Retry

Your LLM application will hit rate limits. Here's how to handle it gracefully with exponential backoff, token buckets, and priority queuing.