OpenData Vector introduces a new approach to vector search databases, offering MIT-licensed software that runs directly on object storage. The system targets organizations caught between basic pgvector implementations and expensive managed services, promising production-ready performance without vendor lock-in.

Built on SlateDB, OpenData Vector represents the third generation of vector databases with a stateless architecture that eliminates node coordination overhead. Unlike earlier systems that maintained local state, this approach stores all metadata and data in object storage, enabling any node to serve any request. The design leverages IVF indexing and LSM-based compaction to optimize for object storage's high-latency characteristics.

The project claims remarkable efficiency, supporting 100 million vectors for approximately $350 monthly while maintaining 99.999999999% durability through object storage SLAs. Multiple deployment topologies accommodate different operational requirements, from embedded single-pod setups to distributed configurations with buffered ingestion. This positions OpenData Vector as a compelling option for teams seeking enterprise-grade vector search capabilities without the traditional operational complexity or licensing costs.