ruvectoredge

name: "@ruvector/edge" description: "Edge AI swarms for browsers with P2P networking, vector search, and neural networks. Use when the user needs browser-based AI swarms, peer-to-peer vector search, client-side neural network inference, decentralized agent coordination, or edge-deployed AI workloads without server infrastructure."

@ruvector/edge

Free edge-based AI swarms in the browser with peer-to-peer networking, vector search, cryptographic security, and neural network inference -- all running client-side with zero server dependencies.

Quick Command Reference

Installation

Hub install (recommended): npx ruvector@latest includes this package. Standalone: npx @ruvector/edge@latest See Installation Guide for the full ecosystem.

Core API

EdgeSwarm Constructor

import { EdgeSwarm } from '@ruvector/edge';

const swarm = new EdgeSwarm({
  workers: 4,                    // Number of Web Workers
  vectorDimensions: 384,         // Vector DB dimensions
  enableP2P: true,               // Enable peer-to-peer
  enableNeural: true,            // Enable neural inference
  signalingServer: 'wss://...',  // WebRTC signaling server
});

Constructor Options:

Swarm Operations

await swarm.start();                              // Start all workers
await swarm.stop();                               // Stop gracefully
await swarm.submit({ type: 'search', query });    // Submit task
await swarm.broadcast(message);                   // Broadcast to workers
const status = swarm.status();                    // Get status
swarm.addWorker(config);                          // Add worker at runtime
swarm.removeWorker(workerId);                     // Remove worker

Vector Search (Edge)

// Insert vectors into edge DB
await swarm.vectorInsert('doc-1', vector, metadata);

// Search across edge swarm
const results = await swarm.search(queryVector, 10);

// Distributed search across peers
const results = await swarm.distributedSearch(queryVector, {
  topK: 10,
  peerTimeout: 5000,
});

P2P Networking

await swarm.connect(peerId);                       // Connect to peer
await swarm.disconnect(peerId);                    // Disconnect
await swarm.sendToPeer(peerId, data);              // Send data
swarm.onPeerMessage((peerId, data) => { ... });    // Listen
const peers = swarm.getPeers();                    // List peers

Neural Inference

// Load ONNX model
await swarm.loadModel('model.onnx');

// Run inference
const output = await swarm.infer(inputTensor);

// Distributed inference across workers
const output = await swarm.distributedInfer(inputTensor, { splitStrategy: 'layer' });

Common Patterns

Browser-Based RAG

const swarm = new EdgeSwarm({ workers: 4, vectorDimensions: 384 });
await swarm.start();
// Load pre-computed embeddings
for (const doc of documents) {
  await swarm.vectorInsert(doc.id, doc.embedding, { text: doc.text });
}
// Search on user query
const results = await swarm.search(queryEmbedding, 5);

Peer-to-Peer Collaborative Search

const swarm = new EdgeSwarm({ enableP2P: true, signalingServer: 'wss://signal.example.com' });
await swarm.start();
// Each browser tab becomes a peer
const results = await swarm.distributedSearch(query, { topK: 20, peerTimeout: 3000 });

Offline-First AI

const swarm = new EdgeSwarm({ workers: 2, enableNeural: true });
await swarm.start();
await swarm.loadModel('/models/classifier.onnx');
const prediction = await swarm.infer(featureVector);
// Works completely offline

Key Options

RAN DDD Context

Bounded Context: Edge/WASM Runtime

References

• API reference: See references/commands.md
• Full README
• npm