NeuronDB

What is NeuronDB?

NeuronDB is a PostgreSQL extension that adds vector search, machine learning inference, and RAG pipeline capabilities directly in your database. Implemented in pure C with zero dependencies.

How does NeuronDB work?

NeuronDB extends PostgreSQL with 665+ SQL functions for vector operations, ML inference, and RAG workflows. It uses HNSW indexing for fast vector search and supports GPU acceleration for batch operations.

Why is NeuronDB different?

NeuronDB provides a complete AI stack in PostgreSQL: vector search, 52 ML algorithms, RAG pipelines, and agent infrastructure. No external services required. Everything runs in your database.

Where does NeuronDB run?

NeuronDB runs as a PostgreSQL extension on PostgreSQL 16, 17, and 18. Supports Linux, macOS, and Windows. GPU acceleration available for CUDA, ROCm, and Metal.

Do I need a separate vector database or ML service?

No. NeuronDB provides vector search, 52 ML algorithms, embedding generation, and a full RAG pipeline inside PostgreSQL. You can keep all data and logic in one database and optionally use NeuronMCP and NeuronAgent for MCP tools and agent runtime.