Realtime visualizes GitHub activity in real time.
The architecture works as follows:
- A celery instance kicks off a
scrapetask every second which queries the GitHub API endpoint at
https://github.com/timeline.json. A human-readable string of the event is generated
A forked repository B from C/Bwith a link to A/B. That JSON object is then sent to a Redis Pub/Sub which is consumed by
- a small NodeJS instance with Socket.IO which pushes the live data to all connected clients. It also caches a couple hundred events so clients can "pre-load" events to fill their screen upon first connecting.
- The front-end then has 2 views of said data-stream
- One is to simply display a global "list-view" of all events that describes what's happening.
- Another is one that queries all available gravatars (this used to happen on the server until gravatar banned my server IP) and simply lists the event's
author's avatar along with the programming language associated with the event.
This allows you to efficiently reduce every programmer to their physical appearance and choice of programming language.
Please note that the live web app at http://realtime.doda.co/ is currently offline. It was fun to demo this to fellow dev buddies while I was building it, though I don't really see a reason to run 3,600 requests / hour against GitHub's API just for the fun of it.