SwanBot was a vague project that has no real goal or intention. In its earlier forms, SwanBot was used to monitor a single IRC channel and provide services like weather, Last.fm status, etc. At one point I decided to write a module that allowed SwanBot to fetch pages from Freebase, all while doing its own research on related topics which was stored in something called the "node mesh."

I decided I wanted to practice with Client/Server applications again, so I set out to do just that by moving all the "node mesh"-related code into a server program, which accepted commands from clients to search the mesh and return data. This proved to be an interesting idea, and I started to invest more time in the server application.

At one point I started wanting to automate stuff in my dorm room, so I set out to add stuff like OpenCV support to look for me as I was coming into my room, and scanning for my phone in the proximity of my computer's Bluetooth adapter. I ended up combining those two features, and then report their success /failure to SwanBot.

But what happens next?

For a while I've been interested in the idea of a "DLA" (a term that I believe was coined by the creator of Project Jarvis.) DLA simply stands for "Digital Life Assistant", and that says it all: a DLA operates passively or interactively, but at the end of the day, is designed entirely to adapt to the way you operate in your day-to-day life.

So what does SwanBot do?

Two key terms get tossed around a lot when I speak about SwanBot: Inputs and Outputs. Inputs are just that- ways for SwanBot to accept incoming data which is then stored in the Node Mesh. What kind of data, though? Well, that's up to the user. I've placed no limitations on what can be stored in the Node Mesh. Even things like images can be stored inside the mesh, then processed by OpenCV or any other user-specified library.

Outputs provide a way for SwanBot to communicate with the user in some way. I've also placed no limit on how outputs should be properly structured or implemented, and have also placed that into the hands of the user.

Here's a working example: You're out and about, let's say 100+ miles from your home with friends or on vacation. Suddenly you receive a text message:

Severe weather warning for ***** is in effect until 6PM.

This is a message sent by SwanBot, which has magically found out your current location and has warned you of possible severe weather. Whoa!

Except, of course, it isn't magic: SwanBot barely has any work to do, because chances are you're updating Twitter, Facebook, or Google+ on the go. Taking pictures? Connecting via Wifi? That works too. It's redundant to explicitly tell SwanBot where you're at; It already has enough data to deduce that on its own.

Once SwanBot is configured to read your various inputs for data, no user interaction is required- it uses the already existing data to provide useful information to you in whatever way possible.

In the above example, a text message was sent to the user's phone. However, this isn't always the best way to communicate with the user. Again, this is where inputs come into play: by providing an input that fires when you enter your house/room, SwanBot can then adjust where it outputs its data. In my case, SwanBot users the text-to-speech function on my tablet to fill me in on things that are relevant to what's going on.

You're probably wondering what kind of information SwanBot outputs. For that, read the modules section below.

Sometimes data just isn't relevant. You want a break, you're in a class, etc; although SwanBot tries its best to avoid bothering you with alerts, it can be really hard to figure out what time is best. Let's use the example of my morning routine:

On Tuesdays I wake up at 9 AM, start breakfast, and check Reddit's technology subreddit for breaking/interesting news. I then check my email and eat while browsing the various channels I follow on YouTube for new uploads. It's consistent- I do this exact routine every Tuesday and Thursday, and have similar patterns for the other weekdays.

With SwanBot, I wake up and start breakfast, and am soon greeted by my tablet's robotic voice:

Good morning. News for <date>:
  Apple accused of plagiarizing iconic Swiss clock design
  ...

Modules provide a way for SwanBot to fetch information from sources specified by the user(s) via the node mesh.

Let's say the mod_twitter module was installed on the server. When it runs, it scans each user's node mesh for twitter-details, grabs the login credentials in the node's data key, and watches for mentions, tweets, or direct messages. If this information is relevant to you at all hours of the day, SwanBot can be configured to alert you by whatever means possible.

See data/api.md

While I have my own idea for this project, I'm interested in seeing what other people do it. And yeah, I understand it's nothing special or groundbreaking, but I feel like I'd be saving someone some time by publishing this.

If you write anything interesting/helpful, file a pull request and I'll send it through (within reason) to the main repo for other people to use.

Happy coding!

-flags