RAG-09 · Knowledge base setup and refresh¶
Teams build the first index and leave refresh until later, so changed or deleted documents keep appearing in answers and adoption quietly drops. When a reindex, embedding upgrade or indexer failure happens, nobody knows the safe procedure, the risks or what to watch.
- Who reads it: GenAI engineers and DevOps engineers running RAG knowledge bases on Azure AI Search in production
- Outcome: Zero-downtime reindex and fresh content, every time
- Pack owner: Yogeshwaran (catalog owner role: RAG Lead)
Master catalog entry¶
- Catalog name: Knowledge Base Setup & Refresh Guide (P2)
- Why it matters: Stale knowledge bases quietly destroy adoption after go-live
- What it removes: Incremental refresh and change detection designed in from the start
- Stack: Azure AI Search indexers, change feed, Service Bus
- Maturity target: Level 2 — full, incremental and event-driven refresh patterns
Playbooks¶
| ID | Story | Title | Kind | Complexity | Urgency | Status | Owner | Planned publish |
|---|---|---|---|---|---|---|---|---|
| PB-RAG-09-01 | RAG-09-S1 | Reindex an Azure AI Search index without downtime | Runbook | Medium | 5 | draft | Yogeshwaran | 2026-11-05 |
| PB-RAG-09-02 | RAG-09-S2 | Upgrade the embedding model and re-embed the corpus | Runbook | High | 5 | draft | Yogeshwaran | 2026-11-13 |
| PB-RAG-09-04 | RAG-09-S3 | Set up incremental and event-driven refresh with change and deletion detection | How-to | Medium | 4 | draft | Kandiah Chockalingam | 2026-12-01 |
| PB-RAG-09-05 | RAG-09-S4 | Deletion and permission change propagation checklist | Checklist | Low | 4 | draft | Kandiah Chockalingam | 2026-11-24 |