Inspiration
We've all had that moment when we can't find something important. Now imagine your TV breaks, your laptop gets stolen, or your home gets damaged by a flood. Suddenly, you need to prove that you actually owned those things.
Most of us have hundreds of photos, old receipts and invoices sitting somewhere on our NAS. But finding the right one when you really need it can be a nightmare.
We started thinking: our NAS already stores so many memories and important files, so why can't it help us remember what we own too?
That's where i-paid-for-that came from. We wanted to build something that could save people the stress of searching through years of files at the worst possible time.
What it does
i-paid-for-that turns the photos, receipts and invoices you already have into a searchable home inventory and an insurance evidence report.
You can go room by room, let the Skill suggest items from your photos, and approve what belongs in your inventory. Later, if you forget where you put something, you can simply search for it.
And if something gets damaged or stolen, that's where the real help begins. The Skill looks for the item's inventory record, receipts or invoices, and older photos that might show its condition before the incident. It then puts everything together in an organized insurance evidence report.
The best part? It doesn't make things up just to fill in the gaps. If a receipt is missing or a document can't be read, it tells you.
How we built it
We built the project using four specialized UGREEN NAS expert agents, each with its own job:
- Photo Agent: Looks through room photos, identifies household items, finds older photos and reads photographed receipts.
- File Agent: Searches for receipts and invoices and saves approved reports.
- Document Agent: Handles PDF receipts and invoices.
- Knowledge Agent: Stores and retrieves the inventory items that users have approved.
We connected everything through a central SKILL.md file that coordinates the agents and manages the workflow using UGOS CLI capabilities.
We also spent time thinking about what could go wrong. We didn't want the Skill saving items without permission or treating a possible match as a confirmed one. So users approve suggested belongings, choose between ambiguous matches, and confirm before anything is saved. Searches are read-only, and saving also requires platform approval.
Challenges we ran into
One thing we quickly realized is that finding a file and having actual proof are two very different things.
For example, finding a picture of a TV doesn't necessarily prove that you owned it. A receipt might belong to a different model, and a photo taken on the day something was damaged doesn't necessarily show what its condition was before.
Getting the Skill to handle these situations honestly was one of our biggest challenges.
We also had to think carefully about how four different agents would work together without stepping outside their responsibilities.
Instead of trying to make the AI guess when information was missing, we designed it to be honest about what it could and couldn't find.
Accomplishments that we're proud of
We're proud that we took something most people probably don't think about until it's too late and turned it into a practical idea.
What makes us especially happy is that this isn't just another tool that searches through photographs. It brings photos, receipts, invoices and household inventory together to make finding evidence easier.
We're also proud of the safety measures we included, especially user approval, careful handling of uncertain matches and honest reporting when evidence is missing.
Our initial Skill package has passed the Uliya static check with zero errors and zero warnings. We haven't tested it on a physical UGREEN NAS yet, but the initial implementation is ready for that next step.
What we learned
We learned that AI doesn't always need to create something new to be useful. Sometimes the most helpful thing it can do is make sense of information we already have.
We also learned that when AI handles personal files and something as important as insurance evidence, being honest matters more than sounding confident. It's much better to say "I couldn't find it" than to give someone information that might not be true.
Building this also gave us a better understanding of how specialized agents can work together and why permissions and clear responsibilities matter.
What's next for i-paid-for-that
There's still more we'd love to do with this project.
Our next step is to test everything on a real UGREEN AI NAS, from creating a room inventory to finding an item and putting together a complete evidence report.
We also want to test less-than-perfect situations, like blurry photos, unreadable receipts, missing documents and items that have several possible matches.
Ultimately, we want to make keeping track of your belongings something you don't have to worry about. Because when something goes wrong, the last thing you should have to do is spend hours digging through old folders trying to prove what was yours.
You paid for it. Your NAS should remember it.
