High-performance vector database for the browser
npm install surgedb-wasm
SurgeDB WASM brings high-performance, persistent vector search directly to the browser.
It is a lightweight WebAssembly binding for the SurgeDB engine, optimized for edge devices and client-side applications. Run semantic search, recommendation systems, and RAG pipelines entirely offline without sending user data to a server.
---
* 🔒 Privacy First: All data stays on the client. No API calls, no data leaks.
* ⚡ Sub-millisecond Search: Powered by SIMD-accelerated HNSW indices (Rust).
* 📦 Tiny Footprint: ~100KB gzipped (WASM + JS).
* 💾 Memory Efficient: Built-in SQ8 quantization reduces memory usage by 4x.
* 🛠️ TypeScript: Fully typed API for seamless development.
---
``bash`
npm install surgedb-wasm
---
SurgeDB is designed to work seamlessly with modern bundlers (Vite, Webpack, etc.).
`javascript
import init, { SurgeDB } from 'surgedb-wasm';
async function main() {
// 1. Initialize the WASM module
await init();
// 2. Create the database (384 dimensions for MiniLM)
const db = new SurgeDB(384);
// 3. Add documents
db.insert("doc_1", new Float32Array([0.1, 0.2, 0.3, ...]), {
title: "Client-side AI",
category: "Web"
});
db.insert("doc_2", new Float32Array([0.4, 0.5, 0.6, ...]), {
title: "Rust in Browser",
category: "Web"
});
// 4. Search
const query = new Float32Array([0.1, 0.2, 0.3, ...]);
const results = db.search(query, 5); // Start with top 5
console.log(results);
// Output: [{ id: "doc_1", score: 1.0, metadata: {...} }, ...]
// 5. Clean up memory when component unmounts
db.free();
}
main();
`
---
For larger datasets (10k+ vectors), use SurgeDBQuantized to enable SQ8 compression. This slightly reduces precision but drastically lowers memory usage.
`javascript
import init, { SurgeDBQuantized } from 'surgedb-wasm';
await init();
const db = new SurgeDBQuantized(384);
// Usage is identical to standard SurgeDB
db.insert("id", vector, metadata);
`
`tsx
import { useEffect, useState } from 'react';
import init, { SurgeDB } from 'surgedb-wasm';
export function VectorSearch() {
const [db, setDb] = useState
useEffect(() => {
init().then(() => {
const database = new SurgeDB(384);
setDb(database);
});
return () => db?.free();
}, []);
const handleSearch = (queryVec: Float32Array) => {
if (!db) return;
const results = db.search(queryVec, 10);
console.log(results);
};
return
---
Performance Benchmark
Benchmarks run on an M2 MacBook Air (Chrome):
| Operation | Metric |
| :--- | :--- |
| Search Latency | ~0.5ms (1k vectors) |
| Insert Speed | ~50ms (1k vectors) |
| Memory (Quantized) | ~400KB per 1k vectors |
| Memory (Standard) | ~1.5MB per 1k vectors |
---
Development & Building
If you are contributing to the project or building from source:
bash
Install wasm-pack
cargo install wasm-packBuild for npm (generates pkg/ directory)
wasm-pack build --target web --release
``