With their own extension module, I would assume so. Edit: Duh, I can just look it up in 10 seconds. " Shelly Plus Add-on: Galvanically isolated sensor interface for all Shelly Plus relays". So, yes.
Shellys are not grounded, so you have to assume that anything electrically connected can have line voltage unless specially insulated with an optocoupler. They are not double-insulated devices you plug into an outlet. (Aside from those that are, like the plugs, of course…)
Considering the warning of high voltage, I would expect the galvanic isolation is provided by Add-on, not Shelly itself. So in that case adding “diy” hardware directly to Shelly might be not advisable. Or maybe the warning is just to sell Add-ons…
It is. You can google for its insides and there are some threads about the ESPHome support that go into details. Most of the addon is galvanic separation, with only a little bit of IO extension on the other side.
I’d say it’s a nice solution if you have to combine line voltage and sensors in one device. It may even be worth it to use it just for the extra sensors with the base Shelly acting as a power supply—getting a dev board supplied with power when mounted in tight spaces (inside a mains-powered appliance or in a distribution box) can be an issue.
The latter is the reason I used a blue Shelly (dry contacts) for my door opener, even though it only needs to switch 9V AC. But it’s inside a distribution board where 230V is readily available in abundance but there’s no outlet for a USB power supply in sight.
This is cool. Now I have a reason to flash some of mine. 
I didn’t know the 1% brightness limit or the course brightness transition was due to their firmware.
Is there an OTA way to flash them or did you open them up?
Dang! I think i just found my long lost brother! I completely agree with that strategy and use it myself too! Over the years i’ve seen way to many get burned because some companies went out of business and their app was useless after that and required people to have to go out and buy new devices or they do like so many others and dont play nice with others like HA and block methods people would use to integrate them and all because they want people to use their dumb ecosystem that had a pitiful product catwlog available so, you cant even find the things you want in their restricted ecosystem…
The simple fact that Shelly doesnt try to dictate what the users of its products can do with that product after sold and owned by the customer, its extremely refreshing and should be praised and appreciated IMO.
They even outdo themselves by putting access to the pins needed to reflash the device firmware so that they’re accessible on the outside and dont require useres to open them up and create access! Or at least for some Shelly models, im not sure if all do that currently but, i know they dont try to prevent people doing it and for that, they’ve earned my business and my recommendation.
I just tested it.
Shelly’s Esp32 GND is directly connected to mains live.
So for whoever here thinking of connecting some components to shelly without Add-on (or optocouplers): Be aware…
Could you elaborate a bit more on this? Are you saying DC 0V is connected to 110V/220V AC?
Yes, DC GND is connected to mains L
That’s impossible (not without some sparks flying). How did you establish that?
Just put your tongue to the pin header (or use multimeter).
No, burden of proof is on you. Show me your circuit diagram or an actual device under testing, and exactly what you are measuring. I’m telling you, what you’re saying is physically impossible.
Physically impossible, why? If there’s no return for the L, there is no problem.
I haven’t done any experiments or research with the subject since I’m not interested in connecting any diy circuits to Shelly. The whole case popped up to my eyes few hours ago from some post on this topic about connecting diy components to shelly with esphome.
If you want to verify what I’m saying (and what Shelly is warning), just measure the continuity between L and Esp GND (from Add-on header for example).
I see now that you’re referring to the DC circuit floating on AC (I got nudged in the right direction by someone else), not that the grounds are connected. My apologies. I stand corrected.
Obviously it can’t be closed circuit 
But it means that if some idiot here decides to solder a temp sensor to “Esphome Shelly”, he gets fried by touching the sensor when powered from mains…
That’s why I wrote: Be aware…
The reason of my posts here is simply to give awareness about these smart switches. You can “hack” them on software side but don’t think you can use them like Esp32 dev board.
Which one was L? Isn’t that the one thats safe to touch while standing in a puddle of water? Im pretty sure it is… Probably pat myself on the back for getting another one right!!!
Haha… (could be) guilty as charged…
But, then again: don’t EVER connect anything to any module while it’s on power. Beside above danger there’s additional danger of frying module at the moment of connecting. You never connect anything live, unless specifically specified so; if you do you’re just asking for problems. Always disconnect power, connect sensor, connect power.
Back those old days (when small switching chips weren’t available yet) pretty much all small low power circuits were live: 200-470nF capacitor (depending on current needed), 2 diodes, zener and electrolytic capacitor and voila! Thus, depending on how we connected L and N live voltage could be directly on GND, yes.
Yes, i’m among olders and i’m aware of these things.
Exactly, and deeper you press your feet in to the mud, safer it is. It’s called PE - protective earth… 
BTW, to understand why DC GND is connected to AC L and not AC N as one would assume, you need to have a look at those Shellys that can take AC and DC.
Th input is pulled high and is low-active in typical microcontroller style. For this to work with an AC wall switch that’s wired to switch L, L needs to be connected to DC GND. This also allows it to have two different DC+ inputs with a common switch input.
I would have to look at the actual circuit, but I think the switch input actually doesn’t react to line voltage but to the DC GND connection to the internally produced DC the switch and L wire provide. One could test that by adding a diode in series with the switch—that should make the switch invisible to the Shelly.
This also explains why using wall switches with built-in signal lights doesn’t work, no matter if the signal is wired “always on”, “on on on”, or “on on off”…
