Ragas
Evaluation library for RAG and agentic RAG pipelines—retrieval quality, faithfulness, and grounding.
Open source
Best when the core quality risk is retrieval—measuring faithfulness, answer relevancy, context precision, and retrieval quality in RAG-based agents.
Selection advice
Choose Ragas when your agent's value depends on retrieval quality and you need metrics that isolate retrieval problems from generation problems.
Best for
- RAG evaluation
- faithfulness metrics
- retrieval quality
- grounding checks
Not ideal for
- teams evaluating non-RAG agents
- projects that need a full LLMOps platform
Core concepts
faithfulnessrelevancycontext precisionretrieval metricsgrounding
Minimal implementation shape
Run Ragas on your RAG pipeline outputs, measure faithfulness and context precision, identify retrieval gaps, and iterate on chunking or retrieval strategy.