hf_upload.py

"""
Hugging Face Upload Setup and Helper
This script helps you push your Singtel Bill Scanner to Hugging Face
"""

import os
import subprocess
from huggingface_hub import HfApi

def check_huggingface_setup():
    """Check if Hugging Face is properly set up"""
    print("🔍 Checking Hugging Face setup...")
    
    # Check if huggingface-hub is installed
    try:
        import huggingface_hub
        print(f"✅ huggingface-hub installed (version: {huggingface_hub.__version__})")
    except ImportError:
        print("❌ huggingface-hub not installed")
        return False
    
    # Check for token
    token = os.getenv("HF_TOKEN")
    if token:
        print("✅ HF_TOKEN environment variable found")
        # Test token validity
        try:
            api = HfApi(token=token)
            user_info = api.whoami()
            print(f"✅ Token valid for user: {user_info['name']}")
            return True
        except Exception as e:
            print(f"❌ Token invalid: {e}")
            return False
    else:
        print("❌ HF_TOKEN environment variable not set")
        return False

def setup_huggingface_token():
    """Guide user through setting up HF token"""
    print("\n🔑 HUGGING FACE TOKEN SETUP")
    print("=" * 40)
    print("1. Go to: https://huggingface.co/settings/tokens")
    print("2. Create a new token (or copy existing one)")
    print("3. Choose 'Write' permissions")
    print("4. Copy the token")
    print()
    
    token = input("Paste your Hugging Face token here: ").strip()
    
    if token:
        # Set environment variable for this session
        os.environ["HF_TOKEN"] = token
        
        # Try to set it permanently in PowerShell
        try:
            print("\n🔧 Setting token in environment...")
            cmd = f'[Environment]::SetEnvironmentVariable("HF_TOKEN", "{token}", "User")'
            subprocess.run(["powershell", "-Command", cmd], check=True)
            print("✅ Token saved to user environment variables")
            print("💡 You may need to restart VS Code to see the change")
        except Exception as e:
            print(f"⚠️  Could not save permanently: {e}")
            print("💡 Token is set for this session only")
        
        # Test the token
        try:
            api = HfApi(token=token)
            user_info = api.whoami()
            print(f"✅ Token works! Logged in as: {user_info['name']}")
            return True
        except Exception as e:
            print(f"❌ Token test failed: {e}")
            return False
    else:
        print("❌ No token provided")
        return False

def create_model_card():
    """Create a README.md for Hugging Face"""
    readme_content = """---
title: Singtel Bill Scanner
emoji: 📱
colorFrom: red
colorTo: orange
sdk: streamlit
sdk_version: 1.28.0
app_file: app.py
pinned: false
tags:
- computer-vision
- ocr
- trocr
- bill-processing
- singtel
- document-ai
---

# Singtel Bill Scanner 📱💡

An AI-powered optical character recognition (OCR) system specifically designed for processing Singtel telecommunications bills. This project uses Microsoft's TrOCR (Transformer-based OCR) model to extract text from bill images and parse key information.

## Features

- 🔍 **Text Extraction**: Uses TrOCR for accurate text recognition from handwritten and printed text
- 📊 **Bill Parsing**: Automatically extracts key information like:
  - Total amount due
  - Due date
  - Account number
  - Service charges
  - Billing period
- 🚀 **Easy to Use**: Simple pipeline interface
- ⚡ **Fast Processing**: Cached models for instant subsequent runs
- 🎯 **Singtel Specific**: Optimized patterns for Singtel bill formats

## Models Used

- **TrOCR**: `microsoft/trocr-base-handwritten` - For text extraction
- **LayoutLMv3**: `microsoft/layoutlmv3-base` - For document structure understanding

## Quick Start

```python
from transformers import pipeline
from PIL import Image

# Initialize the OCR pipeline
pipe = pipeline("image-to-text", model="microsoft/trocr-base-handwritten")

# Process your bill image
image = Image.open("your_singtel_bill.jpg")
result = pipe(image)
extracted_text = result[0]['generated_text']

print(f"Extracted text: {extracted_text}")
```

## Advanced Usage

Use the `SingtelBillScanner` class for comprehensive bill processing:

```python
from singtel_scanner import SingtelBillScanner

# Initialize scanner
scanner = SingtelBillScanner()

# Process bill and get structured data
result = scanner.process_bill("bill_image.jpg")

print(f"Total Amount: ${result['total_amount']}")
print(f"Due Date: {result['due_date']}")
print(f"Account: {result['account_number']}")
```

## Installation

```bash
pip install torch transformers Pillow requests huggingface-hub
```

## Files

- `singtel_scanner.py` - Main scanner class with bill parsing
- `test_model.py` - Comprehensive testing and examples
- `quick_test.py` - Simple test script
- `working_example.py` - Basic functionality demonstration
- `requirements.txt` - Package dependencies

## Performance

- **Model Size**: ~1.3GB (downloaded once, cached forever)
- **Processing Time**: ~2-5 seconds per image (after initial load)
- **Accuracy**: High accuracy for clear, well-lit bill images

## Use Cases

- 📱 **Personal Finance**: Track Singtel bills automatically
- 🏢 **Business Automation**: Process multiple bills in batch
- 📊 **Expense Management**: Extract data for accounting systems
- 🔍 **Document Digitization**: Convert physical bills to digital records

## Requirements

- Python 3.8+
- ~2GB free disk space (for models)
- Good internet connection (for initial model download)

## Contributing

Contributions welcome! Areas for improvement:
- Additional bill format support
- Enhanced parsing accuracy
- Mobile app integration
- Batch processing optimization

## License

This project is open source. Models are subject to their respective licenses:
- TrOCR: MIT License
- LayoutLMv3: MIT License

---

*Created with ❤️ for the Singtel community*
"""
    
    with open("README.md", "w", encoding="utf-8") as f:
        f.write(readme_content)
    
    print("✅ Model card (README.md) created successfully!")

def upload_to_huggingface():
    """Upload the project to Hugging Face"""
    print("\n🚀 UPLOADING TO HUGGING FACE")
    print("=" * 40)
    
    try:
        token = os.getenv("HF_TOKEN")
        api = HfApi(token=token)
        
        print("📁 Preparing files for upload...")
        
        # Create model card if it doesn't exist
        if not os.path.exists("README.md"):
            create_model_card()
        
        print("📤 Starting upload...")
        api.upload_folder(
            folder_path=".",
            repo_id="Cosmo125/Singtel_Bill_Scanner",
            repo_type="space",  # Changed to 'space' for better visibility
            ignore_patterns=[
                "*.pyc", 
                "__pycache__/", 
                ".venv/", 
                "*.jpg", 
                "*.png", 
                "*.jpeg",
                ".git/",
                "test_*.png",
                "sample_*.jpg"
            ],
            commit_message="Upload Singtel Bill Scanner - AI OCR for bill processing"
        )
        
        print("\n🎉 SUCCESS! Upload completed!")
        print("🔗 Your project is available at:")
        print("   https://huggingface.co/spaces/Cosmo125/Singtel_Bill_Scanner")
        print()
        print("💡 It may take a few minutes to build and become available")
        return True
        
    except Exception as e:
        print(f"❌ Upload failed: {e}")
        print("\n🔧 Troubleshooting:")
        print("1. Check your internet connection")
        print("2. Verify your HF token has write permissions")
        print("3. Make sure the repository name is available")
        return False

def main():
    """Main function to handle the upload process"""
    print("SINGTEL BILL SCANNER - HUGGING FACE UPLOAD")
    print("=" * 50)
    
    # Check setup
    if check_huggingface_setup():
        print("\n✅ Setup looks good!")
        
        choice = input("\nDo you want to upload to Hugging Face now? (y/n): ").lower()
        if choice in ['y', 'yes']:
            upload_to_huggingface()
        else:
            print("Upload cancelled.")
    else:
        print("\n🔧 Setup needed!")
        choice = input("Do you want to set up your HF token now? (y/n): ").lower()
        if choice in ['y', 'yes']:
            if setup_huggingface_token():
                print("\n✅ Token setup complete!")
                upload_choice = input("Upload to Hugging Face now? (y/n): ").lower()
                if upload_choice in ['y', 'yes']:
                    upload_to_huggingface()
            else:
                print("❌ Token setup failed")
        else:
            print("Setup cancelled.")

if __name__ == "__main__":
    main()