rag-evaluator
Installation
SKILL.md
RAG Evaluation
Evaluation is what separates a RAG demo from a RAG product. Build evaluation before tuning the system, not after.
Three levels of evaluation
| Level | What it measures |
|---|---|
| Retrieval | Did we fetch the right evidence? |
| Answer | Did the model generate a correct, grounded response? |
| Operational | Latency, cost, abstention rate, drift |
You need all three. Skipping retrieval evaluation is the most common mistake — teams measure answers and miss that retrieval was the actual problem.