thanks, but not what im looking for. im talking about a fallback for when openai's api is down. so i can use a different llm.