utils.py

from paddleocr import PaddleOCR
import re

# Initialize OCR model only once
ocr = PaddleOCR(use_angle_cls=True, lang='en')  # lang='en' for English documents

def extract_kyc_fields(file_path):
    try:
        # Run OCR
        result = ocr.ocr(file_path, cls=True)

        all_text = ""
        for line_group in result:
            for line in line_group:
                all_text += line[1][0] + "\n"

        # Aadhaar number (format with or without space/dash)
        aadhaar_match = re.search(r'\b\d{4}[\s\-]?\d{4}[\s\-]?\d{4}\b', all_text)

        # DOB (any DD/MM/YYYY or similar)
        dob_match = re.search(r'\b\d{2}[\/\-]\d{2}[\/\-]\d{4}\b', all_text)

        # Name: try to detect a line with 'Name' or fallback to top line
        name = "Not found"
        for line in all_text.split("\n"):
            if re.search(r'\b(name|naam|namf)\b', line, re.IGNORECASE):
                name = line.split(":")[-1].strip() if ":" in line else line.strip()
                break

        if name == "Not found":
            name = all_text.split("\n")[0].strip()

        return {
            "aadhaar_number": aadhaar_match.group(0) if aadhaar_match else "Not found",
            "dob": dob_match.group(0) if dob_match else "Not found",
            "name": name
        }

    except Exception as e:
        return {"error": f"PaddleOCR failed: {str(e)}"}