A few weeks later, a buddy and I took this to its logical conclusion.
But even with novelty projects, there are things to be learned, and this project has exposed me to npm modules I probably otherwise never would have used. I’m going to go over a few of the architectural highlights (npm and not) in no particular order.
After you npm install lefnire, typing lefnire runs him.
If you said, “ASCII art?” you would be correct. This was a friend’s idea, and it’s become a staple of the application.
API integration with the superagent and github modules
Currently, lefnire.js is lightly integrated with the HabitRPG and GitHub APIs. The HabitRPG API integration uses straight superagent requests. The GitHub API calls use github.
When you say something containing “habit down,” he makes an API call to https://habitrpg.com/api/v1/status to check.
When you ask him by his IRC name (I’ll be getting to IRC stuff in a second), he queries the GitHub API to find out the number of issues labeled “critical” for HabitRPG on GitHub.
IRC integration with the node-irc module
What would an automaton be without being able to log on to IRC and interact? lefnire.js uses node-irc to accomplish this. The library is pretty robust and follows Node.js patterns well. This is a fantastic example of an application that thrives on asynchronous execution. When you ask him if “habit is down,” and it is, that thread of execution will block. However, it won’t stop the bot from responding, thanks to the event listener model implemented by node-irc.
Mood and the Sentimental module
Another interesting feature is a rudimentary implementation of mood, which is basically just a number between 0 and 30 (0’s the best). This doesn’t do too much yet. It affects his behavior a little bit. The number of GitHub criticals is the baseline metric.
This is then adjusted up or down by how positive or negative Sentimental analyzes the up to the 30 most recent GitHub Events from lefnire to be. This is mostly issue comments, pull requests comments, and commit messages.
When lefnire is in-channel, it also adjusts mood based on Sentimental text analysis of what he says.
No database yet, so everything except what I can retrieve again through API calls is lost if I ctrl-C the bot and stop it.
Those are the main highlights of the current implementation. A lot of things are planned. I doubt anyone would want to contribute to this, but I’d certainly be excited to receive some pull requests.
GitHub repository: https://github.com/litenull/lefnire.js