src/streamlit_app.py

import streamlit as st
import joblib
from textblob import TextBlob
import mammoth
import pdfplumber
import io
import re
import os

# --- 1. MODEL FUNCTION ---
def ekkok(text):
    try:
        words = TextBlob(str(text)).words
        return [word.lemmatize() for word in words]
    except:
        return str(text).split()

# --- PAGE CONFIG ---
st.set_page_config(page_title="AI Resume Analyzer", layout="centered", page_icon="🎯")

# --- SECTOR KEYWORDS ---
SECTOR_KEYWORDS = {
    "Hospitality & Management": ["hospitality", "restaurant", "hotel", "waiter", "bartender", "bar manager", "chef", "tourism"],
    "Marketing / Advertising": ["marketing", "advertising", "social media", "branding", "seo", "store manager", "salesman"],
    "Information Technology": ["software", "developer", "java", "python", "javascript", "cloud", "data science"],
    "Finance & Accounting": ["finance", "accounting", "audit", "banking", "tax", "budget"],
    "Human Resources": ["recruitment", "hr", "payroll", "onboarding", "talent acquisition"]
}

# --- FILE EXTRACTOR ---
def extract_text(uploaded_file):
    try:
        if uploaded_file.name.lower().endswith('.docx'):
            return mammoth.extract_raw_text(io.BytesIO(uploaded_file.getvalue())).value
        elif uploaded_file.name.lower().endswith('.pdf'):
            full_text = ""
            with pdfplumber.open(io.BytesIO(uploaded_file.getvalue())) as pdf:
                for page in pdf.pages:
                    page_text = page.extract_text()
                    if page_text: full_text += page_text + "\n"
            return full_text if full_text.strip() else "ERR_SCAN"
    except: return None

# --- LOAD ASSETS ---
@st.cache_resource
def load_assets():
    m_p, v_p = "best_model.pkl", "tfidf_vectorizer.pkl"
    if os.path.exists(m_p) and os.path.exists(v_p):
        try: return joblib.load(m_p), joblib.load(v_p)
        except: return None, None
    return None, None

model, vectorizer = load_assets()

# --- BILINGUAL TITLE (İNGİLİZCE & TÜRKÇE BAŞLIK) ---
st.markdown("""
    <div style="text-align: center;">
        <h1 style="color: #1E3A8A; margin-bottom: 0;">🎯 AI Resume Classifier / Akıllı CV Sınıflandırıcı</h1>
        <p style="color: #666; font-size: 1.1rem; margin-top: 5px;">Automated Department Prediction System / Otomatik Bölüm Tahmin Sistemi</p>
    </div>
""", unsafe_allow_html=True)

st.divider()

# --- INPUT SECTION ---

# 1. Metin Girişi
manual_input = st.text_area("✍️ Paste CV Text / CV Metnini Yapıştırın:", height=150, key="m_input")

# 2. Dosya Yükleme
uploaded_file = st.file_uploader("📂 Upload CV (PDF/DOCX) / Dosya Yükleyin:", type=['pdf', 'docx'], key="f_input")

# --- VALIDATION ---
final_cv_text = ""

if manual_input.strip() and uploaded_file:
    st.error("⚠️ Please use ONLY ONE method! Delete text OR remove file. / Lütfen SADECE BİR yöntem kullanın! Metni silin VEYA dosyayı kaldırın.")
elif manual_input.strip():
    final_cv_text = manual_input
elif uploaded_file:
    with st.spinner('Reading...'):
        res = extract_text(uploaded_file)
        if res == "ERR_SCAN":
            st.error("❌ This PDF is an image. Please paste text instead. / Bu PDF resimden oluşuyor, lütfen metni kopyalayıp kutuya yapıştırın.")
        elif res:
            final_cv_text = res
            st.success(f"✅ {uploaded_file.name} ready!")

st.divider()

# --- ANALYSIS BUTTON ---
if st.button("🚀 START ANALYSIS / ANALİZİ BAŞLAT", use_container_width=True):
    if not final_cv_text or len(final_cv_text.strip()) < 10:
        st.warning("⚠️ Please provide CV content! / Lütfen CV içeriği sağlayın!")
    else:
        with st.spinner('Processing...'):
            # Email Identity
            email_id = "NOT FOUND"
            match = re.search(r'([a-zA-Z0-9_.+-]+)@[a-zA-Z0-9-]+\.[a-zA-Z0-9-.]+', final_cv_text)
            if match: email_id = match.group(1).upper()

            # Prediction
            low_txt = final_cv_text.lower()
            scores = {s: sum(1 for k in kw if k in low_txt) for s, kw in SECTOR_KEYWORDS.items()}
            prediction = max(scores, key=scores.get)
            
            if scores[prediction] == 0 and model:
                try: prediction = model.predict(vectorizer.transform([final_cv_text]))[0]
                except: prediction = "Unclassified"

            st.balloons()
            st.success("### Results / Sonuçlar")
            c1, c2 = st.columns(2)
            c1.metric("Email Identity / E-posta", email_id)
            c2.metric("Department / Bölüm", prediction)

# --- FOOTER ---
st.markdown("<br><br><br><hr>", unsafe_allow_html=True)
footer_html = """
<div style="display: flex; justify-content: space-between; align-items: center; font-family: sans-serif;">
    <div style="text-align: left;">
        <h4 style="margin:0; color: #1E3A8A;">Developed by Data Science Dept.</h4>
        <p style="margin:0; font-size: 0.8rem; color: #666;">Advanced HR Analytics Solutions</p>
    </div>
    <div style="text-align: right;">
        <h4 style="margin:0; color: #1E3A8A;">Veri Bilimi Departmanı</h4>
        <p style="margin:0; font-size: 0.8rem; color: #666;">Gelişmiş İK Analitik Çözümleri</p>
    </div>
</div>
<div style="text-align: center; margin-top: 15px; border-top: 1px solid #eee; padding-top: 10px;">
    <span style="color: #444; font-size: 0.9rem;">Developer / Geliştirici: <b>EsmaTuğba MERGEN</b> | <b>v37.0</b></span>
</div>
"""
st.markdown(footer_html, unsafe_allow_html=True)