I’m trying to evaluate retrieval quality using RAGAS, specifically with context precision and context recall. According to the documentation (and the prompt used inside the functions), the dataset format should contain: question context answer The error says my dataset is not formatted correctly, it expects: user_input reference retrieved_contexts When I change the format and run the evaluation, I keep getting RagasOutputParserException. This seems inconsistent: the functions themselves use a prompt that assumes question/context/answer, but the evaluation fails unless the dataset is structured differently. I’m fairly sure the issue is caused by this mismatch. Has anyone run into this? is there a way to make RAGAS accept the question/context/answer format that its own functions seem to expect? Continue reading...
