Unfortunately, it doesn't seems to support ollama...which means we can't use it for self-hosted model with our own RAG for automated responses...plus we will leak prompts to OpenAI