import pandas as pd
import plotly.express as px
import plotly.graph_objects as go
import gradio as gr
import re

custom_theme = gr.themes.Base().set(
    body_background_fill="linear-gradient(120deg, #f9fcff 0%, #fef6ff 100%)",
    block_background_fill="white",
    block_shadow="0px 5px 15px rgba(0,0,0,0.06)",
    block_border_width="1px",
    block_border_color="#e6e6e6",
)





# Run predictions
results_df = run_batch_predictions(df, pipe)
# Create a new DataFrame without 'Clusters' and 'Recommendation'
export_df = results_df.drop(columns=[col for col in ['Clusters', 'Recommendation'] if col in results_df.columns])

# Save to CSV
export_df.to_csv('results_summary.csv', index=False)


# Add the constant column
results_df["RecordCount"] = 1


# Extract actions
''''
def extract_actions(html_text):
    if not isinstance(html_text, str):
        return ["No action"]
    text = html_text.replace("<br>", "\n")
    actions = re.findall(r"• (.+)", text)
    return actions if actions else ["No action"]
'''
# Drilldown DataFrame
drilldown_data = []
for _, row in results_df.iterrows():
    churn_pred = row["Churn Prediction"]
    prob= row["Probability"]
    clusters = row["Clusters"]
    if isinstance(clusters, dict):
        for cluster_name, rec_list in clusters.items():
            for rec in rec_list:
                drilldown_data.append({
                    "Churn Prediction": churn_pred,
                    "Cluster": cluster_name,
                    "Recommendation": rec,
                    "Probability": prob
                })

drilldown_df = pd.DataFrame(drilldown_data)

# Chart defaults
px.defaults.template = "plotly_white"
px.defaults.color_discrete_sequence = px.colors.qualitative.Set2

# Helper for numeric KPI cards
def make_numeric_card(title, value, suffix=""):
    fig = go.Figure(go.Indicator(
        mode="number",
        value=value,
        number={'font': {'size': 36}, 'suffix': suffix},
        title={'text': title, 'font': {'size': 16}}
    ))
    fig.update_layout(
        height=140, width=220,
        margin=dict(l=10, r=10, t=25, b=10),
        paper_bgcolor="white"
    )
    return fig

# Build Analytics Figures
def generate_analytics_figs(df):
    figs = {}
    figs["avg_tenure"] = make_numeric_card("Avg Tenure (Months)", df["Tenure in Months"].mean())
    figs["avg_charge"] = make_numeric_card("Avg Monthly Charge ($)", df["Monthly Charge"].mean(), suffix="$")
    figs["avg_data"] = make_numeric_card("Avg Monthly GB Download", df["Avg Monthly GB Download"].mean())

    figs["tenure_hist"] = px.histogram(df, x="Tenure in Months", nbins=20, title="Tenure Distribution")
    figs["age_hist"] = px.histogram(df, x="Age", nbins=20, title="Age Distribution")

    figs["contract_pie"] = px.pie(df, names="Contract", hole=0.4, title="Contract Types")
    figs["dependents_donut"] = px.pie(df, names="Number of Dependents", hole=0.4, title="Dependents Distribution")
    figs["senior_pie"] = px.pie(df, names="Senior Citizen", hole=0.4, title="Senior Citizen Ratio")
    figs["streaming_pie"] = px.pie(df, names="Streaming Movies", hole=0.4, title="Streaming Movies Adoption")
    figs["security_pie"] = px.pie(df, names="Online Security", hole=0.4, title="Online Security Subscription")
    figs["unlimited_pie"] = px.pie(df, names="Unlimited Data", hole=0.4, title="Unlimited Data Adoption")

    referrals_df = df["Number of Referrals"].value_counts().reset_index()
    referrals_df.columns = ["Number of Referrals", "count"]
    figs["referrals_bar"] = px.bar(
        referrals_df,
        x="Number of Referrals",
        y="count",
        title="Referrals Count Distribution",
        labels={"Number of Referrals": "Number of Referrals", "count": "Count"}
    )

    for f in figs.values():
        f.update_layout(margin=dict(l=20, r=20, t=40, b=20))

    return figs

analytics_figs = generate_analytics_figs(results_df)

# Drilldown Chart
def generate_drilldown_chart():
    if drilldown_df.empty:
        fig = px.treemap(names=["No data"], parents=[""], values=[1], title="No recommendations yet")
        return fig
    fig = px.treemap(
        drilldown_df,
        path=['Churn Prediction', 'Cluster', 'Recommendation'],
        values=None,
        title="Recommendations by Churn Prediction & Cluster",
    )
    fig.update_traces(
        textinfo='label+percent parent',
        hovertemplate="<b>%{label}</b><br>Customers: %{value}<br>Percentage of parent: %{percentParent:.1%}<extra></extra>"
    )
    fig.update_layout(
        margin=dict(l=20, r=20, t=50, b=20),
        title_font=dict(size=18)
    )
    return fig

# Customer Details


def get_customer_details(customer_id):
    row = results_df[results_df["Customer ID"] == customer_id].iloc[0]
    prediction = row["Churn Prediction"]
    clusters = row["Clusters"]
    prob= row["Probability"]
    # Convert decimal probability → percentage
    pct_value = float(prob) * 100    # numeric percentage, e.g. 82.3
    pct = f"{pct_value:.0f}%"        # string percentage, e.g. "82%"

    # Categorize risk
    if pct_value < 30:
        risk_level = "Low"
    elif pct_value < 60:
        risk_level = "Medium"
    else:
        risk_level = "High"

    # Corrected line
    churn_risk = f"{risk_level} ({pct})"


    # Build recommendations with cluster names
    if isinstance(clusters, dict):
        rec_lines = []
        for cluster_name, rec_list in clusters.items():
            for rec in rec_list:
                rec_lines.append(f"<b>{cluster_name}:</b> {rec}")
        recommendations_html = "<br>• ".join(rec_lines) if rec_lines else "No recommendations"
    else:
        recommendations_html = "No recommendations"

    selected_features = [
        'Contract', 'Tenure in Months', 'Age', 'Number of Referrals',
        'Monthly Charge', 'Number of Dependents', 'Senior Citizen',
        'Streaming Movies', 'Online Security', 'Avg Monthly GB Download',
        'Unlimited Data'
    ]
    
    # Table with customer features
    table_html = "<table style='width:100%; border-collapse: collapse;'>"
    table_html += "<tbody>"
    for col in selected_features:
        table_html += f"<tr><td style='padding:8px; border:1px solid #ddd; font-weight:600;'>{col}</td>"
        table_html += f"<td style='padding:8px; border:1px solid #ddd;'>{row[col]}</td></tr>"
    table_html += "</tbody></table>"

    # Build full HTML with recommendations **outside** the table
    full_html = (
        f"<b>Churn Risk:</b> {churn_risk}<br><br>"
        f"<b>Recommendations:</b><br>• {recommendations_html}<br><br>"
        f"<b>Customer Details:</b><br>{table_html}"
    )
    return full_html



# Gradio App
with gr.Blocks(
    theme=custom_theme,
    title="CRM Churn Dashboard",
    css="""
        /* FORCE LIGHT MODE */
        :root {
            --color-background: #ffffff !important;
            --color-border: #e6e6e6 !important;
            --color-text: #000000 !important;
            --body-background-fill: linear-gradient(120deg, #f9fcff 0%, #fef6ff 100%) !important;
    
            /* Override dark mode variables */
            --dark-mode: 0 !important;
            --neutral-900: #000000 !important;
            --neutral-800: #1a1a1a !important;
            --neutral-700: #2a2a2a !important;
            --neutral-600: #3a3a3a !important;
            --neutral-500: #777 !important;
            --neutral-400: #aaa !important;
            --neutral-300: #ccc !important;
            --neutral-200: #eee !important;
            --neutral-100: #fafafa !important;
        }

        body {
            background: linear-gradient(120deg, #f9fcff 0%, #fef6ff 100%) !important;
        }

        /* Force blocks to be light */
        .gr-block, .gr-panel, .gr-box, .gr-container {
            background: white !important;
            color: black !important;
        }

        /* Dropdowns, inputs, etc. */
        input, select, textarea {
            background: white !important;
            color: black !important;
            border: 1px solid #ddd !important;
        }
    """
) as demo:



    gr.Markdown("## 📊 CRM Analytics Dashboard")

    with gr.Tabs():
        # Actions Insights tab
        with gr.TabItem("Actions Overview"):
            gr.Markdown("### Drilldown of Recommendations by Cluster")
            drilldown_plot = gr.Plot(value=generate_drilldown_chart())

        # Analytics Dashboard tab (donut style, original font)
        with gr.TabItem("Analytics Dashboard"):
            gr.Markdown("### Customer Base Analytics Overview")
            with gr.Row():
                gr.Plot(value=analytics_figs["avg_tenure"])
                gr.Plot(value=analytics_figs["avg_charge"])
                gr.Plot(value=analytics_figs["avg_data"])
            with gr.Row():
                gr.Plot(value=analytics_figs["tenure_hist"])
                gr.Plot(value=analytics_figs["age_hist"])
            with gr.Row():
                gr.Plot(value=analytics_figs["contract_pie"])
                gr.Plot(value=analytics_figs["dependents_donut"])
            with gr.Row():
                gr.Plot(value=analytics_figs["senior_pie"])
                gr.Plot(value=analytics_figs["streaming_pie"])
            with gr.Row():
                gr.Plot(value=analytics_figs["security_pie"])
                gr.Plot(value=analytics_figs["unlimited_pie"])
                gr.Plot(value=analytics_figs["referrals_bar"])

        # Customer Insights tab
        with gr.TabItem("CRM Details"):
            gr.Markdown("### Customer-Level Predictions & Recommendations")
            customer_dropdown = gr.Dropdown(
                choices=results_df["Customer ID"].tolist(),
                value=results_df["Customer ID"].iloc[0],
                label="Select Customer"
            )
            customer_output = gr.HTML()
            customer_dropdown.change(
                fn=get_customer_details,
                inputs=customer_dropdown,
                outputs=customer_output
            )

demo.launch(share=True)