# app.py
"""
Yahoo Finance Sentiment Analysis with Gemma LLM
Hugging Face Space Application
"""

import gradio as gr
import pandas as pd
from datetime import datetime
from utils import YahooFinanceScraper, SentimentAnalyzer, LLMAnalyzer
from config import POPULAR_STOCKS
import plotly.graph_objects as go

# Initialize components
print("Initializing application...")
scraper = YahooFinanceScraper()
sentiment_analyzer = SentimentAnalyzer()
llm_analyzer = LLMAnalyzer()
print("Application ready!")


def analyze_stock_news(symbol: str, num_articles: int = 10):
    """
    Main function to analyze stock news
    
    Args:
        symbol: Stock ticker symbol
        num_articles: Number of articles to analyze
        
    Returns:
        Tuple of (summary, dataframe, chart, llm_insights)
    """
    try:
        # Fetch news
        articles = scraper.get_stock_news(symbol, num_articles)
        
        if not articles:
            return "No news found for this symbol.", None, None, "No articles to analyze."
        
        # Analyze sentiments
        sentiments = []
        for article in articles:
            text = f"{article['title']}. {article.get('summary', '')}"
            sentiment = sentiment_analyzer.analyze_comprehensive(text)
            sentiments.append(sentiment)
        
        # Generate LLM insights
        market_summary = llm_analyzer.summarize_news(articles)
        investment_insight = llm_analyzer.generate_investment_insight(symbol, articles, sentiments)
        
        # Create dataframe
        df_data = []
        for article, sentiment in zip(articles, sentiments):
            df_data.append({
                'Title': article['title'],
                'Publisher': article['publisher'],
                'Sentiment': sentiment['sentiment_label'],
                'Score': f"{sentiment['combined_score']:.3f}",
                'Confidence': f"{sentiment['confidence']:.2%}",
                'VADER': f"{sentiment['vader']['compound']:.3f}",
                'FinBERT +': f"{sentiment['finbert']['positive']:.3f}",
                'FinBERT -': f"{sentiment['finbert']['negative']:.3f}",
            })
        
        df = pd.DataFrame(df_data)
        
        # Create visualization
        sentiment_counts = df['Sentiment'].value_counts()
        fig = go.Figure(data=[
            go.Bar(
                x=sentiment_counts.index,
                y=sentiment_counts.values,
                marker_color=['#00cc66' if x=='Positive' else '#ff6666' if x=='Negative' else '#999999' 
                              for x in sentiment_counts.index]
            )
        ])
        fig.update_layout(
            title=f"Sentiment Distribution for {symbol}",
            xaxis_title="Sentiment",
            yaxis_title="Number of Articles",
            height=400
        )
        
        # Calculate statistics
        avg_score = sum(s['combined_score'] for s in sentiments) / len(sentiments)
        positive_pct = (sentiment_counts.get('Positive', 0) / len(sentiments)) * 100
        negative_pct = (sentiment_counts.get('Negative', 0) / len(sentiments)) * 100
        
        summary = f"""
## 📊 Analysis Summary for {symbol}

**Total Articles Analyzed:** {len(articles)}

**Sentiment Distribution:**
- 🟢 Positive: {sentiment_counts.get('Positive', 0)} ({positive_pct:.1f}%)
- 🔴 Negative: {sentiment_counts.get('Negative', 0)} ({negative_pct:.1f}%)
- ⚪ Neutral: {sentiment_counts.get('Neutral', 0)} ({100-positive_pct-negative_pct:.1f}%)

**Average Sentiment Score:** {avg_score:.3f}

**Overall Sentiment:** {"🟢 Positive" if avg_score > 0.05 else "🔴 Negative" if avg_score < -0.05 else "⚪ Neutral"}
        """
        
        llm_insights = f"""
## 🤖 AI-Generated Insights (Powered by Gemma)

### Market Summary:
{market_summary}

### Investment Perspective:
{investment_insight}

---
*Note: These insights are generated by AI and should not be considered as financial advice.*
        """
        
        return summary, df, fig, llm_insights
        
    except Exception as e:
        return f"Error: {str(e)}", None, None, "Error generating insights."


def analyze_single_headline(headline: str):
    """
    Analyze a single headline
    
    Args:
        headline: News headline text
        
    Returns:
        Analysis results
    """
    try:
        sentiment = sentiment_analyzer.analyze_comprehensive(headline)
        
        # Create a dummy article dict for LLM analysis
        article = {'title': headline, 'summary': ''}
        explanation = llm_analyzer.analyze_sentiment_context(article, sentiment)
        
        result = f"""
## Sentiment Analysis Results

**Headline:** {headline}

**Overall Sentiment:** {sentiment['sentiment_label']} (Score: {sentiment['combined_score']:.3f})
**Confidence:** {sentiment['confidence']:.2%}

### Detailed Scores:
- **VADER Compound:** {sentiment['vader']['compound']:.3f}
- **FinBERT Positive:** {sentiment['finbert']['positive']:.3%}
- **FinBERT Negative:** {sentiment['finbert']['negative']:.3%}
- **FinBERT Neutral:** {sentiment['finbert']['neutral']:.3%}

### AI Explanation:
{explanation}
        """
        
        return result
        
    except Exception as e:
        return f"Error analyzing headline: {str(e)}"


# Create Gradio Interface
with gr.Blocks(title="Yahoo Finance Sentiment Analyzer", theme=gr.themes.Soft()) as demo:
    gr.Markdown("""
    # 📈 Yahoo Finance Sentiment Analyzer
    ### Powered by FinBERT + Gemma LLM
    
    Analyze market sentiment from Yahoo Finance news using advanced NLP and AI.
    """)
    
    with gr.Tabs():
        # Tab 1: Stock Analysis
        with gr.Tab("📊 Stock Sentiment Analysis"):
            gr.Markdown("### Analyze sentiment of news for any stock symbol")
            
            with gr.Row():
                with gr.Column(scale=2):
                    stock_input = gr.Textbox(
                        label="Stock Symbol",
                        placeholder="e.g., AAPL, GOOGL, TSLA",
                        value="AAPL"
                    )
                with gr.Column(scale=1):
                    num_articles = gr.Slider(
                        minimum=5,
                        maximum=20,
                        value=10,
                        step=1,
                        label="Number of Articles"
                    )
            
            gr.Markdown("**Quick Select:**")
            quick_buttons = []
            with gr.Row():
                for stock in POPULAR_STOCKS[:5]:
                    btn = gr.Button(stock, size="sm")
                    quick_buttons.append(btn)
            with gr.Row():
                for stock in POPULAR_STOCKS[5:10]:
                    btn = gr.Button(stock, size="sm")
                    quick_buttons.append(btn)
            
            analyze_btn = gr.Button("🔍 Analyze News", variant="primary", size="lg")
            
            summary_output = gr.Markdown(label="Summary")
            insights_output = gr.Markdown(label="AI Insights")
            chart_output = gr.Plot(label="Sentiment Distribution")
            table_output = gr.Dataframe(label="Detailed Results")
            
            # Button actions
            analyze_btn.click(
                fn=analyze_stock_news,
                inputs=[stock_input, num_articles],
                outputs=[summary_output, table_output, chart_output, insights_output]
            )
            
            # Quick select buttons
            for btn in quick_buttons:
                btn.click(
                    fn=lambda x: x,
                    inputs=[btn],
                    outputs=[stock_input]
                )
        
        # Tab 2: Single Headline Analysis
        with gr.Tab("📰 Single Headline Analyzer"):
            gr.Markdown("### Analyze sentiment of a single news headline")
            
            headline_input = gr.Textbox(
                label="News Headline",
                placeholder="Enter a financial news headline...",
                lines=3
            )
            
            gr.Markdown("**Example Headlines:**")
            example_headlines = [
                "Apple reaches all-time high as iPhone sales surge",
                "Tesla stock plummets amid production concerns",
                "Fed maintains interest rates, markets remain stable"
            ]
            
            with gr.Row():
                for example in example_headlines:
                    gr.Button(example[:50] + "...", size="sm").click(
                        fn=lambda x: x,
                        inputs=[gr.Textbox(value=example, visible=False)],
                        outputs=[headline_input]
                    )
            
            analyze_headline_btn = gr.Button("🔍 Analyze Headline", variant="primary")
            headline_output = gr.Markdown(label="Analysis Results")
            
            analyze_headline_btn.click(
                fn=analyze_single_headline,
                inputs=[headline_input],
                outputs=[headline_output]
            )
        
        # Tab 3: About
        with gr.Tab("ℹ️ About"):
            gr.Markdown("""
            ## About This Application
            
            This application analyzes sentiment from Yahoo Finance news using multiple advanced techniques:
            
            ### 🛠️ Technologies Used:
            
            1. **VADER Sentiment Analysis**
               - Rule-based sentiment analysis
               - Good for general text sentiment
            
            2. **FinBERT**
               - BERT model fine-tuned for financial text
               - Specialized in financial sentiment analysis
               - Model: `ProsusAI/finbert`
            
            3. **Gemma LLM**
               - Google's Gemma language model
               - Generates human-like insights and summaries
               - Model: `google/gemma-2-2b-it`
            
            ### 📊 Features:
            
            - Real-time news scraping from Yahoo Finance
            - Multi-model sentiment analysis
            - AI-generated market insights
            - Interactive visualizations
            - Batch and single headline analysis
            
            ### 📝 Sentiment Scores:
            
            - **Positive**: Score > 0.05
            - **Negative**: Score < -0.05
            - **Neutral**: -0.05 ≤ Score ≤ 0.05
            
            ### ⚠️ Disclaimer:
            
            This tool is for educational and research purposes only. 
            The sentiment analysis and AI-generated insights should NOT be used as financial advice. 
            Always do your own research and consult with financial professionals before making investment decisions.
            
            ---
            
            **Created with ❤️ using Hugging Face Spaces**
            """)

# Launch the app
if __name__ == "__main__":
    demo.launch()