RAG
These metrics evaluate retrieval augmented generation: retrieved context, its use in the answer, and related quality signals. Each page lists its shortname, inputs, optional metric_args, an Evaluation metadata section describing eval_metadata on success, and an example payload.
Metrics
The same pages appear under RAG in the docs sidebar:
- Context Faithfulness (
ctx_faith) - Context Ranking Precision (
ctx_rank_prec) - Context Recall (
ctx_recall) - Context Relevancy (
ctx_rel) - Context Sufficiency (
ctx_suff) - Context Waste (
ctx_waste)