Last year, Rightpoint Labs launched Pourcast - a webapp + Arduino + flow sensor setup to monitor the amount of beer in the kegs in our Chicago office in real-time. In the year since then, we've tinkered with it a bit to improve things. (Original announcement can be found here if you want the info on the setup)
First up, reliability. As you could see in the (badly-lit) picture from the original post, it consisted of a daughter-board with a bunch of connectors and crappy wiring into the Arduino Ethernet driving things. With the help of Upverter (an online circuit board design tool), we designed a custom PCB that will mount on top like a real Arduino shield. https://oshpark.com/ then turned that board design into reality in a couple of weeks for about $20:
This improved our reliability noticeably, as I think I had done a pretty bad soldering job on some of the connections on the previous iteration. Here’s the old one on the Arduino vs. the new one for comparison:
Next up: UI flexibility + Movember. As happens with these kinds of projects, cool ideas for something to show lead to interesting technical approaches for how to pull it off. As part of Rightpoint's participation in Movember, we wanted to let Pourcast in on the fun. To make it work, Sean Sartell made changes to the way the Pourcast UI was built - instead of a relative static set of HTML/CSS, he used the Knockout Components (new in 3.2) to split the UI into many separate pieces. The pieces could then be pulled together on the fly by the UI to achieve whatever look a particular keg was going for - from interchangeable keg and beer models, to extra bits of flair on or around the keg. We now had the ability to heavily customize the keg for events without replacing the whole UI. With those changes and some help from Rebecca LaRue from our creative team on the faces and flair, Pourcast was able to join in on the Movember fun.
Finally, Pourcast got with the social media times through the addition of a webcam, a Yammer account, and SkyBiometry. If the site is launched using the “?camera=true” query string, it will use the HTML5 camera APIs to access the computer's webcam and take a picture to go with each pour. The backend application then uses SkyBiometry’s facial-recognition APIs to guess which Rightpointer is in the picture. On Fridays after 3pm, Pourcast starts tracking the biggest pour of the day on each tap and announces each successive biggest-pour to a group on Yammer, including the picture it took, and the name of the people it recognized (if it recognized them). We did decide that posting to Twitter may be taking it a bit far though.
What interesting things will we dream up for Pourcast next year? Who knows? Maybe you have an idea (or even better, some code) to contribute - join in with us at https://github.com/RightpointLabs/Pourcast or tweet us. Want to use this software and design to drive your own? You can do that too - all the code is at the above github link, including links to the PCB diagram to connect the sensors to the Arduino.