Very beginning. A pointer to architecture for learning system?

All points to “read the documentation”. Understood, but maybe someone of patience can point me to which section I am looking for, to evaluate whether openhab is what I am looking for.
My brief read tells me of things, and channels through bindings, which through rules cause actions to be taken. The jargon, I’m sure I’ll get right if I get into it.
All the rule examples seem admirably simple. If this then that. Comms by anything, mqtt &c.
The forum is abundant with tips, fixes and recipes to get things going. Admirable.
Even a quick read tells me that I can automate with this.
But where do I put the “smart”? Or how do I link to it?
For example, if rather than hard-code rules, I want to bring in and apply some of my learners to learn when to turn those lights on & when to leave them off, can someone point me to where it fits in the architecture?
Or the section that talks about how to link out to my machine & back?
Do I create bindings to give my learners access to the real sensor states from openhab and have them manage fake soft-sensors that publish to actually cause actions?
Sorry. Newbi question. Can someone point me the the section to read?
Many thanks.

1 Like

Welcome to the openHAB Community :heart:

Yes, RTFM! :smiley: Jokes beside let’s get you started.

You struck a nerve there :wink:
Where to put the Smart in SmartHome is question that needs to be answered.
@luckow Held a talk about that question at FrOSCon 14 (in german but the YouTube auto translation does quite a good job)

Thanks from the bottom of my heart :heart:
So many people here try to give the best support they can offer.
I am truly grateful for this awesome community.

I think I can read from your text that you want a system that can learn from your behavior and support your behavior.
Buzzword cannon watch out “Machine Learning”.
As @luckow points out in his talk, the Smart in SmartHome is YOU.

I think you might did not get a response so far since the thing you are asking for, is a dream of many people here and we are all working on it but are a little embarrassed we did not make it yet. :thinking:
At least I feel a little like that.
Or it’s just TL;DR („too long; didn’t read“) :man_shrugging: who knows.

I am sure openHAB can do what you expect it to do, but only with a lot of time and effort.
Maybe I misinterpreted your question completely wrong :smiley: Maybe you can state what your goal is, so we can give a more Yes/No answer with look at this or that response.


Hello and welcome!

I’m not sure I understand your goal and what you mean by “bring in my learners”.
I will try to summarize OH in a very simplistic way.

Tipically in OH a Thing could be (but not restricted to) your phisical device.
To be able to manage it OH must know how to talk to it (protocol?): that’s the binding .
Once OH knows your device it can interact with it via channels (they can be readonly channels for status or “command” channels or both).

Now you can link those channels to items that are an “abstraction” to read and send data to channels (and to be shown on the UI you choose to use).

Rules are the way to automate action on your items (behavior?) in an hardcoded way.

If you have something external that can “learn” and act differently, it can interact with openhab via the REST API, so you can create all your things/channels/items inside openhab an leave the automation to some external “brain” that reads and writes via REST API.

I hope I answered some of your questions.

Many thnaks for the recap. Yes, as I suspected. Thanks for pointing to the link being restful.
I guess being a little lazy, I was looking for a pointer to the best page to start on for creating that link to external.
Many thnaks for confirming abstractions (my soft-sensors) and some a usual/only route through a restful API.
I’ll search for it.

Thanks for the info.
Sorry, I asked/risked the blunt question, not because I wanted to highlight, but I didn;t want to miss any parts aimed at the same.
I’m quite happy with the statistical inference, prediction, supervised learning, semi-supervised learning, machine learning, even the nowadays often touted “AI”, though hardly fluent in the cloud-based black boxes so popular now.
I’m also comfortable linking to an external platform if it must run there, I just wanted to ask whether there were architectural features or even normal practice for how it connects, interworks.
The platform here seems admirable as the basis.
Ponder: Do such have layers as in comms? Never looked at mL interacts with real world before froman abstract perspective.
There’s obviously a physical layer and macs & such in the comms. Seems as if there should be such defined layers in the system as a whole. Rambling. Must do some homework to get to the standard of others’ contributions.
Many thanks. I’ll set a system up and see how it goes. And check for other topics.

I think this is a good starting point:

Once you’ve installed openhab, you can install a component called REST API documentation that can guide you exploring the API and make examples of calls through an UI.
You can even create dummy items not linked to anything and test them with the api

A lot about the working of the rest api and especially see event streams and co can be learned here: