Inspiration
Home-office and small-team NAS owners dump scans, invoices, contracts and downloads into one folder. Finding anything later means opening files one by one; nothing is summarised, tagged or filed, so the NAS becomes write-only storage.
What it does
NAS Inbox Digest turns a chaotic NAS inbox folder into a filed, summarised and searchable document library, automatically. It watches one user-authorised inbox folder. For every new file it extracts text, classifies the document, writes a short summary with tags, files the original into a dated category folder, and maintains a single searchable digest index, plus a weekly digest note. The user drops files into the inbox folder and later asks, "What changed this week?" or "Find my electricity invoices." The Skill reports what arrived, where each file was filed, and answers from the digest index with links to the originals.
How it differs
Cloud AI organisers require uploading private documents off-device. Generic chatbots cannot act on the NAS file system. This Skill runs on the NAS itself, keeps every byte local, files documents rather than only describing them, and logs every action so any move can be undone.
How we built it
A Uliya-standard Skill orchestrates three expert agents: an Extractor (text and metadata), a Classifier-Summariser (category, summary, tags) and a Filer (moves, index updates, digest). Inputs are files in the authorised folder. Permissions are read and write only inside user-designated folders; no deletion, ever. The AI model is the NAS-provided on-device model. Data never leaves the NAS; the index stores summaries and paths, not document copies.
UGREEN AI NAS features used
- File capability: folder watching, listing and moving files (intake, filing).
- Document capability: text extraction from PDFs and office files (understanding).
- Knowledge capability: the digest index and semantic search (retrieval).
- System capability: scheduled runs and run logs (weekly digest, audit trail).
Built so far, and what needs NAS testing
Built so far: the complete Skill package to the Uliya standard — workflow, orchestration, permissions model, failure handling and test plan — designed and written during this hackathon. Not yet built: the live binding to UGOS CLI capabilities. Needs NAS testing: extraction quality on real scans, move reliability on large folders, and scheduled-run behaviour.
Test plan
Core test 1: fifty mixed documents land in the inbox; expect correct filing, summaries and index entries for at least ninety percent, and zero files lost. Core test 2: natural-language retrieval ("find the March invoices") returns the right originals. Edge case: a corrupt or unsupported file is quarantined to a Needs Review list with a clear message — never deleted, never silently skipped.
Challenges and limits
Handwriting and poor scans reduce extraction quality; very large folders need batched runs; classification can misfire, so filing is logged and reversible rather than destructive. Dependencies are UGOS file, document and knowledge capabilities and the NAS on-device AI model. No external services, accounts or network access are required.
Declarations
Solo entrant: Hassan Dosunmu, Lagos, Nigeria. He consents to the UGREEN Hackathon 2026 official rules and confirms this Skill and its core workflow were newly developed during the Hackathon. Third-party components: none bundled in the Round 1 Skill package; specified against UGREEN's UGOS capabilities and the Uliya Skill Development Standard using the UGREEN Skill Developer Package as documentation; no third-party licences apply. AI tools used: an AI assistant was used to help draft and structure this proposal and the Skill documentation; all design decisions were reviewed by the entrant.
The full Round 1 proposal (PDF) and the Skill package (ZIP) are attached in the Additional info step.
Built With
- ugos
- uliya
