Skip to content
Turns a Web Browser into a WebRTC Web Application Server
Branch: master
Clone or download
Latest commit fbfa547 Jun 7, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
packages docs | build instructions Jun 6, 2019
.gitignore dev dependency update | windows support Jun 6, 2019
.npmrc publish Apr 12, 2019
.travis.yml publish Apr 12, 2019
license publish Apr 12, 2019
package.json update pack tasks Jun 6, 2019

NPM package Build Status


A framework for building Web Server applications in the browser over WebRTC.

$ npm install smoke-node --save
import { Node } from 'smoke-node'

const node = new Node()

const app =

app.get('/', (req, res) => {

  res.send('hello world')

const text = await'/').then(n => n.text())


Smoke is an experimental peer to peer networking framework that allows Web Browsers to run as lightweight Web Servers that operate over WebRTC. It offers a set of APIs to run both HTTP and Web Socket server like functionality in the browser as well as a set of Web like APIs to consume content hosted in remote browsers.

Communication between browsers operates entirely peer to peer with each network Node able to support hundreds of concurrent connections. New server nodes may be deployed when users load web pages, with some potential to scale node infrastructure proportional to the number of users loading pages.

Additionally, this library provides two storage mechanisms for persisting object and file data by leveraging IndexedDB. Nodes can host file and data in much the same way as one would with a traditional file or api server, with IndexedDB offering gigabytes of storage at each node.

This framework was written primarily as a tool to prototype various peer to peer networking architectures. It aims to offer a baseline for exploring various decentralized and distributed technologies using the network and storage capabilites available in modern browsers.

This framework is offered as is to anyone who finds it of use. Built with and tested with Chrome 72, Firefox 65 and Electron 4.0.4.

Released under MIT



This repository contains both Node and Hub projects as well as a Workbench project that can be used to test and script around the smoke Node and Hub.

To build locally.

$ git clone
$ cd ./smoke
$ npm install

# start node | hub | workbench in watch mode.
$ npm run bench 

This will start the workbench web app running a smoke Node. The workbench is accessible on http://localhost:5000 and the Hub signalling server is accessible on http://localhost:5001.


A Hub is the name given to Smoke's WebRTC signalling infrastructure. It is a forwarding channel that Nodes use to relay messages to other Nodes (primarily WebRTC SDP and Candidate exchange messages). A Hub can loosely be thought of as a network Router or sorts. All Nodes connect to at least one Hub, and by doing so, the Node will be joining a network of other Nodes also connected to that Hub.

Smoke Hubs provide the following functionality:

  • They provide ICE configuration for Nodes (STUN / TURN)
  • They provide a each Node a unique address. (DHCP)
  • They provide SDP, ICE Candidate forwarding services for WebRTC (ICE)

From a application standpoint, Hubs are intended to be fairly transparent to applications (in much the same way as one doesn't usually give much thought to a home router when connecting to devices in a local area network), but they are important; forming the backbone of a peer network and playing a central role in describing the overall topology and partitioning of a network.

Smoke provides two built in Hub types:


A PageHub is an in memory in-page signalling Hub. It allows for multiple Nodes to connect to each other so long as those Nodes all belong to the same page. This is the default Hub used when passing no arguments when creating a Node. It can be used for testing, scripting and general demonstration purposes.

import { Node } from 'smoke-node'

const node = new Node() // uses the page hub

node.sockets.createServer(socket => { ... })


A NetworkHub provides over the network signalling for Nodes. This is a more traditional signalling service where ICE messages are able to be exchanged over the public networks. This project provides a reference web socket based hub server implementation that can be installed and run with the following.

$ npm install smoke-hub -g

$ smoke-hub --port 5000

The following assumes the above hub is started on localhost.

import { Node, NetworkHub } from 'smoke-node'

const node = new Node({ hub: new NetworkHub('ws://localhost:5000') })

The PageHub and NetworkHub are both implementations of Hub. Smoke supports implementation of custom Hub types by creating new Hubs that implementation the Hub interface, allowing Hubs to be implemented around existing server infrastructure, or within a peer to peer network itself.


A Node can be thought of as a process running somewhere on the network. All Nodes have network addresses (provided by their Hub), and each may expose zero or more ports to the network. Services are bound to ports in much the same way network servers bind to ports to receive connections.

New Nodes can be created by calling the Node constructor. The following code creates 3 Nodes to match the network above, and makes a socket connection from node2 to node3.

import { Node } from 'smoke-node'

const node1 = new Node() //
const node2 = new Node() //
const node3 = new Node() //

// start server on node3 and listen on port 5000

node3.sockets.createServer(socket => {



// connect to node2 from node3 server

const socket = node2.sockets.connect('', 5000)

socket.on('message', message => {


Note that the addresses allocated from the PageHub are predictable and reset to on page refresh. Each new node entering the network will be 0.0.0.[n+1] up to 255. Applications should not place any special meaning on these addresses other than them being plain random strings. The scheme used by the PageHub and smoke-hub implementations are for convenience only.

Note that Nodes can expose one or more ports. Ports in nodes are inferred from the RTCDataChannel label property. Connection attempts to non open ports on a smoke node result in the immediate closing of that connection.


Nodes house several APIs that allow them to function as network application servers. Many of the APIs provided by this library are based on NodeJS core and community modules that have been rebuilt from the ground up to operate over WebRTC. The following sections provide a high level overview of the APIs available on each Node instance.

const { system, network, hub, sockets, rest, media } = new Node()


Provides access to this Nodes uptime, network and storage metrics.


Provides access to the lowest levels of the node network stack. Allows for the binding and unbinding of ports, creating and listening for RTCDataChannels and provides direct access to the pool of RTCPeerConnections managed for this node.


Provides access to the signalling hub this node is connected to. Allows one to resolve their address within the signalling network.


Provides an interface to create and connect to socket server endpoints within the peer network. The sockets provided by this API are designed to function as typical Web Sockets. They layer RTCDataChannel to offer predictable network timeout, address resolution and sending and receiving message payloads that exceed RTCDataChannel limits.


Provides an interface to create and connect to HTTP like endpoints over WebRTC. The RestServer and Fetch APIs implement full request response semantics allowing for the transmission of data using familiar mechanisms used to send and receive data over HTTP. The RestServer also allows for the hosting of MediaStream content.


Provides an interface to create mediastreams from readable byte streams, such as those read from the IDB file system or received over the network. It also provides a test pattern mediastream source that can be used to test pass through without needing to setup stream sources (such as web camera feeds, etc)

Files and Data

Smoke provides two storage mechanisms for persisting both file and object data in the browser. Both of these mechanisms operate over IndexedDB, with one modelled on data persistense and query (Database), and the other binary file persistence and streaming (Buckets)


A transactional object store over IndexedDB.

import { Database } from 'smoke-node'

// creates a new IDB database named 'my-database'
const database = new Database('my-database')

// generates a new key.
const key = database.key()

// stages a record for insertion.
database.insert('my-table', {
  key, foo: 1, bar: 'hello' 

// commits the record.
await database.commit()

// gets the record via key
const record = await database.get('my-table', key)

// gets the record via idb scan | query.
const record = await database.query('my-table').where(n => n.key === key).first()

The Database type is a transactional object store built over IDB. It manages some of the complexities around working with IDB, offering a simplified interface for reading and writing object records to IDB object stores.


A file persistence store for IndexedDB.

import { Bucket, Buffer } from 'smoke-node'

// Creates a new IDB database named 'my-bucket'
const bucket = new Bucket('my-bucket')

// Writes some content to 'index.html'.
await bucket.write('index.html', `<h1>hello world</h1>`)

// Creates a writable stream and streams content to IDB.
const writable = bucket.writable('source.dat')

// Creates a readable for the above.
const readable = bucket.readable('source.dat')
for await (const buffer of readable) {

// Pipes from 'source.dat' to 'target.dat'
const readable = bucket.readable('source.dat')
const writable = bucket.writable('target.dat')
await readable.pipe(writable)

A Bucket is a specialized implementation of the Database type that supports streaming files to and from IDB. It is intended to be a source for file content transmitted over a peer network. The interface of the bucket shares parallels with Amazon S3 in terms of general functionality. Files stored within the bucket are given simple keys, with hierarchical directory trees able to be emulated in much the same way as one would with S3.

Using Buckets with Rest

import { Node, Bucket } from 'smoke-node'

// Creates a IDB database named 'my-files'
const bucket = new Bucket('my-files')

// Writes this content to 'index.html'
await bucket.write('index.html', '<h1>hello</h1>')

// Creates a new network node.
const node = new Node()

// Create a new rest server.
const app =

// Serves the content from the bucket.
app.get('/index.html', (req, res) => {



The following are a few examples of applications that can be implemented with this framework.

Sockets and Loopback

The following code creates a simple socket server and listens on port 7000. It is connected to in a variety of ways.

import { Node } from 'smoke-node'

const node0 = new Node() //

const node1 = new Node() //

// listen on port 7000

node0.sockets.createServer(socket => console.log('have socket')).listen(7000)

const socket0 = node0.sockets.connect('localhost', 7000) // ok

const socket1 = node0.sockets.connect('', 7000)   // ok

const socket2 = node1.sockets.connect('localhost', 7000) // fail - not localhost

socket2.on('error', console.log)

const socket3 = node1.sockets.connect('', 7000)   // ok

Rest Server and Addresses

The following code demonstrates setting up a node running a rest server. This code sets a host node that runs a small rest server. A seconary node is then created and uses the hosts address to make a fetch request to download content.

import { Node } from 'smoke-node'

(async () => {
  // Server

  const host = new Node({  })

  const app =

  app.get('/', async (req, res) => {

    res.headers['Content-Type'] = 'text/html'
    res.send(`<h1>hello world</h1>`)


  // Client

  const node = new Node({ })

  const response = await`rest://${await host.address()}/`)
  const content = await response.text()


}, 1000)

MediaStream and Media Proxy

The Rest server offers the ability to stream live media feeds (such as those given by getUserMedia) as Rest responses. In addition, Smoke supports mediastream pass-through / proxy using familiar request / response semantics, leading to mediastream fan-out from a single source.

The following code sets up 3 network nodes and pipelines a mediastream in the following way.

source > node0 > node1 > node2 > video element 

Note, this component of smoke is highly experimental and subject to API changes.

import { Node } from 'smoke-node'

const node0 = new Node() //

const node1 = new Node() //

const node2 = new Node() //

// node0 - The video source.'/mediastream', (req, res) => {
  const mediastream =



// node1 - The video proxy.

const app1 =

app1.get('/mediastream', async (req, res) => {
  const response = await'rest://')

  res.mediastream(await response.mediastream())


// node2 - the client

;(async () => {

  const response = await'rest://')
  const video = document.getElementById('video-0') as HTMLVideoElement
  video.srcObject = await response.mediastream()


Database and Network Query

The Rest server framework supports streaming record sets directly from IndexedDB out to rest responses. The Rest server may act as a direct database pass-through, or projected with LINQ style queries. Its designed to allow nodes to function as standalone database servers.

Record iteration with queries means records are only pulled when the receiver moves 'next'. This is handled at a network protocol level and expressed with JavaScript async iterators.

The following code writes some database records and serves them on a Rest endpoint. The server filters the results (view logic) which are ordered and projected when received by the client.

import { Node, Record, Database } from 'smoke-node'

(async () => {

  // Use the rest namespace.
  const { rest } = new Node()
  // Creates a IDB database with this name.
  const db = new Database('users')

  // Insert some users into the database.
  db.insert('users', { 
    key: db.key(), 
    name: 'smith', 
    role: 'admin' 

  db.insert('users', { 
    key: db.key(), 
    name: 'mike', 
    role: 'admin'  
  db.insert('users', { 
    key: db.key(), 
    name: 'dave', 
    role: 'user'  
  db.insert('users', { 
    key: db.key(), 
    name: 'jones', 
    role: 'user'  
  // Commit users to database.
  await db.commit()

  // Create a Rest server.
  const app = rest.createServer()

  // Setup '/users' route.
  app.get('/users', (req, res) => {

    const query = db.query('users').where(n => n.role === 'user')

  // Request query from server.
  const query = await rest.fetch('rest://localhost:5433/users').then(n => n.query<User>())

  // Apply order and iterate, new records pulled on each iteration.
  for await (const user of query.orderBy(n => => {


You can’t perform that action at this time.