Find file
Fetching contributors…
Cannot retrieve contributors at this time
560 lines (430 sloc) 21 KB

Numenta Web Platform

Source code, generators, tooling, and content for Numenta websites and platform.


This is a monorepo, managed with Lerna.js. This monorepo contains several child sub-repos (websites, shared components, utils) located in the /packages directory.


.                     #
├── .babelrc          # Babel ES6 transpiler configuration file
├── .eslintignore     # ES lint files and paths to ignore during run
├── .eslintrc.json    # JS ES6 lint rules, AirBnB defaults + tweaks
├── .jestrc.json      # Jest test framework setup, assets under __tests__/ dirs
├── .stylelintignore  # CSS style lint files to ignore
├── .stylelintrc      # CSS style lint rules
├── LICENSE.txt       # Open Source MIT License information
├──         # This file, welcome documentation
├── __tests__/        # Shared Tests and Test Setup
├── coverage/         # Target for Test Code Coverage reports (not in git)
├── lerna.json        # Lerna.js monorepo tool config file.
├── node_modules/     # `npm` module install target (not in git)
├── package.json      # Skeleton monorepo project config
└── packages/         # Child sub-repos of our Lerna monorepo, details below
    ├── components/   # Shared React View Components to be used amongst sites
    ├──  # Numenta Company website source code
    ├──  # HTM Community website source code
    └── utils/        # Shared helper utils to be used amongst sites.


These instructions assume latest Mac OS/X with homebrew package manager installed. Other platforms should come up just as easily.

brew install git node
git clone
cd numenta-web/
npm install

You should now be ready to run any of the child Website repos, further details below.


If you update the child repos and need to re-sync everything back together again, just re-run npm install from the monorepo root.


To run All Linting checks against JS, CSS, etc:

npm run lint


To run All Tests (unit, etc):

npm run test

# For older versions of Node.js (< 7, having less harmony)
npm run test --harmony_proxies


Run only unit tests:

npm run test:unit

Unit tests take and use snapshots in order to perform. Make sure new or updated snapshots are correct before committing! To update these snapshots, try running:

npm run test:unit:update

Run the command below to generate a Unit Test Code Coverage report, which will be saved in your local ./coverage directory.

npm run test:unit:coverage

# open nice coverage report in browser (Mac OS/X)
open ./coverage/lcov-report/index.html

Run unit tests while watching filesystem for changes in realtime (requires 3rd-party Watchman to be installed first):

brew install watchman
npm run test:unit:watch


# Test hyperlinks on
#   You probably have `npm run serve` ready in another terminal window or
#   background job process. Or, you could do `npm run build` and run an http
#   server on the contents of `public/`.
npm run test:links -- http://localhost:8000


Please change directories to one of the individual website sources, i.e.:

cd packages/

The rest of this section will assume you are in one of the website sub-directories.


These sites are an attempt at getting back to the roots of the web: reading. Inspired by the classic design of Wikipedia, we hope our sites will be easy to read, helpful for finding information quickly, simple to navigate, and a delight to interact with.

Moving from full statically-generated multi-page websites, to single one-page web applications, is a tricky proposition. Search engine ranking can be affected negatively, as important internal pages and links would disappear. Navigating the user to one of many collapsable sections on a long page can be difficult. Accessible, legacy, and unique browsers cannot usually support the modern technology required for single-page web applications.

To solve these challenges, while still accomplishing our goals, we have built these as hybrid websites. They are both classic statically-generated multi-page websites, and also modern single-page web applications. All clients load the classic static sites initially, and then modern clients will be augmented with the single-page applications. This is accomplished with the magic of "universal" or "isomorphic" javascript, which means that the exact same code is used to generate the static sites on the server, and then is loaded on the clients to continue the applications live.

Gatsby.js is the amazing tool that helps us accomplish all of this.




The contents of each individual website source directory (i.e. cd packages/ will look like the following:

.                       # Inside `packages/` or similar
├── __tests__/          # Site-specific Tests
├── components/         # Site-specific custom React view Components, etc.
├── config.toml         # Configuration setings for Gatsby static site generator
├── gatsby-browser.js   # Browser-specific code, tied to React Router by Gatsby
├── gatsby-node.js      # Build-time, Node.js, Webpack & Server specific code
├── html.jsx            # Main HTML Document Component
├── package.json        # Project scripts, settings, and `npm` dependency lists
├── pages/              # Webpage Documents and URL Tree Structure
├── public/             # STATIC OUTPUT. Static website generated built files
├── static/             # Static html, styles, assets, etc. Copied after build.
├── utils/              # Local Helpers, utils, client, and misc code
└── wrappers/           # Document-type (.html, .md, etc.) wrapper Components

These websites use shared modules that exist outside of the specific website source directories, located under projects/components and projects/utils. The sites also can have their own custom components/, unique to each site, under their own subdirectory.

The shared Components start out as a base for, where instead uses a few more of it's own custom Components. When making changes to shared Components or Utils, don't forget to run npm run compile to convert down to ES5 first (npm install from root also handles this).


Gatsby starts Webpack looping over and loading files under pages/. .jsx page code is run as normal. Other file extensions (.md, etc.), which have matching wrappers/ (either Gatsby or custom), are translated accordingly. A few examples of page build paths follow.

Homepage Advanced Section/Pages build path:

    • ⇒ (pages/contact/_Section.jsx)
      • pages/contact/index.jsx
        • ⇒ (pages/contact/_template.jsx - if exists)
          • ⇒ (pages/_MainSections.jsx)
            • pages/_template.jsx
              • ./html.jsx
                • public/contact/index.html

Regular Page build paths:


    • pages/sitemap/index.jsx
      • ⇒ (pages/sitemap/_template.jsx - if exists)
        • pages/_template.jsx
          • ./html.jsx
            • public/sitemap/index.html

    • pages/legal/
      • wrappers/md.jsx
        • ⇒ (pages/legal/_template.jsx - if exists)
          • pages/_template.jsx
            • ./html.jsx
              • public/legal/privacy/index.html




All browsers and clients will start by loading a classic, simple, statically-generated, multi-page site. This guarantees that all users will receive a decent baseline experience, especially on older, accessible, or unique browsers. This also gives robots and search engine crawlers to have many internal pages to link and index.

Advanced / Targets

After loading the baseline experience, modern clients (listed in the support table below) will then be augmented with an advanced single-page style web application. This lets us simulate a large single main page, with many collapsable sections, much like Wikipedia. This also provides fast "no-refresh" page transitions, a full client-side text search system, etc.

Note: All versions are "Latest Stable" unless otherwise noted.

Chrome Firefox Safari Edge IE/11 IE/10 Android


Builds should be run from inside each websites individual source directory (cd packages/

Average build wait times:

Environments 4x 2.8ghz 16gb 2x 2.6ghz 8gb
Dynamic Dev ~10s ~20s
Static (Any) ~6m ~12m

Additional Notes:

  • Config: Unless otherwise stated below, the config.toml file should have the default empty linkPrefix setting:

    # config.toml
    linkPrefix = ""
  • Disqus Comments: Disqus Comments only work on Production by default, not under Development, Staging, etc. You can test/verify outside Prod by sending an override production Post URL to Disqus via the component, here.



npm run dev
# Visit http://localhost:8000

Additional Notes:

  • Static Assets: Static assets won't load correctly from static/assets during dynamic dev, but they will be fine later on staging and production after static build. Try the static dev (next section below) if you need the static asset links to work during development.

  • Specific Local IP: If you need to host via a specific local IP (testing on a VM, etc.), gatsby options can be passed to npm run like this:

    npm run dev -- --host --port 8765
    # Visit
  • Search: To get search working under local dynamic development mode, you first need to build the search index, and then copy the index over to where it will be accessible by the dev webserver:

    npm run build
    cp public/_searchIndex.json pages/.


npm run serve
# Visit http://localhost:8000


Update Gatsby config for GitHub Pages (gh-pages) integration targets. Modify config.toml and set (but DO NOT commit):

# config.toml
linkPrefix = "/numenta-web"

And then run an integration build:

npm run build


Push build to personal fork origin:gh-pages integration hosting branch:

(Note: This will remove any other gh-pages site staged on that branch).

npm run deploy:gh-pages
npm run deploy:gh-pages -- --remote origin  # same thing
# Visit


Push build to shared upstream:gh-pages integration hosting branch:

(Note: This will remove any other gh-pages site staged on that branch).

# Deploy master branch public/ to upstream:gh-pages branch
npm run deploy:gh-pages -- --remote upstream

# Visit

Staging and Production

Numenta handles builds and deployments for the Staging/Production runways in our internal Continuous Integration tooling.


Code standards are handled by their respective linters and lint configs (for example, see the local file .eslintrc.json).


  • Pages and Main Sections:
    • Main Sections (pages grouped on our Wikipedia-like single-page app homepage) have their actual content in each pages' pages/*/_Section.jsx file
      • These main Sections are exposed as individual Pages via each pages' default pages/*/index.jsx component
    • Regular Pages (non-Main Sections) are fully contained in the usual default pages/*/index.jsx location
  • Each React Component should return 1 small element
  • Include spaces manually around React Elements in JSX with: {' '}
  • Custom React context which is available (in addition to Gatsby defaults):
    • config = Site config object (see: config.toml).
    • manifest = Site package.json manifest object vars like repo, 'version`
    • stamp = Build start timestamp string (for cache-busting)
  • Make sure to use the prefixLink() helper function on all internal links. This should be handled auto-magically for you already, but if you have trouble with links on staging, this may be the problem. Try to keep this working so that each developer can run their own demo sites on gh-pages.


  • Markdown link target filenames with spaces must be quoted: [link text]("spaces in filename.txt")
  • Markdown YAML headmatter: be careful with quotes and escaping titles, etc.
  • Beware extra Markdown typographer shortcuts
  • Internal flat markdown auto-links ( may break on staging servers, use markdown style instead: [](/htm-studio/)
  • Search Engine Optimization (SEO) notes:
    • Make sure to follow the Link and URL Format notes further below.

Links and URLs

Link Text

  • Use meaningful and semantic text for links, which describes what will be experienced at the destination of the link.
  • Avoid over-generalized link text, like "click here".

URL Format

  • URLs should end with / (slash)
  • Examples of ideal URLs detailed below:

Possible Link Situations

  • Google Analytics should be tracking pages and events for all Situations in the matrix below.
Host Type .jsx Source .md Source
Internal Path View View
Internal Anchor View View
Internal Asset (/assets/*) View View
Internal Asset (*/.pdf) N/A View
External Path View View
External Anchor View View
External Protocol View View


  • Use exact versions for dependencies in package.json
  • Keep packages as up-to-date as possible: npm outdated -depth 0
  • Update and test 1 package at a time for safety
  • These packages have breaking updates which it seems we'll be skipping for now (at least until Gatsby, etc., update to work with them):
    • history@3 history@4
    • react-router@3
    • react-g-analytics@0.3


  • CSS Module Composition order:

    1. Compose from Tachyons classes
    2. Compose from Site Theme classes (Numenta colors)
    3. Compose from _local_file_custom classes
    4. Specific custom CSS overrides


    ._visited:visited {
      color: #1aa0db;
    .iconlink {
      composes: dim from 'tachyons';  /* 1. */
      composes: link from 'tachyons';
      composes: color-blue from '../../assets/css/theme.css';  /* 2. */
      composes: _visited;  /* 3. */
      opacity: 0.5;  /* 4. */


These projects are open source, and Pull Requests are welcome. Contributors, please sign and submit our Contributor License.

If you'd like to help, please make your own fork of this repo, and work from branches in your fork. Pull Requests should be between your fork, and our main repo. This will keep our main repo clean of working branches.

Please take a look at our Open Issues to see how you can dive in and help.

Before submitting Pull Requsts, please make sure you have successfully run npm run lint and npm run test on your change branch.


Numenta Web Platform and Sites source code
MIT License (see LICENSE.txt)
Copyright © 2005—2017 Numenta <>