app.py

import gradio as gr
import os
import zipfile
import json
import pandas as pd
import datetime
import shutil
import hashlib
from io import StringIO
from pdfminer.pdfinterp import PDFResourceManager, PDFPageInterpreter
from pdfminer.converter import TextConverter
from pdfminer.layout import LAParams
from pdfminer.pdfpage import PDFPage
from PyPDF2 import PdfReader
from openai import OpenAI, RateLimitError
import backoff
import re

class PDFUtils:
    @staticmethod
    def extract_text_with_pdfminer(path):
        rsrcmgr = PDFResourceManager()
        retstr = StringIO()
        codec = 'utf-8'
        laparams = LAParams()
        device = TextConverter(rsrcmgr, retstr, codec=codec, laparams=laparams)

        with open(path, 'rb') as fp:
            interpreter = PDFPageInterpreter(rsrcmgr, device)
            for page in PDFPage.get_pages(fp, check_extractable=True):
                interpreter.process_page(page)

        text = retstr.getvalue()
        return text

    @staticmethod
    def extract_text_with_pypdf2(path):
        reader = PdfReader(path)
        text = ''.join(page.extract_text() for page in reader.pages)
        return text

    @staticmethod
    def convert_pdf_to_text(path):
        text = PDFUtils.extract_text_with_pdfminer(path)
        if text is None:
            print("Processing using PyPDF2")
            text = PDFUtils.extract_text_with_pypdf2(path)
        return text

class OpenAIClassifier:
    def __init__(self, api_key):
        self.client = OpenAI(api_key=api_key)
        self.processed_resumes = set()

    @backoff.on_exception(backoff.expo, RateLimitError) 
    def completions_with_backoff(self, **kwargs):
        return self.client.chat.completions.create(**kwargs)

    def hash_resume_text(self, resume_text):
        return hashlib.md5(resume_text.encode()).hexdigest()

    def classify_resume(self, resume_text, questions_string):
        resume_hash = self.hash_resume_text(resume_text)
        if resume_hash in self.processed_resumes:
            return "Already Processed"
        
        response = self.client.chat.completions.create(
            model="gpt-3.5-turbo",
            messages=[
                {"role": "system", "content": f"As a very strict hiring manager who is evaluating a resume. Answer each question, let questions be keys and answers be values {questions_string} As a hiring manager parse through the resume thoroughly and answer the following questions with the same index in dictionary format. In the experience section, if there is no direct reference to the number of years of experience, count the number of years from the least present year in the experience section."},
                {"role": "user", "content": resume_text}
            ],
            temperature=0.7,
            max_tokens=516
        )

        summary = str(response.choices[0].message)
        self.processed_resumes.add(resume_hash)

        # Extract JSON substring using regular expression
        json_match = re.search(r"\{.*\}", summary)
        if json_match:
            return json_match.group(0)  # Return only the JSON part
        else:
            return "{}"

def read_json_file(file_path):
    with open(file_path, 'r') as json_file:
        json_data = json.load(json_file)
        return json.dumps(json_data, indent=2)

def process_text(text):
    return text.replace('\\n', '<br>') if text else "Error: Text is None"

def process_pdf(file, model_choice, openai_classifier, questions_string):
    data = PDFUtils.convert_pdf_to_text(file)
    if model_choice == "openai":
        classification = openai_classifier.classify_resume(data, questions_string)
        print(f"Debug: classification output:\n{classification}\n")  # Debugging line to confirm JSON structure
        return {"file_path": file, "result": process_text(classification)}
    else:
        print("Only openai can be utilized")

def unzip_and_process(zip_file, model_choice, openai_classifier, questions_string):
    extract_folder, selected_folder, rejected_folder, error_folder = 'extracted_files/', 'extracted_files/Selected/', 'extracted_files/Rejected/', 'extracted_files/Error/'
    os.makedirs(extract_folder, exist_ok=True)
    os.makedirs(selected_folder, exist_ok=True)
    os.makedirs(rejected_folder, exist_ok=True)
    os.makedirs(error_folder, exist_ok=True)

    for root, dirs, files in os.walk(extract_folder):
        for file_name in files:
            os.remove(os.path.join(root, file_name))

    with zipfile.ZipFile(zip_file, 'r') as zip_ref:
        zip_ref.extractall(extract_folder)

    df_results = pd.DataFrame()
    questions_dict = json.loads(questions_string)
    questions_list = list(questions_dict.values())
    df_results = pd.DataFrame(columns=questions_list)

    for root, dirs, files in os.walk(extract_folder):
        for file_name in files:
            file_path = os.path.join(root, file_name)
            if file_path.lower().endswith('.pdf'):
                try:
                    result = process_pdf(file_path, model_choice, openai_classifier, questions_string)

                    # If already processed, skip this file
                    if result['result'] == "Already Processed":
                        print(f"{file_name} is already processed. Skipping.")
                        continue

                    cleaned_data_str = result['result'].replace("<br>", "").replace("\\", "")
                    
                    # Attempt to parse JSON data
                    data_dict = json.loads(cleaned_data_str)
                    df_resume = pd.DataFrame(data_dict.items(), columns=["Key", os.path.basename(file_path)])
                    df_resume["Key"] = df_resume["Key"].replace(questions_dict)
                    df_results = pd.concat([df_results, df_resume.set_index(["Key"]).T.reset_index(drop=True)], axis=0, ignore_index=True)

                    # Decide on the destination based on the final selection key
                    selection_rejection = data_dict.get(list(data_dict.keys())[-1])
                    destination = selected_folder if selection_rejection == "Selected" else rejected_folder if selection_rejection == "Rejected" else error_folder
                    shutil.move(file_path, os.path.join(destination, os.path.basename(file_path)))

                except json.JSONDecodeError as e:
                    print(f"JSON decoding error for {file_path}: {e}")
                    shutil.move(file_path, os.path.join(error_folder, file_name))
                except Exception as ex:
                    print(f"Error processing file {file_path}: {ex}")
                    shutil.move(file_path, os.path.join(error_folder, file_name))
            else:
                shutil.move(file_path, os.path.join(error_folder, file_name))
                print(f"File '{file_name}' is not a PDF. Moved to error directory.")
    
    # Save processed results to an Excel file
    timestamp = datetime.datetime.now().strftime("%Y-%m-%d_%H-%M-%S")
    excel_file = f'extracted_files/output_results_{timestamp}.xlsx'
    df_results.to_excel(excel_file, index=False)
    print(f"Processed resumes saved to {excel_file}")
    return excel_file  # Return the path to the Excel file

def run_interface(questions_json, resumes_zip):
    try:
        # Handle the questions.json file content
        if isinstance(questions_json, bytes):
            # Parse bytes directly as JSON string
            questions_string = questions_json.decode('utf-8')
        elif isinstance(questions_json, str):
            # If it's a file path
            with open(questions_json, 'r') as f:
                questions_string = f.read()
        else:
            raise ValueError("Unexpected questions_json type")

        # Handle the ZIP file
        if isinstance(resumes_zip, bytes):
            # Write bytes to a temporary zip file
            temp_zip_path = "temp_uploaded_resumes.zip"
            with open(temp_zip_path, "wb") as f:
                f.write(resumes_zip)
            zip_path = temp_zip_path
        elif isinstance(resumes_zip, str):
            zip_path = resumes_zip
        else:
            raise ValueError("Unexpected resumes_zip type")

        openai_classifier = OpenAIClassifier(api_key="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")
        output_file = unzip_and_process(zip_path, "openai", openai_classifier, questions_string)
        
        # Clean up temporary zip file if it was created
        if isinstance(resumes_zip, bytes) and os.path.exists(temp_zip_path):
            os.remove(temp_zip_path)
            
        return output_file

    except Exception as e:
        print(f"Error in run_interface: {str(e)}")
        # Clean up temporary files in case of error
        if 'temp_zip_path' in locals() and os.path.exists(temp_zip_path):
            os.remove(temp_zip_path)
        raise gr.Error(f"Processing failed: {str(e)}")

# Set up the Gradio interface
interface = gr.Interface(
    fn=run_interface,
    inputs=[
        gr.File(
            label="Upload questions.json", 
            type="binary",
            file_types=[".json"]
        ),
        gr.File(
            label="Upload ZIP of Resumes", 
            type="binary",
            file_types=[".zip"]
        )
    ],
    outputs=gr.File(label="Processed Results Excel"),
    title="Resume Analysis System",
    description="Upload a questions.json file and a ZIP file containing resumes to process.",
    allow_flagging="never",
    examples=[
        ["questions.json", "to_be_screened_2.zip"]
    ],
    cache_examples=False
) 

# Launch with debugging and sharing enabled
interface.launch(debug=True, share=True)