Giving Agents a Memory: Vector Search with PostgreSQL

A hands on demo with example implementation of how to build a completely local AI memory system using bx-ai and postgresql. This talk covers many foundational AI development in a fun and approachable with broad application.

Large Language Models are stateless, which means building a truly useful AI agent requires giving it a long-term memory. While purpose-built vector databases get a lot of attention, you might already have the perfect tool in your stack: PostgreSQL. In this session, we will bridge the gap between traditional relational data and AI embeddings. We will explore how to architect an agentic memory system using Postgres and pgvector, discussing how to generate embeddings, perform efficient similarity searches, and seamlessly blend traditional SQL filtering with AI retrieval.

Jacob Beers

Jacob Beers

Find out more