app.py

# from flask import Flask, request, jsonify, render_template
# import pdfplumber
# import io
# from transformers import T5ForConditionalGeneration, T5Tokenizer
# import torch

# app = Flask(__name__)

# # Load the T5 model and tokenizer
# model_name = "t5-large"
# tokenizer = T5Tokenizer.from_pretrained(model_name)
# model = T5ForConditionalGeneration.from_pretrained(model_name)

# @app.route('/')
# def index():
#     return render_template('index.html')  # Ensure index.html exists in templates/

# @app.route('/upload-resume', methods=['POST'])
# def upload_resume():
#     try:
#         file = request.files['resume']
#         if not file.filename.endswith('.pdf'):
#             return jsonify({"error": "Only PDF files are supported"}), 400

#         # Extract text from PDF
#         text = ""
#         with pdfplumber.open(io.BytesIO(file.read())) as pdf:
#             for page in pdf.pages:
#                 page_text = page.extract_text()
#                 if page_text:
#                     text += page_text + "\n"

#         # Create a prompt dynamically based on resume content
#         prompt = f"Generate 5 technical and 3 behavioral interview questions based on this resume:\n{text.strip()}"

#         # Tokenize and generate questions
#         inputs = tokenizer(prompt, return_tensors="pt", max_length=1024, truncation=True)
#         output = model.generate(**inputs, max_length=512, num_beams=4, early_stopping=True)

#         questions = tokenizer.decode(output[0], skip_special_tokens=True)
#         questions_list = [q.strip() for q in questions.split("\n") if q.strip()]

#         return jsonify(questions_list)

#     except Exception as e:
#         print("Error:", e)
#         return jsonify({"error": str(e)}), 500

# if __name__ == '__main__':
#     app.run(debug=True)




from flask import Flask, request, jsonify, render_template, send_from_directory
import pdfplumber
import io
import re
import nltk
from nltk.corpus import stopwords
from nltk.tokenize import word_tokenize
import spacy
from transformers import GPT2Tokenizer, GPT2LMHeadModel
import torch
import logging
from collections import Counter
import os


os.environ["TRANSFORMERS_CACHE"] = os.path.join(os.getcwd(), "models")

# Configure logging
logging.basicConfig(level=logging.INFO, 
                    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
logger = logging.getLogger(__name__)

app = Flask(__name__, static_folder='static')

# Download required NLTK data
try:
    nltk.download('punkt', quiet=True)
    nltk.download('stopwords', quiet=True)
    stop_words = set(stopwords.words('english'))
except Exception as e:
    logger.warning(f"NLTK download error: {str(e)}")
    stop_words = set()

# Load spaCy model for entity recognition
try:
    nlp = spacy.load("en_core_web_sm")
except Exception as e:
    logger.warning(f"SpaCy model loading error: {str(e)}")
    
    # Optional fallback if spaCy model isn't installed
    def download_spacy_model():
        import subprocess
        subprocess.call(["python", "-m", "spacy", "download", "en_core_web_sm"])
        
    try:
        download_spacy_model()
        nlp = spacy.load("en_core_web_sm")
    except:
        logger.error("Failed to load spaCy model")
        nlp = None

# Load GPT-2 model and tokenizer for better text generation
try:
    model_name = "distilgpt2"  
    tokenizer = GPT2Tokenizer.from_pretrained(model_name)
    model = GPT2LMHeadModel.from_pretrained(model_name)
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    model.to(device)
except Exception as e:
    logger.error(f"Model loading error: {str(e)}")
    model = None
    tokenizer = None

def extract_skills(text):
    """Extract technical skills from resume text"""
    # Common technical skills to look for
    common_skills = {
        'programming': ['python', 'java', 'javascript', 'c++', 'c#', 'ruby', 'php', 'swift', 'kotlin', 'go', 'rust', 'typescript', 'scala', 'perl', 'shell', 'bash', 'sql', 'html', 'css'],
        'frameworks': ['react', 'angular', 'vue', 'django', 'flask', 'spring', 'express', 'rails', 'asp.net', 'laravel', 'node.js', 'bootstrap', 'jquery', 'tensorflow', 'pytorch', 'numpy', 'pandas'],
        'databases': ['mysql', 'postgresql', 'mongodb', 'oracle', 'sql server', 'sqlite', 'redis', 'cassandra', 'dynamodb', 'firebase', 'elasticsearch'],
        'tools': ['git', 'docker', 'kubernetes', 'jenkins', 'aws', 'azure', 'gcp', 'terraform', 'ansible', 'jira', 'confluence', 'notion', 'figma', 'photoshop', 'illustrator'],
        'methodologies': ['agile', 'scrum', 'kanban', 'devops', 'ci/cd', 'test driven development', 'tdd', 'behavior driven development', 'bdd', 'rest', 'soap', 'microservices', 'serverless']
    }
    
    # Flatten the list
    all_skills = [skill for category in common_skills.values() for skill in category]
    
    # Find matches in the text
    found_skills = []
    text_lower = text.lower()
    
    for skill in all_skills:
        # Check for whole word matches
        pattern = r'\b' + re.escape(skill) + r'\b'
        if re.search(pattern, text_lower):
            found_skills.append(skill)
    
    # If spaCy is available, also look for named entities that might be technologies
    if nlp:
        doc = nlp(text)
        for ent in doc.ents:
            if ent.label_ in ["ORG", "PRODUCT"] and len(ent.text) > 2:
                entity = ent.text.lower()
                # Check if entity might be a technology
                if any(tech_word in entity for tech_word in ["tech", "software", "platform", "system", "framework", "api", "cloud"]):
                    found_skills.append(ent.text)
    
    # Count occurrences to identify most important skills
    skill_counter = Counter(found_skills)
    top_skills = [skill for skill, _ in skill_counter.most_common(10)]
    
    return top_skills

def extract_experience(text):
    """Extract work experience information from resume"""
    experience_data = []
    
    # Look for common experience section headers
    experience_headers = ["experience", "work experience", "employment history", "professional experience"]
    
    # Simple pattern matching for job titles and dates
    job_title_pattern = r"(?:^|\n)(?:Senior |Lead |Junior |Staff |Principal )?\b(?:Developer|Engineer|Designer|Manager|Director|Analyst|Consultant|Administrator|Architect|Specialist)\b"
    date_pattern = r"\b(?:Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec)[a-z]* \d{4}\s*(?:-|–|to)\s*(?:(?:Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec)[a-z]* \d{4}|Present|Current|Now)"
    
    # Find experience section
    text_lower = text.lower()
    section_start = None
    for header in experience_headers:
        if header in text_lower:
            section_start = text_lower.find(header)
            break
    
    if section_start is not None:
        # Extract the section (assume it ends at the next major section)
        next_section_start = float('inf')
        for next_header in ["education", "skills", "projects", "certifications", "references"]:
            pos = text_lower.find(next_header, section_start + 1)
            if pos > section_start and pos < next_section_start:
                next_section_start = pos
        
        experience_section = text[section_start:next_section_start] if next_section_start < float('inf') else text[section_start:]
        
        # Extract job titles
        job_titles = re.findall(job_title_pattern, experience_section, re.IGNORECASE)
        
        # Extract date ranges
        date_ranges = re.findall(date_pattern, experience_section)
        
        # Combine the information
        for i, title in enumerate(job_titles[:3]):  # Limit to top 3 positions
            date = date_ranges[i] if i < len(date_ranges) else "Unknown date"
            experience_data.append({"title": title.strip(), "date": date})
    
    return experience_data

def extract_education(text):
    """Extract education information from resume"""
    education_data = []
    
    # Look for degrees and institutions
    degree_pattern = r"\b(?:Bachelor|Master|PhD|Doctorate|BSc|MSc|BA|MA|MBA|MD|JD|BS|MS|B\.S\.|M\.S\.|B\.A\.|M\.A\.)['\s\w]*\b"
    institution_pattern = r"\b(?:University|College|Institute|School) of [\w\s]+\b"
    
    # Find matches
    degrees = re.findall(degree_pattern, text)
    institutions = re.findall(institution_pattern, text)
    
    # Combine the information
    for i, degree in enumerate(degrees[:2]):  # Limit to top 2 degrees
        institution = institutions[i] if i < len(institutions) else "Unknown institution"
        education_data.append({"degree": degree.strip(), "institution": institution})
    
    return education_data

def preprocess_resume(text):
    """Extract structured information from resume text"""
    # Basic text cleaning
    text = text.replace('\n\n', ' [BREAK] ')
    text = re.sub(r'\s+', ' ', text)
    text = text.replace(' [BREAK] ', '\n\n')
    
    # Extract key information
    skills = extract_skills(text)
    experience = extract_experience(text)
    education = extract_education(text)
    
    # Create structured resume data
    resume_data = {
        "skills": skills,
        "experience": experience,
        "education": education,
        "full_text": text
    }
    
    return resume_data

def generate_interview_questions(resume_data):
    """Generate interview questions based on resume data"""
    # Default set of questions if model fails
    default_questions = [
        # Technical questions
        "What challenges have you faced when working with databases, and how did you overcome them?",
        "Describe a project where you had to optimize code for performance. What approach did you take?",
        "How do you ensure your code is maintainable and follows best practices?",
        "What software development methodologies are you familiar with, and which do you prefer?",
        "How do you approach testing your code?",
        
        # Behavioral questions
        "Tell me about a challenging project you worked on and how you approached it.",
        "Describe a situation where you had to learn a new technology quickly.",
        "How do you handle tight deadlines and pressure?"
    ]
    
    # If no model is loaded, return default questions
    if model is None or tokenizer is None:
        return default_questions
    
    # Extract key resume information
    skills_str = ", ".join(resume_data["skills"])
    
    experience_str = ""
    for exp in resume_data["experience"]:
        experience_str += f"{exp['title']} ({exp['date']}), "
    
    # Build a prompt for the model
    prompt = f"""Generate 8 interview questions based on this resume information:
Skills: {skills_str}
Experience: {experience_str}

Include 5 technical questions specific to the candidate's skills and 3 behavioral questions.
Format each question on a new line and make them realistic interview questions.
"""
    
    try:
        # Generate text
        input_ids = tokenizer.encode(prompt, return_tensors="pt").to(device)
        attention_mask = torch.ones(input_ids.shape, dtype=torch.long, device=device)
        
        # Generate with better parameters for coherent questions
        output = model.generate(
            input_ids,
            attention_mask=attention_mask,
            max_length=1024,
            num_return_sequences=1,
            no_repeat_ngram_size=2,
            do_sample=True,
            top_p=0.92,
            top_k=50,
            temperature=0.85
        )
        
        generated_text = tokenizer.decode(output[0], skip_special_tokens=True)
        
        # Process the output to extract questions
        text_split = generated_text.replace(prompt, "").strip().split("\n")
        
        # Clean up and format questions
        questions = []
        for line in text_split:
            # Remove question numbers and clean
            line = re.sub(r'^\d+[\.\)]\s*', '', line.strip())
            
            # Only keep lines that look like questions
            if line and ('?' in line or any(q_word in line.lower() for q_word in ["how", "what", "why", "when", "where", "describe", "tell", "explain"])):
                # Ensure questions end with question mark
                if not line.endswith('?') and any(q_word in line.lower() for q_word in ["how", "what", "why", "when", "where"]):
                    line += '?'
                questions.append(line)
        
        # Check if we got enough questions
        if len(questions) >= 5:
            return questions[:8]  # Return up to 8 questions
        else:
            # Fallback to default questions
            logger.warning("Generated questions insufficient, using defaults")
            return default_questions
            
    except Exception as e:
        logger.error(f"Question generation error: {str(e)}")
        return default_questions

@app.route('/')
def index():
    return render_template('index.html')

@app.route('/static/<path:path>')
def serve_static(path):
    return send_from_directory('static', path)

@app.route('/upload-resume', methods=['POST'])
def upload_resume():
    try:
        if 'resume' not in request.files:
            return jsonify({"error": "No file part"}), 400
            
        file = request.files['resume']
        
        if file.filename == '':
            return jsonify({"error": "No selected file"}), 400
            
        if not file.filename.endswith('.pdf'):
            return jsonify({"error": "Only PDF files are supported"}), 400

        # Extract text from PDF
        text = ""
        try:
            with pdfplumber.open(io.BytesIO(file.read())) as pdf:
                for page in pdf.pages:
                    page_text = page.extract_text()
                    if page_text:
                        text += page_text + "\n"
        except Exception as e:
            logger.error(f"PDF extraction error: {str(e)}")
            return jsonify({"error": "Unable to extract text from PDF. Is the file corrupt?"}), 500

        if not text.strip():
            return jsonify({"error": "No text could be extracted from the PDF"}), 400

        # Process resume and generate questions
        resume_data = preprocess_resume(text)
        questions = generate_interview_questions(resume_data)

        # Add tailored question based on skills
        if resume_data["skills"]:
            top_skill = resume_data["skills"][0]
            skill_question = f"Tell me about your experience with {top_skill} and how you've applied it in your projects."
            questions.append(skill_question)

        return jsonify(questions)

    except Exception as e:
        logger.error(f"General error: {str(e)}")
        return jsonify({"error": "An error occurred processing your request. Please try again."}), 500

if __name__ == '__main__':
    # Make sure templates and static directories exist
    for directory in ['templates', 'static']:
        os.makedirs(directory, exist_ok=True)
    
    # Create index.html in templates if needed
    templates_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'templates')
    if not os.path.exists(os.path.join(templates_dir, 'index.html')):
        with open(os.path.join(templates_dir, 'index.html'), 'w') as f:
            f.write('''<!DOCTYPE html>
<html>
<head>
    <title>Resume Question Generator</title>
    <meta http-equiv="refresh" content="0;url=/" />
</head>
<body>
    <p>Redirecting...</p>
</body>
</html>''')
    
    # Start the Flask app
    app.run(debug=True)