Out of curiosity, I have asked Deep Seek to write various types of virtual device drivers for Hubitat, to see how it does. It does a pretty good job, and you can even get specific about the variable types for the attributes you want it to have, and even multiple capabilities.
It makes me wonder how many Rule Machine automations that people have now could be changed into Hubitat apps fairly easily using AI with the right prompts?
I asked Deep Seek today:
“Write an app for hubitat that controls two dimmable lights in a kitchen. The two dimmable lights are "Sink", "Island". There is also a switch for "Under Cabinet" lights. Allow the user to choose the dimmable lights in settings, and the under cabinet switch, and also choose a motion sensor device. Have the app use the motion sensor device to turn on the lights when motion changes to active. Turn on the Sink Light to level 50 when active, turn on the Island lights to 75 when active, and turn on Under Cabinet when active. When the motion sensor changes to inactive, wait for one minute and then turn off all the lights in the kitchen.”
It gave me code for a totally working app and I did not touch the code. No errors, clear code, with debug statements and a way to turn off debug. The specific values I asked for were set to be settings instead, so they can be changed. It also set the inactive timeout to be a changeable setting.
I just thought some people here might find that AI custom automation apps are easier than figuring out how to do things in Rule Machine (or Webcore), if you don’t like writing code. I've been writing all my automations into custom apps and drivers for over a year now, after being solid Webcore for years before that, and I would rather manually write them because programming is fun, but for some people, AI generated Apps might be easier than the rule apps.
Here are my testing screenshots, (tested it in my office instead of the kitchen):
Log of it working: