app.py

import os
import glob
import gradio as gr
import numpy as np
import pandas as pd
from sklearn.metrics.pairwise import cosine_similarity
from sentence_transformers import SentenceTransformer
from PyPDF2 import PdfReader
import docx
import re
from collections import Counter

# -------------------------
# CONFIG
# -------------------------
DATASET_FOLDER = "resumes"  # Folder with 288 resumes
TOP_K = 3                   # Top 3 recommendations

# -------------------------
# HELPER FUNCTIONS
# -------------------------

def extract_text_from_pdf(file):
    text = ""
    try:
        reader = PdfReader(file)
        for page in reader.pages:
            page_text = page.extract_text()
            if page_text:
                text += page_text + " "
    except:
        pass
    return text

def extract_text_from_docx(file):
    text = ""
    try:
        doc = docx.Document(file)
        text = " ".join([para.text for para in doc.paragraphs])
    except:
        pass
    return text

def extract_text(file):
    ext = file.name.split('.')[-1].lower() if hasattr(file, "name") else "txt"
    if ext == "pdf":
        return extract_text_from_pdf(file)
    elif ext == "docx":
        return extract_text_from_docx(file)
    elif ext == "txt":
        try:
            file.seek(0)
            return file.read().decode("utf-8")
        except:
            return ""
    else:
        return ""

def load_resume_dataset(folder_path):
    resumes = []
    names = []
    paths = glob.glob(os.path.join(folder_path, "*"))
    for path in paths:
        text = ""
        ext = path.split('.')[-1].lower()
        try:
            if ext == "pdf":
                text = extract_text_from_pdf(path)
            elif ext == "docx":
                text = extract_text_from_docx(path)
            elif ext == "txt":
                with open(path, 'r', encoding='utf-8', errors='ignore') as f:
                    text = f.read()
        except:
            continue
        if text.strip():
            resumes.append(text)
            names.append(os.path.basename(path))
    return names, resumes

# -------------------------
# DYNAMIC JOB ROLE EXTRACTION
# -------------------------

def infer_job_from_text(text):
    """
    Extract probable job/role from resume text.
    Uses heuristics: first lines, capitalized phrases, or frequent nouns.
    """
    lines = text.split("\n")
    # Try first 5 lines for capitalized phrases
    candidate_lines = lines[:5]
    pattern = re.compile(r'\b[A-Z][a-zA-Z &/-]{2,}\b')
    roles = []
    for line in candidate_lines:
        matches = pattern.findall(line)
        roles.extend(matches)
    # Fallback: top 1 frequent capitalized word
    if roles:
        most_common = Counter(roles).most_common(1)
        return most_common[0][0]
    else:
        # fallback: "Other"
        return "Other"

# -------------------------
# LOAD MODEL & DATASET
# -------------------------

st_model = SentenceTransformer('all-MiniLM-L6-v2')
resume_names, resume_texts = load_resume_dataset(DATASET_FOLDER)
resume_embeddings = st_model.encode(resume_texts, convert_to_numpy=True)

# Build dynamic job roles dict
resume_roles = {name: infer_job_from_text(text) for name, text in zip(resume_names, resume_texts)}

# -------------------------
# MATCH FUNCTION
# -------------------------

def match_resume(file):
    input_text = extract_text(file)
    if not input_text.strip():
        return pd.DataFrame([{"Error": "Could not extract text from this resume."}])
    
    input_emb = st_model.encode([input_text], convert_to_numpy=True)
    sims = cosine_similarity(input_emb, resume_embeddings)[0]
    top_indices = sims.argsort()[-TOP_K:][::-1]
    
    results = []
    for idx in top_indices:
        matched_resume_name = resume_names[idx]
        similarity_score = sims[idx]
        recommended_job = resume_roles[matched_resume_name]
        results.append({
            "Matched Resume": matched_resume_name,
            "Recommended Job": recommended_job,
            "Confidence Score": f"{similarity_score*100:.2f}%"
        })
    return pd.DataFrame(results)

# -------------------------
# GRADIO UI
# -------------------------
css = """
body {background-color: #f7f9fc;}
h1 {color: #333; text-align: center;}
.gr-button {background-color: #4CAF50; color: white;}
"""

title = "<h1>AI Resume Analyzer & Job Matcher</h1>"

iface = gr.Interface(
    fn=match_resume,
    inputs=gr.File(label="Upload Your Resume (PDF, DOCX, TXT)"),
    outputs=gr.Dataframe(label="Top Job Matches"),
    title="AI Resume Analyzer & Job Matcher",
    description="Upload a resume to get top 3 job recommendations with confidence scores.",
    css=css,
)

if __name__ == "__main__":
    iface.launch()