A home NAS holds a lifetime: some 216,000 photos, documents, videos, songs and clips. But nothing connects a subject to its records, so finding everything about one daughter, one house or one tax year means hunting file names kind by kind — and family material must never be handed to a cloud assistant.
µTAG Life Archive gives every subject — a person, a pet, the house, the car, a trip, a year — one symbol. Items live in per-kind bundles, each served by its registered domain expert (documents, photos, media, surveillance, files). Across those bundles, every item joins its subjects' chains at the moment it is filed. Pull one symbol and the whole story follows, across kinds, each item re-verified against its stored fingerprint. Names stay in a protected table; chains carry only addresses, kinds, dates and fingerprints. Nothing leaves the NAS.
The package includes a platform layer for the UGREEN team: storage placement, per-step expert contracts, invariants, per-item index cost, a back-filing migration path, and a proposal to adopt the subject key as a shared reference other Skills can use.
Status, stated plainly: designed, packaged and measured off-device (household-scale rehearsal: 216,000 items, 400 subjects, 241,741 chain links; largest subject retrieved fully verified in about 15 ms; a full verification sweep caught both planted exceptions). On-device execution has not happened yet; the first finalist-day checks are named in the test plan, and anything the device cannot do will be reported as blocked, never simulated.
This project was conceived, designed, directed and approved by the entrant, Dukha Kim. AI assistants were used under his direction for drafting, checking and simulation judging: Muse (Meta), a local Qwen3-14B (Alibaba, via llama.cpp), and Groq-hosted models (gpt-oss-120b, gpt-oss-20b, Qwen3.8-27B). The full disclosure list, including third-party components and design references, is in the Skill package (references/disclosures.md). The submitted Skill itself embeds no AI; it orchestrates only the platform's registered experts.
Built With
- kit
- skill
- ugreen
- uliya
