ML engineers operating production RAG systems that are returning inaccurate or irrelevant answers need to diagnose whether failures originate in retrieval (wrong chunks fetched), generation (model ignoring retrieved content), or indexing (relevant content not in the index at all). This agent runs a systematic quality audit across the full RAG pipeline.