utils.py

from paddleocr import PaddleOCR
import re

# Initialize OCR
ocr = PaddleOCR(use_angle_cls=True, lang='en')

# Helper regex patterns
PAN_REGEX = r'\b[A-Z]{5}[0-9]{4}[A-Z]\b'
AADHAAR_REGEX = r'\b\d{4}[\s-]?\d{4}[\s-]?\d{4}\b'
DOB_REGEX = [
    r'\b\d{2}[./-]\d{2}[./-]\d{4}\b',  # 12/04/1980 or 12-04-1980
    r'\b\d{4}-\d{2}-\d{2}\b',          # 1980-04-12
    r'\b\d{2}[./-](JAN|FEB|MAR|APR|MAY|JUN|JUL|AUG|SEP|OCT|NOV|DEC)[./-]\d{4}\b',  # 12-APR-1980
    r'\b(19|20)\d{2}\b'                # Year only
]

GENDERS = ["MALE", "FEMALE", "TRANSGENDER"]

def extract_kyc_fields(file_path, force_type=None):
    try:
        result = ocr.ocr(file_path, cls=True)
        lines = []

        # Normalize OCR text
        for block in result:
            for line in block:
                text = re.sub(r'\s+', ' ', line[1][0].strip())
                if text:
                    lines.append(text)

        full_text = "\n".join(lines)

        # Determine card type
        if force_type:
            card_type = force_type.upper()
        else:
            pan_match = re.search(PAN_REGEX, full_text)
            aadhaar_match = re.search(AADHAAR_REGEX, full_text)
            card_type = "UNKNOWN"
            if pan_match:
                card_type = "PAN"
            elif aadhaar_match:
                card_type = "AADHAAR"

        response = {"card_type": card_type}

        if card_type == "PAN":
            pan_number = re.search(PAN_REGEX, full_text)
            response["pan_number"] = pan_number.group(0) if pan_number else "Not found"
            response["dob"] = extract_dob(lines)
            response["name"] = extract_pan_name(lines)

        elif card_type == "AADHAAR":
            aadhaar_number = re.search(AADHAAR_REGEX, full_text)
            response["aadhaar_number"] = aadhaar_number.group(0) if aadhaar_number else "Not found"
            response["dob"] = extract_dob(lines)
            response["gender"] = extract_gender(lines)
            response["name"] = extract_aadhaar_name(lines)

        else:
            response["error"] = "Could not identify document as PAN or Aadhaar."

        return response

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


def extract_dob(lines):
    for line in lines:
        for pattern in DOB_REGEX:
            match = re.search(pattern, line, re.IGNORECASE)
            if match:
                return match.group(0)
    return "Not found"


def extract_gender(lines):
    for line in lines:
        for gender in GENDERS:
            if gender in line.upper():
                return gender
    return "Not found"


def extract_pan_name(lines):
    # Heuristic: Name is after "INCOME TAX DEPARTMENT" and contains only letters/spaces
    for i, line in enumerate(lines):
        if "INCOME TAX DEPARTMENT" in line.upper():
            for j in range(i + 1, len(lines)):
                candidate = lines[j].strip()
                if re.match(r'^[A-Z\s.]+$', candidate) and not re.search(r'\d', candidate):
                    # Skip words like INDIA, GOVT, DEPARTMENT
                    if not any(x in candidate.upper() for x in ["INDIA", "GOVT", "DEPARTMENT"]):
                        return candidate
    return "Not found"


def extract_aadhaar_name(lines):
    # Heuristic: Name is usually above DOB
    for i, line in enumerate(lines):
        if any(re.search(p, line) for p in DOB_REGEX):
            if i > 0:
                candidate_name = lines[i - 1].strip()
                if not re.search(r'\d', candidate_name) and len(candidate_name.split()) >= 2:
                    if not any(x in candidate_name.upper() for x in ["DOB", "INDIA", "MALE", "FEMALE", "GOVERNMENT"]):
                        return candidate_name
    # Fallback: First line with >=2 words and no digits
    for line in lines:
        candidate_name = line.strip()
        if len(candidate_name.split()) >= 2 and not re.search(r'\d', candidate_name):
            if not any(x in candidate_name.upper() for x in ["DOB", "INDIA", "MALE", "FEMALE", "GOVERNMENT"]):
                return candidate_name
    return "Not found"