Evaluating RAG: Benchmark Your System in 5 Mins

A copy-paste evaluation pipeline to measure retrieval quality and iterate with confidence.

Don’t have a RAG assistant yet? If you want to set one up from scratch, the RAG Blueprint has step-by-step implementation instructions to deploy a RAG on your content from zero to a live URL in ~60 minutes.

These scripts are designed to slot into the existing repo pipeline. You should already have data ingested to supabase and your assistant up and running(Streamlit, agent). The eval scripts plug in after ingestion: generate the eval set from the same site_pages supabase table your assistant queries, then run the benchmark using the same retrieval path.

The eval harness runs seamlessly on your content: one system, one data source.