HelixDB: Graph Database with Vector and FTS on Object Storage

HelixDB, a graph database developed by a team from college, has been launched, offering an OLTP (Online Transaction Processing) solution built on object storage. This innovative database integrates native vector search and full-text search capabilities, addressing the common challenge of combining these functionalities across disparate systems in AI-driven applications. The core innovation lies in its architecture, which utilizes object storage like S3 as its persistence layer. This design allows for virtually unlimited graph size and relationships, providing a cost-effective solution for storing terabytes of data. It also enables horizontal scaling by spinning up nodes and caching relevant data subsets, achieving low latency for frequently accessed "hot" data while keeping "cold" data in cheap storage. HelixDB is particularly suited for AI memory systems and applications requiring extensive data traversal and search, such as GraphRAG and HybridRAG. The developers are also working on an open-source generalized AI memory layer powered by HelixDB, pre-filtering for vector search, and an upcoming GA cloud offering.