We need to add Langfuse observability to our LLM execution flow so that all LLM provider calls are automatically traced, logged, and grouped under a unified observability layer. This will enable better debugging, performance analytics, and visibility into usage and errors across all LLM calls.