Family Memory Curator turns scattered photo and document dumps on a UGREEN AI NAS into dated, captioned albums and a searchable family timeline, entirely on-device.
The target user is a family archivist, often a parent or grandparent, whose photos sit in mixed phone dumps, scans, and downloads spread across folders. Finding one memory means scrolling thousands of files, and cloud organizing tools require uploading private family media.
The Skill inventories authorized folders, groups media by time, location, and recognized people, flags duplicates, and presents a curation plan for approval. After confirmation it creates albums, writes short captions and a timeline summary using the local model, and saves everything back to the NAS. Nothing is deleted; duplicates move only after explicit confirmation.
Built-in album tools organize an already tidy photo library, and cloud AI requires uploads. This Skill starts from messy mixed dumps that include documents, produces an auditable plan before changing anything, and keeps every step local and reversible.
Round 1 status, stated honestly as in the proposal: the complete Skill package (SKILL.md workflow, capability mapping, confirmation rules, failure handling) is built and checked against supported capabilities only. Live execution against UGOS CLI capabilities and on-device model calls is not yet built; it is the finalist-stage work.
AI disclosure (from the proposal): this proposal and Skill package were drafted with AI assistance (Muse, by Anthropic), as the Hackathon expressly contemplates. No AI-generated media, code executables, or binaries are included in the package.
Built With
- ai
- idx6011
- nas
- pro
- skill.md
- ugos
- ugreen
- uliya
Log in or sign up for Devpost to join the conversation.