app.py

import gradio as gr
from tempfile import TemporaryDirectory, NamedTemporaryFile
from pdf2image import convert_from_path
from PIL import Image
import os
from io import BytesIO
import base64
import requests
import pandas as pd
import json

os.system("apt-get update")
os.system("apt-get install poppler-utils")

# Function to convert PDF to images or open a single image
def get_images(file_path):
    images = []
    extension = os.path.splitext(file_path)[-1].lower()
    if extension == ".pdf":
        images = convert_from_path(file_path)
    elif extension in [".tiff", ".tif", ".png", ".jpg", ".jpeg"]:
        image = Image.open(file_path)
        images.append(image)
    return images

# Function to encode image to base64
def encode_image_to_base64(image):
    # Ensure the image is in a format compatible with JPEG
    if image.mode in ["P", "RGBA"]:
        image = image.convert("RGB")
    buffered = BytesIO()
    image.save(buffered, format="JPEG")
    return base64.b64encode(buffered.getvalue()).decode("utf-8")

def process_files_fixed(image_path, page_identifier, error_pages):
    api_key= os.getenv('OPENAI_API_KEY')
    headers = {
        "Content-Type": "application/json",
        "Authorization": f"Bearer {api_key}"
    }

    # read the image and create image object
    image = Image.open(image_path)
    base64_image = encode_image_to_base64(image)

    prompt = """Perform OCR on this image. Analyze the table in the provided image, focusing on the first five columns labeled S.No, Reg #, Roll No. and Marks. In case the table headers are not visible or not present, assume the mentioned order for the columns. Extract and list the data only from these columns, omitting any additional columns that may be present. But DO NOT skip any row from the table, extract all the rows present in the table. The obtained marks are written in both integral and written format as well. Verrify both for better ocr in integers.

Return the response in the following JSoN response format:
```
{
    "data": [
        {
            "S_No": "1",
            "Reg": "059287",
            "Roll_No": "2345234",
            "Marks": "20"
        },
        {
            "S_No": "2",
            "Reg": "059288",
            "Roll_No": "2345235",
            "Marks": "25"
        },
        ...
    ]
}
```"""

    payload = {
        "model": "gpt-4-vision-preview",
        "messages": [
            {
                "role": "user",
                "content": [
                    {
                        "type": "text",
                        "text": prompt
                    },
                    {
                        "type": "image_url",
                        "image_url": {
                            "url": f"data:image/jpeg;base64,{base64_image}",
                            "detail": "high",
                        }
                    }
                ]
            }
        ],
        "max_tokens": 4096,
    }

    response = requests.post("https://api.openai.com/v1/chat/completions", headers=headers, json=payload)

    try:
        if response.status_code == 200:
            json_response = response.json()
            response_content = json_response["choices"][0]["message"]["content"]

            if response_content:
                json_string = response_content[response_content.find("{"): response_content.rfind("}") + 1]
                json_data = json.loads(json_string)
                if "data" in json_data and json_data["data"]:
                    return json_data["data"]
                else:
                    print(f"No records found in page/file: {page_identifier}")
                    error_pages.append(page_identifier)
                    return []
            else:
                print(f"No content in JSON response for page/file: {page_identifier}")
                error_pages.append(page_identifier)
                return []
        else:
            print(f"Error in API call for page/file: {page_identifier}")
            error_pages.append(page_identifier)
            return []
    except Exception as e:
        print(f"Exception processing page/file {page_identifier}: {e}")
        error_pages.append(page_identifier)
        return []

def process_pdf_and_generate_csv(file_path):
    error_pages = []  # Initialize the list to track error pages or files
    images = get_images(file_path)
    structured_data = []
    
    for i, image in enumerate(images, start=1):
        with TemporaryDirectory() as temp_dir:
            image_path = os.path.join(temp_dir, "image.jpg")
            image.save(image_path)
            data = process_files_fixed(image_path, i, error_pages)
            structured_data.extend(data or [])
    
    if structured_data:
        df = pd.DataFrame(structured_data)
        # Save to a temporary file to return through Gradio
        tmp_file = NamedTemporaryFile(delete=False, suffix='.csv')
        df.to_csv(tmp_file.name, index=False)
        return tmp_file.name, f"Errors or no records found in {len(error_pages)} pages/files: {error_pages}"
    else:
        return None, "No data to save or an error occurred."

def gradio_interface(pdf_file):
    result_csv, message = process_pdf_and_generate_csv(pdf_file.name)
    if result_csv:
        return result_csv, message
    else:
        return None, message

iface = gr.Interface(fn=gradio_interface,
                     inputs=gr.File(label="Please upload your PDF file"),
                     outputs=[gr.File(label="Download the generated CSV file"), gr.Textbox(label="Messages")],
                     title="PDF to CSV Table Extractor",
                     description="Upload a PDF file to extract tables into a CSV format.")

iface.launch(share=False)