- LiteLLM-based LLM wrapper does not provide an AutoGen-compatible model client for different frameworks. - Need to implement an explicit AutoGen model client that wraps the existing LiteLLM client.