RAG-03 · Index design, schema change and cut-over¶
Index schema mistakes such as a wrong analyser, a missing filterable flag or the wrong vector dimensions only show up late and most of them need a full rebuild. Teams then reindex in place and take an outage because nobody wrote down the safe way.
- Who reads it: GenAI engineers designing Azure AI Search indexes and DevOps engineers operating them in production
- Outcome: Zero-downtime reindex, every time
- Pack owner: Yogeshwaran (catalog owner role: RAG Lead)
Master catalog entry¶
- Catalog name: Index Design & Schema Guide (P1)
- Why it matters: Index design mistakes surface late and are expensive to reverse
- What it removes: Field, analyser, vector profile and scoring decisions pre-made
- Stack: Azure AI Search, hybrid search, semantic ranker
- Maturity target: Level 3 — schema templates for four corpus archetypes
Playbooks¶
| ID | Story | Title | Kind | Complexity | Urgency | Status | Owner | Planned publish |
|---|---|---|---|---|---|---|---|---|
| PB-RAG-03-01 | RAG-03-S1 | Choose fields, analysers, vector profiles and ranking for four corpus archetypes | Decision guide | High | 3 | draft | Yogeshwaran | 2026-12-07 |
| PB-RAG-03-04 | RAG-03-S2 | Index go-live and capacity checklist | Checklist | Medium | 4 | draft | Kandiah Chockalingam | 2026-11-20 |