Repository files navigation

HelixDB: an open-source graph-vector database written in Rust built for RAG and AI applications.

Homepage |Docs |Discord |X

HelixDB is a high-performance graph-vector database designed with a focus on developer experience and performance. Built in Rust and powered by LMDB as its storage engine, it combines the reliability of a proven storage layer with modern features tailored for AI and vector-based applications.

We are currently using LMDB via Heed3, a rust wrapper built by the amazing team over atMeilisearch.

• Fast & Efficient: Built for performance we're currently 1000x faster than Neo4j, 100x faster than TigerGraph and on par with Qdrant for vectors.
• RAG-First: Native support for graph and vector data types, making it ideal for RAG (Retrieval Augmented Generation) and AI applications
• Graph-Vector: Easiest database for storing relationships between nodes, vectors, or nodes AND vectors.
• Reliable Storage: Powered by LMDB (Lightning Memory-Mapped Database) for robust and efficient data persistence
• ACID Compliant: Ensures data integrity and consistency

The Helix CLI tool can be used to check, compile and deploy Helix locally.

1. Install CLI

   curl -sSL"https://install.helix-db.com"| bash

2. Install Helix

   helix install

3. Setup

   helix init --path<path-to-project>

4. Write queries

   Open your newly created.hx files and start writing your schema and queries.Head over toour docs for more information about writing queries

   QUERYaddUser(name:String,age:I64)=>user<-AddN<User({name:name,age:age})RETURNuserQUERYgetUser(user_name:String)=>user<-N<User::WHERE(_::{name}::EQ(user_name))RETURNuser

5. Check your queries compile before building them into API endpoints (optional)

   # in ./<path-to-project>helix check

6. Deploy your queries

   # in ./<path-to-project>helix deploy

7. Start calling them using ourTypeScript SDK orPython SDK. For example:

   importHelixDBfrom"helix-ts";// Create a new HelixDB client// The default port is 6969constclient=newHelixDB();// Query the databaseawaitclient.query("addUser",{name:"John",age:20});// Get the created userconstuser=awaitclient.query("getUser",{user_name:"John"});console.log(user);

Other commands:

• helix instances to see all your local instances.
• helix stop <instance-id> to stop your local instance with specified id.
• helix stop --all to stop all your local instances.

Our current focus areas include:

• Expanding vector data type capabilities for RAG applications
• Implementing a test suite to enable end-to-end testing of queries before deployment
• Building a Deterministic Simulation Testing engine enabling us to robustly iterate faster
• Binary quantisation for even better performance

Long term projects:

• In-house graph-vector storage engine (to replace LMDB)
• In-house network protocol & serdes libraries (similar to protobufs/gRPC)

• All code should be well-documented so no additional documentation is needed.
• All code should be well-tested (DST coming soon).
• If you can build it yourself, do it. Reduce dependencies.
• If it doesn't need to exist, don't build it.
• Don't lose users' data.
• Use asserts in production: make an effort towards putting asserts anywhere you see fit even if the assert should always obviously be true.
• If there is a more functional way to write a piece of code, do so. ie. pattern matching over if statements, iterators+combinators over loops, closures over small 1-2 line functions, etc.
• When optimizing pieces of software, first write an eval/benchmark so that you can measure an exact metric for performance change

HelixDB is licensed under the The AGPL (Affero General Public License).

HelixDB is available as a managed service for selected users, if you're interested in using Helix's managed service or want enterprise support,contact us for more information and deployment options.