RAG Interview Questions
Dashboard
0%
1
Curious builder0 XP earned · 300 to level 2
0 daysFinish a lesson to begin
Badge collection0 of 6 unlocked
51 small wins to finish your pathNext question →

Q27IntermediateScenario

Users complain the chatbot gives outdated answers even though documents were updated. How do you fix it?

30-second answerSay your answer out loud first, then reveal.
A change detector fed by webhooks, change feeds or modified-since polling sends new or updated documents through parse, chunk and embed to an upsert by doc_id and version, sends deleted or superseded documents to a delete of their old chunks, and both update the index, which then invalidates the answer cache for affected docs.

Root causes to check

  1. Sync lag / failures: the nightly job failed silently; the connector missed some folders.
  2. Duplicates: the new version was added, but the old chunks were never deleted (common with append-only pipelines).
  3. Stable IDs missing: without a stable doc_id, you can't find old chunks to replace.
  4. Caches: semantic or answer caches still serving old answers.
  5. Ranking: the old doc is phrased closer to the query.

Fixes

  • Upsert by doc_id: on update, delete all chunks for that doc_id, then insert the new ones (or version them and mark old ones status=superseded).
  • Change detection: webhooks or change feeds where available; otherwise compare content hashes and modified timestamps.
  • Version metadata: effective_date, version, status. Filter to current, or apply a recency boost.
  • Freshness SLA + monitoring: alert when a source hasn't synced in X hours; track index lag; periodic full reconciliation (compare source doc list with indexed doc list).
  • Show dates in citations so users can see how fresh a source is.
PreviousExplain the core RAGAS metrics: faithfulness, answer relevancy, context precision, and context recall.

You understood something today that you didn't yesterday.