Voy
A portable WebAssembly vector similarity search engine written in Rust with a tiny footprint (75KB gzipped). Designed for edge deployment, browsers, and IoT devices with support for k-d tree indexing and optimized for modern web applications.
About this tool
Overview
Voy is a WASM vector similarity search engine written in Rust with an exceptionally small footprint of only 75KB gzipped (69KB brotli). It enables semantic search in browsers, edge environments, and IoT devices without requiring server-side infrastructure.
Key Features
- Tiny Footprint: Only 75KB gzipped, 69KB with brotli compression, reducing overhead for limited devices like mobile browsers and IoT
- Fast: Optimized for creating the best search experience with efficient k-d tree indexing
- Tree Shakable: Optimizes bundle size and enables asynchronous capabilities for modern Web APIs like Web Workers
- Resumable: Generate portable embeddings index anywhere, anytime
- Worldwide: Designed to deploy and run on CDN edge servers
- Browser Support: Runs in non-Node environments including all modern browsers
Technical Implementation
- Data Structure: Uses k-d tree (k-dimensional tree) to organize and store embeddings
- Distance Metric: Internally uses Squared Euclidean distance to calculate nearest neighbors
- WASM: Compiled to WebAssembly from Rust for portability and performance
Design Principles
Voy follows several core design principles:
- Minimize size for limited devices (mobile browsers, IoT)
- Optimize for fast search performance
- Enable tree shaking for bundle optimization
- Support modern Web APIs like Web Workers
- Create portable, resumable indexes
- Enable edge deployment on CDNs
Integration
- LangChain.js: Voy can be used as a vector store with LangChain.js
- npm Package: Available as
voy-searchon npm - Installation: Can be installed via npm, Yarn, or pnpm
Simple Interface
Voy provides a straightforward API with just two main operations:
- Index construction
- Querying an index for nearest neighbors
Development Status
Voy is under active development. The API is not yet stable, and there may be breaking changes before the 1.0 release.
Use Cases
- Client-side semantic search in web applications
- Edge computing vector search
- Mobile browser search functionality
- IoT device similarity search
- Offline-first applications
- CDN-deployed search services
Resources
- GitHub: https://github.com/tantaraio/voy
- npm: voy-search package
- Language: Rust (compiled to WASM)
Surveys
Loading more......