How Much Do Your Conversations With Allan Impact Your Smart Home?

Does anyone else see the abbreviation for Artificial Intelligence, AI, and immediately think it is the shortened version of the man's name Allan? Personally I would like to see AI referred to as Allan, it provides a more personable element to the topic...

Separate to that.... :slight_smile: I'd be interested to know how other people have started to integrate Allan's many features into their smart home.

A good starting point is:

I prefer Kier, like another confused lying pratt

1 Like

Ignoring the political aspect.... you are right, AI does include an element of inaccuracy, in some part by design. I would prefer to start from the point of view of what people do or want to use AI in order to improve their home automation experience.

3 Likes

I need to stop raising these topics in the Lounge.... :wink: But thankyou... Very funny.

:popcorn: - getting the popcorn ready! :popcorn:

For me, I'm looking for Hubitat to be more proactive. ie. It's over 80f in the house, it sends me a message - "hey, it's hot in here. Want to turn on the AC?". I tell it 'Sure'. It checks the windows and doors, if all is good, turn the AC on. Then says, "Want me to make a rule/flow/etc. so this happens automatically in the future?" ...

Basically, I want it to 'think' of things that we haven't thought of yet!

3 Likes

That certainly sounds cool (pun intended :slight_smile: ). I'm not sure how much that could be artificial and reactive to the kinds of rules people might want. I am thinking that specific example is more an example of a list of rules Hubitat could have programmed in to look out for and offer up as something that could be setup in response to a known set of conditions.

I guess my point is that I am not so sure that it is an example where AI could be introduced. that said, it is certainly something that I would like to see...

"Hal, open the garage door"
"I'm sorry, I can't do that"

6 Likes

I already have chatGPT integrated into Flow Engine and can make flows by simply typing. "If front door is open, turn living room ceiling light on". :grin:

Part of my conversation with chatGPT...

1. Natural Language → Automation

  • Let ChatGPT parse plain-English commands like “Turn off the kitchen lights in 20 minutes if the TV is still on” and convert them into Flow Engine JSON or Hubitat Rule Machine rules.
  • You already have your Flow Engine Editor + Generator, so an LLM could be the “front door” to make flows without manually dragging nodes.

2. Conversational Hubitat Control

  • Run a local or cloud-hosted ChatGPT endpoint (OpenAI API, or Ollama on a local server) and expose Hubitat Maker API devices to it.
  • Example: “Is the garage door open?” → ChatGPT queries Maker API → answers naturally.
  • Or “Turn on porch light at sunset” → ChatGPT generates the appropriate Maker API command.
  • This could be voice-enabled (through a tablet, smart speaker, or something like Home Assistant’s Assist-style interface).

3. Context-Aware Notifications

  • Hubitat apps can already send push/speech notifications.
  • ChatGPT could “rewrite” these into friendly, context-aware messages:
    • Raw: “contact sensor = open”
    • AI: “Hey, the back door’s still open — should I lock it for you?”
  • Useful if you want natural language status reports rather than raw device logs.

4. AI Decision Support

  • Some flows are complex (multiple conditions, exceptions, variable states).
  • ChatGPT could be asked to “explain what will happen” if a given flow runs, or “why didn’t this light turn on?” by analyzing your JSON flow logic and Hubitat logs.
  • Basically, AI as a troubleshooting assistant.

5. Vision + Hubitat

  • Tie in camera/vision AI (Blue Iris, Frigate, etc.) with ChatGPT to describe what’s happening and push events to Hubitat:
    • “I saw a person near the driveway at 9:14pm — should I turn on floodlights?”
  • You’ve mentioned Vision AI flows before, so GPT could act as the reasoning layer between detection and action.

6. Dashboard / Voice Companion

  • Add a ChatGPT “tile” to your dashboard — type or speak anything, and it answers with Hubitat context included.
  • For example:
    • “What’s the temperature upstairs?” → queries sensor
    • “Make it 72” → sends thermostat command
  • This makes Hubitat feel like having a live “home AI assistant.”

7. Automation Generator App (Flow Generator++)

  • You’ve already started a Flow Generator app.
  • Next step: have ChatGPT not just output raw JSON, but also explain what it built, let you tweak it conversationally (“actually, make that 15 minutes”), and save straight into Hubitat File Manager.
6 Likes

This topic was brought to you by OpenAI and Flow Engine.... :wink:

4 Likes

I asked gemini if it could edit a pic, it said yes paste the pic' so I did then it said sorry I can't do that, I replied you said you could, it then said.................sorry I forgot that I can't do it

2 Likes

Some basic innacuracies, even those tnat have to ve derived like this one, are often baken into the release of the AI platorm

What are you using to between Hubitat and the LLM/AI service. Do you have a Python server inbetween.

Right now everything is done in the cloud until I decide a direction to go in. Would love to have the AI local but not willing to spend money on it until I can prove it's worth it. Been watching your progress but local AI still seems 'not quite ready'. It's getting better but until someone can come out with a system that works together (hub and satellites), it's easier to just go to the cloud.

Surprised more people haven't chimed in good or bad about how they think AI will effect HA in the very near future! What do you envision?

I see it starting to happen now and I think this will start growing significantly more (especially among new-to-smarthome folks) -- using vibe-coded apps exclusively instead of learning how to use any rule engines (i.e. avoid learning how to formulate a logical argument).

I realize the rich irony of me saying that as someone who couldn't code my way out of a wet paper bag, so although I'm just one small step above those folks on the ladder, at least I'm not going to end up at the complete mercy of AI in my home automation setup.

2 Likes

Personally I am not surprised.

I think the media are right to make a big thing about the possibilities and risks associated with AI. On the flipside I don't think enough has been done (yet) to sell the benefits in a tangible sense for consumers to properly action.

That's not to say there aren't examples where people can be making advances or obtaining some benefit, but... well....

That all sounds really high level.... I guess my take is if you have a specific use-case then AI can likely offer a solution, but don't expect a well-honed approach,,,,

I want to say yep, go for it, but that response doesn't really support the reality....I guess it is grey....

EDIT - Ok, maybe I could be a little more positive... for those wanting to use local voice control... personally things have picked up and become more promising in recent months.... I'm not going to claim that I am personally astute when it comes to predicting trends in this regard, but others have expressed excitement based on increased hardware specs in local hubs like the Philips Hue Pro and new Aqara hub, indicating these could be to support processing local AI requests.

This does provide a level of optimism for me personally if "big business" are investing in what is expected to be local control. I would personally like to see that translate into similar enhancements in more generic platforms than Philips Hue or Aqara... But am willing to acknowledge the positive at this stage. And to be fair, that is actually the case with the C8 Pro....

I would be very cautious about considering Local AI on anything Arm based. Processing usefull LLM's requires allot of CPU, memory, memory bandwith, and high parrallelism. This is why anything that runs a LLM reasonably well now has a decent GPU behind it. My Ryzen 5090 with 128GB of ram can't run a LLM anywhere close to as well as my tiny 3050 GPU with 8GB of ram. The usefulness of the AI Engine will be sevearly handicapped if you are limited to a very tiny model to work on such small hardware.

I want a self-learning (like self-driving) automatic house that learns what its occupants need/want and then does those things. Most people can't have home automation because they lack the skills to design, and write code, or troubleshoot bugs. If I wake up dead tomorrow, my wife would have to hire an electrician to come out and remove all the smart bits.

2 Likes

My problem is that I've set up so much of this several years ago, that just works, When I go to make a minor edit it can take a while to figure out which rule is managing it. And, what was I thinking when I set it up.

2 Likes