Support more than one llm

[tibroc]

We would want the option to configure the llms used. E.g. we might want to use ollama or something else instead of ChatGPT.

This could look like this:

• A config file on the server that specifies llm options and their APIs, Tokens, etc.
• A menu option for the user where they can choose which model to use in the session/assistant/…