Ragas
Evaluation library that scores retrieval and answer quality separately.
Visão geral
Ragas scores a retrieval pipeline on its own metrics - whether the retrieved context was relevant, whether the answer used it, whether it is faithful to it - instead of only asking whether the final answer looked right. It solves the problem that an end-to-end score cannot tell you which stage failed. It exists because a RAG system has several stages and improving it requires knowing which one is losing the quality.
Grafo de tecnologias
Como Ragas se liga ao resto do stack.
Aplicado a
- RAG technique Techniques Ragas used for RAG High ✓ source
Ferramentas
Software que implementa ou se integra com Ragas.