And I could have saved myself a lot of heartache by avoiding cloud services.
Thank you. I do not want the functionality of my home dependant on someone else's infrastructure. Or my house to stop working just 'cause the Internet is down.
It's the epitome of stupid when something as simple as turning on a lightswitch has to send a message miles across the Internet and back to get the job done. John Henry Holmes is rolling in his grave laughing at us.
There was an article on HN a bit ago where Amazon dumped a customer, which deactivated a good chunk of their house. The author wrote something like, "I may consider removing Amazon from the equation over this." Fucking "consider"??? It's perplexing trying to figure out that perspective.
Ten or so years ago, when Google was still mostly a darling, I never thought they would ever try to pull anything like that. Yet here we are, and my dropcam is just going to brick itself in April. No update for RTSP/Onvif, just FU you are out of luck.
Similarly, at least one device I bought didn't require an account when I first got it, but then all of a sudden there was a new app update and you didn't think twice about it, but now this requires a cloud connection. It sneaks up on you, and a few years ago I wasn't thinking about this stuff.
But yeah, now I have zero tolerance for this stuff. I have had Google/Nest brick my stuff and nag me to get a subscription, Sonos has tried to strong arm me into upgrading my equipment at a cost of over $1k, etc. I am more than willing to pay tons extra for devices I actually own and control, and whose data I own and never leaves my network. Problem is, that its tricky to find stuff that supports open standards. I was more than willing to go in on Ubiquiti's very expensive gear until I learned that they also lock you in and don't support RTSP/Onvif. Then you are mostly in dubious Chinese brands for things and you can't be sure of the quality- though at least I can block them from phoning home or sending anything out of my network.
The smarthome world is just a real mess right now.
Now, a decade later, I have more time than money, the cloud services aren’t so reliable, and I’m increasingly uneasy with the implications of relying on something that can disappear (or change for the worse) at a moment’s notice.
So I’m ripping all of that crap out of my digital life.
That’s at least one example of how someone can be savvy enough to set up something like Home Assistant but end up with a bunch of clown crap in their house.
The other issue is that some cloud services are just better than what I can do on a raspberry pi. I’d rather get voice recognition from Google for ‘free’ than pay electricity and time for a server to always be on.
They get something home, and it is useful, and their spouse or kids like it... but it has to be hooked to the internet. maybe it needs an account to activate. or many of the other convenient/inconvenient tricks.
But I'm happy that we have home assistant and tasmota and esphome and all these other wonderful projects that route around lots of these problems.
Basically, I want smart home things, I have some blind spots in my software and engineering skills, therefore I'll DIY my smarthome features from the ground up.
I wrote a server in python, integrated some ESP32s with temp/humidity sensors. I then integrated them into Vindrikting units. I learned super basic front end stuff to give a page with graphs. I wrote a telegram bot to request the data from anywhere. Then I re-wrote the server in Go(because I wanted to learn go) to where its mostly functional but I lost steam and still rely on the python server.
Then I found some Kasa smart power strips that are individually addressable so I set up a isolated IoT network with a Pi zero w, and started writing a Kasa protocol library in Rust(because I want more rust experience) then used that library on an ESP32 using the embedded rust toolkits(because my ESPs with C code randomly crash and i want more protection against my grug-brain tendencies) to create remote controls for these kasa devices.
This approach is far from efficient but it's less a solution to home automation and more a vehicle to guide various learning projects with fun tech. Since its all on my local network and just for me I can decide what I need and don't need. Basically I'm making a really delicate system bespoke for me but as I learn more, each component gets easier and less delicate and I'm getting pretty good at architecting these things. Only downside is its really hard to return to older aspects when new ideas are circulating(ie I really need to finish the go/vue server and begin more integration stuff with the dedicated IoT network).
If I plug a device into power here, it must abide by electrical safety standards in order to be sold.
If I buy a phone or a laptop, consumers expect that it has wifi that works interchangeably with existing/other devices or it will not sell.
I suppose these rules haven't been applied to IoT stuff simply because the market for it is so small. Technoshit not working together in a common way is what bugs me the most about living in 2024; companies only trying to make a profit and not create a shared platform that betters the consumer, even doing stuff specifically to make competition harder.
We could have it so good!
But when everything works, it’s such a cool experience. I hope Matter begins to make local-first a bigger priority as well. That’s a huge concern I’ve recently begun to develop.
Also I still haven’t figured out how to combine a mechanical and digital solution. Like I want light switches that match my server state for my lights. If I change the switch via my phone, I want the switches to change as well so that the physical switch is always the true state of the light.
Like it's an absolute requirement that they show a different physical position when on or off?
I've got some relatively cheap z-wave switches. They're only momentary, but up is on and down is off. When they're off, I have them configured to show a small white light on the bottom. So my indication of switch state is "small light = off, big light in the middle of the ceiling = on".
The switches _can't_ be out of sync with their state because they are directly controlling the attached lights. If anything were to somehow get out of sync, it would be home assistant.
Mechanical switches aren't even reliable for this because of how 3-way switches work (their state doesn't indicate the state of the light). Unless you remember which are 3-way and which are single, you can't trust any of them. With smart switches in the same locations this is a non-issue.
An interesting thing is switches where you can't see if they're on as easily, such as the outside front lights, are for the most part fully automated (based on timer/door/motion and/or camera object detection) and we basically never touch the switch at all.
It's important to be consistent, though, since you technically can program the switches to do anything. In my house pressing up once turns the light on as you'd expect, but in some rooms the light will be less than 100% or have a warmer color temperature at night. Pressing twice turns on max brightness (or something similar, such as: 5000K color temp; also turning on the lights an adjacent room; turning on _all_ the backyard lights). I rarely use 3, 4 or 5-taps except for a couple special things (eg: 4x off on my office switch puts my PC to sleep; 3x on one of the basement switches activates a "party mode" my kid loves).
This sounds great. Can you provide a pointer or full search term to getting these ?
Updating HA always reminds me of what a friend used to say. "Software is like cathedrals. First you build it, then you pray".
The post-update relief when everything actually keeps working for once is really great though.
If you can from some other location see the light switch you already know the state of the lights.
If your automations are set up well you really should hardly be touching the light switches so they become more of something you only interact with for special reasons.
Not easy to do, short of using switches that can be operated through servos. However you could replace the switches with buttons and step relays that feed back their state to the server, in other words the server doesn't care about the switches or buttons anymore but only reads/writes the relays state.
That would be nice, but in all other areas in tech that's exacly the opposite of what companies want. I haven't looked into matter too much, but I'm sceptical.
For the lights that I want to "keep mechanically in sync" I replaced the switches with buttons and used a re-flashed (ESPhome) sonoff-mini which conveniently has a button input for the relay. Everything is synchronized no matter the operating interface.
As of today, Matter is still basically junk.
I recently bought some Matter smart plugs; I wanted the matter ones to test out Matter and see if it was worth investing more into since I have a SkyConnect in dual-mode.
I paid _double_ for smart plugs with Matter support, only to find that when connected via matter they're just dumb switches; none of the advertised energy monitoring functionality works. When I asked around on the HA Discord channels about _why_, it turns out that the Matter spec doesn't yet support things like energy monitoring.
Until the spec becomes actually useful, it's just a waste of money that will give you fewer features than the device might actually support.
HA supports HomeKit without requiring any Apple devices.
Personally my ideal HA would just be if YoLink made their platform open source. Their hardware is cheap and actually works pretty much perfectly, the only thing missing is open local APIs.
If you would rather have the lights control themselves (ex: rgb lights) you can always just replace your physical switch with something similar to the TX but that's just a button and then use automations on HA to have it tell the light what to do.
When this thing craps out, there is nobody in the world who will maintain it as far as I know.
The thought of hiring someone to replace it with another Crestron seems silly and expensive given I am a software developer and should be able to hack my home automation system.
OTOH I also don't want to spend my life debugging HA and don't really see myself getting around to the months of tinkering this project will require.
OTOOH professional integrators for HA seem to not exist at all. HAOS yields no results on Upwork. There is little discussion of professional installers in their forums.
Ideally I could hire someone to develop a system and give me a walkthrough so I can go from there.
Surprisingly the HA community basically rejects the idea that professional services around the platform makes sense, basically indicating that's its too unreliable.
You won't find a professional integrator to take on a home-assistant based project as the economic incentive is not there. There are some good independent devs but most avoid resi due to volatility and relatively small projects compared to commercial.
In the long run, this business model will just not fly in a world with today's connection requirements.
My rather basic home automation system has devices from more than ten different vendors on two networks (Zigbee, Wifi, soon probably Thread), many of which communicate with custom protocols. It's a wonder that it's working as well as it is.
Also, HA is developing at a breakneck speed. They release new iterations eleven times per year. Reading the release notes is a must, as many releases change things in subtle ways that might break your system.
But other than that, Mrs. Lincoln, how was the theatre?
Maybe it's reliable for you, however why has no one built a professional installer service on the platform?
My "solution" is just to update it infrequently, but that does not feel like a very sustainable solution either.
If they ever introduced a LTS version I would jump on that straight away. My home automation needs to do not change every month. And I'd be willing to pay money for that.
But it might be that it's a bit unrealistic for this ecosystem and the amount of contributions, so I'll just keep putting off updates until I know I have time to fix it if and when it breaks.
That would be a lot of effort so might make sense to have company and subscription service do the maintenance.
Furthermore a lot of this you'll need physical access for which limits the scope of remote support.
Once you're paying someone professional rates than the cost of the parts becomes less relevant.
I don't think HA is that unreliable itself these days but if I were charging £££ to install I'd probably want to be able to have a company with an SLA to fall back on.
I suspect there are people using it professionally however they'll be selling automation solutions and just use HA in the background rather than exposing HA to the end user.
While that settles down, HA needs to focus on standardisation and redundancy.
First, there is so much potential from a BIM perspective, but they really ought to think about putting some frameworks in to allow exchange with industry standard modelling and data formats if they want to attract a professional market.
The Industry’s File Classes (IFC) [1] for example could let you tap into building automation, not just homes.
Second, it’s a single host service, which makes redundancy hard. What if the network goes down? What do? MQTT could allow the database and controllers to be multi-homed, but the architecture of the software itself is still a bit too tightly integrated (even if it is docker containers all the way down…)
https://www.designingbuildings.co.uk/wiki/File_formats_for_B...
I don't think it's a coincidence that a lot of what you like about 'Home Assistant' is Node-RED and Zigbee. They are two of the big building blocks for making a home that is more convenient to live in rather than less (the third being an MQTT server.)
I like the energy dashboard a lot, it has literally all data in a single view including cost and PV. There's nothing to automate: the dashboard is the point.
I also don't automate my lights. I don't want to automate my lights because our life isn't run on such a tight schedule. It always ends up less useful than a light switch in the wall.
What I do want is to turn off all lights in one go when we go out, and quickly glance at the status of lights all over the house.
For these kinda things a dashboard is perfect.
My use case is maybe a little extreme - I live off grid, and I have HA managing our picogrid (dump loads, shifting charge between battery banks, throttling the hydro, etc.), our water pumps, monitoring filters and telling me when I need to change them, and then of course all the usual climate control etc.
Ironically the one use case I have no interest in is the most common one - lighting.
If office computer is used, and "home alone" mode is on, turn off all non-office rooms.
If I hit play on something on the main TV, or my laptop is used and its network MAC is the couch dock, turn off all rooms except living room, etc...
(I have a guest mode that disables these things, which is automatically enabled when my partner uses her door keycode)
Now, if they were incandescent it would be a different story entirely - it would be more like 3kW!
I do actually have one small lighting project in mind - it’s about 500 meters from our house to the mill/guest house, and I’m toying with the idea of having little LoRa/solar fairy lights that turn on and off as you walk from one to the other at night, based on device presence, zone by zone, mostly because it would feel magical.
- Do NOT use a device with a micro SDcard as the main storage!
- Play around with devices, but be picky about what you use long term
- Communicate with your spouse/other people in the house before experimenting :)
After some crashes and hard lessons, I now run HA on a Proxmox that has two boot drives and regular snapshots/backups of the VM. This adds enough redundancy to keep things running and avoid a dark house if a disk breaks down. With great power comes great responsibility!
You should post why. It could be that you don’t know to buy endurance cards for wear levelling? Or you do and that’s still not good enough? Or something else entirely. Basically I can’t tell if you know more or less than I do from just the statement.
Personally I boot my Raspberry's using netboot (DHCP via dnsmasq) and NFS. However I don't run Home Assistant on a Raspberry, it runs on the "server" (headless machine) that's serving NFS .
The SDcard started failing when I was on a prolonged Christmas holiday – and Home Assistant was responsible for keeping my house above freezing temps when it was below -10℃ outside. Luckily I have nice neighbors who helped me out!
Anyway, I had planed to move HA away from the rPi3 for a while, but you know how it is, more important things keep happening, so this crash forced my hand.
SDcards are a gamble, especially in a device you want to keep running for years and years. I feel much better having it running om my Proxmox with dual boot disks and regular backups, so WHEN a disk is calling quits, it still keeps on working and I have time to replace the broken disk.
I started out with the VM image, which was a good low commitment "gateway drug" to try it out and get to know it... But a big downside I witnessed is their "haos" distro handles disk failures very poorly. I hit that twice, and recovery was not fun. It basically put up an infinite spinner on the web interface, until I messed with the disk image somewhere else. Haos does not even have fsck.
Now my docker setup runs out of ZFS and I'm happy. There is no update UI or addons but if you're comfortable with a little mild sysadmin style stuff it's fine.
*EDIT: I meant HACS, HACS is what I meant.
Thanks for the tip.
This is my concern too, having started just a few months ago. A few days after switching from ZHA to z2m, suddenly one of my zigbee TRVs stayed at 22°C. I thought it was better thermostat acting up, every time I tried to lower it directly via Home assistant it would snap right back to 22. So I opened the z2m webif and was greeted with "no network route (205)" every time I tried to set a new target temp on the trv. You see, with ZHA any zigbee errors would show a popup in HA directly, but since now HA just talks to z2m via mqtt, HA never knows anything went wrong. And to add insult to injury, the trv was still sending updates of the measured ambient temperature just fine, so I didn't immediately think of a network issue.
- having HA control your heating's on/off hours, where the watchdog is the heating's thermostat making it stop heating when it's warm enough or forces heating if it gets too cold. So HA crashing and leaving the heating in heat mode might be just a waste but not a serious issue.
- having HA control the wanted temperature on the other hand is more problematic, because for all I know it could misbehave and make the heating want to go to 30 degrees Celcius or to a value so low that there's no frost protection anymore, then crash and never get it out of that state anymore. And there's no watchdog anymore correcting for it. Potentially this can cause issues. Chances are small, but I don't like the idea that AFAIK these chances are much larger than standalone heating acting up like this.
Likewise we can now opt for a dynamic electricy tarrif, basic idea being that for instance when you now the prices are going to drop below a certain threshold during the night you're going to tell your home battery to charge at that time. Of course the thing acts like a watchdog for itself in that it stops charging when full, but there is no watchdog keeping it from charging continuously. In other words: if it's put into charging mode and HA crashes and leaves it in charging mode then it will happily continue charging during peak hours. Not 'serious' per se but pretty stupid.
I've done it in "pure" HA due to specific requirements and do not recommend it. Reliability has not been a problem but everything else was. You'd probably need a custom thermostat software component (as the HA built in one is limited), a wi-fi connected switch (like a Shelly), a high amp relay (if heating or controlling an AC due to startup load) and a low latency temp sensor (most have a 10+ minute delay or 2 degree F delay, I found Govee Bluetooth sensors to work well). Then you'd still probably want a remote control or a physical control panel/dashboard. The IR remotes for ACs are absurdly difficult to decipher (I never succeeded) and there's no good Zigbee remotes I found (the best I found is the Yolink remote but that's not local). Dashboards either involve running a browser on some LCD on a wall or an e-paper display wired to esp-home. Making a dashboard in esp-home is like going back 40 years as everything is individual graphic components drawn one by one. It might annoy you (like it did me) enough to build an svg bashed dashboard creator, a renderer on your HA box and then use a dead PR's remote image loading support in esp-home.
edit: The only positive about my approach is that I was able to build a custom "feel like" temperature curve that combines temp and humidity. So I no longer wonder all the time why I'm feeling cold or hot despite the thermostat being set to the same thing as last week.
The problem is this disruption makes me feel really disinclined to update but when leaving it multiple months it's harder to find out what breaks beside I have to go hunt for all the intermediate changelogs.
I just had to restart, and everything worked again.
Sounds like that would be a major pain though. Is there a pressing need to keep it up to date? Could it be treated as a set up and forget forever system, like I did with my NAS years ago?
Especially if you leave it multiple months because the info will be scattered over different release notes.
Once I changed my mindset, I counted the number of rooms in my house and bought that many motion sensors. Same with door sensors: every door and closet got one. (I had already replaced all light switches with Lutron Caseta.)
After a weekend of attaching and configuring sensors, the physical part was done, and it was all software after that. Now, if we want a different behavior, we pull out a phone and change an automation, rather than getting out a stepstool and VHB tape. The house feels more like a platform than a bunch of temporary hacks.
I like Z-Wave, but the stuff is expensive and options (for different types of devices) are fewer, so I've generally followed the economic incentives and ended up with many more Zigbee devices (38) than Z-Wave ones (7).
Also, Z-Wave is more power-hungry. I have one Z-Wave motion sensor and I need to change the battery every few months. My double-As in my Zigbee motion sensors last two years.
It just works for me and what i do, all you have to do is occasionally apply updates. Granted i mainly focus on creating my own powersockets and fuses with shelly products and otherwise focus on local homekit devices.
I also have some of those tuya devices and i hope to get rid of them at some point very soon. Zigbee is good for local and as long as the switch runs by clicking it too, i really have no issue nor care what its running on.
I just have a secops feeling that wifi is really not what i want.
Alternatively you boot and run a pi off a USB stick.
Then on install it auto-discovered a ton and I've slowly kept adding more and more smart devices. I went the same way as the OP did where all of the automations are very unobtrusive or simple, or do things like monitor some plants.
Very happy with it, and I'd encourage paying for their cloud service for remote access just to throw the project some cash!
Curious about this. I tried HA a couple of times but kept going back to Homebridge. I find it far simpler than HA, and easy to write my own plugins for devices that don't have functionality yet
For me HB was the lighter brother of HA, and HA the full-fledged thing
I'm loving the results of Year of the Voice, and there's a contest on now about how to best leverage the hardware and software to build voice interactions. It's a pretty exciting time to be using it.
The original one of the pi is for the whole image. For the Jetson series I create partitions so the OS indeed is memory overlayFS but there’s also a rw data partition and a ro re-mountable configuration portion.
In worst case i don't need it.
But controling heating remotly is great to have. Being able to collect the humidity to prevent mold is also great.
I hope that all of this tinkering and pushing for it, will make a much more stable and fun experience in the long term.
I might have a chance to rebuild a old house, i'm still not sure if i want to put cable (some bus system, 12v or PoE/Ethernet) everywere or just using some radio based components.
The biggest annoyance is defintily the energy usage of bigger things like heating valves or potentially a window or a door.
After moving on from X10 I found another poorly known "automation system" called Home Heartbeat as I was away from home a lot. Having some reassurances the doors were closed and the basement was not flooding was good at that time. Seems like the main thing this system was intended to do was to allow remote water main shutoff and then provide sensors (although the water shutoff valve was unobtanium, lol). I picked up the system and a bunch of sensors on clearance as it went discontinued. This was a system that was just a few years too early to market, it had some features that would be considered hokey today, but could be repurposed into a useful tool. It used an early Zigbee wireless mesh for sensors (door open/close, overhead door tilt sensor, water sensors, motion sensors, etc). The base station had an integrated modem, but also had a USB port for computer integration. I wrote some perl to digest the USB status output and that made its sensor network entirely accessible the way I wanted, I had status updates viewable on my phone from there. I was considering integrating it into HA at some point, but another problem the system had was that the sensors were AVR based rather than ti MSP430 with low-power features, the sensors ate CR123's like candy.
At this point I'd just like a page to view all the wireless temp sensors around my house in one place. I don't need another light-switch-with-more-steps.
On another note, any users of Mr. House here? (Pst, hey kid... you like perl AND home automation? I got just the thing for you!).
I have a server I’m sending to someone else’s house, and if it crashes, I want to use a smart plug to cycle it. I have a Eufy smart plug I got for free. Is there a better alternative? This persons house has no automation and no smart hub. Unfortunately I’ll either need something cloud based or something I can statically point to my own Home Assistant running in my house. Is this cloud based Eufy plug my best choice?
There are other ways you could have the S31 do operational checks but ultimately ESPHome is probably an interesting consideration and supported by tons of off the shelf hardware.
There are devices with "watchdog" features that can cycle power automatically if an ICMP ping fails. https://shop.netio.eu/netio-power-sockets/powercable-2pz/
Also, people have rolled their own watchdog solutions using Sonoff devices flashed with Tasmota. https://community.home-assistant.io/t/watchdog-device-using-...
Not that I'd recommend it, but you could probably just port forward an obscure port directly to the Kasa plug (DHCP to a static IP).
[0] https://cpc.farnell.com/tekview/powertxtuk/gsm-power-socket-...
Ended-up building a PLC based system. Its been rock solid reliable & I can do pretty much anything I can imagine with it. Its a little risky, in that if I ever need to sell I've limited potentional buyers. But, I think thats the case with just about any centralized system. Not long after I looked at a lutron system, the local lutron tech passed away. One house under construction using lutron was brought to a dead standstill. AFAIK, there are no nearby lutron programmers to this day.
However, PLCs are easy to obtain, relatively easy to learn (if you're technical) and just about any small+ city is likely to have control engineers that might be willing to support as a side hussle.
Native floorplan would be great, too.
HA is fantastic, but it has a steep learning curve on the frontend part.
As for default dashboards, I think the only real way to do this properly would be to create a wizard to help get things off the from ground.
It really does need a fair understanding of (or at least the existence of) entities, devices, HACS, and the developer tools pane. And that’s just the foundational stuff you need to know about before you can even consider tinkering with platforms and entities in yaml files.
Some jargon-free steps like ‘Select all of your outlets’ > ‘Select the properties you want to display’ etc.
A ChatGPT configurator would be interesting, similar to the one in Pirtainer (which I admin I’ve never used).
I also think a floorplan where you can click a room and see the lights and everything is much more intuitive than a website.
But also I think many things can just be assumed. Everyone has lights, for example. So why not have this be a native solution in stead of having to set it up manually.
Setting up the Energy dashboard, including all the graphs, data and history etc... took less time and effort than it would take to automate a single light bulb.
The biggest pain has been ZWave devices. I've had multiple in wall switches croak for no obvious reason. Also devices dropping out, intermittent problems receiving commands. Recently installed a whole house surge protector in the main panel...we'll see if that helps.
The biggest hit with the household is the random family photo on the dashboard, updated every few minutes from the NAS archive.
Zero dependencies on cloud services. Wherever there are no good choices, instead I make custom ESP8266 devices, all integrated to mqtt. Have deployed one to listen to my weather station, one to control outdoor lighting, a roof snow melt system, even one to control the sprinklers intelligently.
Like writting a program.
I wonder what the exact differences are. As i'm currently using ZHA and it's running fine... (mostly using Ikea devices and a few HUE bulbs.)
https://youtu.be/PR-ugz1uLWc?list=PLnQ5CgUMCvdbb8UHaY3Qy12NY...
(German video)
While I have a few Raspberry Pi and Odroid devices kicking around, they mostly have jobs already. I would like to be able to run something as a package on top, rather than replace what it does now.
$ python -m venv hass.env
$ source hass.env/bin/activate
(hass.env) $ pip install homeassistant
Good developer community and helpful forum
UI could be better but really powerful.
If you use a system like HA to automate a lot of simple things, it just works™. I've been running HA for years now and I'm incredibly happy with it.
Most rooms I don't spend a lot of time in have motion activated lights that sill allow me to override that with a simple wall switch, some lights turn off when I stream something on my TV and I have temperature and moisture monitoring with alerting in case something goes wrong like my washing machine blowing up.
Hear the tale of the two apps:
1. Haier hOn: takes between 5 and 15 seconds to start up. Assuming it didn't randomly log you out, it takes a minute of tapping through bullshit multi-level screens, with multi-second delay between each, to get to the point when you can control a single aircon unit. Switching between units is 2-5 seconds - and you need to be careful, because if you press the phone's "back" button, it'll shut down the app, so the whole 1-minute ordeal restarts. Every 5-10 control interactions, it'll go "oops, something went wrong", and then to get state and control over units, you need to restart the app. You look at it sideways, believe it or not, restart the app.
Total time to set three aircon units: some two minutes, maybe more, and a lot of frustration. My wife flat-out refused to do it, instead always asked me (as I have much more experience in quickly navigating buggy software). Myself, I hated it every time.
2. Home Assistant app: takes 2-5 seconds to cold-start into the app, less than that for warm start. The default dashboard shows all three A/Cs, with controls, temperature graphs - everything at a glance, on a single screen. And then below there are controls for floor heating and other stuff. Zero crashes.
Total time to set three aircon units: 3-10 seconds. I sometimes race to see how fast can I do it. My wife loves it, I never get asked to do anything with A/C for her anymore.
I'm only beginning to integrate and automate more things. Recently I upgraded a bunch of cheap (but stylish, and importantly, well-built) IKEA PM2.5 sensors with ESP8266 (decade-old NodeMCUs I had in a box), and through a bunch of graphs quickly added to another Home Assistant dashboard, we've learned a lot about what does and does not affect air quality around the house. I've also started to set up automated notifications for bad air in different rooms, because it's stupid easy to do. Another stupid easy thing to do was to make my washing machine notify both of us when it's done. Or auto-dimming that one Hue bulb my kids are using as night light. Etc.
If my setup died today, nothing would stop working - a few things would become less convenient, aircon would become stupidly annoying to operate again - but we would feel it, so I'm now trying to beef the hardware durability up a bit.
And yes, I ain't buying anything "smart" that can't talk well with Home Assistant anymore.