A small software update recently stopped camera streams displaying in Apple Home for a number of our customers.
The cameras were still working. They were still recording locally in UniFi Protect. No camera hardware had failed.
But open the Apple Home app, and instead of your driveway or front door, you were greeted with “No Response”.
For the customer, that distinction only goes so far. Something that worked yesterday doesn't work today.
What happened next is a practical example of why we build our automation platforms on open-source software and Linux. And why, in a world of AI Agents, that choice matters more than ever.
The Smart Home Update Problem
Anyone who implements integrated smart homes knows this battle.
Disable updates and you keep the software that works today. You also miss security fixes, bug fixes and improvements, while everything around it continues to change. Eventually, a phone update, a new device or a change to another system can leave your carefully frozen installation behind.
Allow updates and you receive those fixes and improvements. You also accept that one of those updates might break something that was working perfectly.
Update and risk breaking something. Don't update and risk breaking something later.
Testing and managed updates help. Neither removes the problem entirely.
This isn't unique to Apple Home, open-source software, or any particular smart home platform. It's a reality of integrating products and software that evolve independently.
The important question isn't whether your system will ever encounter a software problem.
It's what can be done when it does.
A Small Update, a Broken Connection
In this case, the problem was an update to the UniFi Protect plugin in Scrypted — the software we use to make Protect cameras available in Apple Home.
The update changed how the plugin identified cameras internally. On affected systems, Scrypted treated existing cameras as new devices, while their Apple Home pairings remained attached to the old device records.
Same physical cameras. Same local recording system. A broken connection between the software presenting the cameras and the app displaying them.
The solution was to reconnect the existing pairing information to the correct camera records, preserving the Apple Home relationship instead of asking customers to remove and re-add their cameras.
Straightforward to describe. More involved to diagnose and repair across different customer installations.
From Support Calls to a Fleet-Wide Response
When customers started reporting that their cameras weren't responding in Apple Home, we used our AI Agent to check across our fleet of customer systems.
It connected to systems one at a time and checked whether the specific problem was present.
Where the fault matched the repair conditions, it backed up the relevant data, applied a code-based repair and ran checks to confirm that camera streaming and pairing functionality had been restored. Systems with different conditions were flagged for separate review, rather than blindly applying the same change everywhere.
Then it moved on to the next system.
We weren't limited to fixing the homes whose owners had already called. We could identify the same fault elsewhere and address it without waiting for another support ticket.
The process also brought together the list of repaired customer systems and sent those customers an email through our CRM:
Hi xxxx,
A recent software update affected camera streams displaying in Apple Home on several customer systems. We identified that your system was likely affected, and you may have noticed your cameras showing “No Response” over the last few days.
Your cameras continued working and recording locally in UniFi Protect throughout.
We've remotely fixed the Apple Home issue, and your camera streams should now be displaying correctly again.
If this isn't the case, or you have any questions, please lodge a support ticket.
Identify the problem. Check who else is affected. Repair it. Verify the result. Let the customer know.
That's a much better support experience than waiting for every customer to discover and report the same fault.
AI Is Only Half the Picture
An AI Agent, connected to a capable model from a leading AI provider, can coordinate this kind of work from a single request.
But a prompt isn't the same thing as access.
The agent can only investigate and repair what the underlying platform allows it to reach.
Our open-source automation software runs in Linux environments. That gives us access to the source code, logs, configuration, stored data and standard tools needed to understand what has happened and implement a repair.
Open source makes the software inspectable and changeable. Linux provides an environment where diagnostic and repair tools can be run, scripted and repeated.
With authorised remote access and the right tools, an agent can work with those systems directly.
It isn't magic. It's capable software working on an accessible platform.
And it doesn't mean the home suddenly needs an AI subscription to operate. The agent is a maintenance tool, not a requirement for turning on the lights or recording your cameras.
You Can't Prompt Your Way Through a Closed Door
Proprietary home control platforms such as Control4, Crestron and Savant operate within boundaries set by their manufacturers.
They may offer APIs, dealer tools and integrations. They may also introduce their own AI capabilities.
But access to an API isn't the same as access to the software underneath it.
You cannot assume that your preferred AI Agent will be allowed to inspect internal data, trace a fault through the platform's source code and apply a repair. You remain dependent on what the manufacturer exposes and permits.
A smarter AI model doesn't remove that restriction.
Open-source platforms give us a fundamentally different starting point. We can investigate the system itself, not just the controls a vendor has chosen to provide.
Open Systems Have a Head Start
AI Agents aren't making open-source software immune to bugs. This incident proves that clearly enough.
What they are changing is how quickly accessible software can be understood, maintained and improved.
Developers and integrators can use agents to investigate faults, write fixes, test changes and contribute improvements. Those improvements can then benefit everyone using the project.
The same openness that lets us repair an installation also lets the wider community improve the platform.
For the homeowner, the benefit isn't Linux or source code. It's a system that can be supported more effectively, with fewer barriers between finding a problem and fixing it.
We've always believed the equipment in your home should be yours.
In a world of AI Agents, being able to understand and repair the software running it is becoming just as important.
The best user experience isn't just how your home works when everything is fine. It's how effectively we can put things right when it isn't.