README_HF.md

metadata

title: AP Agent Demo
emoji: 📄
colorFrom: blue
colorTo: purple
sdk: docker
pinned: false
license: mit

AP Agent Demo - Invoice ↔ PO Matching

AI-powered Accounts Payable processing application that automatically matches invoices to purchase orders, detects backorders, validates suppliers, and identifies duplicates.

Features

• 🤖 AI-Powered Matching: Uses Azure OpenAI to dynamically match invoice lines to PO lines, handling any column names
• 📊 Backorder Tracking: Automatically calculates remaining quantities per PO line
• ✅ Supplier Validation: Validates suppliers against approved supplier lists
• 🔍 Duplicate Detection: Detects duplicate invoices by file hash and business key
• 📈 History & Analytics: Tracks processing history with search and filter capabilities
• 🎨 Modern UI: Beautiful React frontend with real-time processing feedback

How to Use

1. Upload Invoice: Upload a PDF, DOCX, or TXT invoice file
2. Upload PO File: Upload a CSV or Excel file containing purchase order data
3. Review Results: The AI agent will:
   • Extract invoice fields and line items
   • Match invoice lines to PO lines dynamically
   • Calculate backorders
   • Validate supplier
   • Detect duplicates
   • Provide approval decision (Approved/Partially Approved/Rejected)

Environment Variables

Set these in your Hugging Face Space settings:

• AZURE_OPENAI_ENDPOINT: Your Azure OpenAI endpoint
• AZURE_OPENAI_API_KEY: Your Azure OpenAI API key
• AZURE_OPENAI_DEPLOYMENT: Your Azure OpenAI deployment name
• AZURE_OPENAI_API_VERSION: (Optional) API version, defaults to 2024-10-21

Optional (for real OCR on scanned invoices):

• AZURE_DOCUMENT_INTELLIGENCE_ENDPOINT: Azure Document Intelligence endpoint
• AZURE_DOCUMENT_INTELLIGENCE_KEY: Azure Document Intelligence key

Technical Details

• Backend: FastAPI (Python)
• Frontend: React + TypeScript + Tailwind CSS
• AI: Azure OpenAI (GPT-4)
• Database: SQLite (persistent storage)
• Deployment: Docker on Hugging Face Spaces

License

MIT