Retrieval quality and latency
RAG support bot
- Retrieve before you answer
- Hybrid search and a reranker
- Fresh answers when docs change
- Run evals off the live path
- Only two GPUs
- Final test
Answer support questions from 10,000 help articles with a language model. Make retrieval find the right text, keep cached answers fresh when articles change, test releases without hurting users, and fit it all on two GPUs.
Final testEditors flood the docs: edits x6 for 30 s at 0:40, so cached answers drop fast. Your final board runs once more with a new traffic seed.
Stage 5 of 5 · 7 parts given · 38 requests/s
Progress on this page stays on this device.