Generative AI Use Cases with pgvector, Aurora PostgreSQL, and Amazon Bedrock
A hands-on workshop of labs that build semantic search, recommendations, retrieval-augmented generation and agentic incident remediation using pgvector, Aurora PostgreSQL and Amazon Bedrock.
This repository holds the hands-on labs for a workshop on generative AI with pgvector and Aurora PostgreSQL, building from vector-search fundamentals into retrieval-augmented generation, in-database ML inference, semantic caching and agentic operational patterns. Labs can be run through a hosted Workshop Studio experience for the fastest path through the material, or self-paced in a participant's own AWS account to inspect and adapt the code.
The labs build semantic search over structured and unstructured content using HNSW indexes and cosine distance, product and movie recommendation flows powered by embeddings, RAG applications built with LangChain that retrieve context from Aurora PostgreSQL and generate answers through the Bedrock Converse API, an e-commerce catalog with semantic search and insights, and agentic incident remediation implemented two ways, managed Bedrock Agents and a code-first Strands Agents implementation. A separate lab adds semantic caching with ElastiCache for Valkey to cut latency and model cost on repeated queries.
The labs include Streamlit apps, notebooks and Lambda functions.