Inspiration
You remember the moment, but you cannot find the photo.
Maybe it was a camping trip, a birthday, or a family picnic. You know the picture is somewhere on your NAS, yet every search leaves you scrolling through the wrong results.
PhotoTrace grew from a simple question: when photo search comes up empty, what should happen next?
We wanted to turn that frustrating dead end into a guided investigation—with useful questions, possible matches, and a clear explanation of what was checked.
What it does
PhotoTrace is a photo-finding assistant built as a Uliya Skill for UGREEN AI NAS.
Choose a folder and describe what you remember. PhotoTrace starts with Photos search, then checks a limited set of accessible files when needed. If there are too many candidates, it asks for a helpful clue, such as an approximate date, a filename fragment, or a narrower folder.
You confirm which image is the one you wanted. PhotoTrace never treats a visual guess as a confirmed match.
Every investigation ends with four clear answers:
- Found: Your confirmed photo or possible matches.
- Checked: The searches and files actually examined.
- Still unknown: What could not be established.
- Next step: One useful action to take.
PhotoTrace is designed to leave your library untouched. It does not move or delete photos, change permissions, or modify indexing settings.
How we built it
We built PhotoTrace using the official UGREEN Skill Developer Package. The six-file Markdown Skill coordinates three registered Uliya experts:
- File-agent checks the selected folder and finds accessible files.
- Photo-agent performs supported photo searches, reads metadata, and inspects candidate images.
- Knowledge-agent retrieves official troubleshooting guidance when it can help.
The workflow keeps track of the chosen folder, actual file references, search limits, and your confirmations throughout the conversation. Private-content investigation pauses if the required local processing route cannot be established.
The package passes the official static checker with zero errors and warnings. Testing on a UGREEN NAS is the next development milestone.
Challenges we ran into
The biggest challenge was making the results useful without overstating what they mean.
A photo missing from one search does not mean it was deleted. A file found in a folder does not prove that indexing is broken. Two files with the same name are not necessarily the same image.
We built these distinctions into the workflow. PhotoTrace tracks files discovered separately from images inspected, preserves uncertainty when metadata is missing, and waits for your confirmation before declaring success.
We also had to make limited inspection useful. Instead of silently scanning more images, PhotoTrace asks a focused question that can help narrow the search.
Accomplishments that we're proud of
We turned PhotoTrace into a focused Skill package with a complete investigation flow, clear stopping conditions, and an understandable result summary.
We prepared 22 acceptance scenarios covering successful searches, duplicate filenames, missing metadata, permission failures, cancellation, and unexpected content. All eight automated packaging tests pass.
The result is a product with a clear purpose: help someone find a remembered photo and understand what to do when the search is still incomplete.
What we learned
A helpful search assistant needs to do more than return matches. It needs to explain the limits of its search and help the user make the next decision.
We also learned that one useful question can be more valuable than a longer list of results. The person looking for the photo often remembers something that can make the search much more focused.
What's next for PhotoTrace
Next, we will validate PhotoTrace on a UGREEN AI NAS, including folder boundaries, native search behavior, image inspection, and local processing.
We will compare it with manual Photos/Files searches and a general Uliya prompt using known target images. We will measure confirmed matches, time taken, and whether the next step helps the user move forward.
Those results will guide improvements to candidate selection and the conversation—bringing PhotoTrace closer to its goal: less searching, more finding.
