node package manager




Cerebellum.js is a powerful set of tools that help you structure your isomorphic apps, just add your preferred view engine. Cerebellum works great in conjunction with React.

Cerebellum is designed for single-page apps that need search engine visibility. Same code works on server and client.

Introductory blog post, published on February 5, 2015.

Cerebellum React helpers

What does it do?

  • Fully shared GET routes between server and client
  • Fully shared data stores between server and client, uses Vertebrae's Collection & Model with Axios adapter, so you can use the same REST APIs everywhere.
  • Stores the server state snapshot to JSON. Browser client will automatically bootstrap from snapshot, you don't need to do any extra requests on client side.
  • Uses express.js router on server and page.js router on browser. Both use same route format, so you can use named parameters, optional parameters and regular expressions in your routes.
  • Data flows from models/collections to views and views can dispatch changes with change events. All rendering happens through router.
  • Automatic SEO, no hacks needed for server side rendering.
  • You can easily make apps that work even when JavaScript is disabled in browser
  • Fast initial load for mobile clients, browser bootstraps from server state and continues as a single-page app without any extra configuration.
  • Store state is maintained in Immutable.js
  • Can be used with any framework that provides rendering for both server and client. React.js recommended, see examples/static-pages.

Data flow

Cerebellum's data flow is similar to Flux architecture in many ways, but it has some key differences.

Diagram below shows the data flow for client side. Server side is identical, except that there are naturally no interaction triggered updates (green arrows).


Cerebellum data flow


In a nutshell, route handler asks stores for data and renders a view with the response.

When you want to change things, you send a change event to central Store instance. Store will perform the API call and trigger a success event when it's ready. You can then act on that event by invoking a route handler again.

All rendering happens through route handlers in Cerebellum.

Server and client data flow example

  1. User requests a page at /posts/1

  2. Server or client will ask from router to check for a route that matches "/posts/1".

  3. Router finds a matching route handler at "/posts/:id" and queries Store's post store (which is a model) with parameter {id: id}.

  4. Store will either invoke GET /api/posts/1 call or use cached post data (if available).

  5. Store returns post data to route handler

  6. Route handler passes data to Post view component

  7. Server or client renders the returned view component

Triggering changes with client side change events (green arrows)

Views can trigger change events (create, update, delete or expire) with Store's dispatch method. Store delegates change event to corresponding store and invokes API request. When request is completed, Store triggers completion event (create:storeId, update:storeName, delete:storeName or expire:storeName).

Client can listen for these events. In store event callbacks you can clear caches and re-render current route (or invoke another route handler). There's also an option to automatically clear caches for stores.

Change events example

Let's say that reader wants to add a comment to a blog post. We want to persist that comment to server and re-render the blog post.

  1. When our avid reader clicks "Send comment" button, view triggers change event with
store.dispatch("create", "comments", {id:}, {name: "Hater", comment: "This example sucks!"})
  1. Store makes a API call POST /api/posts/1/comments to create a new comment with our data

  2. Store automatically clears the client side cache for this particular post's comments and triggers create:comments event

  3. We have a event handler in client.js which invokes the /posts/1 route handler"create:comments", function(err, data) {
  // document.location.pathname is the current url, 
  // you could also use "/posts/" +; 
  1. Route handler re-fetches comments collection for this post as its cache was cleared

  2. When comments collection has been fetched, the blog post gets re-rendered with the new comment


Store is responsible for handling all data operations in Cerebellum. It receives change events for individual stores, performs changes and notifies client by triggering success events.

You register your collections and models (stores) to Store by passing them to server and client constructors in options.stores (see "Stores (stores.js)" section below for more details).

Store will automatically snapshot its state on server and client will bootstrap Store from that state. Client will also cache all additional API requests, but you can easily clear caches when fresh data is needed.

Models and Collections are immutable

All your stores are read only, state is stored in Immutable.js. All mutations are handled by Store with create, update, delete and expire events.

Fetching data inside routes

You can retrieve data from a store using fetch(storeId, options). Route's this context includes Store instance ( that is used for all data retrieval.

When fetching collections, you don't usually need any parameters, so you can do:"posts").then(...);

If your store needs to fetch dynamic data (models usually do), pass options to fetch as second parameter. For example, if you need to fetch model by id, your options would be {id: id}."post", {id: id}).then(...);

You can also fetch multiple stores at once with fetchAll:{"post": {id: id}, "comments": {id: id}}).then(...);

fetch returns Immutable.js Cursors. If you're using React, Omniscient's shouldUpdate mixin works great with these cursors.

Caches and cacheKeys

Store will populate its internal cache when calling fetch(). So when you request same data in different route on client, Store will return the cached data.

Your models and collections can have cacheKey method, it defines the path where data will be cached in Store's cache. Store will automatically generate the cacheKey for collections and models if you don't provide one. Note that the model needs to be fetched with id parameter for automatic cacheKey generation.

If you want to use fetch options as part of cacheKey, you can access them using this.storeOptions.

var Model = require('cerebellum').Model;
var Post = Model.extend({
  cacheKey: function() {
    return "posts/" +;
  url: function() {
    return "/posts/" + + ".json";

Triggering changes

Pass router's store instance to your view components and call store.dispatch with create, update, delete or expire.

For example, you would create a new post to "posts" collection by calling:

store.dispatch("create", "posts", {title: "New post", body: "Body text"});

You can update a model with:

store.dispatch("update", "post", {id: id}, {
  title: "New post",
  body: "New body text"

Store will then execute the API call to url defined in given store and fire an appropriate callback when it finishes.

Expiring caches and re-rendering routes

You can listen for store events in your client.js. Make sure to wait for client's initialize callback to finish before placing event handlers.

options.initialize = function(client) {
  var store =;
  var router = client.router;
  store.on("create:posts", function(err, data) {
    console.log( // => posts 
    console.log(data.result); // => {id: 3423, title: "New post", body: "Body text"} 
    router("/posts"); // navigate to posts index, will re-fetch posts from API as cache was automatically cleared 
  store.on("update:post", function(err, data) {
    // explicitly clear posts collection in addition to automatically cleared post model 
    // you could also handle this in post model with relatedCaches method 
    // re-render route, posts collection & post model will be re-fetched 
    router.replace("/posts/" +;


Options below are processed by both server.js & client.js, it usually makes sense to create a shared options.js for these shared options. Server and Client specific options are documented in their respective sections in documentation.

Options (options.js)

example options.js with default values, these options are shared with client & server constructors.

var stores = require('./stores');
var routes = require('./routes');
module.exports = {
  routes: routes,
  storeId: "store_state_from_server",
  stores: stores,
  initStore: true


Object of route paths and route handlers. Best practice is to put these to their own file instead of bloating options.js, see "Routes (routes.js)" documentation below.


DOM ID in index.html where server stores the JSON snapshot that client will use for bootstrapping Store.


Object containining store ids and stores. Best practice is to put these to their own file as well, see "Stores (stores.js)" documentation below.


Initialize store for route handlers ( Defaults to true. Disable if you want to perform the data retrieval elsewhere. For example, when using framework like Omniscient you would perform the data fetching in server.js & client.js and pass cursor to immutable data structure in routeContext.


Method that decides how route handlers are being called. Default behaviour is that route handler gets applied with route params. With React, you can use cerebellum-react/route-handler.

Routes (routes.js)

Example routes.js

var Index = React.createFactory(require('./components/index'));
var Post = React.createFactory(require('./components/post'));
module.exports = {
  '/': function() {
    return"posts").then(function(posts) {
      return {title: "Front page", component: Index({posts: posts})};
  '/posts/:id': function(id) {
    return"post", {id: id}).then(function(post) {
      return {title: post.get("title"), component: Post({post: post})};

Your routes will get picked by client.js and server.js and generate exactly same response in both environments (provided you implement your render functions in that manner).

Your route handlers can return either promises or strings, cerebellum will handle both use cases.

In route handler's this scope you have which is the reference to Store instance. It contains all your stores and fetch for getting the data.

On the server Store is initialized for every request and on the client it's created only once, in the application's initialization phase.

Server serializes all Store content to a JSON snapshot at the end of a request and client then deserializes that JSON and bootstraps itself.

Stores (stores.js)

Example stores.js

var PostsCollection = require('./stores/posts');
var AuthorModel = require('stores/author');
module.exports = {
  posts: PostsCollection,
  author: AuthorModel

Return an object of store ids and stores. These will be registered to be used with Store.

Server (server.js)

Server is responsible for rendering the first page for the user. Under the hood server creates an express.js app and server constructor returns reference to that express app instance.

Server is initialized by calling:

var app = cerebellum.server(options, routeContext);

If you want to customize route handler's this context, pass your own context as routeContext. routeContext must be an object or a promise that resolves with an object. Cerebellum will automatically add store to the context. If you don't want this, use the initStore: false option.

See "Options (options.js)" section for shared options (routes, storeId, stores ...), options below are server only.

options.render(document, options={}, request={params: {}, query: {}})

Route handler will call server's render with document, its options and request object. document is a cheerio instance containing the index.html content.

Render method is invoked with route handler's this context.

Render method can return either a string or a promise resolving with string.

Example server render function:

options.render = function(document, options) {
  return document.html();


Path to static files, index.html will be served from there.

options.staticFiles = __dirname + "/public"


Array of middleware functions, each of them will be passed to express.use(). You can also include array with route & function.

var compress = require('compression');
var auth = require('./lib/auth');
options.middleware = [
  ["/admin", auth()]


You can define entry files per route pattern, e.g. you want to include different .js bundle & .css in admin section. If entries option is not defined, server will default to path.join(options.staticFiles, "index.html"). If routes object is empty or any route pattern does not match, server will default to path.join(options.entries.path, "index.html").

options.entries = {
  path: "assets/entries",
  routes: {
    "/admin": "admin.html"


Instance method for cerebellum.server instance. Registers express.js static file handling, you usually want to call this after executing cerebellum.server constructor, so Cerebellum routes take precedence over static files.

var app = cerebellum.server(options);

Client (client.js)

Client is responsible for managing the application after getting the initial state from server.

Client is initialized by calling:

cerebellum.client(options, routeContext);

If you want to customize route handler's this context, pass your own context as routeContext. routeContext must be an object or a promise that resolves with an object. Cerebellum will automatically add store to the context. If you don't want this, use the initStore: false option.

See "Options (options.js)" section for shared options (routes, storeId, stores ...), options below are client only.

options.render(options={}, request={params:{}, query:{}})

Route handler will call client's render with its options and request object when it gets resolved.

Render method is invoked with route handler's this context.

options.render = function(options) {
  document.getElementsByTagName("title")[0].innerHTML = options.title;
  return React.render(options.component, document.getElementById("app"));


This callback will be executed after client bootstrap is done. Returns client object with router and store instances.

You can listen for store events, expire store caches and render routes here.

options.initialize = function(client) {


With this option Store will automatically clear cache for matching cacheKey after create, update or delete. Defaults to true.

options.autoClearCaches = true;


With instantResolve you can make the fetch promises to resolve immediately with empty data. When fetch calls actually finish, they will fire fetch:storeId events that you can use to re-render the routes. This is really useful when you want to render the view skeleton immediately and show some loading spinners while the data retrieval is ongoing. instantResolve will only affect client side fetch calls, it has no effect on server side.

options.instantResolve = true;

Models & Collections

Cerebellum comes with models & collections from CommonJS version of Backbone, Vertebrae. However, you could roll your own implementations as Cerebellum's Store has no dependencies to any model or collection libraries.

Model options

Model documentation for Backbone applies to Cerebellum models, but there are some extra options that can be utilized.


Optional property or method, should return the cache key for model. This will be generated automatically if not provided (you must provide for automatic generation).

cacheKey: function() {
    return "myCustomPrefix:"; // defaults to 


With relatedCaches method you can define additional cache sweeps that happen when model's cache gets cleared.

relatedCaches: function() {
    return {"comments":};

Collection options

Collection documentation for Backbone applies to Cerebellum collections, but the are some extra options that can be utilized.


Optional property or method, should return the cache key for collection. This will be generated automatically if not provided.

cacheKey: function() {
    return "myCustomPrefix:"; // defaults to 


With relatedCaches method you can define additional cache sweeps that happen when collection's cache gets cleared.

relatedCaches: function() {
    return {"posts": "/"};

Usage with React

Cerebellum works best with React.

React makes server side rendering easy with React.renderToString and it can easily initialize client from server state. All code examples in this documentation use React for view generation.

Running tests

Start test server for client tests

npm start

Running client tests (requires test server to be up and running)

npm run test_client

Running server tests

npm run test_server

Running all tests (server & client)

npm run test

Browser support

Internet Explorer 9 and newer, uses ES5 and needs pushState.

Apps using Cerebellum


Sample app for saving & tagging urls, demonstrates CRUD & authorization


Sample app on how to declare needed data directly in view components and keep the router clear of store.fetch calls.


Stats site for Finnish hockey league (Liiga)

Source available at:



Copyright (c) 2014-2015 Lari Hoppula, SC5 Online