# HelixDB — Graph-Vector Database Built for AI Workloads

> A combined graph and vector database written in Rust, designed for RAG and AI applications that need both relationship traversal and similarity search.

## Install

Save in your project root:

# HelixDB — Graph-Vector Database Built for AI Workloads

## Quick Use
```bash
curl -fsSL https://install.helix-db.com | bash
helix init my_project
helix deploy
```
```
# Define a schema in schema.hx
N::Document {
  content: String,
  embedding: F32
}
```

## Introduction
HelixDB is an OLTP graph-vector database written in Rust that combines graph traversal with vector similarity search in a single engine. It targets AI and RAG applications where data has both relational structure and embedding-based semantics.

## What HelixDB Does
- Stores nodes, edges, and vector embeddings in a unified data model
- Supports graph traversal queries alongside vector similarity search in a single query language
- Provides a schema definition language (`.hx` files) for declaring node and edge types
- Deploys as a serverless function or self-hosted service via a CLI tool
- Exposes a TypeScript/JavaScript client SDK for application integration

## Architecture Overview
HelixDB stores graph data and vector indexes on object storage with an in-memory cache layer for hot data. The query engine processes a custom query language (HelixQL) that combines graph pattern matching with vector search operators. Vector indexes use HNSW for approximate nearest neighbor search. The Rust implementation provides memory safety and consistent low-latency performance. The CLI manages schema migrations, deployment, and local development.

## Self-Hosting & Configuration
- Install the CLI with the official install script or from source via Cargo
- Define your data model in `.hx` schema files with typed node and edge declarations
- Run `helix deploy` to deploy to HelixDB Cloud or configure a self-hosted instance
- Use `helix dev` for local development with a built-in development server
- Integrate via the TypeScript SDK or the HTTP API for other languages

## Key Features
- Unified graph + vector model eliminates the need to sync data between separate databases
- HelixQL query language combines graph traversal and vector search in one syntax
- Schema-first design with typed nodes and edges catches errors at definition time
- Built in Rust for memory safety and consistent performance under concurrent load
- Serverless deployment option reduces operational overhead for smaller workloads

## Comparison with Similar Tools
- **Neo4j** — Mature graph database without native vector support; HelixDB combines both natively
- **Milvus** — Pure vector database; HelixDB adds graph relationships to vector search
- **Qdrant** — Vector search with payload filtering; HelixDB provides full graph traversal
- **Dgraph** — GraphQL-native graph DB; HelixDB integrates vector similarity alongside graph queries
- **Weaviate** — Vector DB with cross-references; HelixDB offers richer graph modeling primitives

## FAQ
**Q: When should I use HelixDB over a standalone vector database?**
A: When your data has meaningful relationships (user-document, document-topic) that should influence retrieval alongside embedding similarity.

**Q: What languages are supported for the client SDK?**
A: TypeScript/JavaScript is the primary SDK. Other languages can use the HTTP API directly.

**Q: Can it handle large-scale production workloads?**
A: HelixDB is designed for OLTP workloads. Performance scales with the underlying object storage and cache configuration.

**Q: Does it support standard graph query languages like Cypher or Gremlin?**
A: No. HelixDB uses its own HelixQL language optimized for combined graph and vector operations.

## Sources
- https://github.com/HelixDB/helix-db
- https://docs.helix-db.com/

---
Source: https://tokrepo.com/en/workflows/asset-2a45f020
Author: AI Open Source