model.py

import os
import fitz
from docx import Document
from sentence_transformers import SentenceTransformer, util
from transformers import pipeline
import pandas as pd
from datetime import datetime
import zipfile
import shutil
import re

# ================== Models ==================
MODELS = {
    "Fast (MiniLM)": "sentence-transformers/all-MiniLM-L6-v2",
    "Balanced (Recommended)": "sentence-transformers/all-mpnet-base-v2",
    "High Accuracy": "sentence-transformers/multi-qa-mpnet-base-dot-v1"
}

loaded_models = {}
skills_classifier = None

def get_model(model_name: str):
    if model_name not in loaded_models:
        print(f"🔄 Loading model: {model_name}")
        loaded_models[model_name] = SentenceTransformer(MODELS[model_name])
    return loaded_models[model_name]

def get_skills_classifier():
    global skills_classifier
    if skills_classifier is None:
        skills_classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")
    return skills_classifier

def extract_text_from_file(file_path: str) -> str:
    ext = os.path.splitext(file_path)[1].lower()
    try:
        if ext == ".pdf":
            doc = fitz.open(file_path)
            text = "\n".join([page.get_text("text") for page in doc])
            doc.close()
            return text.strip()
        
        elif ext in [".docx", ".doc"]:
            doc = Document(file_path)
            full_text = [para.text.strip() for para in doc.paragraphs if para.text.strip()]
            for table in doc.tables:
                for row in table.rows:
                    for cell in row.cells:
                        if cell.text.strip():
                            full_text.append(cell.text.strip())
            return "\n".join(full_text).strip()
        return ""
    except Exception as e:
        print(f"Extraction error: {e}")
        return ""

def extract_qualifications(text: str):
    """Extract only Degrees / Qualifications"""
    if not text:
        return "Not mentioned"
    
    text_lower = text.lower()
    degree_pattern = r'\b(bba|bs|be|bit|bsc|ba|b\.?a\.?|b\.?sc\.?|bachelor|master|mba|msc|m\.?sc\.?|mphil|phd|doctorate|intermediate|matric)\b'
    
    degrees = re.findall(degree_pattern, text_lower)
    degrees = list(set([d.upper().replace('.', '') for d in degrees if d not in ["ms office", "microsoft"]]))
    
    return ", ".join(degrees) if degrees else "Not mentioned"

def extract_courses(text: str):
    """Extract Courses, Certifications & Trainings"""
    if not text:
        return "Not mentioned"
    
    text_lower = text.lower()
    course_keywords = ["course", "certification", "certificate", "certified", "training", "workshop", "bootcamp", "specialization", "diploma"]
    
    courses = []
    sentences = re.split(r'[.!?;\n]', text)
    
    for sentence in sentences:
        sent = sentence.strip()
        sent_lower = sent.lower()
        if any(kw in sent_lower for kw in course_keywords):
            clean_sent = re.sub(r'\s+', ' ', sent)
            if 15 < len(clean_sent) < 110:   # reasonable length
                courses.append(clean_sent)
    
    courses = list(dict.fromkeys(courses))[:8]  # max 8 courses
    return " | ".join(courses) if courses else "No courses/certifications mentioned"

def get_required_qualifications(job_description: str):
    jd_lower = job_description.lower()
    degree_keywords = ["bba", "bachelor", "bs", "be", "master", "mba", "phd", "degree", "graduation", "qualification", "engineering"]
    low_edu_keywords = ["guard", "security", "watchman", "peon", "driver", "helper", "labor"]
    
    if any(kw in jd_lower for kw in low_edu_keywords):
        return False, "Degree not required"
    
    if any(kw in jd_lower for kw in degree_keywords):
        return True, "Degree required"
    return False, "Qualification preferred"

def extract_skills(resume_text: str):
    if len(resume_text) < 100:
        return "Insufficient content"
    classifier = get_skills_classifier()
    labels = ["Python", "Machine Learning", "Deep Learning", "Data Science", "SQL", "FastAPI", "AWS", "Docker", "Communication", "Leadership"]
    try:
        result = classifier(resume_text[:2000], labels, multi_label=True)
        skills = [label for label, score in zip(result['labels'], result['scores']) if score > 0.40]
        return ", ".join(skills[:12]) if skills else "Basic skills"
    except:
        return "Skills extraction failed"

# ================== Main Function ==================
def screen_resumes_backend(job_description: str, cv_folder, model_option: str, threshold: float = 0.65):
    if not cv_folder:
        return [], "No CVs found.", None, None

    files = cv_folder if isinstance(cv_folder, list) else [cv_folder]
    model = get_model(model_option)
    results = []
    
    os.makedirs("outputs/selected_cvs", exist_ok=True)
    os.makedirs("outputs/reports", exist_ok=True)

    job_embedding = model.encode(job_description, convert_to_tensor=True)
    degree_required, qual_note = get_required_qualifications(job_description)

    for file_obj in files:
        source_path = file_obj.name if hasattr(file_obj, 'name') else str(file_obj)
        filename = os.path.basename(source_path)

        resume_text = extract_text_from_file(source_path)

        if len(resume_text) < 100:
            results.append({
                "Candidate": filename,
                "Score (%)": 0.0,
                "Qualifications": "❌ Extraction Failed",
                "Courses": "N/A",
                "Status": "❌ Extraction Failed",
                "Extracted Skills": "N/A",
                "Model Used": model_option
            })
            continue

        qualifications = extract_qualifications(resume_text)
        courses = extract_courses(resume_text)
        skills = extract_skills(resume_text)

        resume_embedding = model.encode(resume_text, convert_to_tensor=True)
        similarity = util.cos_sim(job_embedding, resume_embedding)[0][0].item()

        final_status = "✅ Shortlisted"
        if degree_required and qualifications == "Not mentioned":
            final_status = "❌ Rejected (Missing Qualification)"
        elif similarity < threshold:
            final_status = "❌ Rejected (Low Score)"

        if final_status == "✅ Shortlisted":
            shutil.copy2(source_path, os.path.join("outputs/selected_cvs", filename))

        results.append({
            "Candidate": filename,
            "Score (%)": round(similarity * 100, 2),
            "Qualifications": qualifications,
            "Courses": courses,
            "Status": final_status,
            "Extracted Skills": skills,
            "Model Used": model_option
        })

    df = pd.DataFrame(results)
    df.to_csv(f"outputs/reports/report_{datetime.now().strftime('%Y%m%d_%H%M')}.csv", index=False)

    shortlisted = sum(1 for r in results if "Shortlisted" in r["Status"])

    summary = f"""
**✅ Screening Completed Successfully!**
**Model Used:** {model_option}  
**Total CVs Processed:** {len(files)}  
**Shortlisted:** {shortlisted}  
**Threshold:** {threshold*100}%
**Note:** Qualifications and Courses are now shown in separate columns.
"""

    zip_path = None
    if shortlisted > 0:
        zip_path = f"outputs/shortlisted_cvs_{datetime.now().strftime('%Y%m%d_%H%M')}.zip"
        with zipfile.ZipFile(zip_path, 'w', zipfile.ZIP_DEFLATED) as zipf:
            for f in os.listdir("outputs/selected_cvs"):
                zipf.write(os.path.join("outputs/selected_cvs", f), f)

    return results, summary.strip(), "outputs/selected_cvs/" if shortlisted > 0 else "", zip_path