Design the ingestion pipeline for a RAG system with frequent document updates.

Instruction: Explain how you would ingest, update, and retire documents for a rapidly changing knowledge base.

Context: Assesses whether the candidate can design a practical architecture and explain the main tradeoffs. Explain how you would ingest, update, and retire documents for a rapidly changing knowledge base.

Official answer available

Preview the opening of the answer, then unlock the full walkthrough.

I would build the ingestion pipeline as an event-driven system with explicit document identity and versioning. The core idea is that updates should propagate incrementally. A changed file or record should trigger parsing, normalization, chunking, embedding, and index updates only for the affected content, not a...

Upgrade to view official answer

Related Questions