Skip to main content
Vector is a semantic search database built on SlateDB and object storage. Each database stores documents conforming to a user-defined schema. The schema defines fields with various types, along with a “vector” field that stores dense vectors with a fixed number of dimensions. Vector serves search queries that find the nearst N documents to a given query vector. Because all data is durably stored on object storage, a single replica provides full durability with no operational overhead for data replication or backups. Writes are limited to a single database instance, but queries can be scaled horizontally across multiple stateless reader instances.

Quickstart

Install Vector and write your first documents in under five minutes

API Reference

Browse the full REST API for writing and searching vectors

Data Model

Understanding Vector’s Data Model

Configuration

Vector configuration reference

Storage Design

How Vector indexes documents to support efficient similarity search

Getting to Production

Deploy, monitor, and secure Vector for production workloads

GitHub

View the source code, open issues, and contribute