I wanted to share a project I’ve been working on called Chatomation.
It’s a Hubitat app that lets you use AI (ChatGPT or Claude) to interact with and automate your Hubitat system using natural language.
Instead of building rules manually, you can describe what you want and let the system help create or trigger automations.
GitHub:
What It Does
The idea behind Chatomation is to add an AI layer on top of Hubitat automations so you can:
Control devices using natural language
Generate or trigger automations from prompts
Experiment with AI-driven home automation workflows
Extend Hubitat with external AI models or APIs
Features
Natural language — describe automations the way you'd explain them to a person
Conversational — each automation has its own chat thread; the AI asks clarifying questions if needed
OpenAI & Claude — choose GPT or Claude and plug in your own API key
Runs locally — the AI is only used when creating or modifying an automation; once built, the rule executes entirely on your hub with zero API calls
Full device support — lights, dimmers, color bulbs, motion sensors, contact sensors, locks, thermostats, fans, shades, garage doors, valves, alarms, speakers, and more
Conditions — time of day, hub modes, device states, days of the week
Delays — "turn off after 10 minutes of no motion" just works
Sunrise / Sunset — schedule actions relative to sun events with optional offsets
Enable / Disable — pause and resume automations like Rule Machine
Modify by chat — jump back into the conversation to change anything
Error notifications — logs everything and sends push alerts when something fails
Why I Built It
I’ve been experimenting with ways to make home automation more conversational and dynamic, rather than relying entirely on static rule creation.
Hubitat is already incredibly powerful, so the goal here is to explore what happens when you combine it with modern AI tools.
Current Status
This is an early project / experiment, but it’s working well enough that I figured I’d share it with the community and see what others think.
Feedback Welcome
I’d love feedback from anyone interested in:
AI-assisted automations
Improving the architecture
Feature ideas
Testing it in different Hubitat setups
If anyone wants to contribute or try it out, feel free to jump in on GitHub.
Hi there! I built something siimlar as an integrated feature of my HousePanel dashboard application. After building it I had the idea of generalizing it and making an app separate from my dashboard so I started with a search to see if anyone had done this already, and that led me to you. So wow - great minds think alike.
My current implementation is intricately linked in with my dashboard app and on the back end calls the Open AI api. It uses the Responses api and sends context based on prior smart home interactions as captured by subscriptions. I also wrote a function to query all devices on the hub and tied that into the Responses api and it works great. I assume you are doing something similar.
I look forward to checking out what you've done. Let's compare notes so we can learn from each other as both of our projects evolve. You can read about my work here:
I just checked out your code... nicely written. We are solving similar but different problems. I like how you capture the context write there in the groovy app. I wrote all my stuff in Python and push device info to the python app. My main dashboard has an integrated Rule engine similar to what you've written but with more flexibility. I haven't tried tying the AI into the Rule engine but it should be straight forward to do so. I tried making a Responses function to have chatGPT call Hubitat hub directly to do stuff but it didn't work well. I like your approach much better where you control the hub api call directly. Anyway just wanted to give you a thumbs up on a nicely done app. I didn't install it yet to try because I'm traveling but will do so when I get home.
Thanks so much! I’m a smart home integrator / installer and grew tired of putting together complicated rules for automations. I expose devices to apple home. So they’re able to do a lot of simple scenes and control of devices individually, or even tell Siri what to do. But this app also makes things more streamlined for them if they want to take some reigns and write their own automations.
I'm very excited about this, happy to help test and share feedback. Thanks for creating and sharing it.
Did a few trial runs today, and couple notes/questions:
The chat did not return a rule, instead showed "[Automation Rule Generated]". I needed to prompt to return the JSON coding in plain text to actually see what it did
Confirming the rules are housing within Chatomation app - they are not intended/able to be ported over to the Rule Machine app? It would be a nice QOL improvement to have all rules located a central location, provided its feasible to do so, and maintain the ability to modify using the API/chat function you built!
So stoked you’re checking it out!!! I thought more people would have by now That’s a great question. I believe Hubitat sandboxes plugins. But let me see what I can do about importing them to rule machine. Maybe an import export with some form of automating that. Thanks Ross!
Just a thought - make it a separate step, AFTER the user can do some testing/troubleshooting of the rule operation within the Chatomation app. Just like any rule built by hand, it'll probably take a bit of trial and error to validate its functioning how you want.
Would be great if it prompted "do you want to publish to Rule Machine, or continue to work on it?"
That’s a great question. I believe Hubitat sandboxes plugins. But let me see what I can do about importing them to rule machine. Maybe an import export with some form of automating that. Thanks Ross!