import gradio as gr
import pandas as pd
import numpy as np
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
import matplotlib.dates as mdates
import seaborn as sns
from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer
import statsmodels.api as sm
from itertools import product as iter_product
import warnings
import io
warnings.filterwarnings("ignore")

# ============================================================
# LOAD DATA
# ============================================================
df_feedback = pd.read_csv("hotel_guest_feedback.csv")
df_bookings = pd.read_csv("synthetic_booking_data.csv")
df_reviews = pd.read_csv("synthetic_hotel_reviews.csv")
df_bookings["month"] = pd.to_datetime(df_bookings["month"])

# ============================================================
# VADER SENTIMENT
# ============================================================
analyzer = SentimentIntensityAnalyzer()

def get_sentiment_label(text):
    score = analyzer.polarity_scores(str(text))["compound"]
    if score >= 0.05:
        return "positive"
    elif score <= -0.05:
        return "negative"
    else:
        return "neutral"

df_reviews["vader_sentiment"] = df_reviews["review_text"].apply(get_sentiment_label)
df_reviews["vader_compound"] = df_reviews["review_text"].apply(
    lambda x: analyzer.polarity_scores(str(x))["compound"]
)

HOTEL_NAMES = sorted(df_bookings["hotel_name"].unique().tolist())

# ============================================================
# TAB 1: PIPELINE RUNNER
# ============================================================
def run_pipeline(step):
    log = []
    if step in ["Step 1: Data Loading", "Full Pipeline"]:
        log.append("✅ Step 1: Data loaded successfully")
        log.append(f"   - Guest feedback: {len(df_feedback)} rows")
        log.append(f"   - Bookings: {len(df_bookings)} rows")
        log.append(f"   - Reviews: {len(df_reviews)} rows")

    if step in ["Step 2: Sentiment Analysis", "Full Pipeline"]:
        pos = (df_reviews["vader_sentiment"] == "positive").sum()
        neu = (df_reviews["vader_sentiment"] == "neutral").sum()
        neg = (df_reviews["vader_sentiment"] == "negative").sum()
        log.append("✅ Step 2: VADER sentiment analysis complete")
        log.append(f"   - Positive: {pos} | Neutral: {neu} | Negative: {neg}")

    if step in ["Step 3: Pricing Decisions", "Full Pipeline"]:
        log.append("✅ Step 3: Pricing decisions generated")
        avg_bookings = df_bookings.groupby("hotel_name")["bookings"].mean()
        sent_ratios = df_reviews.groupby(["hotel_name", "vader_sentiment"]).size().unstack(fill_value=0)
        sent_ratios["total"] = sent_ratios.sum(axis=1)
        sent_ratios["positive_ratio"] = sent_ratios.get("positive", 0) / sent_ratios["total"]
        sent_ratios["negative_ratio"] = sent_ratios.get("negative", 0) / sent_ratios["total"]
        for hotel in HOTEL_NAMES:
            avg_b = avg_bookings.get(hotel, 0)
            pos_r = sent_ratios.loc[hotel, "positive_ratio"] if hotel in sent_ratios.index else 0
            neg_r = sent_ratios.loc[hotel, "negative_ratio"] if hotel in sent_ratios.index else 0
            if avg_b >= 200 and pos_r >= 0.6:
                action = "📈 Increase Price"
            elif avg_b <= 100 and neg_r >= 0.4:
                action = "📉 Decrease Price"
            else:
                action = "➡️ Keep Price"
            log.append(f"   {hotel}: {action}")

    return "\n".join(log)

# ============================================================
# TAB 2: DASHBOARD
# ============================================================
def generate_dashboard(chart_type):
    fig, ax = plt.subplots(figsize=(12, 6))

    if chart_type == "Revenue Trends":
        colors = sns.color_palette("tab10", len(HOTEL_NAMES))
        for i, hotel in enumerate(HOTEL_NAMES):
            data = df_bookings[df_bookings["hotel_name"] == hotel]
            ax.plot(data["month"], data["revenue"], label=hotel, color=colors[i], linewidth=1.5)
        ax.set_title("Monthly Revenue Trends by Hotel", fontsize=14)
        ax.set_ylabel("Revenue (€)")
        ax.legend(bbox_to_anchor=(1.05, 1), loc="upper left", fontsize=7)
        ax.xaxis.set_major_formatter(mdates.DateFormatter("%Y-%m"))
        plt.xticks(rotation=45)

    elif chart_type == "Sentiment Distribution":
        sent_counts = df_reviews.groupby(["hotel_name", "vader_sentiment"]).size().unstack(fill_value=0)
        sent_counts = sent_counts.reindex(columns=["positive", "neutral", "negative"], fill_value=0)
        sent_counts.plot(kind="bar", stacked=True, ax=ax, color=["#2ecc71", "#f1c40f", "#e74c3c"])
        ax.set_title("Sentiment Distribution by Hotel", fontsize=14)
        ax.set_ylabel("Number of Reviews")
        plt.xticks(rotation=45, ha="right")

    elif chart_type == "Bookings Trends":
        colors = sns.color_palette("tab10", len(HOTEL_NAMES))
        for i, hotel in enumerate(HOTEL_NAMES):
            data = df_bookings[df_bookings["hotel_name"] == hotel]
            ax.plot(data["month"], data["bookings"], label=hotel, color=colors[i], linewidth=1.5)
        ax.set_title("Monthly Bookings by Hotel", fontsize=14)
        ax.set_ylabel("Number of Bookings")
        ax.legend(bbox_to_anchor=(1.05, 1), loc="upper left", fontsize=7)
        ax.xaxis.set_major_formatter(mdates.DateFormatter("%Y-%m"))
        plt.xticks(rotation=45)

    elif chart_type == "Correlation Heatmap":
        hotel_avg_sent = df_reviews.groupby("hotel_name")["vader_compound"].mean().reset_index()
        hotel_avg_sent.columns = ["hotel_name", "avg_sentiment"]
        booking_summary = df_bookings.groupby("hotel_name").agg(
            avg_bookings=("bookings", "mean"),
            avg_rate=("avg_nightly_rate", "mean"),
            avg_revenue=("revenue", "mean")
        ).reset_index()
        merged = booking_summary.merge(hotel_avg_sent, on="hotel_name")
        sns.heatmap(merged[["avg_bookings", "avg_rate", "avg_revenue", "avg_sentiment"]].corr(),
                    annot=True, cmap="coolwarm", center=0, fmt=".2f", ax=ax)
        ax.set_title("Correlation Heatmap", fontsize=14)

    plt.tight_layout()
    return fig

# ============================================================
# TAB 3: AI DASHBOARD (Hotel Profile + Forecast)
# ============================================================
def hotel_profile(hotel_name):
    # Sentiment summary
    hotel_revs = df_reviews[df_reviews["hotel_name"] == hotel_name]
    total = len(hotel_revs)
    pos = (hotel_revs["vader_sentiment"] == "positive").sum()
    neu = (hotel_revs["vader_sentiment"] == "neutral").sum()
    neg = (hotel_revs["vader_sentiment"] == "negative").sum()
    avg_comp = hotel_revs["vader_compound"].mean()

    # Booking summary
    hotel_book = df_bookings[df_bookings["hotel_name"] == hotel_name]
    avg_bookings = hotel_book["bookings"].mean()
    avg_rate = hotel_book["avg_nightly_rate"].mean()
    avg_revenue = hotel_book["revenue"].mean()

    # Pricing decision
    pos_ratio = pos / total if total > 0 else 0
    neg_ratio = neg / total if total > 0 else 0
    if avg_bookings >= 200 and pos_ratio >= 0.6:
        action = "📈 INCREASE PRICE"
    elif avg_bookings <= 100 and neg_ratio >= 0.4:
        action = "📉 DECREASE PRICE"
    else:
        action = "➡️ KEEP PRICE"

    report = f"""
🏨 HOTEL PROFILE: {hotel_name}
{'='*50}

📊 SENTIMENT ({total} reviews):
   Positive: {pos} ({pos_ratio:.0%})
   Neutral:  {neu} ({neu/total:.0%})
   Negative: {neg} ({neg_ratio:.0%})
   Avg Score: {avg_comp:.3f}

📈 BOOKINGS (18 months):
   Avg Monthly Bookings: {avg_bookings:.0f}
   Avg Nightly Rate: €{avg_rate:.0f}
   Avg Monthly Revenue: €{avg_revenue:,.0f}

💰 PRICING RECOMMENDATION: {action}
"""

    # ARIMA forecast chart
    revenue_series = hotel_book.sort_values("month").set_index("month")["revenue"]
    fig, ax = plt.subplots(figsize=(12, 5))

    try:
        best_aic = float("inf")
        best_order = None
        best_model = None
        for p, d, q in iter_product(range(0, 4), range(0, 3), range(0, 2)):
            try:
                model = sm.tsa.ARIMA(revenue_series, order=(p, d, q))
                fitted = model.fit()
                if fitted.aic < best_aic:
                    best_aic = fitted.aic
                    best_order = (p, d, q)
                    best_model = fitted
            except:
                continue

        if best_model:
            forecast = best_model.forecast(steps=6)
            forecast_idx = pd.date_range(start=revenue_series.index[-1] + pd.DateOffset(months=1),
                                         periods=6, freq="ME")
            ax.plot(revenue_series.index, revenue_series.values, "b-", label="Historical")
            ax.plot(forecast_idx, forecast.values, "r--", label=f"Forecast ARIMA{best_order}")
            ax.set_title(f"Revenue Forecast — {hotel_name}", fontsize=14)
            ax.set_ylabel("Revenue (€)")
            ax.legend()
            plt.xticks(rotation=45)

            report += f"\n🔮 FORECAST (ARIMA{best_order}):\n"
            for date, val in zip(forecast_idx, forecast.values):
                report += f"   {date.strftime('%Y-%m')}: €{val:,.0f}\n"
    except Exception as e:
        report += f"\n⚠️ Forecast error: {str(e)}"
        ax.text(0.5, 0.5, "Forecast unavailable", ha="center", transform=ax.transAxes)

    plt.tight_layout()
    return report, fig

# ============================================================
# BUILD GRADIO APP
# ============================================================
with gr.Blocks(title="Hotel Pricing Optimizer", theme=gr.themes.Soft()) as app:
    gr.Markdown("""
    # 🏨 Hotel Pricing Optimizer
    *AI-powered luxury hotel pricing and service quality analysis*
    """)

    with gr.Tab("Pipeline Runner"):
        step_dropdown = gr.Dropdown(
            choices=["Step 1: Data Loading", "Step 2: Sentiment Analysis",
                     "Step 3: Pricing Decisions", "Full Pipeline"],
            label="Select Step", value="Full Pipeline"
        )
        run_btn = gr.Button("Run Pipeline", variant="primary")
        log_output = gr.Textbox(label="Execution Log", lines=20)
        run_btn.click(fn=run_pipeline, inputs=step_dropdown, outputs=log_output)

    with gr.Tab("Dashboard"):
        chart_dropdown = gr.Dropdown(
            choices=["Revenue Trends", "Sentiment Distribution",
                     "Bookings Trends", "Correlation Heatmap"],
            label="Select Chart", value="Revenue Trends"
        )
        chart_btn = gr.Button("Generate Chart", variant="primary")
        chart_output = gr.Plot(label="Chart")
        chart_btn.click(fn=generate_dashboard, inputs=chart_dropdown, outputs=chart_output)

    with gr.Tab("AI Dashboard"):
        hotel_dropdown = gr.Dropdown(choices=HOTEL_NAMES, label="Select Hotel", value=HOTEL_NAMES[0])
        profile_btn = gr.Button("Generate Hotel Profile & Forecast", variant="primary")
        profile_text = gr.Textbox(label="Hotel Profile", lines=25)
        forecast_plot = gr.Plot(label="Revenue Forecast")
        profile_btn.click(fn=hotel_profile, inputs=hotel_dropdown, outputs=[profile_text, forecast_plot])

app.launch()