Rag And Retrieval medium
In a chat assistant, a user asks 'How do I reset my VPN token?' and then follows up with 'And on Linux?' The follow-up retrieves random Linux documentation unrelated to VPNs. What is the best fix?
- A Embed the entire conversation transcript as the query vector so all prior turns influence what gets retrieved.
- B Raise top-k for follow-up turns so that VPN-related Linux chunks have a better chance of being included.
- C Use a small LLM call to rewrite the follow-up into a standalone query such as 'reset VPN token on Linux'.
- D Skip retrieval on follow-up turns and let the model answer from the chunks retrieved for the first question.
Answer: C. Use a small LLM call to rewrite the follow-up into a standalone query such as 'reset VPN token on Linux'.
**Query rewriting** resolves references against the chat history. It costs an extra model call, typically a few hundred milliseconds, and produces a focused retrieval query. Embedding the whole transcript dilutes the query with older, unrelated topics as the conversation grows. Skipping retrieval works only when the follow-up is a clarification fully covered by the earlier context.
One system design question a day, with spaced repetition so the ones you miss come back.
Get the appRelated questions
- You must index 100M chunks using a 3,072-dimension embedding in fp32, served from RAM via HNSW. The raw vectors alone take about 1.2 TB before graph overhead. You need to cut memory several-fold while losing at most a point or two of recall@10. Which approach fits?
- A multi-tenant SaaS app stores all tenants in one vector index with a tenant_id field. Queries fetch the top 20 chunks by similarity and then drop chunks from other tenants. Small tenants often get 0-2 results. Why does this happen, and what is the fix?
- Runbooks are split into fixed 300-token chunks with no overlap. When a procedure step starts near the end of one chunk and finishes in the next, answers come back incomplete. What is the lowest-cost fix?
- Your offline eval shows recall@5 of 92% on a labeled set, so the gold passage is in the context for nearly every question. Yet human graders mark 35% of answers wrong, and the wrong answers often ignore or contradict the gold passage. Where should the team invest next?
- Your product requires every RAG answer to include clickable source citations that users can verify. How should those citations be produced?
- A runbook RAG system uses fixed 500-token chunks with no overlap. Users who ask for a multi-step recovery procedure get answers that stop at step 4 of 9, and retrieval logs show the top hit ends mid-procedure. What is the most targeted fix?