Cross-Doc

Cross Doc

1. What does your Skill do? Cross-Doc automatically extracts metadata from newly uploaded financial or operational documents on the NAS and logically cross-references them against existing spreadsheets to flag operational anomalies.
2. Who is your target user, and what is their problem? Operations managers and small business owners who waste hours manually verifying that unstructured incoming files (like raw PDF invoices) exactly match structured database files (like CSV purchase logs).
3. What is your solution? A multi-agent background workflow that chains text-extraction and tabular-logic to validate files across completely different formats without user intervention.
4. What is the main user flow? The user drops an invoice PDF into a monitored folder; the Skill automatically reads it, queries a Master Ledger CSV elsewhere on the NAS, and outputs a localized Markdown discrepancy report.
5. How is your idea different from existing tools or common AI solutions? Unlike cloud-based tools, it processes confidential operational files entirely locally without data leaks, and unlike standard LLM search, it specifically cross-references data across different file formats.
6. Which UGREEN AI NAS features does your Skill use, and at which step? Uses the UGOS File Notification API to trigger on folder uploads, reads target files using native NAS file paths, and utilizes the integrated local LLM/OCR compute engine for processing.
7. What is your technical approach? An event-driven script monitors a designated directory. Upon file entry, an OCR specialist agent extracts text, passing structured JSON payload to a Data-Matching agent. This agent reads local CSV files via standard CLI file tools, performs a deterministic Python logic check, writes local summaries via markdown, and requires standard Read/Write storage permissions.
8. What have you built so far? What is not built yet? What still needs testing on a NAS? The localized text extraction logic and CSV parsing python scripts are built. The UGOS-specific directory monitoring trigger and the native agent orchestration configuration are not built. Remote NAS testing is needed to evaluate real-time file trigger lag.
9. How will you test your Skill if you become a Finalist? Test Case 1: Upload a standard, matching PDF invoice to confirm instant verification. Test Case 2: Upload an invoice with a mismatched total amount to verify anomaly flagging. Edge Case: Upload a highly corrupted, unreadable scanned image PDF to test structural graceful degradation.
10. What are the known limits, risks, or outside dependencies of your Skill? It depends strictly on consistent layout structures for data matching; highly erratic invoice formats may require manual mapping adjustments.
    • Third-Party Components: pandas (BSD-3-Clause)/polars for tabular data processing.

Built With

• nas