HelixDB
An open-source graph and vector database for AI memory and RAG, with full-text search and embedded or self-hosted deployment under Apache 2.0.
HelixDB is a database for developers building AI agent memory and retrieval-augmented generation (RAG). It keeps entities, relationships, embeddings, and text together so applications can search both content and its connections.
- Graph and semantic search: Combine relationship traversal with vector similarity to retrieve connected information for agent memory, document search, or code intelligence.
- Keyword and temporal search: Search node and relationship properties with BM25 full-text search, query by time, and filter candidates before ranking keyword or vector results.
- Atomic writes: Commit graph and vector changes in a single transaction, or roll them back together, for applications that need consistent updates.
- Tenant isolation: Keep each tenant's data separate while using shared infrastructure.
You can embed HelixDB in an application on macOS, Linux, or Windows, with data in memory or on disk. A self-hosted Docker server also supports object storage. Local development instances use memory by default, with disk persistence available. The project is open source under Apache 2.0.
Rust, TypeScript, Python, and Go SDKs let applications query the database. Source builds also support Cypher reads and data changes. Helix Cloud is a separate managed service with high availability, durable object storage, and reader nodes that scale with query demand.
