Make LiteLLM the default llm api

[zsimjee]

Hi @krrishdholakia my comment on parity is incorrect. I'm not sure what I meant, but I might have meant the parity within our codebase - we didn't support streaming for litellm.

I've also changed the next steps here.

For 0.5.0, I want to make litellm the DEFAULT llm handler we provide first class support for. We will change all runbooks etc to use litellm. We will make the llm_api parameter in guard calls optional, and pass through all args provided in that call directly to an internal litellm chat client.

Originally posted by @zsimjee in #680 (reply in thread)

• Use the openai interface for our main callable - Guard.call. We do not need to explicitly do this. Instead, we can take all args and kwargs and pass them through to the litellm sdk
• use the same interface within validators that use LLMs
• support batch and async litellm workflows
• Make the llm_callable param in Guard.call optional. When not provided, but an arg is passed that litellm uses to determine the model (the model arg) then automatically create and use a LiteLLM client. For async, use acreate and attach to the event loop if it exists by now.
• make changes in the Guardrails API that lets users pass the same params over the wire, automatically use a generated LiteLLM client to make llm requests on the server
• make sure custom callables still work