app.py

import streamlit as st
import pandas as pd
import numpy as np
import plotly.express as px
import plotly.graph_objects as go
from sklearn.decomposition import PCA
import joblib
import os
from datetime import datetime

# Import paths
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
SCALED_DATA_PATH = os.path.join(BASE_DIR, "data", "processed", "scaled_rfm_data.pkl")
KMEANS_MODEL_PATH = os.path.join(BASE_DIR, "models", "kmeans_model.pkl")
CUSTOMER_SEGMENTS_PATH = os.path.join(BASE_DIR, "outputs", "customer_segments.csv")
SEGMENT_PRODUCTS_PATH = os.path.join(BASE_DIR, "outputs", "segment_products.csv")

# Set Page Config
st.set_page_config(
    page_title="SegmentX | Customer Intelligence Portal",
    page_icon="💎",
    layout="wide",
    initial_sidebar_state="expanded"
)

# --- Industry-Grade UI Refinement ---
st.markdown("""
<style>
    @import url('https://fonts.googleapis.com/css2?family=Inter:wght@300;400;600;700&display=swap');
    
    html, body, [class*="css"] {
        font-family: 'Inter', sans-serif;
    }
    
    .main {
        background-color: #0f172a;
        color: #f8fafc;
    }
    
    /* Stabilized Content Wrapper */
    .block-container {
        max-width: 1400px;
        padding: 2rem 5rem !important;
    }
    
    .stMetric {
        background: rgba(30, 41, 59, 0.7);
        backdrop-filter: blur(8px);
        padding: 24px;
        border-radius: 16px;
        border: 1px solid rgba(148, 163, 184, 0.1);
        box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1);
    }
    
    .stMetric label {
        color: #94a3b8 !important;
        font-weight: 500 !important;
    }
    
    h1, h2, h3 {
        color: #f8fafc;
        letter-spacing: -0.025em;
    }
    
    /* Brand Header */
    .brand-header {
        display: flex;
        align-items: center;
        gap: 12px;
        padding-bottom: 2rem;
        margin-bottom: 2rem;
        border-bottom: 1px solid rgba(148, 163, 184, 0.1);
    }
    
    .brand-tag {
        background: #3b82f6;
        color: white;
        padding: 4px 12px;
        border-radius: 20px;
        font-size: 0.8rem;
        font-weight: 600;
        text-transform: uppercase;
    }
</style>
""", unsafe_allow_html=True)

@st.cache_data
def load_data():
    if not os.path.exists(CUSTOMER_SEGMENTS_PATH):
        return None, None
    df = pd.read_csv(CUSTOMER_SEGMENTS_PATH, index_col='Customer ID')
    
    # Load raw cleaned data for time-series analysis
    RAW_CLEANED_PATH = os.path.join(BASE_DIR, "data", "processed", "cleaned_retail_data.csv")
    if os.path.exists(RAW_CLEANED_PATH):
        df_raw = pd.read_csv(RAW_CLEANED_PATH, parse_dates=['InvoiceDate'])
    else:
        df_raw = None
    return df, df_raw

@st.cache_resource
def load_model():
    if not os.path.exists(KMEANS_MODEL_PATH) or not os.path.exists(SCALED_DATA_PATH):
        return None, None
    model = joblib.load(KMEANS_MODEL_PATH)
    data_dict = joblib.load(SCALED_DATA_PATH)
    return model, data_dict

@st.cache_data
def get_pca_data(scaled_data, labels):
    pca = PCA(n_components=2)
    X_pca = pca.fit_transform(scaled_data)
    pca_df = pd.DataFrame(X_pca, columns=['PCA1', 'PCA2'], index=scaled_data.index)
    pca_df['Segment'] = labels
    return pca_df

def main():
    df, df_raw = load_data()
    model, data_dict = load_model()

    if df is None or model is None:
        st.error("Project data or models not found. Please run the pipeline scripts first.")
        return

    # Modern Sidebar
    st.sidebar.markdown("<h2 style='color:#3b82f6'>SegmentX</h2>", unsafe_allow_html=True)
    st.sidebar.markdown("---")
    page = st.sidebar.radio("Console Navigation", ["Overview", "Segment Profiles", "Customer Lookup"])
    
    segments_list = df['Segment'].unique().tolist()
    selected_segments = st.sidebar.multiselect("Global Segment Filter", segments_list, default=segments_list)
    df_filtered = df[df['Segment'].isin(selected_segments)]

    st.markdown("""
        <div class="brand-header">
            <span class="brand-tag">Intelligence Console</span>
            <h1 style="margin:0">Behavioral Portal <span style="color:#3b82f6; font-weight:300">v2.0</span></h1>
        </div>
    """, unsafe_allow_html=True)

    if page == "Overview":
        # Interactive Overview
        c1, c2, c3, c4 = st.columns(4)
        c1.metric("Revenue Impact", f"£{df_filtered['Monetary'].sum():,.0f}")
        c2.metric("Customer Scale", f"{len(df_filtered):,}")
        c3.metric("Retention Risk", f"{(len(df_filtered[df_filtered['Recency'] > 90]) / len(df_filtered) * 100):.1f}%")
        c4.metric("Avg. Order Value", f"£{df_filtered['Monetary'].mean():,.1f}")

        st.markdown("<br>", unsafe_allow_html=True)
        
        # Interactive Row 1
        r1_c1, r1_c2 = st.columns([1, 1.2])
        
        with r1_c1:
            st.markdown("### Segment Distribution")
            counts = df['Segment'].value_counts()
            fig = px.pie(
                values=counts.values, 
                names=counts.index, 
                hole=0.5,
                color_discrete_sequence=px.colors.sequential.ice_r
            )
            fig.update_layout(
                paper_bgcolor='rgba(0,0,0,0)', 
                plot_bgcolor='rgba(0,0,0,0)',
                font_color="#f8fafc",
                margin=dict(t=0, b=0, l=0, r=0),
                legend=dict(orientation="h", yanchor="bottom", y=-0.1, xanchor="center", x=0.5)
            )
            st.plotly_chart(fig, use_container_width=True)

        with r1_c2:
            st.markdown("### 2D Projection Topology")
            pca_df = get_pca_data(data_dict['rfm_scaled'], df['Segment'])
            fig = px.scatter(
                pca_df, x='PCA1', y='PCA2', color='Segment',
                opacity=0.6,
                color_discrete_sequence=px.colors.sequential.ice_r,
                hover_data=[pca_df.index]
            )
            fig.update_layout(
                paper_bgcolor='rgba(0,0,0,0)', 
                plot_bgcolor='rgba(15, 23, 42, 0.5)',
                font_color="#f8fafc",
                margin=dict(t=10, b=10, l=10, r=10),
                xaxis=dict(showgrid=False),
                yaxis=dict(showgrid=False)
            )
            st.plotly_chart(fig, use_container_width=True)

        st.markdown("---")
        st.markdown("### 📈 Revenue Benchmarking")
        if df_raw is not None:
            df_raw['Month'] = df_raw['InvoiceDate'].dt.to_period('M').astype(str)
            df_raw['Revenue'] = df_raw['Quantity'] * df_raw['Price']
            monthly_rev = df_raw.groupby('Month')['Revenue'].sum().reset_index()
            
            fig = px.line(
                monthly_rev, x='Month', y='Revenue',
                color_discrete_sequence=['#3b82f6'],
                render_mode='svg'
            )
            fig.update_layout(
                paper_bgcolor='rgba(0,0,0,0)', 
                plot_bgcolor='rgba(15, 23, 42, 0.5)',
                font_color="#f8fafc",
                xaxis_title=None,
                yaxis_title="Total Revenue (GBP)",
                margin=dict(t=20, b=20, l=20, r=20)
            )
            fig.update_traces(line_width=3, fill='tozeroy', fillcolor='rgba(59, 130, 246, 0.1)')
            st.plotly_chart(fig, use_container_width=True)

        st.markdown("<br>", unsafe_allow_html=True)
        st.markdown("### 📊 Revenue Concentration (Pareto)")
        seg_rev = df_filtered.groupby('Segment')['Monetary'].sum().sort_values(ascending=False).reset_index()
        fig_bar = px.bar(
            seg_rev, x='Segment', y='Monetary',
            color='Monetary',
            color_continuous_scale='ice'
        )
        fig_bar.update_layout(paper_bgcolor='rgba(0,0,0,0)', plot_bgcolor='rgba(15, 23, 42, 0.5)', font_color="#f8fafc")
        st.plotly_chart(fig_bar, use_container_width=True)

        st.markdown("---")
        st.subheader("🚨 Risk Analytics")
        high_risk = len(df_filtered[df_filtered['Recency'] > 90])
        risk_pct = (high_risk / len(df_filtered)) * 100
        if risk_pct > 30:
            st.warning(f"**Critical Warning**: {risk_pct:.1f}% of selected customers are churn-risk (90+ days inactive).")
        else:
            st.success(f"**Healthy Signal**: Retention is stable with only {risk_pct:.1f}% churn-risk.")

        st.markdown("<br>", unsafe_allow_html=True)
        st.subheader("📥 Data Export & Actions")
        csv = df_filtered.to_csv().encode('utf-8')
        st.download_button("Export Intelligence Report (CSV)", data=csv, file_name='segmentx_report.csv', mime='text/csv')

    elif page == "Segment Profiles":
        st.subheader("Cluster Behavioral Heatmap")
        profile_stats = df.groupby('Segment')[['Recency', 'Frequency', 'Monetary']].mean()
        profile_norm = (profile_stats - profile_stats.min()) / (profile_stats.max() - profile_stats.min())
        
        fig = px.imshow(
            profile_norm.T,
            labels=dict(x="Segment", y="Metric", color="Score"),
            color_continuous_scale='Blues'
        )
        fig.update_layout(paper_bgcolor='rgba(0,0,0,0)', font_color="#f8fafc")
        st.plotly_chart(fig, use_container_width=True)
        
        st.markdown("#### Mean Values Matrix")
        st.table(profile_stats.style.format(lambda x: f"£{x:,.2f}" if x > 100 else f"{x:.2f}"))

        st.markdown("---")
        st.subheader("🛍️ Segment Affinity: Top 10 Products")
        if os.path.exists(SEGMENT_PRODUCTS_PATH):
            all_top_prods = pd.read_csv(SEGMENT_PRODUCTS_PATH)
            display_segs = selected_segments[:3]
            cols = st.columns(len(display_segs)) if display_segs else [st.container()]
            
            for i, seg in enumerate(display_segs):
                with cols[i]:
                    st.markdown(f"**{seg}**")
                    seg_prods = all_top_prods[all_top_prods['Segment'] == seg].head(10)
                    if not seg_prods.empty:
                        seg_prods['Description'] = seg_prods['Description'].str.slice(0, 30) + '...'
                        st.table(seg_prods[['Description', 'Quantity']].set_index('Description'))
                    else: st.info("No data.")
        else:
            st.info("Run product pipeline to see affinities.")

    elif page == "Customer Lookup":
        st.subheader("🔍 Intelligent Query")
        
        if st.button("🎲 Randomized ID Picker"):
            random_id = np.random.choice(df.index)
            st.session_state.customer_lookup_id = int(random_id)
        
        all_ids = sorted(df.index.unique().tolist())
        if 'customer_lookup_id' not in st.session_state:
            st.session_state.customer_lookup_id = all_ids[0]
            
        customer_id = st.selectbox(
            "Target Customer ID", 
            options=all_ids, 
            index=all_ids.index(st.session_state.customer_lookup_id) if st.session_state.customer_lookup_id in all_ids else 0
        )
        st.session_state.customer_lookup_id = customer_id
        
        cust_data = df.loc[customer_id]
        
        l1, l2, l3 = st.columns(3)
        l1.metric("Segment Identity", cust_data['Segment'])
        l2.metric("Orders", f"{cust_data['Frequency']:.0f}")
        l3.metric("LTV GBP", f"£{cust_data['Monetary']:,.2f}")
        
        st.markdown("---")
        st.markdown("### 🛡️ Strategic Intelligence")
        
        ci1, ci2 = st.columns(2)
        # Churn Probability Logic
        avg_rec = df['Recency'].mean()
        churn_prob = 1 - np.exp(-cust_data['Recency'] / (avg_rec * 1.5))
        churn_pct = min(max(churn_prob * 100, 0), 100)
        ci1.metric("Churn Risk Score", f"{churn_pct:.1f}%")
        
        # Predicted CLV
        avg_order = cust_data['Monetary'] / cust_data['Frequency']
        projected_clv = cust_data['Monetary'] + (avg_order * cust_data['Frequency'])
        ci2.metric("Projected 1Y-LTV", f"£{projected_clv:,.2f}")

        st.markdown("<br>", unsafe_allow_html=True)
        recommendations = {
            "Champions": "High Value, Low Churn. Goal: Retention. Strategy: Early Access, Loyalty Rewards.",
            "Loyal Customers": "Consistent Value. Goal: Growth. Strategy: Cross-sell related categories.",
            "At-Risk": "Recent Inactivity. Goal: Re-activation. Strategy: Limited-time win-back discounts.",
            "Lost/Hibernating": "Historical only. Goal: Win-back or Pause. Strategy: Reactivate only high LTV types."
        }
        st.info(f"**Execution Strategy**: {recommendations.get(cust_data['Segment'], 'Maintain baseline engagement.')}")

if __name__ == "__main__":
    main()