What it does
memory-curator is a UGOS Skill that turns the photos, videos, and documents stored on a UGREEN AI NAS into a searchable memory collection. Ask in natural language — for example, "find last summer's beach photos" — and it retrieves matching items, groups them, and builds albums.
The problem
Photos and documents keep piling up on the NAS, but finding a specific memory means scrolling through folders and filenames. Search does not understand what is in the photos or documents.
How it works
memory-curator uses UGOS RAG over captions and document text: it indexes what photos, videos, and documents contain, retrieves the most relevant items for a natural-language query, groups related items, and builds albums from them. Private items such as receipts are excluded from shared results by privacy rules, so a shared album does not leak sensitive documents.
How we built it
Built as a Skill following the Uliya Skill Standard, with UGOS, RAG, and Markdown. The Round 1 package was developed and verified in mock NAS mode, without requiring a physical NAS, so the retrieval, grouping, and album-building flow can be checked locally before testing on real hardware.
What's next
Validate the Skill on a real UGREEN AI NAS in Round 2, including retrieval accuracy, album quality, and privacy exclusions with real NAS data.
Built With
- markdown
- rag
- skill
- standard
- ugos
- uliya
