5_page_answer_parse

import os
import base64
from dotenv import load_dotenv
from groq import Groq

os.environ['GROQ_API_KEY'] = "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"
load_dotenv()
client = Groq()

GROQ_API_KEY = os.getenv("GROQ_API_KEY")



# ----------------------------------
# Encode image as base64 for API
# ----------------------------------
def encode_image(image_path):
    if not os.path.isfile(image_path):
        raise FileNotFoundError(f"Image not found: {image_path}")
    with open(image_path, "rb") as f:
        return base64.b64encode(f.read()).decode("utf-8")

# ----------------------------------
# Extract and format visible text using Groq vision model
# ----------------------------------
def extract_text_from_images(image_paths):
    base64_images = []
    for path in image_paths:
        base64_images.append(encode_image(path))

    prompt = (
    "Extract only the visible text from these images, and organize it by question number.\n"
    "- Identify each question based on its number (e.g., Q1, 1., 2., etc.).\n"
    "- Group each answer under its respective question number using clear headings like 'Question 1:', 'Question 2:', etc.\n"
    "- Do NOT generate or assume any new content—only extract what's actually visible in the image.\n"
    "- Correct any spelling mistakes.\n"
    "- Preserve logical structure (e.g., headings, bullet points, tables, equations) within each answer.\n"
    "- Use clean and consistent formatting so the output is both human-readable and machine-readable.\n"
    "- Ignore decorative elements, arrows, or icons unless they contain actual text.\n"
    "- Ensure each answer appears immediately after its corresponding question number."
    )

    message_content = [{"type": "text", "text": prompt}]
    for base64_img in base64_images:
        message_content.append({
            "type": "image_url",
            "image_url": {"url": f"data:image/png;base64,{base64_img}"}
        })

    response = client.chat.completions.create(
        model="meta-llama/llama-4-scout-17b-16e-instruct",
        messages=[{"role": "user", "content": message_content}],
        temperature=0.2,
        max_completion_tokens=2048,
        top_p=1,
        stream=False
    )

    extracted_text = response.choices[0].message.content
    return extracted_text

# ----------------------------------
# Example usage
# ----------------------------------
if __name__ == "__main__":
    image_paths = [
        "image1.jpeg",
        "image2.jpeg",
        "image3.jpeg",
        "image4.jpeg",
        "image5.jpeg"
    ]

    print("🔍 Extracting text from images...")

    try:
        result = extract_text_from_images(image_paths)
        print("\n📄 Extracted Text:\n")
        print(result)
    except Exception as e:
        print(f"❌ Error: {e}")