PostgreSQL has a long history of handling multidimensional data, with geospatial searches through the PostGIS extension. This layed the ground for the jump from a few dimensions (2D or 3D) to thousands of dimensions with the `pgvector` extension. pg_vector allows AI/ML systems to run directly on PostgreSQL and is an example of the richness of PostgreSQL’s open-source ecosystem, offering robust support for storing and querying high-dimensional embeddings.

In this talk, we’ll explore how vector-based **retrieval augmented generation (RAG)** is a natural progression from standard SQL queries and full-text searches. We will demonstrate how large language models (LLMs) interact with `pgvector`, how embeddings are stored, and which indexing strategies work best for high-dimensional data.

Join me to discover how `pgvector` can help future-proof your organization by powering recommendation systems and more.