Vector Stores & Index Operations

Practise and Check

Deliverable

Living index: stand up pgvector with Docker, ingest the M11 corpus, then change, add and delete source files and show that the record manager re-indexes only what changed.

Assessment

Answer the five Interview questions without notes, then run the lab.

Interview probe

Larkspur cut the annual refund window from 30 days to 14 and the nightly sync ran. A week later the assistant still tells some customers "30 days". The new chunk is in the index. What happened, and how do you fix it?

Common mistakes

• Minting a fresh UUID for every chunk on each ingestion run
• Relying on a tenant metadata filter for access control
• Running full cleanup when the loader returned only some files

Related Modules

• T5.1Why RAG & the RAG Lifecycle
• T5.2Document Ingestion & Metadata Design
• T5.3Chunking Strategies
• T5.4Embeddings & Model Selection