Large Villa Renovation – Shelly + Home Assistant + Wall Displays Architecture (Need Advanced Advice)

Hardware
1

I’m currently renovating a large multi-level villa (roughly 1000 sqm / ~10,700 sq ft) and I’m defining the final control architecture before walls are closed.

The switching layer will be based almost entirely on Shelly devices installed centrally on DIN rails inside zone panels. No in-wall relays behind switches. Every floor will have one or more electrical cabinets where lighting, shutters and power circuits converge.

Home Assistant will be the main automation brain. Shelly Cloud and Ubiquiti Cloud will remain enabled as secondary fallback layers, not as primary logic controllers.

The design goal is long-term reliability, structured redundancy and clean wall integration.

Room-Level Interface

In each room I’m planning to install a Shelly Wall Display XL to manage local lighting, shutters and scene triggers.

The idea is that:

  • Local control of Shelly relays must always work, even if Home Assistant is offline.
  • The Shelly Wall Display provides direct device-level control through Shelly ecosystem.
  • At the same time, it can expose a small dedicated Home Assistant dashboard for extended functionality.

However, the Shelly Wall Display runs an embedded Android version that is relatively old and not upgradable. From my testing, it handles basic HA dashboards, but struggles with more complex Lovelace layouts, advanced graphics, animated cards, energy dashboards, etc.

This means I would likely need:

  • A lightweight HA dashboard specifically designed for Shelly Wall Displays.
  • A separate, more advanced dashboard for larger wall panels.

This is technically acceptable, but not ideal from a long-term maintainability perspective.

Floor-Level Central Displays

For each floor (or cluster of rooms), I want to install a larger wall-mounted touchscreen acting as a central control point. This display should be capable of running a fully featured Home Assistant dashboard with modern UI components, graphs, camera feeds, energy flows and system status.

I’m aiming for 15" minimum, preferably 21" or even 27". The device should be designed for continuous operation, wall-mounted, ideally PoE powered, and not battery-dependent.

Here are the options I’ve evaluated so far for the common areas:

  • Ubiquiti Digital Signage Displays Since my network and surveillance stack will be primarily Ubiquiti, I tested their digital signage products. The hardware is clean but extremely limited: no microphone, no camera, no sensors, no flexibility in app switching, and occasional freezing. The ecosystem is closed and not designed for interactive HA use.
  • Tablets (iPad / Android tablets) From a reliability standpoint, this is the safest route. However, size is limited (rarely above 15"), and embedding a battery-powered device permanently in a wall concerns me. Even with charge-cycling automation to preserve battery health, long-term thermal and safety considerations are not ideal in a permanent installation.
  • Android PoE wall panels (various OEM / AliExpress models, Android 12-14) Some models are attractive: larger formats available, PoE power, sometimes integrated RGB LED frames that can be controlled via HA for alarm or system status indication. That LED feature is actually very compelling architecturally. The concern is long-term reliability, firmware updates, security posture and vendor support. Many reviews are inconsistent and I would be installing multiple units across the house.

Architectural Intent

The overall architecture would be:

  • Centralized Shelly switching layer (DIN mounted).
  • Home Assistant as the primary logic engine.
  • Local device-level autonomy via Shelly ecosystem.
  • Simplified dashboards on Shelly Wall Displays.
  • Advanced dashboards on larger central panels per floor.
  • Ubiquiti network backbone with VLAN segmentation.

The objective is a system that remains stable and maintainable for at least 10 years, without depending on fragile consumer hardware.

Questions for Advanced Installers

  1. From a large-scale residential perspective, does this layered architecture make sense?
  2. Would you keep the Shelly Wall Display (XL) per room, or would you simplify and standardize everything on generic Android panels?
  3. What would you recommend for 21-27" permanently wall-mounted touch displays that:
    • Are industrial-grade or commercial-grade
    • Support PoE
    • Have no internal battery
    • Can run modern HA dashboards smoothly
  4. Has anyone deployed large Android wall panels long-term in a residential context?
  5. Are there industrial HMI panels or commercial touchscreen solutions that integrate well with HA via browser mode?

I’m not looking for a quick solution - I’m looking for a structured, future-proof design.

Any experience from large installs would be extremely valuable.

Thanks to all!!

2

Your presumption that the cloud and wireless control are reliable may be false, or proven to be a formidable challenge in hindsight.

RS485, Ethernet, CBUS, ModBus, Fiber Optics, etc, anything hard wired should carefully be considered as forming part of the backbone to interconnect as many devices as possible. Sure, it may add structural support to your building with the amount of copper, and cost a bundle, but you will have assurance that your devices are talking to each other consistently, not when the phase of the moon and conjunction of the stars align to allow a ZigBee packet to traverse your IPV6 VLAN segment you spend days configuring to keep hackers out and control local, and every time there is a weekly software update everything stops working till a fix is released, often next month. Consider a vertical backbone to connect each floor, and a locked utilities cupboard on each floor where everything terminates.

Then of course, there is that new-fangled stuff called Matter. Plan carefully.

With the expansive choices you seem to be minutely focusing on, it would appear you need expertise to focus in the bigger picture. Call for quotes from local electricians and home automation experts, or hire a consultant for advice. The small upfront expense will save you considerable grief and time down the track, and may avoid having to rewire everything when you finally realize that the solution you chose initially was based on random advice from random strangers off the internet. If you want professional advice, talk to professionals, and be prepared to generously recompense them for their wisdom and experience.

As far as Android display panels, consider future support and upgrades as critical for long term, and the dismal track record, based on phones, for Google to update them to retain HomeAssistant compatibility. You may find a cheap Raspberry Pi, connected to a vertical HDMI panel such as a large screen TV, or touch screen display may be an option you may wish to explore. At least you will have some upgrade path for software updates, and room for lots of futzing.

Your need for displays in a lot of rooms may also prove to be unnecessary as your mobile device, your phone, can control all your needs with custom dashboards, unless the w*nk factor to show off to guests is a big requirement.

3

To me, no. It only adds unnecessary complications. More points of failure.

Again, my opinion. No and no. Why do you need a panel in every room? My view of Home Assistant is that things that can be automated should be, but the control panel is for managing Home Assistant, not the primary function controls. (OTOH, I have seen some cool-looking Star Trek style dashboards). My experience with Android is that the processor is underpowered for anything but basic phone or tablet functions. Side-loading apps is maybe it will work and maybe it will brick.

Look at wall-mount touch-screen HDMI displays with a Raspberry Pi behind them. You could run in Kiosk mode or just run a browser that would also permit guests to browse the Internet. No, POE is not feasible.

You can never have too much Ethernet.

I don’t see much in these forums about redundancy. In fact, very little, and what I see is remarkably complicated to implement. But is it necessary? I run my Home Assistant, bare metal, on an Intel NUC i3 and I measure my uptime in years. A few years ago, I upgraded my original NUC to one with an SSD slot and the complete changeover took less than an hour. It’s not instant redundancy but not an ordeal, either.

IMHO- Matter is still premature and not ready for wide use. Matter guarantees only the baseline cluster set; anything outside that stays locked behind the vendor’s own bridge. For example, Philips Hue. A Hue Bridge is required for gradient control, scenes and firmware updates, Without the Hue Bridge you only get basic light control.

4

Touch screens for large displays with visiting toddlers, and avid tinkerers suffering from too much alcohol, using sharp objects and greasy fingers and both having vomit issues - whatever could go wrong?

You know HomeAssistant does not have robust security that can lock down things to a read-only format.

How long do you think people would just look without touching?

What could they do, either innocently, or deliberately, to really muck up your carefully crafted yaml code and settings? Expose passwords? Peer in other rooms where you have sensors and cameras.

Yeah/naah…

Behind solid perspex, not touch screen, display only, with a presence sensor to activate the display, and all the control behind a dashboard on your phone would be my advice.

5

Hi @IOT7712,
thanks again for your very detailed reply – I really appreciate the time and experience you put into it.

My original idea is actually very close to what you describe as a “serious” hard-wired backbone: I’m planning proper zone panels on each floor where all lighting and shutter circuits converge, and in those panels I want to use almost exclusively Ethernet-based Shelly devices (DIN-rail) as the main switching layer. Wi-Fi-only Shellys would be reserved for a few very specific edge cases where cabling is not practical, and I’d keep Zigbee only for bulbs, LED strips or truly secondary devices, not as part of the core control backbone.

One non-negotiable design goal for me is that every light and every shutter remains wired to physical wall switches in the rooms, with the Shellys acting as “normal” relays behind them. This way, even if Home Assistant, the network, or any software layer goes down or becomes buggy, there is always a fallback to basic on/off control directly from the wall switches, without depending on any automation or cloud service.

On the implementation side I’m in a fairly good position: I’m the CEO of two companies in Italy, and one of them does electrical installations and security systems, so I have full access to experienced electricians and low-voltage technicians. I’m also using this house project as a sort of lab to experiment and “play” with home automation, which so far we have only touched marginally in our client projects.

What was attracting me to Android tablets as wall panels are the integrated sensors, especially the camera and the microphone:

  • The camera would be used for presence detection – e.g. wake the dashboard only when someone is standing in front of the panel, not just generic motion in the room.
  • The microphone is important because I’m hoping to leverage Home Assistant’s native voice control directly from these panels.

With “pure” touchscreen monitors (industrial or commercial PoE displays without a full Android stack and without onboard sensors), I find it much harder to solve presence detection in front of the panel and to get a clean, always-on audio + mic path suitable for voice commands. Also worth noting: HA via Amazon/companion integrations doesn’t expose everything and does not fully replace HA’s own voice pipeline, especially for custom intents and deeper local control.

Given all of that – hard-wired Shelly Ethernet backbone, physical switches everywhere as a fail-safe, Zigbee only for secondary loads, and a strong need for camera + mic on the wall panels for presence detection and native HA voice – would you be willing to revisit or refine your recommendations for the wall and floor-level interfaces?

Thanks a lot once more for sharing your insights! :pray:

6

Hi @stevemann, thanks for your honest and direct take – I appreciate the no-nonsense approach.

Let me go through your points one by one.

On the layered architecture adding unnecessary complexity

I understand the concern, and in general I agree: more layers = more failure points. That said, my setup is designed so that the physical layer (wired switches → Shelly relays) always works standalone, regardless of HA, the network, or any software. HA is the intelligence layer on top, not the thing keeping the lights on. So in practice the “layers” are additive, not dependencies in series.

On panels in every room

Fair point that full automation should reduce the need for manual interaction. However, in a ~1000 sqm villa with guests and family, having a local control point per area is also a UX decision, not just a technical one. People should be able to control the room they’re in without pulling out a phone. The “Star Trek” factor is real, I won’t deny it, but it’s also about convenience and independence from personal devices.

On Android being underpowered

Agreed for older or cheap hardware. That’s exactly why I’m evaluating proper Android wall panels with modern SoCs (Android 12-14, not locked firmware), not repurposed phones. The Shelly Wall Display XL is a different beast from a sideloaded tablet.

On Raspberry Pi + HDMI touchscreen

This is actually the direction I’m leaning for the larger floor-level panels (21-27"). It gives a proper upgrade path (swap the Pi, keep the screen), runs a real OS, and I can manage it centrally. Your point about PoE not being feasible is noted – power + HDMI over separate runs is totally manageable since I have full access to the walls during renovation.

On redundancy

Years of uptime on a NUC is solid. On my side, HA will run on a Proxmox cluster, so I’ll have proper VM-level redundancy and live migration if a node needs maintenance. That said, the redundancy I care about most is at the physical layer: if HA goes down for any reason, wall switches still work, Shelly devices still respond via their own local API, and the house keeps functioning. Zigbee is only used for truly secondary devices (decorative lights, LED strips) that can simply stay off in a failure scenario – no critical function depends on it.

On Matter

Fully agree with you and @IOT7712 on this. Matter is not in scope for this project. I’ll avoid it until it matures significantly.

One thing I’d like your input on, given all of the above

The main reason I was considering Android panels (tablets or dedicated wall units) over pure HDMI + RPi setups is the integrated sensors: specifically the front-facing camera (for presence detection to wake the screen only when someone is actually standing in front of it) and the microphone (for HA native voice control directly from the panel, without relying on Amazon or Google assistant intermediaries).

With a standard HDMI touchscreen + RPi combo, solving those two things requires external hardware (a USB mic, a separate camera module, a PIR or mmWave sensor). Doable, but it adds complexity right back in.

Do you have a preferred way to handle presence + voice on a RPi-based panel setup? That would really help me decide between the two approaches. Thanks again!

7

I don’t, but I know it’s been done. Yes, my only exposure to Android tablets was limited to old models, but I haven’t felt the need to revisit it. I run most of my interaction in my house with Alexa in every room. My interface with Alexa is in Node Red because I have much more control and versatility.

I am keenly interested in your progress, so keep us posted. Also, you can never have too much Ethernet.