This is cool. So I seem to remember that

[rose stratus]

We are implementing an option to bring your own LLM, then our server would communicate with your backend to generate agent's answer request.

Is there a preffered way you would wish to hook up your llm? E.g. Https requests vs stable ws connection?