PayGuard

PayGuard is an intelligent SAP invoice analysis tool that uses SBERT and LLM models to identify potential duplicate invoices.

PayGuard Website: you can manage your JSON SAP Prediction File at our website https://payguard-duplicate-detect.lovable.app/

Prerequisites:

• Access to a Gemini API key

Clone the Repository (if you haven't already):

git clone https://github.com/FelixSBuehrm/payguardml.git
cd payguardml

1. Project Architecture

PayGuard follows a modular architecture with clear separation between ML backend and Electron frontend:

├── app/                         # Electron frontend application
│   ├── main.js                  # Main Electron process
│   ├── renderer.js              # Renderer process for UI
│   ├── index.html               # Main UI layout
│   └── *.css                    # Styling files
├── assets/                      # Images and icons
├── backend/                     # Python backend for ML processing
│   ├── main.py                  # Main processing pipeline
│   ├── llm_classifier.py        # Gemini LLM integration
│   ├── pair_predictor.py        # Invoice pair prediction
│   ├── progress_monitor.py      # UI progress bar
│   └── utils.py                 # Utility functions
├── build/                       # Build assets for packaging
├── content/                     # Core ML models and data
│   ├── invoice_sbert/           # Pre-trained SBERT model
│   └── predict_pairs.py         # SBERT prediction functionality
├── output/                      # Generated analysis outputs
├── scripts/                     # Build and packaging scripts
├── SBERT_V3_GPU.ipynb           # SBERT Notebook Pipeline
├── val_pairs.csv                # Test data
├── training.csv                 # Training data
├── README.md
└── package.json                 # Project configuration

2. Setup Backend:

• Create .env file: Navigate to the backend/ directory and create a file named .env. Add your Gemini API key to it:

  # payguardml/backend/.env
  GEMINI_API_KEY="YOUR_GEMINI_API_KEY_HERE"
  

• Install Python Dependencies: It's highly recommended to use a virtual environment.

  cd backend # if not already there
  python3 -m venv venv
  source venv/bin/activate  # On macOS/Linux
  # venv\Scripts\activate    # On Windows
  
  # Navigate to the content directory to install requirements
  cd ../content
  pip install -r requirements.txt
  cd ../backend # Go back to backend directory

3. Setup Frontend (Electron App):

• Install Node.js Dependencies: Navigate to the project root directory (payguardml/) if you aren't already there.

  npm install

4. Running the Application Locally:

• From the project root directory (payguardml/):

  npm start

  This will launch the PayGuard Electron application.

5. Building the Application Locally:

• Ensure all dependencies are installed (npm install).

• From the project root directory (payguardml/):

  • To build for your current platform (e.g., macOS):

    npm run dist

    Alternatively, for specific platforms:
  • For macOS (creates .dmg and .zip):

    npm run dist:mac-arm
    or 
    npm run dist:mac-intel

  • For Windows (creates NSIS installer and .zip): (Requires a Windows environment or appropriate cross-compilation setup)

    npm run dist:win
    # or
    npm run dist:win-x64

  • The distributable files will be located in the dist/ directory.