Azure Computer Vision Setup

Use Azure Computer Vision OCR (Read API) for receipt extraction; mock mode stays available for tests.

Option 1: Mock mode (default)

No setup needed. Uses fake receipt data.

Option 2: Real Azure Computer Vision

1. Create resource in Azure Portal → Computer Vision (choose region, name, pricing tier).
2. Get credentials: endpoint + key.
3. Configure env:

AZURE_CV_ENDPOINT=https://your-resource.cognitiveservices.azure.com/
AZURE_CV_KEY=your-key
OCR_MODE=azure

1. Install deps: pip install -e . (from repo root).
2. Test: python run_demo.py (or your receipt workflow).

Pricing (Read API) - Free F0: 5,000 tx/mo, 20/min. - S1: ~$1 per 1k (first 1M), cheaper at higher volumes.

Using your own images Update the input plan with your image URL (JPG/PNG/BMP/PDF first page, <=50 MB).

Security best practices - Keep .env out of git; rotate keys; prefer Key Vault in production; enable Azure Monitor + spending alerts.

Troubleshooting - "credentials not configured": check env vars + OCR_MODE. - Module missing: install requirements. - Access denied: verify key/region and subscription status. - Local images: host via blob SAS or any temporary HTTPS URL.

Switching modes - Mock: OCR_MODE=mock (or unset) - Azure: OCR_MODE=azure