Retrieval Patterns for RAG Pipelines
The lesson demonstrates how to reproduce a search index as a PostgreSQL table of document chunks with embeddings and token counts, covering chunk size selection, source document references, adding preceding and following chunks for better context, and testing the RAG setup with accuracy, precision, and recall metrics before integrating the LLM.
Previous Module
Introducing Azure Database for PostgreSQL
This Module
AI Vector Search on PostgreSQL
Introduction to pgvector
Store and Query Embeddings with pgvector
Vector Similarity Search Indexes
Vector Similarity Search for Semantic Retrieval
Retrieval Patterns for RAG Pipelines
Optimize Vector Search in Azure Database for PostgreSQL
Tune PostgreSQL for pgvector
Choose and Configure Vector Indexes
Optimize Data Layout for Filtered Search
Scale PostgreSQL for High-Volume Workloads