UP_OCR_data/ocr.py

import os
import json
import mimetypes
import time
import logging
import pandas as pd
import requests
import google.generativeai as genai
import pypdf
from tabulate import tabulate

gemini_api_key = "AIzaSyDC5D6SFk4SRlPzBGmXGwQZBtFd5jXr384" ###Use own API key

# Configure logging
logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(levelname)s] %(message)s")

# Altered project path structure
# Assuming the script is located in the project root:
BASE_DIR = os.path.abspath(os.path.dirname(__file__))
DATA_DIR = os.path.join(BASE_DIR, "data")
RESULTS_DIR = os.path.join(BASE_DIR, "results")
CSV_PATH = os.path.join(DATA_DIR, "output_new.csv")

def extract_text_from_proper_pdf(pdf_path):
    """
    Extracts text from a proper (digitally generated) PDF using pypdf.
    """
    logging.info(f"Extracting text from proper PDF: {pdf_path}")
    try:
        with open(pdf_path, "rb") as f:
            reader = pypdf.PdfReader(f)
            text = ""
            for page in reader.pages:
                page_text = page.extract_text()
                if page_text:
                    text += page_text
            if not text:
                logging.warning(f"No text found in proper PDF: {pdf_path}")
            return text
    except Exception as e:
        logging.error(f"Error extracting text from proper PDF {pdf_path}: {e}")
        return ""

def extract_text_with_gemini_ocr(pdf_path, gemini_api_key):
    """
    Extracts text from a scanned image PDF using Gemini multimodal OCR capabilities.
    """
    logging.info(f"Attempting OCR extraction for scanned PDF: {pdf_path}")
    try:
        genai.configure(api_key=gemini_api_key)
        mime_type, _ = mimetypes.guess_type(pdf_path)
        if mime_type is None:
            mime_type = "application/pdf"
            logging.warning(f"Could not guess MIME type for {pdf_path}, assuming {mime_type}")

        logging.info(f"Uploading file {pdf_path} with MIME type {mime_type}...")
        pdf_file = genai.upload_file(path=pdf_path, mime_type=mime_type)
        logging.info(f"File uploaded successfully: {pdf_file.name}")

        model = genai.GenerativeModel('gemini-1.5-flash-latest')
        max_attempts = 30
        attempts = 0
        while pdf_file.state.name == "PROCESSING" and attempts < max_attempts:
            print('.', end='', flush=True)
            time.sleep(10)
            pdf_file = genai.get_file(pdf_file.name)
            attempts += 1

        if attempts == max_attempts:
            logging.error("Error: File processing timed out.")
            return ""
        if pdf_file.state.name == "FAILED":
            logging.error(f"Error: File processing failed for {pdf_path}")
            return ""

        logging.info("\nFile processed. Sending prompt to Gemini for OCR text extraction...")
        prompt = "Extract all the text content from the provided document. Preserve formatting like paragraphs and tables as best as possible."
        response = model.generate_content([prompt, pdf_file])
        if response.parts:
            extracted_text = response.text
            logging.info(f"OCR text extraction successful (length: {len(extracted_text)} chars).")
            return extracted_text
        else:
            logging.warning("Gemini response contained no parts for text extraction.")
            logging.debug(f"Full Response: {response}")
            return ""
    except Exception as e:
        logging.error(f"An error occurred during Gemini OCR for {pdf_path}: {e}")
        return ""

def extract_tender_info(extracted_text, gemini_api_key):
    """
    Extracts structured tender information using Gemini API.
    """
    genai.configure(api_key=gemini_api_key)
    model = genai.GenerativeModel('gemini-1.5-flash-latest')
    prompt = f"""
Okay, here are prompt templates designed to extract the specified data points from State Transport Corporation (STC) tender documents. Each template includes a standardized name, a definition, and a prompt for extraction.
________________________________________Basic Tender Information
1.	Name: TenderBasic_Id
   Prompt: Extract the unique Tender ID.
2.	Name: TenderBasic_Title
   Prompt: Extract the complete Tender Title.
3.	Name: TenderBasic_IssuingStu
   Prompt: Extract the issuing State Transport Undertaking (STU) name.
4.	Name: TenderBasic_IssuingDepartment
   Prompt: Extract the Issuing Department name, if mentioned.
5.	Name: TenderBasic_State
   Prompt: Extract the State of operation.
6.	Name: TenderBasic_Type
   Prompt: Extract the Tender Type.
7.	Name: TenderBasic_Category
   Prompt: Extract the Tender Category.
8.	Name: TenderBasic_ProcurementMethod
   Prompt: Extract the Procurement Method and any bid conditions.
________________________________________Product/Service Details
9.	Name: Product_ItemCategory
   Prompt: Extract the main Item Category.
10.	Name: Product_ItemSubcategory
   Prompt: Extract the Item Subcategory, if specified.
11.	Name: Product_ItemCode
   Prompt: Extract the Item Code(s) or Part Number(s), if available.
12.	Name: Product_ItemDescription
   Prompt: Extract the detailed description for each required item.
13.	Name: Product_QuantityRequired
   Prompt: Extract the Quantity Required for each item.
14.	Name: Product_UnitOfMeasurement
   Prompt: Extract the Unit of Measurement.
15.	Name: Product_QualityStandards
   Prompt: Extract the Quality Standards or Certifications.
16.	Name: Product_WarrantyRequirements
   Prompt: Extract the details of the Warranty Period and Coverage.
17.	Name: Product_DeliveryLocation
   Prompt: Extract the Delivery Location(s).
18.	Name: Product_InstallationRequirements
   Prompt: Extract the installation requirements.
________________________________________Timeline Information
19.	Name: Timeline_PublicationDate
   Prompt: Extract the Tender Publication Date.
20.	Name: Timeline_BidSubmissionStartDate
   Prompt: Extract the Bid Submission Start Date and Time.
21.	Name: Timeline_BidSubmissionEndDate
   Prompt: Extract the Bid Submission End Date and Time.
22.	Name: Timeline_BidOpeningDate
   Prompt: Extract the Bid Opening Date and Time.
23.	Name: Timeline_DocDownloadStartDate
   Prompt: Extract the Document Download Start Date and Time.
24.	Name: Timeline_DocDownloadEndDate
   Prompt: Extract the Document Download End Date and Time.
25.	Name: Timeline_PreBidMeetingDate
   Prompt: Extract the Pre-Bid Meeting Date, Time, and Venue.
26.	Name: Timeline_ClarificationDeadline
   Prompt: Extract the Clarification Submission Deadline.
27.	Name: Timeline_ContractAwardDate
   Prompt: Extract the Contract Award Date, if mentioned.
28.	Name: Timeline_DeliveryPeriod
   Prompt: Extract the Delivery Timeline or Period.
________________________________________Financial Information
29.	Name: Financial_EstimatedValue
   Prompt: Extract the Estimated Contract Value or Cost.
30.	Name: Financial_EmdAmount
   Prompt: Extract the EMD Amount.
31.	Name: Financial_EmdExemption
   Prompt: Extract details about EMD Exemption eligibility.
32.	Name: Financial_TenderFee
   Prompt: Extract the Tender Fee amount and payment method.
33.	Name: Financial_TenderFeeExemption
   Prompt: Extract details about Tender Fee Exemption.
34.	Name: Financial_PerformanceSecurity
   Prompt: Extract the Performance Security details.
35.	Name: Financial_PaymentTerms
   Prompt: Extract the Payment Terms.
36.	Name: Financial_PriceRevisionTerms
   Prompt: Extract the Price Revision Terms.
________________________________________Documentation Requirements
37.	Name: Docs_RequiredGeneral
   Prompt: Extract the list of general technical and financial documents.
38.	Name: Docs_RequiredLegal
   Prompt: Extract the list of legal documents.
39.	Name: Docs_RequiredCompliance
   Prompt: Extract the list of compliance documents.
40.	Name: Docs_SubmissionFormat
   Prompt: Extract the required Format and Method of Submission.
________________________________________Contract Terms
41.	Name: Contract_Duration
   Prompt: Extract the Contract Duration.
42.	Name: Contract_ExtensionProvisions
   Prompt: Extract the Contract Extension provisions.
43.	Name: Contract_PenaltyClauses
   Prompt: Extract the Penalty Clauses.
44.	Name: Contract_DisputeResolution
   Prompt: Extract the Dispute Resolution mechanism.
45.	Name: Contract_ForceMajeure
   Prompt: Extract the Force Majeure conditions.
46.	Name: Contract_TerminationConditions
   Prompt: Extract the Termination Conditions.
47.	Name: Contract_ContinuingObligations
   Prompt: Extract any Continuing Obligations.
________________________________________Contact Information
48.	Name: Contact_PersonName
   Prompt: Extract the Contact Person's name.
49.	Name: Contact_PersonDesignation
   Prompt: Extract the Contact Person's designation.
50.	Name: Contact_PhoneNumber
   Prompt: Extract the Contact Phone Number(s).
51.	Name: Contact_Email
   Prompt: Extract the Contact Email Address(es).
52.	Name: Contact_OfficeAddress
   Prompt: Extract the Tender Office Address.
________________________________________Eligibility and Qualification Criteria
53.	Name: Eligibility_BidderNationality
   Prompt: Extract the required Bidder Nationality.
54.	Name: Eligibility_MinAnnualTurnover
   Prompt: Extract the Minimum Annual Turnover.
55.	Name: Eligibility_MinYearsExperience
   Prompt: Extract the Minimum Years of Experience.
56.	Name: Eligibility_SimilarWorkExperience
   Prompt: Extract the Similar Work Experience requirements.
57.	Name: Eligibility_IsoCertification
   Prompt: Extract any required ISO Certifications.
58.	Name: Eligibility_ManufacturingCapacity
   Prompt: Extract the Manufacturing Capacity requirements.
59.	Name: Eligibility_TechnicalCapability
   Prompt: Extract the required Technical Capabilities.
60.	Name: Eligibility_FinancialRatios
   Prompt: Extract any Financial Ratios requirements.
61.	Name: Eligibility_RegistrationRequirements
   Prompt: Extract the required Registration Requirements.
________________________________________
Tender Document Text:
{extracted_text}
Provide ONLY the extracted information in valid JSON format, without any introductory text, explanations, or markdown formatting.
    """
    try:
        response = model.generate_content(prompt)
        cleaned_text = response.text.strip()
        if cleaned_text.startswith("```json"):
            cleaned_text = cleaned_text[7:]
        if cleaned_text.endswith("```"):
            cleaned_text = cleaned_text[:-3]
        cleaned_text = cleaned_text.strip()
        extracted_data = json.loads(cleaned_text)
        return extracted_data
    except json.JSONDecodeError as e:
        logging.error(f"Error decoding JSON from Gemini response: {e}")
        logging.debug(f"--- Raw Response Text ---:\n{response.text}\n-------------------------")
        return {}
    except Exception as e:
        logging.error(f"An unexpected error occurred during Gemini processing: {e}")
        if 'response' in locals() and hasattr(response, 'text'):
            logging.debug(f"--- Raw Response Text ---:\n{response.text}\n-------------------------")
        return {}

def process_tender_documents():
    """
    Main function to process tender documents. It reads a CSV file that contains
    the file path and PDF type (e.g., 'Proper' for digital PDFs or 'Scanned' for image PDFs),
    extracts text using the appropriate method, and then extracts structured tender information.
    """
    gemini_api_key = input("Enter your Gemini API key: ").strip()
    os.makedirs(RESULTS_DIR, exist_ok=True)
    
    try:
        df_paths = pd.read_csv(CSV_PATH)
        # Ensure that a "PDF Type" column exists; if not, default all to "Scanned"
        if "PDF Type" not in df_paths.columns:
            df_paths["PDF Type"] = "Scanned Image PDF"
        file_info = df_paths[["Local PDF File", "PDF Type"]].dropna()
        file_info = file_info.iloc[:]  # Process only the first 10 files
    except Exception as e:
        logging.error(f"Error reading CSV file: {e}")
        return

    results = []
    for index, row in file_info.iterrows():
        # Assuming PDF paths in the CSV are relative to the data directory
        pdf_path = os.path.join(DATA_DIR, os.path.normpath(row["Local PDF File"]))
        pdf_type = str(row["PDF Type"]).strip().lower()
        if not os.path.exists(pdf_path):
            logging.warning(f"File not found: {pdf_path}. Skipping...")
            continue

        filename = os.path.basename(pdf_path)
        logging.info(f"Processing {filename} with PDF type: {pdf_type}...")

        # Choose extraction method based on PDF type
        if pdf_type in ["Proper PDF", ""]:
            extracted_text = extract_text_from_proper_pdf(pdf_path)
        else:
            extracted_text = extract_text_with_gemini_ocr(pdf_path, gemini_api_key)
        
        # Save extracted text to a .txt file
        text_filename = os.path.join(RESULTS_DIR, f"{os.path.splitext(filename)[0]}.txt")
        try:
            with open(text_filename, 'w', encoding='utf-8') as f:
                f.write(extracted_text)
            logging.info(f"Text extracted and saved to {text_filename}")
        except Exception as e:
            logging.error(f"Error saving extracted text for {filename}: {e}")

        # Extract structured tender information
        logging.info("Extracting tender information using Gemini...")
        tender_info_dict = extract_tender_info(extracted_text, gemini_api_key)
        if isinstance(tender_info_dict, dict) and tender_info_dict:
            tender_info_dict['filename'] = filename
            results.append(tender_info_dict)
            logging.info(f"Successfully extracted data for {filename}.")
        else:
            logging.warning(f"Could not extract structured data or received empty data for {filename}.")

    if results:
        df = pd.DataFrame(results)
        cols = df.columns.tolist()
        if 'filename' in cols:
            cols.remove('filename')
            cols = ['filename'] + cols
            df = df[cols]
        csv_output_path = os.path.join(RESULTS_DIR, "tender_results.csv")
        try:
            df.to_csv(csv_output_path, index=False)
            logging.info(f"Results saved to {csv_output_path}")
        except Exception as e:
            logging.error(f"Error saving results to CSV: {e}")
        print("\nExtracted Tender Information:")
        print(tabulate(df, headers='keys', tablefmt='grid'))
    else:
        logging.info("No results to display.")

def main():
    process_tender_documents()

if __name__ == "__main__":
    main()