Implement deep OCR with Tesseract and RegEx patterns for lab extraction

Optimize extraction questions and add manual hint fallback

Fix IndentationError in parser.py

Remove Blue Dream mock template and use error fallback

Add debug logging to AI extraction

Correct base64 handling in extract endpoint

Refine AI extraction questions and logging

Add PDF to Image conversion for real extraction

Enable real AI extraction using HF Inference API

Add root endpoint so HF doesn't log 404 on health check

Fix 500 error on PDF mock extract and add traceback

Optimize requirements: Use CPU version of PyTorch to reduce image size

Fix Dockerfile: replace libgl1-mesa-glx with libgl1

Add Hugging Face Space configuration to backend README

Initial commit