One of the reasons I was interested in getting into SmartThings was to pull most of what I have off the cloud and directly onto my network instead. I have been extremely pleased with the automations, buttons, switches, “toys” that SmartThings has opened me up to that previously weren’t compatible with just my Echos or my Google Nest Hub Max, but after a few months I am questioning if when I bark a simple command like I had previously, is it the SmartThings Hub that is using my speaker to “hear” the command, then executing the action, and then using my speaker to relay a message to me that it has been done… or is it still my Google Nest Hub Max that is doing the action and responding itself?
If anything, it seems like response times for easy commands like, :“Hey Google, turn off all the lights,” or “Hey Google, play this station on the Bedroom Speaker,” have gotten longer over time than when I was just using Google devices to trigger and execute actions…
In both the SmartThings App and the Google Home App, I see the device status updates correctly so I know they at least recognize what’s going on. Something just seems a little off to me, like the promised reaction times aren’t there.
If I could say, “Hey Dave, turn off the lights,” and then the lights would go off and my Nest Hub Max would then speak, “Dave here, got those lights you were on about Trav, pleasure to assist,” then I would KNOW Google had nothing to do with the actual execution, but I haven’t figured that part out yet.
The voice assistants are not just speakers added to ST, because ST doesn’t understand spoken language. So both platforms have their part to play.
you speak to Google home. It does NLP (natural language processing) to figure out what you said and converts that to a request to be sent to ST, like “turn on the kitchen light.” This is why you can say “Lights on” or “turn on the kitchen” or “turn on the ceiling light” and get the same result for each voice command.
ST receives the request and then acts on it exactly as if you had selected that option in the ST app. So it is the one that sends the instruction to the kitchen light.
(There are a few partnerships where a smart speaker can send a device control command directly to a device, like Google home and Cync lights, but not many, and it’s not how the ST integration works.)
Right now NLP mostly happens in a cloud because it is very complex AI that takes a huge amount of processing power. So the communication is
A. Smart speaker talks to its cloud
B. Smart speaker cloud does a LOT of AI work to figure out what request to send to ST. This is sent to the ST Cloud via the internet.
C. ST cloud receives the request and sends it via the internet to your St/Aeotec hub
D. Your hub then sends the command to your local hub-connected devices.*
So right now voice processing typically requires two systems (one to process speech and one to control devices) and two clouds. And the internet between them. Which is how Google home and ST work together.
The industry would like to eventually have local voice processing, but so far there’s very little of that.
If there is some weirdness, it’s usually an NLP failure, not multiple commands being sent or lost. The process of turning what you said into an actionable command failed because the smart speaker didn’t hear you correctly or the AI came to the wrong conclusion about what command the verbal statement represented.
It can also occur because your smart speaker tells you something is done when it’s done its part and sent the request over to the ST cloud. But if there’s a failure on that side, like your hub being in the middle of a firmware update, the smart speaker may not know that’s happening. So it could tell you something was done when in fact all that really happened was that the request was sent. This comes down to the details of each integration.
- other protocols that don’t require a hub can work a little differently, there may even be 3 clouds involved, but the main point is that the smart speaker platform still does the NLP.
Thank you @JDRoberts for such a thoughtful and clearly expressed response. I really appreciate that since I am teetering on the line of “got a grasp” and “over my head” when I jumped into SmartThings.
Before I had a ST Hub, those simple commands like “Hey Google, turn off the kitchen light,” the Google Nest Hub was “hearing my command,” then performing the NLP in the cloud, then executing the action itself on whatever smart bulb I had identified as Kitchen Light- it would have had to since at the time there was no ST hub involved.
It’s almost as if I would have to remove every item from the Google Home App, make sure every bulb, plug, switch, and button only existed in Smartthings, and then that would prove that ST was the one sending the execute command.
Part of the reason I’m thinking about all of this is because I am finishing a new house in about a month. I’d like to walk in there and set up an efficient and concise smart home that doesn’t have multiple devices doing the same tasks. Just as an example- If I install a Lutron Dimming Switch and assign it to the kitchen light, I can then dim that light on the wall switch, I can walk to my Google Nest Hub Max to select the light and then dim it, I can say “Hey Google, dim the kitchen light to 60 percent” and either my Nest Hub Max, my Google home phone app, watch app, or Nest Speaker in another room will respond (sometimes all of them!)), and if I have a ST Hub I can say the exact same phrase but the Google Home App will send the command to the ST Hub to execute the action. And God forbid the damn speaker thinks it hears anything even remotely like “Hey Google” and acts on its own instead.
I just feel like there is a way to tighten everything up that I am missing. Maybe if I go into the new home and start from a mentality of Smartthings Hub and work outwards, instead of making the hub work with everything else I’ve cobbled together over the years, makes more sense?
First rule of home automation: “the model number matters.“
Most smart bulbs do in fact have their own cloud, even if they don’t have a hub, and that is what is sending command. Google home is still just collecting The voice input and passing it to its cloud to do the NLP.
If you had to set up the bulb with its own app before it would work with Google home, then that’s the most likely architecture.
As I said, there are a few exceptions, like Cync by GE, but at the present time that’s pretty rare.
(This will likely change once matter is introduced, but we aren’t there yet.)
So even though there isn’t a hub, it’s very likely that there is a third-party control system involved, not just your Google home assistant. With a few exceptions.
That’s just a matter of personal preference. When I had a smartthings hub, I went the other way: if the voice assistant offered its own native integration with a device, I always used that for preference over the smartthings voice assistant integration. But at that time, we were allowed to individually approve devices listed in the SmartThings app for use by the voice assistants, and that’s no longer true. Now it’s all or nothing.
For Alexa, you can go into the Amazon app and individually disable the devices that would otherwise be duplicated, but that’s a huge amount of work. I don’t know if there’s a similar option for Google assistant or not.
But basically my assumption is that the device manufacturer is going to do a better job of making a voice assistant integration for their device than SmartThings is. So I only use smartthings for the voice assistant integration if that’s the only option for that particular device.
Thanks again @JDRoberts.
When you had a Smartthings Hub…? Curious- have you moved on to something else? Why? Never looking back? Regrets?
I started using a SmartThings hub back in 2014 and stopped using one about 4 years later because I just could never get the reliability I needed.
As I have mentioned before, I need my home automation systems to have an MFOP (maintenance free operating period) of at least 6 months and preferably 12. I get that easily with Philips Hue, Lutron, Alexa, Apple HomeKit, even IFTTT. But with SmartThings historically I’ve been lucky to go 3 weeks. The fact that we can neither deny nor defer updates also creates instability.
I am quadriparetic. I have to pay someone else to do almost all tech maintenance, including just popping the batteries in a sensor. SmartThings was costing me a lot of money just to deal with minor glitches.
So I ended up moving all my critical use cases to Apple’s HomeKit, while continuing to use SmartThings for complex convenience use cases like getting a notification if the guest room window was left open, the guest was out of the house, and rain was expected.
This combination has worked well for me. But there are many different systems out there, each with its own pluses and minuses. Each person has to find the best match to their own needs and preferences.