Azure Bill Extraction - Quick Start

The Azure extraction functionality is now available as a module in src/backend/.

Setup

1. Install Azure CLI and Login

The easiest way to authenticate with Azure is using Azure CLI:

# Install Azure CLI (macOS)
brew install azure-cli

# Login to Azure (opens browser for authentication)
az login

# Verify you're logged in
az account show

After logging in, Azure credentials will be automatically used by the Python SDK.

2. Environment Variables

Make sure you have your .env file with Azure credentials:

AZURE_DOC_KEY=your_azure_key_here
AZURE_EXISTING_AIPROJECT_ENDPOINT=your_endpoint_here
AZURE_EXISTING_AGENT_ID=your_agent_id_here
AZURE_SUBSCRIPTION_ID=your_subscription_id_here

Note: If you've run az login, you may not need all these variables depending on which Azure services you're using.

3. Install Python Requirements

pip install -r requirements.txt

Usage Options

Option 1: Run directly from command line

python src/backend/extractData.py front.jpg back.jpg

Option 2: Import as a module

from src.backend import get_ocr_lines, parse_front, parse_back, print_results

# Extract data
lines = get_ocr_lines("front.jpg")
front_data = parse_front(lines)

# Access fields
account = front_data.get("account_number")
total = front_data.get("invoice_total_eur")

Option 3: Use the example script

python example_usage.py

Authentication Methods

Method 1: Azure CLI (Recommended for local development)

az login

Then run your scripts normally - authentication is handled automatically.

Method 2: Environment Variables

Set these in your .env file for service principal authentication:

AZURE_TENANT_ID=your_tenant_id
AZURE_CLIENT_ID=your_client_id
AZURE_CLIENT_SECRET=your_client_secret

Method 3: API Key

Some Azure services support direct API key authentication via AZURE_DOC_KEY.

Available Functions

• get_ocr_lines(file_path) - Extract text lines from image using Azure
• parse_front(lines) - Parse front page data
• parse_back(lines) - Parse back page data
• print_results(front_data, back_data) - Pretty print extracted data
• parse_date(s) - Helper to parse dates
• after(lines, keyword, offset) - Helper to find text after keyword
• find_re(text, pattern) - Helper to find regex patterns

What Gets Extracted

Front Page

• Customer info (name, address, account)
• Invoice details (number, dates, total)
• Service period and consumption

Back Page

• Meter readings
• All charge breakdowns (supply, regulated, misc, VAT, municipality)

Troubleshooting

Authentication Errors

If you see DefaultAzureCredential failed to retrieve a token:

1. Run az login to authenticate
2. Verify with az account show
3. Make sure your account has permissions on the Azure resources

Permission Errors (403)

If you see permission errors, you need the appropriate role:

1. Go to Azure Portal
2. Navigate to your resource
3. Access Control (IAM) → Add role assignment
4. Assign "Azure ML Data Scientist" or "Contributor" role

Now all your other code is preserved, and you can easily import and use the extraction functions!