Support for Ollama

[ClarkKentIsSuperman]

I'd like to use this against an Ollama model - which follows the OpenAI formats/spec usually - but I was wondering if the only current way to do this would be to modify the OpenAI version to customize to localhost and any other minor changes, then rebuild?

Also just wondering if anyone has already done this or its in the works to not duplicate work.

[jplhughes]

This seems like the easiest approach. You might come up against problems with how it keeps track of rate limits, costs and context length. You'll have to either bypass that or add a way to get that via ollama. Alternatively you just create a separate class that makes requests to the localhost url. No one else is working on this AFAIK so feel free to pick it up.