Skip to content

Introduction

GhostSQL is a high-performance, PostgreSQL-compatible SQL database written. It is designed for modern applications that require both scalable relational data and fast vector search for AI/ML workloads.

Vision

The goal of GhostSQL is to bridge the gap between traditional relational databases and specialized vector stores. By providing a single, unified interface that speaks the PostgreSQL protocol, GhostSQL allows developers to use their favorite tools while gaining native vector acceleration.

Key Principles

  • Protocol Compatibility: Works with standard psql and PostgreSQL drivers.
  • Vector First: High-performance embeddings storage and HNSW indexing.
  • Relational Integrity: Support for JOINs, Foreign Keys, and ACID properties.
  • Performance: Written in Go with efficient on-disk binary storage.

Use Cases

  • RAG Applications: Store documents and their embeddings in the same table.
  • Semantic Search: Build intelligent search into existing relational applications.
  • Edge Computing: Lightweight, single-binary database for distributed environments.