How would you keep a RAG index fresh through updates and deletes?

Coordinates source changes, versions, tombstones, reconciliation, and query-time freshness without treating the index as authoritative.

1Say itAnswer out loud first

2CompareCompare with Core answer

3RefineRefine with Stronger and Advanced

TL;DR

I consume source change events or scheduled deltas, reprocess changed versions idempotently, and remove or tombstone obsolete chunks when a source is deleted or revoked. Each record carries source and indexed timestamps so freshness can be measured. The source system remains authoritative, and high-risk queries can reject evidence older than an explicit policy allows.

Full answer below

Practice focus

Change propagation, deletion, stale content, index publication, and recovery.

Answer first

Answer path

Core answer first, then checkpoints toward a stronger answer.

Start
Core answer
Start here. This is the direct spoken answer to practice first.
Overview
RAG freshness depends on the delay and correctness of the ingestion path, not on the language model.
I consume source change events or scheduled deltas, reprocess changed versions idempotently, and remove or tombstone obsolete chunks when a source is deleted or revoked. Each record carries source and indexed timestamps so freshness can be measured. The source system remains authoritative, and high-risk queries can reject evidence older than an explicit policy allows.
Checkpoint
Stronger answer
Optional upgrade with examples, trade-offs, and sharper phrasing.
For a policy update, I build the new document version, validate its chunks, and make it queryable before retiring the old version to avoid a gap. A reconciliation job compares the source catalog with the index to catch missed events and orphaned records. Dashboards show ingestion lag, stale-result rate, delete backlog, and failures by source connector.

Review signals

Use these to catch weak phrasing, shallow reasoning, and risky claims.

Common mistakesCandidateCommon mistakesMistakes that weaken the answer or signal shallow understanding.Open details

Details

Updating changed documents without removing old chunks.
Assuming a source connector never misses events.
Caching retrieval results without a freshness policy.

How would you keep a RAG index fresh through updates and deletes?

Practice focus

Answer path

Overview

Review signals

Follow-up prompts

Continue practice

Next in path

Tech stacks

Skill areas

Human editorial direction

Was this drill useful?

Found this answer useful?