streamlit_app.py

import streamlit as st
import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import numpy as np
import os

st.set_page_config(
    page_title="TruthLens — Fake News Detector",
    page_icon="🔍",
    layout="wide",
    initial_sidebar_state="collapsed"
)

st.markdown("""
<style>
@import url('https://fonts.googleapis.com/css2?family=Bebas+Neue&family=Outfit:wght@300;400;500;600&family=JetBrains+Mono:wght@400;500&display=swap');

*, *::before, *::after { box-sizing: border-box; margin: 0; padding: 0; }

html, body, [class*="css"], .stApp {
    background: #080810 !important;
    color: #e8e8f0 !important;
    font-family: 'Outfit', sans-serif !important;
}

.main .block-container { padding: 2rem 2rem 4rem !important; max-width: 900px !important; margin: 0 auto !important; }

#MainMenu, footer, header { visibility: hidden; }
.stDeployButton { display: none; }

h1, h2, h3 { font-family: 'Bebas Neue', sans-serif !important; letter-spacing: 2px !important; }

.stTextArea textarea {
    background: rgba(255,255,255,0.04) !important;
    border: 1px solid rgba(255,255,255,0.08) !important;
    border-radius: 14px !important;
    color: #e8e8f0 !important;
    font-family: 'Outfit', sans-serif !important;
    font-size: 15px !important;
    font-weight: 300 !important;
    line-height: 1.7 !important;
}

.stTextArea textarea:focus {
    border-color: rgba(255,200,0,0.3) !important;
    box-shadow: none !important;
}

.stTextArea textarea::placeholder { color: rgba(255,255,255,0.2) !important; }
.stTextArea label { display: none !important; }

.stButton > button {
    font-family: 'Bebas Neue', sans-serif !important;
    letter-spacing: 2px !important;
    border-radius: 12px !important;
    border: none !important;
    transition: all 0.2s !important;
}

.main-btn .stButton > button {
    background: #ffc800 !important;
    color: #080810 !important;
    width: 100% !important;
    font-size: 18px !important;
    padding: 16px !important;
}

.fake-btn .stButton > button {
    background: rgba(255,50,80,0.1) !important;
    color: #ff3250 !important;
    border: 1px solid rgba(255,50,80,0.2) !important;
    font-size: 12px !important;
    width: 100% !important;
}

.real-btn .stButton > button {
    background: rgba(0,220,130,0.1) !important;
    color: #00dc82 !important;
    border: 1px solid rgba(0,220,130,0.2) !important;
    font-size: 12px !important;
    width: 100% !important;
}

.stButton > button:hover { transform: translateY(-2px) !important; filter: brightness(1.1) !important; }

.verdict-fake {
    background: linear-gradient(135deg, rgba(255,30,60,0.08), rgba(255,30,60,0.03));
    border: 1px solid rgba(255,30,60,0.25);
    border-radius: 20px; padding: 28px; margin: 20px 0;
}

.verdict-real {
    background: linear-gradient(135deg, rgba(0,220,130,0.08), rgba(0,220,130,0.03));
    border: 1px solid rgba(0,220,130,0.25);
    border-radius: 20px; padding: 28px; margin: 20px 0;
}

.verdict-uncertain {
    background: linear-gradient(135deg, rgba(255,180,0,0.08), rgba(255,180,0,0.03));
    border: 1px solid rgba(255,180,0,0.25);
    border-radius: 20px; padding: 28px; margin: 20px 0;
}

.big-label {
    font-family: 'Bebas Neue', sans-serif;
    font-size: 52px;
    letter-spacing: 4px;
    line-height: 1;
}

.metric-box {
    background: rgba(255,255,255,0.03);
    border: 1px solid rgba(255,255,255,0.07);
    border-radius: 14px;
    padding: 18px;
    text-align: center;
}

.metric-num {
    font-family: 'Bebas Neue', sans-serif;
    font-size: 36px;
    letter-spacing: 1px;
    line-height: 1;
}

.metric-lbl {
    font-family: 'JetBrains Mono', monospace;
    font-size: 9px;
    color: rgba(255,255,255,0.3);
    letter-spacing: 2px;
    text-transform: uppercase;
    margin-top: 4px;
}

.signal-box {
    background: rgba(255,255,255,0.02);
    border: 1px solid rgba(255,255,255,0.06);
    border-radius: 14px;
    padding: 18px;
    margin-top: 16px;
}

.sig-row {
    display: flex;
    justify-content: space-between;
    align-items: center;
    padding: 8px 0;
    border-bottom: 1px solid rgba(255,255,255,0.04);
    font-size: 13px;
}

.sig-row:last-child { border-bottom: none; }

.badge-high { color: #ff1e3c; font-family: monospace; font-size: 11px; }
.badge-med  { color: #ffb400; font-family: monospace; font-size: 11px; }
.badge-low  { color: #00dc82; font-family: monospace; font-size: 11px; }

.footer-txt {
    font-family: 'JetBrains Mono', monospace;
    font-size: 11px;
    color: rgba(255,255,255,0.15);
    text-align: center;
    margin-top: 48px;
    padding-top: 20px;
    border-top: 1px solid rgba(255,255,255,0.05);
}

.mono { font-family: 'JetBrains Mono', monospace; }
</style>
""", unsafe_allow_html=True)

# ── Load model ───────────────────────────────────────────────
@st.cache_resource
def load_model():
    from transformers import pipeline
    # Using a well-trained public fake news detection model
    classifier = pipeline(
        "text-classification",
        model="hamzab/roberta-fake-news-classification",
        device=-1
    )
    return classifier

classifier = load_model()

# ── Predict ──────────────────────────────────────────────────
def predict(text):
    result = classifier(text, truncation=True, max_length=512)[0]
    label_raw = result['label'].upper()
    score = result['score']

    if 'FAKE' in label_raw or label_raw == 'LABEL_0':
        label = "FAKE"
        fake_prob = round(score * 100, 1)
        real_prob = round((1 - score) * 100, 1)
    else:
        label = "REAL"
        real_prob = round(score * 100, 1)
        fake_prob = round((1 - score) * 100, 1)

    confidence = round(score * 100, 1)
    risk_score = int(fake_prob)
    return label, confidence, fake_prob, real_prob, risk_score
    return label, confidence, round(fake_prob*100,1), round(real_prob*100,1), risk_score

def get_signals(text):
    lower = text.lower()
    signals = []
    cap = sum(1 for c in text if c.isupper()) / max(len(text),1)
    signals.append(("Capitalization", f"{cap*100:.0f}%", "HIGH" if cap>0.2 else "LOW"))
    excl = text.count('!')
    signals.append(("Exclamation marks", str(excl), "HIGH" if excl>=3 else ("MED" if excl else "LOW")))
    sens = ['shocking','breaking','secret','hidden','suppressed','miracle','exposed','bombshell']
    found = [w for w in sens if w in lower]
    signals.append(("Sensational words", str(len(found)), "HIGH" if len(found)>=2 else ("MED" if found else "LOW")))
    factual = ['according to','study','research','published','reported','announced']
    fc = [w for w in factual if w in lower]
    signals.append(("Factual language", f"{len(fc)} found", "LOW" if len(fc)>=2 else "HIGH"))
    return signals

# ── Samples ──────────────────────────────────────────────────
SAMPLES = {
    "fake1": "SHOCKING: Scientists CONFIRM that drinking lemon juice cures ALL cancers overnight! Big Pharma desperately hiding this miracle cure! Share before it gets deleted!!!",
    "fake2": "BREAKING: The moon landing was STAGED in Hollywood! Leaked NASA documents CONFIRM what conspiracy theorists have known. Government LYING to us for decades!!!",
    "real1": "The Intergovernmental Panel on Climate Change released its assessment report indicating global temperatures have risen 1.1 degrees Celsius above pre-industrial levels, compiled by 230 scientists.",
    "real2": "Apple reported quarterly revenue of $89.5 billion, with its services segment growing 16 percent year-over-year according to the company's official earnings call."
}

# ── UI ───────────────────────────────────────────────────────

# Header
st.markdown("""
<div style="display:flex;align-items:center;justify-content:space-between;
     padding-bottom:20px;border-bottom:1px solid rgba(255,255,255,0.06);margin-bottom:40px;">
    <div style="display:flex;align-items:center;gap:14px;">
        <div style="width:44px;height:44px;background:#ffc800;border-radius:12px;
             display:flex;align-items:center;justify-content:center;font-size:22px;
             box-shadow:0 0 30px rgba(255,200,0,0.3);">🔍</div>
        <div style="font-family:'Bebas Neue',sans-serif;font-size:28px;letter-spacing:3px;">
            TRUTH<span style="color:#ffc800;">LENS</span>
        </div>
    </div>
    <div style="font-family:'JetBrains Mono',monospace;font-size:10px;
         color:rgba(255,255,255,0.3);border:1px solid rgba(255,255,255,0.1);
         padding:4px 12px;border-radius:20px;">v1.0 · PORTFOLIO PROJECT</div>
</div>
""", unsafe_allow_html=True)

# Hero
st.markdown("""
<div style="text-align:center;margin-bottom:40px;">
    <div style="display:inline-flex;align-items:center;gap:8px;
         font-family:'JetBrains Mono',monospace;font-size:11px;color:#ffc800;
         letter-spacing:3px;text-transform:uppercase;margin-bottom:16px;
         padding:6px 16px;background:rgba(255,200,0,0.08);
         border:1px solid rgba(255,200,0,0.2);border-radius:99px;">
        ● AI-Powered Detection
    </div>
    <div style="font-family:'Bebas Neue',sans-serif;font-size:72px;
         letter-spacing:4px;line-height:0.95;color:#fff;margin-bottom:16px;">
        DETECT<br><span style="color:#ffc800;">FAKE</span> <span style="color:rgba(255,255,255,0.2);">/</span> REAL<br>NEWS
    </div>
    <p style="font-size:15px;font-weight:300;color:rgba(255,255,255,0.4);
       max-width:480px;margin:0 auto;line-height:1.7;">
        Fine-tuned DistilBERT model trained on 70,000+ articles.
        Paste any headline or article to instantly analyze it.
    </p>
</div>
""", unsafe_allow_html=True)

# Stats
st.markdown("""
<div style="display:flex;align-items:center;justify-content:center;gap:40px;
     margin-bottom:40px;padding:18px;background:rgba(255,255,255,0.02);
     border:1px solid rgba(255,255,255,0.05);border-radius:16px;">
    <div style="text-align:center;">
        <div style="font-family:'Bebas Neue',sans-serif;font-size:28px;
             letter-spacing:2px;color:#ffc800;">98.9%</div>
        <div style="font-family:'JetBrains Mono',monospace;font-size:9px;
             color:rgba(255,255,255,0.3);letter-spacing:2px;text-transform:uppercase;">Accuracy</div>
    </div>
    <div style="width:1px;height:32px;background:rgba(255,255,255,0.08);"></div>
    <div style="text-align:center;">
        <div style="font-family:'Bebas Neue',sans-serif;font-size:28px;
             letter-spacing:2px;color:#ffc800;">70K+</div>
        <div style="font-family:'JetBrains Mono',monospace;font-size:9px;
             color:rgba(255,255,255,0.3);letter-spacing:2px;text-transform:uppercase;">Training Articles</div>
    </div>
    <div style="width:1px;height:32px;background:rgba(255,255,255,0.08);"></div>
    <div style="text-align:center;">
        <div style="font-family:'Bebas Neue',sans-serif;font-size:28px;
             letter-spacing:2px;color:#ffc800;">0.989</div>
        <div style="font-family:'JetBrains Mono',monospace;font-size:9px;
             color:rgba(255,255,255,0.3);letter-spacing:2px;text-transform:uppercase;">F1 Score</div>
    </div>
    <div style="width:1px;height:32px;background:rgba(255,255,255,0.08);"></div>
    <div style="text-align:center;">
        <div style="font-family:'Bebas Neue',sans-serif;font-size:28px;
             letter-spacing:2px;color:#ffc800;">&lt;1s</div>
        <div style="font-family:'JetBrains Mono',monospace;font-size:9px;
             color:rgba(255,255,255,0.3);letter-spacing:2px;text-transform:uppercase;">Inference Time</div>
    </div>
</div>
""", unsafe_allow_html=True)

# Sample buttons
st.markdown('<p style="font-family:JetBrains Mono,monospace;font-size:10px;color:rgba(255,255,255,0.3);letter-spacing:2px;text-transform:uppercase;margin-bottom:10px;">Try a sample →</p>', unsafe_allow_html=True)

c1, c2, c3, c4 = st.columns(4)
with c1:
    st.markdown('<div class="fake-btn">', unsafe_allow_html=True)
    if st.button("🚨 Fake Sample 1"): st.session_state.text = SAMPLES["fake1"]
    st.markdown('</div>', unsafe_allow_html=True)
with c2:
    st.markdown('<div class="fake-btn">', unsafe_allow_html=True)
    if st.button("🚨 Fake Sample 2"): st.session_state.text = SAMPLES["fake2"]
    st.markdown('</div>', unsafe_allow_html=True)
with c3:
    st.markdown('<div class="real-btn">', unsafe_allow_html=True)
    if st.button("✅ Real Sample 1"): st.session_state.text = SAMPLES["real1"]
    st.markdown('</div>', unsafe_allow_html=True)
with c4:
    st.markdown('<div class="real-btn">', unsafe_allow_html=True)
    if st.button("✅ Real Sample 2"): st.session_state.text = SAMPLES["real2"]
    st.markdown('</div>', unsafe_allow_html=True)

st.markdown("<div style='height:12px'></div>", unsafe_allow_html=True)

# Input
text_input = st.text_area(
    "input",
    value=st.session_state.get('text', ''),
    height=160,
    placeholder="Paste any news article, headline, or suspicious text here..."
)

char_color = "rgba(255,200,0,0.7)" if len(text_input) > 0 else "rgba(255,255,255,0.2)"
st.markdown(f"<p style='font-family:JetBrains Mono,monospace;font-size:11px;color:{char_color};margin-top:6px;letter-spacing:1px;'>{len(text_input)} CHARACTERS</p>", unsafe_allow_html=True)

st.markdown("<div style='height:12px'></div>", unsafe_allow_html=True)

col_btn, _ = st.columns([1, 2])
with col_btn:
    st.markdown('<div class="main-btn">', unsafe_allow_html=True)
    analyze = st.button("⚡  ANALYZE NOW", use_container_width=True)
    st.markdown('</div>', unsafe_allow_html=True)

# ── Result ───────────────────────────────────────────────────
if analyze:
    if not text_input.strip():
        st.markdown("""<div style='background:rgba(255,30,60,0.08);border:1px solid rgba(255,30,60,0.2);
            border-radius:12px;padding:14px 18px;margin-top:16px;
            font-family:JetBrains Mono,monospace;font-size:12px;color:#ff1e3c;'>
            ⚠ Please enter some text to analyze.</div>""", unsafe_allow_html=True)
    else:
        with st.spinner("Analyzing with DistilBERT..."):
            label, confidence, fake_prob, real_prob, risk_score = predict(text_input)
            signals = get_signals(text_input)

        if label == "FAKE":
            vc, icon, color = "verdict-fake", "⚠️", "#ff1e3c"
            verdict_text = "LIKELY FAKE"
            desc = "High probability of misinformation detected"
        elif label == "REAL":
            vc, icon, color = "verdict-real", "✅", "#00dc82"
            verdict_text = "LIKELY REAL"
            desc = "Low misinformation indicators detected"
        else:
            vc, icon, color = "verdict-uncertain", "🔍", "#ffb400"
            verdict_text = "UNCERTAIN"
            desc = "Mixed signals — verify with trusted sources"

        st.markdown(f"""
        <div class="{vc}">
            <div style="font-size:48px;line-height:1;margin-bottom:8px;">{icon}</div>
            <div class="big-label" style="color:{color};">{verdict_text}</div>
            <div style="font-family:'JetBrains Mono',monospace;font-size:12px;
                 color:rgba(255,255,255,0.35);margin-top:6px;">{desc}</div>
            <div style="margin-top:20px;">
                <div style="display:flex;justify-content:space-between;
                     font-family:'JetBrains Mono',monospace;font-size:10px;
                     color:rgba(255,255,255,0.25);margin-bottom:6px;">
                    <span>RISK LEVEL</span><span>{risk_score}/100</span>
                </div>
                <div style="height:4px;background:rgba(255,255,255,0.06);border-radius:99px;overflow:hidden;">
                    <div style="height:100%;width:{risk_score}%;background:linear-gradient(90deg,{color},{color}aa);border-radius:99px;"></div>
                </div>
            </div>
        </div>
        """, unsafe_allow_html=True)

        # Metrics
        m1, m2, m3 = st.columns(3)
        with m1:
            st.markdown(f'<div class="metric-box"><div class="metric-num" style="color:{color};">{confidence:.1f}%</div><div class="metric-lbl">Confidence</div></div>', unsafe_allow_html=True)
        with m2:
            st.markdown(f'<div class="metric-box"><div class="metric-num" style="color:#ff1e3c;">{fake_prob:.1f}%</div><div class="metric-lbl">Fake Probability</div></div>', unsafe_allow_html=True)
        with m3:
            st.markdown(f'<div class="metric-box"><div class="metric-num" style="color:#00dc82;">{real_prob:.1f}%</div><div class="metric-lbl">Real Probability</div></div>', unsafe_allow_html=True)

        # Signals
        sig_rows = ""
        for name, val, risk in signals:
            badge = f'<span class="badge-{"high" if risk=="HIGH" else "med" if risk=="MED" else "low"}">{risk}</span>'
            sig_rows += f'<div class="sig-row"><span>{name}: <span style="color:rgba(255,255,255,0.3);font-family:monospace;font-size:11px;">{val}</span></span>{badge}</div>'

        st.markdown(f"""
        <div class="signal-box">
            <div style="font-family:'JetBrains Mono',monospace;font-size:10px;
                 color:rgba(255,255,255,0.3);letter-spacing:2px;text-transform:uppercase;
                 margin-bottom:14px;">Signal Breakdown</div>
            {sig_rows}
        </div>
        """, unsafe_allow_html=True)

# Footer
st.markdown("""
<div class="footer-txt">
    TRUTHLENS v1.0 · BUILT BY VISHAL · DISTILBERT · WELFAKE · 98.9% ACCURACY<br>
    Portfolio Project #1 of 5 · MSc AI Engineering Applicant
</div>
""", unsafe_allow_html=True)