RAGEngineering Hub
LAB · mini-rag-labRuns in your browser

Mini RAG Lab

One question through the whole pipeline: keyword search and vector search side by side, rank fusion, the prompt, and the answer. Retrieval runs in your browser; a live answer needs your own API key.

Used on RAG-02 and RAG-05

Index: 0 chunks, recursive 512 tokens. The top 3 fused chunks (teal edge) go into the prompt.

Prompt sent to the model (0 tokens)

Reference answer

Your own question has no reference answer. Add a key to get a live one.

Live answer (your own key)

The key stays in this browser and goes only to api.anthropic.com. This site has no server. Without a key, everything else on this page still works.

What to notice

Pick g-01 (error E4012). Keyword search finds the exact code; compare where vector search ranks the same chunk.

Surprise

The prompt is short. Three chunks and an instruction are often all a good answer needs.

Try next

Ask something the documents do not cover and check whether the live answer says it does not know.