app.py

import streamlit as st
import pandas as pd
import numpy as np
from GoogleNews import GoogleNews
from transformers import pipeline
import plotly.express as px
import plotly.graph_objects as go
import dateparser
from datetime import datetime, timedelta

# --- SYSTEM CONFIG ---
st.set_page_config(
    page_title="VBI Terminal: Strategic AI", 
    page_icon="🛡️", 
    layout="wide",
    initial_sidebar_state="collapsed"
)

# --- PROFESSIONAL UI (Cyberpunk/Terminal Style) ---
st.markdown("""
    <style>
    /* Dark Deep Background */
    .stApp { background-color: #0f172a; color: #e2e8f0; }
    
    /* Metrics Cards */
    [data-testid="stMetric"] {
        background-color: #1e293b !important;
        border: 1px solid #334155 !important;
        border-left: 5px solid #0ea5e9 !important;
        border-radius: 8px;
        padding: 15px !important;
        box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.5);
    }
    [data-testid="stMetricLabel"] p { color: #94a3b8 !important; font-size: 0.9rem !important; }
    [data-testid="stMetricValue"] div { color: #f8fafc !important; font-size: 2rem !important; }
    
    /* Advisor Box */
    .advisor-box {
        background-color: #334155; 
        border: 1px solid #10b981; 
        border-radius: 10px; 
        padding: 20px; 
        margin-top: 20px;
        color: #f0fdf4;
    }
    .warning-box {
        background-color: #450a0a; 
        border: 1px solid #ef4444; 
        border-radius: 10px; 
        padding: 20px; 
        margin-top: 20px;
        color: #fef2f2;
    }
    
    /* Tables */
    .stDataFrame { border: 1px solid #334155; border-radius: 5px; }
    </style>
    """, unsafe_allow_html=True)

# --- AI ENGINE ---
@st.cache_resource
def load_neural_engine():
    """Load Zero-Shot Classification Model"""
    return pipeline("zero-shot-classification", model="facebook/bart-large-mnli")

analyzer = load_neural_engine()

# --- INTELLIGENCE ENGINE ---
def calculate_source_weight(source_name):
    """Assigns weight based on source credibility (Heuristic)"""
    source_name = str(source_name).lower()
    if any(x in source_name for x in ['reuters', 'bloomberg', 'forbes', 'tengrinews', 'zakon', 'kapital']):
        return 2.0  # High Impact Tier 1
    elif any(x in source_name for x in ['blog', 'reddit', 'twitter', 'post']):
        return 0.8  # Low Impact / High Noise
    return 1.0  # Standard

def fetch_intelligence(query, region, depth, mode):
    """Fetches real-time data using OSINT techniques."""
    lang_set = 'en'
    gn = GoogleNews(lang=lang_set, region=region)
    gn.clear()
    
    # Advanced Query Logic
    final_query = query
    if mode == "Social Buzz (Risk)":
        final_query = f'{query} AND ("scandal" OR "fail" OR "crash" OR "complaint" OR "leak" OR "reddit" OR "opinion")'
    elif mode == "Financial Intel":
        final_query = f'{query} AND ("stock" OR "profit" OR "loss" OR "quarter" OR "ipo" OR "revenue")'
    
    gn.search(final_query)
    results = gn.result()
    
    if len(results) < depth:
        gn.getpage(2)
        results += gn.result()
    
    clean_data = []
    seen_titles = set()
    
    for item in results[:depth]:
        title = item.get('title', '')
        if title in seen_titles: continue 
        seen_titles.add(title)
        
        raw_date = item.get('date', '')
        parsed_date = dateparser.parse(raw_date)
        if not parsed_date:
            parsed_date = datetime.now()
            
        source_label = item.get('media', 'Unknown Node')
        weight = calculate_source_weight(source_label)
        
        clean_data.append({
            "Timestamp": parsed_date,
            "Source": source_label,
            "Headline": title,
            "Link": item.get('link', '#'),
            "Weight": weight
        })
    return clean_data

# --- STRATEGIC ADVISOR MODULE ---
def generate_ai_advice(rep_index, top_risk, volatility, sentiment_ratio):
    """Generates text-based strategy based on metrics."""
    advice = []
    status = "STABLE"
    css_class = "advisor-box"
    
    # 1. Reputation Assessment
    if rep_index < 40:
        status = "CRITICAL"
        css_class = "warning-box"
        advice.append(f"🚨 **CRISIS MODE ACTIVATED:** Reputation Index ({round(rep_index)}%) is critically low. Immediate PR intervention required.")
    elif rep_index < 60:
        status = "VOLATILE"
        advice.append(f"⚠️ **Caution:** Brand sentiment is mixed ({round(rep_index)}%). Monitor closely.")
    else:
        advice.append(f"✅ **Healthy:** Strong market perception ({round(rep_index)}%). Focus on maintaining momentum.")

    # 2. Risk Specific Strategy
    if top_risk == "Legal/Compliance":
        advice.append("⚖️ **Legal Vector:** High volume of legal discussions detected. Prepare official statements regarding compliance/lawsuits immediately.")
    elif top_risk == "Technical Failure":
        advice.append("🔧 **Ops Vector:** Technical complaints are trending. Issue a transparency report on uptime/fixes to reassure customers.")
    elif top_risk == "Financial Risk":
        advice.append("📉 **Market Vector:** Financial anxiety detected. IR (Investor Relations) should release clarifying data.")
    elif top_risk == "PR Crisis":
        advice.append("📣 **PR Vector:** Viral negativity detected. Do not ignore. Use 'Recall & Reframe' strategy.")

    # 3. Volatility Check
    if volatility > 0.3:
        advice.append("🌊 **High Volatility:** Opinions are shifting rapidly. Avoid controversial statements for 48h.")

    return status, "\n\n".join(advice), css_class

# --- SIDEBAR ---
with st.sidebar:
    st.header("🛰️ VBI: COMMAND")
    target_region = st.selectbox("Geo-Node", ["KZ", "US", "GB", "RU"], index=0)
    source_mode = st.radio("Signal Mode:", ["Corporate News", "Social Buzz (Risk)", "Financial Intel"])
    scan_depth = st.slider("Depth", 10, 60, 30)
    st.divider()
    st.markdown("Created by **VBI Intelligence**")

# --- MAIN LAYOUT ---
st.title("🛡️ VBI: Strategic Brand Intelligence")
st.markdown("`v2.0` | Neural OSINT & Automated Strategy Generation")

col_search, col_btn = st.columns([4, 1])
with col_search:
    target_query = st.text_input("TARGET ENTITY:", placeholder="e.g. FlyArystan, Kaspi, KazMunayGas")
with col_btn:
    st.write("")
    st.write("")
    start_btn = st.button("🚀 INITIATE SCAN", use_container_width=True)

if start_btn and target_query:
    with st.spinner(f"🛰️ Intercepting signals for '{target_query}'..."):
        # 1. Fetch
        raw_data = fetch_intelligence(target_query, target_region, scan_depth, source_mode)
        
        if not raw_data:
            st.error("No signals detected. System Standby.")
        else:
            # 2. Process
            processed_data = []
            risk_vectors = ["Legal/Compliance", "Financial Risk", "Technical Failure", "Market Expansion", "PR Crisis", "Customer Service"]
            sentiment_cats = ["Positive", "Negative", "Neutral"]
            
            prog = st.progress(0)
            for i, item in enumerate(raw_data):
                # Neural Analysis
                risk_out = analyzer(item['Headline'], candidate_labels=risk_vectors)
                sent_out = analyzer(item['Headline'], candidate_labels=sentiment_cats)
                
                # Logic: If Risk Score is low, it might be just General News
                top_risk = risk_out['labels'][0]
                if risk_out['scores'][0] < 0.4:
                    top_risk = "General Noise"

                # Weighted Score Calculation
                sent_score = 1 if sent_out['labels'][0] == 'Positive' else -1 if sent_out['labels'][0] == 'Negative' else 0
                weighted_impact = sent_score * item['Weight']

                processed_data.append({
                    "Time": item['Timestamp'],
                    "Source": item['Source'],
                    "Headline": item['Headline'],
                    "Risk Category": top_risk,
                    "Risk Conf": risk_out['scores'][0],
                    "Sentiment": sent_out['labels'][0],
                    "Impact Score": weighted_impact
                })
                prog.progress((i + 1) / len(raw_data))
            prog.empty()

            df = pd.DataFrame(processed_data).sort_values(by='Time')

            # --- ANALYTICS ENGINE ---
            total_impact = df['Impact Score'].sum()
            total_weight = df['Impact Score'].abs().sum() if df['Impact Score'].abs().sum() != 0 else 1
            
            # Reputation Index (0-100)
            rep_index = 50 + ((total_impact / len(df)) * 50)
            rep_index = max(0, min(100, rep_index)) # Clamp 0-100

            # Volatility (Std Dev of Impact)
            volatility = df['Impact Score'].std()
            
            # Top Risk
            top_risk_cat = df['Risk Category'].value_counts().idxmax()
            
            # Advice Generation
            status, strategy_text, css_style = generate_ai_advice(rep_index, top_risk_cat, volatility, 0)

            # --- DASHBOARD ---
            
            # 1. STRATEGIC ADVISOR BLOCK
            st.markdown(f"""
            <div class="{css_style}">
                <h3>🧠 AI Strategic Advisor: {status}</h3>
                <div style='white-space: pre-wrap; font-size: 1.1rem;'>{strategy_text}</div>
            </div>
            """, unsafe_allow_html=True)
            
            st.divider()

            # 2. KEY METRICS
            m1, m2, m3, m4 = st.columns(4)
            m1.metric("Reputation Index", f"{round(rep_index, 1)}%", delta="Weighted Impact")
            m2.metric("Signal Volume", len(df))
            m3.metric("Primary Threat", top_risk_cat, delta_color="off")
            m4.metric("Volatility", f"{round(volatility, 2)}", help="Higher means less stable opinions")

            # 3. VISUALIZATIONS
            st.subheader("📊 Visual Intelligence")
            
            tab1, tab2 = st.tabs(["📈 Trend Dynamics", "🛡️ Risk Matrix"])
            
            with tab1:
                # Rolling Average Trend
                df['MA'] = df['Impact Score'].rolling(window=3).mean()
                fig_trend = go.Figure()
                fig_trend.add_trace(go.Scatter(x=df['Time'], y=df['Impact Score'], mode='markers', name='Raw Signal', marker=dict(color='#94a3b8')))
                fig_trend.add_trace(go.Scatter(x=df['Time'], y=df['MA'], mode='lines', name='Reputation Trend', line=dict(color='#38bdf8', width=3)))
                fig_trend.update_layout(title="Reputation Stability Over Time", template="plotly_dark", paper_bgcolor='rgba(0,0,0,0)', plot_bgcolor='rgba(0,0,0,0)')
                st.plotly_chart(fig_trend, use_container_width=True)

            with tab2:
                c1, c2 = st.columns(2)
                with c1:
                    fig_pie = px.pie(df, names='Sentiment', title="Sentiment Distribution", color='Sentiment',
                                     color_discrete_map={"Positive": "#10b981", "Negative": "#ef4444", "Neutral": "#64748b"})
                    fig_pie.update_layout(template="plotly_dark", paper_bgcolor='rgba(0,0,0,0)')
                    st.plotly_chart(fig_pie, use_container_width=True)
                with c2:
                    risk_counts = df['Risk Category'].value_counts().reset_index()
                    risk_counts.columns = ['Risk', 'Count']
                    fig_bar = px.bar(risk_counts, x='Count', y='Risk', orientation='h', title="Risk Vectors",
                                     color='Count', color_continuous_scale='Redor')
                    fig_bar.update_layout(template="plotly_dark", paper_bgcolor='rgba(0,0,0,0)')
                    st.plotly_chart(fig_bar, use_container_width=True)

            # 4. INTELLIGENCE LOG
            st.subheader("📡 Decrypted Signal Log")
            
            def highlight_risk(val):
                color = '#ef4444' if val in ['Legal/Compliance', 'Technical Failure', 'PR Crisis'] else '#cbd5e1'
                return f'color: {color}'

            display_df = df[['Time', 'Source', 'Headline', 'Risk Category', 'Sentiment', 'Impact Score']].copy()
            st.dataframe(
                display_df.style.map(highlight_risk, subset=['Risk Category']),
                use_container_width=True,
                height=400
            )