama provides an auth route officially. Still see this open issue: ollama/ollama#849

We will provide Open AI compatible endpoint so that could be reused to have Ollama with auth token
#885

[xorinzor]

Given that their issue is closed I'm not expecting them to do so anytime soon.

Having an option to provide our own headers along with the http requests that goose makes would allow us to implement our own authentication system in front of ollama.

This could be something simple like OLLAMA_HTTP_HEADERS which is an object where each key is the header name, and value the header value.