One query, every stage.
One real query, walked through every stage of a production RAG pipeline. Colors answer the question that matters: what actually computes this step?
- Plain code — no model at all
- Small encoder — embedder / cross-encoder, not an LLM
- Small LLM — fast, cheap
- Large LLM — the expensive call
Feel the precision/recall tradeoff
The one interactive exception this product allows (ADR-0009): a deterministic, client-side simulator — drag the sliders and watch precision, recall, and F1 move against each other.
Illustrative simulation — not a live retrieval system.
- Precision
- 0.83
- Recall
- 0.63
- F1
- 0.72
Sign in to read the full stage-by-stage walkthrough. Sign in