LiteLLM support

Hi, thanks a lot for this demo repo.
It'd be great if we didn't need to use our OpenAI key for running it.
Could you consider taking in a model name as input, and then putting that in a `litellm.chat_completion.create()` call instead of `openai...` in the same format?

That'll allow us to set our own LLMs (over 100 options!): https://litellm.vercel.app/docs