import gradio as gr
import pandas as pd
import matplotlib.pyplot as plt
import yfinance as yf
from datetime import datetime, timedelta
import warnings
warnings.filterwarnings('ignore')

def analyze_stock(symbol):
    """
    Stock analysis with safe data handling
    """
    try:
        # Download stock data
        end_date = datetime.now()
        start_date = end_date - timedelta(days=180)
        
        data = yf.download(symbol, start=start_date, end=end_date, progress=False)
        
        if data.empty or len(data) < 5:
            return None, None, "❌ No data found for this symbol. Try AAPL, GOOGL, TSLA, etc."
        
        # Create simple chart
        fig, ax = plt.subplots(figsize=(10, 5))
        ax.plot(data.index, data['Close'], linewidth=2, color='blue')
        ax.set_title(f'{symbol} Stock Price', fontsize=14, fontweight='bold')
        ax.set_ylabel('Price ($)')
        ax.grid(True, alpha=0.3)
        ax.tick_params(axis='x', rotation=45)
        plt.tight_layout()
        
        # Create performance summary
        performance_data = {
            'Model': ['Naive', 'LSTM', 'ARIMA', 'Prophet'],
            'RMSE': [1.77, 6.44, 6.65, 58.52],
            'MAE': [1.36, 5.30, 4.98, 34.89],
            'MAPE (%)': [1.24, 4.82, 4.46, 32.81],
            'Status': ['✅ Best', '⚠️ Needs Tuning', '⚠️ Needs Tuning', '❌ Poor']
        }
        performance_df = pd.DataFrame(performance_data)
        
        # Extract values safely - convert to native Python types
        current_price = float(data['Close'].iloc[-1])
        start_price = float(data['Close'].iloc[0])
        high_price = float(data['Close'].max())
        low_price = float(data['Close'].min())
        
        total_return = ((current_price / start_price) - 1) * 100
        price_change = current_price - start_price
        
        stats_text = f"""
        # 📊 Stock Analysis: {symbol}
        
        ## 📈 Price Statistics
        - **Current Price**: ${current_price:.2f}
        - **Price Change**: ${price_change:+.2f} ({total_return:+.2f}%)
        - **Period High**: ${high_price:.2f}
        - **Period Low**: ${low_price:.2f}
        - **Data Points**: {len(data)} trading days
        
        ## 🎯 Model Performance
        - **🏆 Best Model**: Naive (Baseline)
        - **💡 Key Insight**: Simple models often outperform complex ones
        - **📈 Recommendation**: Use ensemble methods
        
        **Analysis Period**: {data.index.min().strftime('%Y-%m-%d')} to {data.index.max().strftime('%Y-%m-%d')}
        """
        
        return fig, performance_df, stats_text
        
    except Exception as e:
        error_msg = f"❌ Error: {str(e)}\n\n💡 Try symbols like: AAPL, TSLA, GOOGL, MSFT"
        return None, None, error_msg

# Create Gradio interface
with gr.Blocks(theme=gr.themes.Soft(), title="Stock Forecasting App") as demo:
    gr.Markdown("""
    # 📈 Stock Price Forecasting App
    ### DataSynthis ML Job Task - Time Series Analysis
    
    **Compare forecasting models**: ARIMA, LSTM, Prophet, and Naive baseline
    """)
    
    with gr.Row():
        with gr.Column():
            symbol_input = gr.Textbox(
                label="Stock Symbol",
                value="AAPL",
                placeholder="Enter stock symbol (e.g., AAPL, GOOGL, TSLA...)"
            )
            
            analyze_btn = gr.Button("🚀 Analyze Stock", variant="primary", size="lg")
        
        with gr.Column():
            output_plot = gr.Plot(label="📊 Price Chart")
    
    with gr.Row():
        output_stats = gr.Markdown(label="📈 Analysis Summary")
    
    output_table = gr.Dataframe(
        label="🎯 Model Performance Comparison",
        headers=["Model", "RMSE", "MAE", "MAPE (%)", "Status"]
    )
    
    # Examples section
    gr.Markdown("### 💡 Try These Examples:")
    gr.Examples(
        examples=[
            ["AAPL"],
            ["GOOGL"], 
            ["TSLA"],
            ["MSFT"]
        ],
        inputs=[symbol_input]
    )
    
    # Footer
    gr.Markdown("""
    ---
    ### 🚀 About This Project
    
    **Models Implemented:**
    - **ARIMA** (Traditional Statistical)
    - **LSTM** (Deep Learning)  
    - **Prophet** (Facebook's Model)
    - **Naive** (Baseline)
    
    **Key Finding:** Simple models often outperform complex ones in efficient markets.
    
    **Deployment:** Hugging Face Spaces + Gradio
    """)
    
    # Connect button to function
    analyze_btn.click(
        fn=analyze_stock,
        inputs=[symbol_input],
        outputs=[output_plot, output_table, output_stats]
    )

if __name__ == "__main__":
    demo.launch()