Inspiration
Home archives grow into thousands of files nobody can search — photos, documents, and media that can only be found by opening files one by one.
What it does
ArchiveMind is a NAS-resident media/document intelligence Skill that turns a home archive into a searchable, self-describing library — auto-tagging, summarization, and natural-language search over photos, documents, and media, running locally on the NAS.
How it works
An ingest watcher queues new files in watched shares for analysis automatically. Auto-tagging covers images (scene/object content tags, face-free), documents (type, entities, dates), and media (duration/format/series metadata). Tags are written as sidecar metadata plus a searchable index. Summarization produces one-paragraph summaries for documents and long media (transcript-first where audio exists). Natural-language search — for example, "the insurance PDFs from last spring" — resolves over tags, summaries, and full-text, ranked, returning matching files as UGOS file references.
Privacy
All inference is on-device or on the owner's own hardware; nothing leaves the NAS without an explicit user action. This is the headline differentiator vs cloud photo services.
Status
Proposal and pre-test package stage. The package (SKILL.md manifest, pipeline definition, and a demo walkthrough over a sample archive) is in active development and being verified locally on November 11, 2026. It has not yet been validated on UGREEN NAS hardware; that validation is the finalist-phase step (remote NAS access November 22–30).
Built With
- ai
- auto-tagging
- nas
- ugos
- ugreen
