Home Insurance Inventory

Turns the photos, receipts and invoices already on your NAS into an insurance-ready inventory of your valuables, with photo evidence and serial numbers.

Comment

Inspiration

After a burglary, a fire or a flood, insurers ask for a list of everything you owned and proof that you owned it. Most of us already have that proof, scattered across years of phone photos, receipt screenshots and PDF invoices, but nobody can dig it out on the worst day of their year. Claims get reduced or delayed simply because the evidence is buried.

What it does

Home Insurance Inventory is a UGREEN AI NAS Skill that builds that list ahead of time, privately, on your own NAS:

  1. Finds photos that show valuable items (electronics, appliances, furniture, bikes, instruments, watches, jewellery, art) with AI object classification.
  2. Describes each item with the vision model and reads serial numbers or model numbers that are visible, using OCR.
  3. Reads receipt screenshots and PDF invoices and matches them to items by serial number, model number, or brand and product name, recording why each match was made.
  4. Shows a draft inventory table for the user to correct before anything is saved.
  5. Saves the confirmed report in the folder the user chooses and, optionally, tags the evidence photos with home-inventory.

An update mode compares a new run with the previous report and marks items as new, unchanged or missing.

How we built it

The Skill was created with the official UGOS Skill Creation Tool (uliya-skill-kit). It is an orchestration Skill that delegates to photo-agent, document-agent, knowledge-agent and file-agent. Every step receives only the real results of the previous step, all five delegation states are handled, empty results stop the flow honestly, and saving or tagging always waits for user confirmation and platform approval.

Challenges

Keeping it trustworthy: an insurance inventory is useless if it contains guesses. The Skill never invents serial numbers, prices or items. Anything unreadable is marked "not visible", items without a receipt are marked "no receipt found", and prices are copied exactly as written.

What's next

As a finalist, test detection and OCR quality on a real UGREEN AI NAS with a realistic photo library, and tune the item categories and matching rules from those results.

Built With

Share this project:

Updates

Submission history