app.py

import gradio as gr
import requests
import json
import pandas as pd
from datetime import datetime, timedelta
import yfinance as yf
import numpy as np
from typing import Dict, List, Optional
import time
import os
import google.generativeai as genai
from textblob import TextBlob
import re
from concurrent.futures import ThreadPoolExecutor, as_completed
import asyncio
import aiohttp
import random
from io import BytesIO
import base64
import tempfile
import speech_recognition as sr
from gtts import gTTS
import pygame
import io

# Configure Gemini API
GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")
if GEMINI_API_KEY:
    genai.configure(api_key=GEMINI_API_KEY)
    model = genai.GenerativeModel('gemini-2.0-flash-exp')

class APIAgent:
    """Handles real-time market data retrieval with better error handling"""
    
    def __init__(self):
        self.session = requests.Session()
        self.session.headers.update({
            'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36'
        })
        # Fallback data for demo purposes
        self.fallback_data = {
            'AAPL': {'price': 175.84, 'change': 2.1},
            'GOOGL': {'price': 142.56, 'change': -0.8},
            'MSFT': {'price': 378.85, 'change': 1.5},
            'NVDA': {'price': 875.28, 'change': 3.2},
            'TSM': {'price': 92.45, 'change': -1.1},
            'ASML': {'price': 756.32, 'change': 0.7}
        }
    
    def get_stock_data(self, symbol: str, period: str = "1d") -> Dict:
        """Fetch stock data with multiple fallback methods"""
        try:
            # Method 1: Try yfinance with better error handling
            ticker = yf.Ticker(symbol)
            
            # Add delay to avoid rate limiting
            time.sleep(0.5)
            
            # Try to get basic info first
            try:
                info = ticker.info
                current_price = info.get('currentPrice') or info.get('regularMarketPrice', 0)
                prev_close = info.get('previousClose', current_price)
                
                if current_price and current_price > 0:
                    change_percent = ((current_price - prev_close) / prev_close) * 100 if prev_close else 0
                    
                    return {
                        'symbol': symbol,
                        'current_price': round(float(current_price), 2),
                        'change_percent': round(change_percent, 2),
                        'volume': info.get('volume', 0),
                        'market_cap': info.get('marketCap', 'N/A'),
                        'pe_ratio': info.get('trailingPE', 'N/A'),
                        'source': 'yfinance_info'
                    }
            except:
                pass
            
            # Method 2: Try historical data
            try:
                hist = ticker.history(period="5d")
                if not hist.empty:
                    current_price = hist['Close'].iloc[-1]
                    prev_price = hist['Close'].iloc[-2] if len(hist) > 1 else current_price
                    change_percent = ((current_price - prev_price) / prev_price) * 100 if prev_price else 0
                    
                    return {
                        'symbol': symbol,
                        'current_price': round(float(current_price), 2),
                        'change_percent': round(change_percent, 2),
                        'volume': int(hist['Volume'].iloc[-1]) if 'Volume' in hist.columns else 0,
                        'market_cap': 'N/A',
                        'pe_ratio': 'N/A',
                        'source': 'yfinance_history'
                    }
            except:
                pass
            
        except Exception as e:
            print(f"yfinance failed for {symbol}: {e}")
        
        # Method 3: Use fallback data with some randomization for demo
        if symbol in self.fallback_data:
            base_data = self.fallback_data[symbol]
            # Add some random variation to make it look live
            price_variation = random.uniform(-0.02, 0.02)
            change_variation = random.uniform(-0.5, 0.5)
            
            return {
                'symbol': symbol,
                'current_price': round(base_data['price'] * (1 + price_variation), 2),
                'change_percent': round(base_data['change'] + change_variation, 2),
                'volume': random.randint(1000000, 50000000),
                'market_cap': f"${random.randint(500, 3000)}B",
                'pe_ratio': round(random.uniform(15, 35), 1),
                'source': 'fallback_demo'
            }
        
        # Method 4: Return error case
        return {
            'symbol': symbol,
            'current_price': 0,
            'change_percent': 0,
            'volume': 0,
            'market_cap': 'N/A',
            'pe_ratio': 'N/A',
            'error': f'Unable to fetch data for {symbol}',
            'source': 'error'
        }
    
    def get_multiple_stocks(self, symbols: List[str]) -> List[Dict]:
        """Fetch data for multiple stocks with better concurrency control"""
        results = []
        
        # Sequential processing to avoid rate limits
        for symbol in symbols:
            try:
                result = self.get_stock_data(symbol)
                results.append(result)
                # Small delay between requests
                time.sleep(0.3)
            except Exception as e:
                results.append({
                    'symbol': symbol,
                    'error': str(e),
                    'source': 'exception'
                })
        
        return results

class ScrapingAgent:
    """Handles news and sentiment scraping with better reliability"""
    
    def __init__(self):
        self.session = requests.Session()
        self.session.headers.update({
            'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36'
        })
        # Fallback news for demo
        self.fallback_news = [
            {
                'title': 'Tech Stocks Rally on AI Optimism',
                'summary': 'Major technology stocks gained ground as investors showed renewed confidence in artificial intelligence developments and cloud computing growth prospects.',
                'publisher': 'Market News',
                'symbol': 'TECH',
                'sentiment': 'Positive'
            },
            {
                'title': 'Semiconductor Demand Remains Strong',
                'summary': 'Global semiconductor companies report continued strong demand driven by AI chips and data center expansion, despite geopolitical concerns.',
                'publisher': 'Tech Today',
                'symbol': 'SEMI',
                'sentiment': 'Positive'
            },
            {
                'title': 'Market Volatility Expected Ahead of Earnings',
                'summary': 'Analysts warn of potential market volatility as major tech companies prepare to report quarterly earnings amid mixed economic signals.',
                'publisher': 'Financial Times',
                'symbol': 'MARKET',
                'sentiment': 'Neutral'
            }
        ]
    
    def get_market_news(self, query: str = "tech stocks") -> List[Dict]:
        """Get market news with fallback to demo data"""
        news_items = []
        
        # Try to get real news from yfinance
        search_terms = ["AAPL", "GOOGL", "MSFT", "NVDA"]
        
        for symbol in search_terms[:2]:  # Limit to avoid rate limits
            try:
                ticker = yf.Ticker(symbol)
                time.sleep(0.5)  # Rate limiting
                news = ticker.news[:1]  # Get latest 1 news item
                
                for item in news:
                    news_items.append({
                        'title': item.get('title', 'No title'),
                        'summary': item.get('summary', 'No summary')[:150] + "...",
                        'publisher': item.get('publisher', 'Unknown'),
                        'symbol': symbol,
                        'sentiment': self.analyze_sentiment(item.get('title', '') + ' ' + item.get('summary', ''))
                    })
                    
            except Exception as e:
                print(f"News fetch failed for {symbol}: {e}")
                continue
        
        # Add fallback news if we don't have enough real news
        while len(news_items) < 3:
            remaining_fallback = [n for n in self.fallback_news if n not in news_items]
            if remaining_fallback:
                news_items.append(random.choice(remaining_fallback))
            else:
                break
        
        return news_items[:5]
    
    def analyze_sentiment(self, text: str) -> str:
        """Enhanced sentiment analysis"""
        try:
            # Check for specific keywords first
            positive_words = ['rally', 'gain', 'surge', 'optimism', 'strong', 'growth', 'beat', 'exceed']
            negative_words = ['fall', 'drop', 'decline', 'concern', 'weak', 'miss', 'disappoint', 'volatility']
            
            text_lower = text.lower()
            pos_count = sum(1 for word in positive_words if word in text_lower)
            neg_count = sum(1 for word in negative_words if word in text_lower)
            
            if pos_count > neg_count:
                return "Positive"
            elif neg_count > pos_count:
                return "Negative"
            
            # Fallback to TextBlob
            blob = TextBlob(text)
            polarity = blob.sentiment.polarity
            
            if polarity > 0.1:
                return "Positive"
            elif polarity < -0.1:
                return "Negative"
            else:
                return "Neutral"
                
        except Exception as e:
            return "Neutral"

class RetrieverAgent:
    """Enhanced data indexing and retrieval"""
    
    def __init__(self):
        self.knowledge_base = {}
        self.embeddings_cache = {}
    
    def index_data(self, data: Dict, category: str):
        """Improved in-memory indexing with timestamps"""
        if category not in self.knowledge_base:
            self.knowledge_base[category] = []
        
        self.knowledge_base[category].append({
            'timestamp': datetime.now(),
            'data': data,
            'id': f"{category}_{len(self.knowledge_base[category])}"
        })
        
        # Keep only last 50 entries per category
        if len(self.knowledge_base[category]) > 50:
            self.knowledge_base[category] = self.knowledge_base[category][-50:]
    
    def retrieve_relevant_data(self, query: str, top_k: int = 5) -> List[Dict]:
        """Enhanced retrieval with better matching"""
        relevant_data = []
        query_words = set(query.lower().split())
        
        for category, entries in self.knowledge_base.items():
            for entry in entries[-10:]:  # Get recent entries
                data_str = str(entry['data']).lower()
                data_words = set(data_str.split())
                
                # Calculate simple word overlap score
                overlap = len(query_words.intersection(data_words))
                if overlap > 0:
                    relevant_data.append({
                        'category': category,
                        'data': entry['data'],
                        'timestamp': entry['timestamp'],
                        'relevance_score': overlap
                    })
        
        # Sort by relevance and recency
        relevant_data.sort(key=lambda x: (x['relevance_score'], x['timestamp']), reverse=True)
        return relevant_data[:top_k]

class AnalysisAgent:
    """Enhanced quantitative analysis with better metrics"""
    
    def __init__(self):
        self.metrics_cache = {}
        self.risk_thresholds = {
            'low': 1.5,
            'medium': 3.0,
            'high': 5.0
        }
    
    def calculate_portfolio_metrics(self, stocks_data: List[Dict]) -> Dict:
        """Enhanced portfolio analysis"""
        try:
            valid_stocks = [s for s in stocks_data if 'error' not in s and s.get('current_price', 0) > 0]
            
            if not valid_stocks:
                return {
                    'error': 'No valid stock data available',
                    'total_stocks': 0,
                    'data_quality': 'Poor'
                }
            
            # Calculate comprehensive metrics
            prices = [s.get('current_price', 0) for s in valid_stocks]
            changes = [s.get('change_percent', 0) for s in valid_stocks]
            
            total_value = sum(prices)
            positive_movers = len([c for c in changes if c > 0])
            negative_movers = len([c for c in changes if c < 0])
            neutral_movers = len(valid_stocks) - positive_movers - negative_movers
            
            avg_change = np.mean(changes) if changes else 0
            volatility = np.std(changes) if len(changes) > 1 else 0
            max_gain = max(changes) if changes else 0
            max_loss = min(changes) if changes else 0
            
            # Risk assessment
            if volatility <= self.risk_thresholds['low']:
                risk_level = 'Low'
                risk_color = '🟢'
            elif volatility <= self.risk_thresholds['medium']:
                risk_level = 'Medium'
                risk_color = '🟡'
            else:
                risk_level = 'High'
                risk_color = '🔴'
            
            # Data quality assessment
            sources = [s.get('source', 'unknown') for s in valid_stocks]
            real_data_count = len([s for s in sources if s not in ['fallback_demo', 'error']])
            data_quality = 'Good' if real_data_count > len(valid_stocks) * 0.7 else 'Mixed' if real_data_count > 0 else 'Demo'
            
            return {
                'total_stocks': len(valid_stocks),
                'positive_movers': positive_movers,
                'negative_movers': negative_movers,
                'neutral_movers': neutral_movers,
                'avg_change_percent': round(avg_change, 2),
                'volatility': round(volatility, 2),
                'max_gain': round(max_gain, 2),
                'max_loss': round(max_loss, 2),
                'total_portfolio_value': round(total_value, 2),
                'risk_level': risk_level,
                'risk_color': risk_color,
                'data_quality': data_quality,
                'timestamp': datetime.now().strftime("%H:%M:%S")
            }
            
        except Exception as e:
            return {
                'error': f'Analysis failed: {str(e)}',
                'total_stocks': 0,
                'data_quality': 'Error'
            }
    
    def detect_earnings_surprises(self, stocks_data: List[Dict]) -> List[Dict]:
        """Enhanced earnings surprise detection"""
        surprises = []
        
        for stock in stocks_data:
            if 'error' not in stock and stock.get('current_price', 0) > 0:
                change = stock.get('change_percent', 0)
                symbol = stock.get('symbol', 'Unknown')
                
                # Define surprise thresholds
                if abs(change) > 5:  # Major movement
                    surprise_type = 'Major Beat' if change > 5 else 'Major Miss'
                    impact = 'High'
                elif abs(change) > 2:  # Moderate movement
                    surprise_type = 'Beat' if change > 2 else 'Miss'
                    impact = 'Medium'
                else:
                    continue
                
                surprises.append({
                    'symbol': symbol,
                    'change_percent': change,
                    'type': surprise_type,
                    'impact': impact,
                    'direction': '📈' if change > 0 else '📉'
                })
        
        # Sort by absolute change
        surprises.sort(key=lambda x: abs(x['change_percent']), reverse=True)
        return surprises

class LanguageAgent:
    """Enhanced LLM-based synthesis"""
    
    def __init__(self):
        self.model = model if 'model' in globals() else None
    
    def synthesize_market_brief(self, portfolio_data: Dict, news_data: List[Dict], 
                              analysis_data: Dict, query: str) -> str:
        """Generate comprehensive market brief"""
        
        if not self.model:
            return self._generate_fallback_brief(analysis_data, news_data, query)
        
        try:
            # Prepare concise data for the prompt
            key_metrics = {
                'total_stocks': analysis_data.get('total_stocks', 0),
                'risk_level': analysis_data.get('risk_level', 'Unknown'),
                'avg_change': analysis_data.get('avg_change_percent', 0),
                'volatility': analysis_data.get('volatility', 0),
                'positive_movers': analysis_data.get('positive_movers', 0),
                'negative_movers': analysis_data.get('negative_movers', 0)
            }
            
            news_headlines = [n.get('title', 'N/A') for n in news_data[:3]]
            news_sentiment = [n.get('sentiment', 'Neutral') for n in news_data[:3]]
            
            prompt = f"""
            As a professional financial analyst, provide a concise market brief for this query: "{query}"
            Current Portfolio Metrics:
            - Analyzed {key_metrics['total_stocks']} stocks
            - Risk Level: {key_metrics['risk_level']} (Volatility: {key_metrics['volatility']}%)
            - Average Change: {key_metrics['avg_change']}%
            - Positive Movers: {key_metrics['positive_movers']}, Negative: {key_metrics['negative_movers']}
            Recent Headlines: {', '.join(news_headlines[:2])}
            Market Sentiment: {', '.join(set(news_sentiment))}
            Provide a professional response that:
            1. Directly addresses the query
            2. Highlights key portfolio insights
            3. Notes significant market movements
            4. Offers actionable insights
            5. Keep it under 150 words and use a confident, professional tone
            Format as a concise market brief.
            """
            
            response = self.model.generate_content(prompt)
            return response.text
            
        except Exception as e:
            return self._generate_fallback_brief(analysis_data, news_data, query)
    
    def _generate_fallback_brief(self, analysis_data: Dict, news_data: List[Dict], query: str) -> str:
        """Fallback brief generation when Gemini is unavailable"""
        
        risk_level = analysis_data.get('risk_level', 'Medium')
        avg_change = analysis_data.get('avg_change_percent', 0)
        total_stocks = analysis_data.get('total_stocks', 0)
        pos_movers = analysis_data.get('positive_movers', 0)
        neg_movers = analysis_data.get('negative_movers', 0)
        
        sentiment_summary = "Mixed"
        if news_data:
            sentiments = [n.get('sentiment', 'Neutral') for n in news_data]
            pos_count = sentiments.count('Positive')
            if pos_count > len(sentiments) / 2:
                sentiment_summary = "Positive"
            elif sentiments.count('Negative') > len(sentiments) / 2:
                sentiment_summary = "Negative"
        
        brief = f"""
        **Market Brief - {datetime.now().strftime('%H:%M')}**
        
        Portfolio Analysis: Analyzed {total_stocks} stocks with {risk_level.lower()} risk exposure. 
        Overall performance shows {avg_change:+.1f}% average change with {pos_movers} positive movers vs {neg_movers} declining positions.
        
        Market Sentiment: Current news flow suggests {sentiment_summary.lower()} sentiment in tech sector. 
        {"Strong buying interest evident" if avg_change > 1 else "Cautious trading patterns observed" if avg_change > -1 else "Risk-off sentiment dominating"}.
        
        **Key Insight**: {"Maintain positions with selective buying opportunities" if risk_level == "Low" else "Monitor volatility and consider risk management" if risk_level == "Medium" else "Exercise caution and review position sizing"}.
        
        *Data Quality: Using {"live market data" if analysis_data.get('data_quality') == 'Good' else "mixed data sources for demonstration"}*
        """
        
        return brief.strip()
    
    def generate_risk_assessment(self, analysis_data: Dict) -> str:
        """Generate risk assessment narrative"""
        
        if not self.model:
            return self._generate_fallback_risk_assessment(analysis_data)
        
        try:
            risk_level = analysis_data.get('risk_level', 'Medium')
            volatility = analysis_data.get('volatility', 0)
            
            prompt = f"""
            Generate a brief risk assessment (2-3 sentences) for a portfolio with:
            - Risk Level: {risk_level}
            - Volatility: {volatility}%
            
            Focus on current risk level, key concerns, and recommended actions.
            Be concise and actionable.
            """
            
            response = self.model.generate_content(prompt)
            return response.text
            
        except Exception as e:
            return self._generate_fallback_risk_assessment(analysis_data)
    
    def _generate_fallback_risk_assessment(self, analysis_data: Dict) -> str:
        """Fallback risk assessment"""
        
        risk_level = analysis_data.get('risk_level', 'Medium')
        volatility = analysis_data.get('volatility', 0)
        risk_color = analysis_data.get('risk_color', '🟡')
        
        if risk_level == 'Low':
            return f"{risk_color} **Low Risk Portfolio**: Current volatility of {volatility:.1f}% indicates stable market conditions. Suitable for maintaining current positions with potential for tactical allocation increases."
        elif risk_level == 'High':
            return f"{risk_color} **High Risk Alert**: Elevated volatility of {volatility:.1f}% suggests heightened market stress. Consider reducing position sizes and implementing stop-loss strategies."
        else:
            return f"{risk_color} **Moderate Risk Profile**: Volatility at {volatility:.1f}% reflects normal market conditions. Monitor closely for trend changes and maintain balanced approach to position management."

class VoiceAgent:
    """Real voice processing with TTS and STT functionality"""
    
    def __init__(self):
        self.recognizer = sr.Recognizer()
        self.microphone = sr.Microphone()
        
        # Initialize pygame mixer for audio playback
        try:
            pygame.mixer.init()
            self.audio_enabled = True
        except:
            self.audio_enabled = False
            print("Audio playback not available")
        
        # Adjust for ambient noise
        try:
            with self.microphone as source:
                self.recognizer.adjust_for_ambient_noise(source, duration=1)
        except:
            print("Microphone not available for ambient noise adjustment")
    
    def text_to_speech(self, text: str, lang: str = 'en') -> str:
        """Convert text to speech and return audio file path"""
        try:
            # Clean text for voice output
            clean_text = self._clean_text_for_speech(text)
            
            # Create TTS object
            tts = gTTS(text=clean_text, lang=lang, slow=False)
            
            # Save to temporary file
            with tempfile.NamedTemporaryFile(delete=False, suffix='.mp3') as temp_file:
                tts.save(temp_file.name)
                return temp_file.name
                
        except Exception as e:
            return f"TTS Error: {str(e)}"
    
    def play_audio(self, audio_file_path: str) -> str:
        """Play audio file using pygame"""
        try:
            if not self.audio_enabled:
                return "Audio playback not available"
            
            pygame.mixer.music.load(audio_file_path)
            pygame.mixer.music.play()
            
            # Wait for playback to finish
            while pygame.mixer.music.get_busy():
                time.sleep(0.1)
            
            return "Audio played successfully"
            
        except Exception as e:
            return f"Audio playback error: {str(e)}"
    
    def speech_to_text(self, audio_data=None, timeout: int = 5) -> str:
        """Convert speech to text from microphone or audio data"""
        try:
            if audio_data is None:
                # Listen from microphone
                with self.microphone as source:
                    print("Listening for speech...")
                    audio = self.recognizer.listen(source, timeout=timeout, phrase_time_limit=10)
            else:
                audio = audio_data
            
            # Recognize speech using Google Speech Recognition
            text = self.recognizer.recognize_google(audio)
            return f"Recognized: {text}"
            
        except sr.WaitTimeoutError:
            return "Listening timeout - no speech detected"
        except sr.UnknownValueError:
            return "Could not understand audio"
        except sr.RequestError as e:
            return f"Speech recognition error: {e}"
        except Exception as e:
            return f"STT Error: {str(e)}"
    
    def process_voice_input(self, audio_file_path: str = None) -> str:
        """Process voice input from uploaded audio file"""
        try:
            if audio_file_path:
                # Load audio file
                with sr.AudioFile(audio_file_path) as source:
                    audio = self.recognizer.record(source)
                return self.speech_to_text(audio)
            else:
                # Use microphone
                return self.speech_to_text()
                
        except Exception as e:
            return f"Voice input processing error: {str(e)}"
    
    def _clean_text_for_speech(self, text: str) -> str:
        """Clean text for better speech synthesis"""
        # Remove markdown formatting
        clean_text = re.sub(r'\*\*([^*]+)\*\*', r'\1', text)  # Remove bold
        clean_text = re.sub(r'\*([^*]+)\*', r'\1', clean_text)  # Remove italic
        clean_text = re.sub(r'#+ ', '', clean_text)  # Remove headers
        
        # Remove emojis and special characters
        clean_text = re.sub(r'[📊📈📉🟢🟡🔴⚠️💡🎯📰🔊🎤🚀✨🔄]', '', clean_text)
        
        # Replace newlines with periods
        clean_text = re.sub(r'\n+', '. ', clean_text)
        
        # Clean up extra spaces
        clean_text = re.sub(r'\s+', ' ', clean_text).strip()
        
        # Limit length for better TTS
        if len(clean_text) > 500:
            sentences = clean_text.split('. ')
            clean_text = '. '.join(sentences[:3]) + '.'
        
        return clean_text
    
    def create_voice_response(self, text: str) -> tuple:
        """Create both audio file and playback status"""
        try:
            # Generate TTS audio
            audio_file = self.text_to_speech(text)
            
            if audio_file.startswith("TTS Error"):
                return None, audio_file
            
            # Return audio file path and success message
            return audio_file, "Voice response generated successfully"
            
        except Exception as e:
            return None, f"Voice response error: {str(e)}"

class MultiAgentOrchestrator:
    """Enhanced orchestrator with real voice capabilities"""
    
    def __init__(self):
        self.api_agent = APIAgent()
        self.scraping_agent = ScrapingAgent()
        self.retriever_agent = RetrieverAgent()
        self.analysis_agent = AnalysisAgent()
        self.language_agent = LanguageAgent()
        self.voice_agent = VoiceAgent()
        
        # Default portfolio - mix of US and Asian tech stocks
        self.default_stocks = ["TSM", "NVDA", "AAPL", "GOOGL", "MSFT", "ASML"]
        self.last_update = None
        self.cache_duration = 30  # seconds
    
    def process_market_query(self, query: str, include_voice: bool = False, 
                           custom_stocks: str = "", voice_input_file=None) -> Dict:
        """Enhanced main processing pipeline with voice integration"""
        start_time = time.time()
        
        try:
            # Process voice input if provided
            voice_input_text = ""
            if voice_input_file is not None:
                voice_input_text = self.voice_agent.process_voice_input(voice_input_file)
                if "Recognized:" in voice_input_text:
                    # Extract recognized text and use as query
                    recognized_query = voice_input_text.split("Recognized: ")[1]
                    query = recognized_query if recognized_query.strip() else query
            
            # Determine stock symbols to analyze
            if custom_stocks.strip():
                symbols = [s.strip().upper() for s in custom_stocks.split(',') if s.strip()]
            else:
                symbols = self.default_stocks
            
            # Limit symbols to prevent timeout
            symbols = symbols[:6]
            
            # Step 1: Get market data
            print(f"Fetching data for {len(symbols)} stocks...")
            stocks_data = self.api_agent.get_multiple_stocks(symbols)
            
            # Step 2: Get news and sentiment
            print("Gathering market news...")
            news_data = self.scraping_agent.get_market_news("tech stocks")
            
            # Step 3: Perform analysis
            print("Analyzing portfolio metrics...")
            analysis_data = self.analysis_agent.calculate_portfolio_metrics(stocks_data)
            earnings_surprises = self.analysis_agent.detect_earnings_surprises(stocks_data)
            
            # Step 4: Index data for retrieval
            self.retriever_agent.index_data(stocks_data, 'stocks')
            self.retriever_agent.index_data(analysis_data, 'analysis')
            
            # Step 5: Generate comprehensive market brief
            print("Generating market brief...")
            market_brief = self.language_agent.synthesize_market_brief(
                stocks_data, news_data, analysis_data, query
            )
            
            # Step 6: Generate risk assessment
            risk_assessment = self.language_agent.generate_risk_assessment(analysis_data)
            
            # Step 7: Process voice output if requested
            voice_output = None
            voice_file_path = None
            if include_voice:
                print("Generating voice response...")
                voice_response_text = f"{market_brief}\n\n{risk_assessment}"
                voice_file_path, voice_status = self.voice_agent.create_voice_response(voice_response_text)
                voice_output = voice_status
            
            # Calculate processing time
            processing_time = round(time.time() - start_time, 2)
            
            # Compile comprehensive results
            results = {
                'query': query,
                'voice_input': voice_input_text,
                'stocks_data': stocks_data,
                'news_data': news_data,
                'analysis_data': analysis_data,
                'earnings_surprises': earnings_surprises,
                'market_brief': market_brief,
                'risk_assessment': risk_assessment,
                'voice_output': voice_output,
                'voice_file_path': voice_file_path,
                'processing_time': processing_time,
                'timestamp': datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
                'symbols_analyzed': symbols,
                'data_sources': list(set([s.get('source', 'unknown') for s in stocks_data]))
            }
            
            self.last_update = datetime.now()
            return results
            
        except Exception as e:
            return {
                'error': f'Processing failed: {str(e)}',
                'query': query,
                'processing_time': round(time.time() - start_time, 2),
                'timestamp': datetime.now().strftime("%Y-%m-%d %H:%M:%S")
            }
    
    def get_real_time_update(self, symbols: List[str] = None) -> Dict:
        """Get real-time market updates with caching"""
        if symbols is None:
            symbols = self.default_stocks
        
        # Check cache
        if (self.last_update and 
            (datetime.now() - self.last_update).seconds < self.cache_duration):
            return {"status": "Using cached data", "cache_valid": True}
        
        # Fetch fresh data
        stocks_data = self.api_agent.get_multiple_stocks(symbols)
        analysis_data = self.analysis_agent.calculate_portfolio_metrics(stocks_data)
        
        return {
            'stocks_data': stocks_data,
            'analysis_data': analysis_data,
            'timestamp': datetime.now().strftime("%H:%M:%S"),
            'cache_valid': False
        }
    
    def format_display_data(self, results: Dict) -> tuple:
        """Format data for Gradio display"""
        if 'error' in results:
            return results['error'], "", "", ""
        
        # Format stock data table
        stocks_df = pd.DataFrame([
            {
                'Symbol': s.get('symbol', 'N/A'),
                'Price': f"${s.get('current_price', 0):.2f}",
                'Change %': f"{s.get('change_percent', 0):+.2f}%",
                'Volume': f"{s.get('volume', 0):,}" if s.get('volume', 0) > 0 else 'N/A',
                'Source': s.get('source', 'unknown')
            }
            for s in results.get('stocks_data', [])
        ])
        
        # Format news summary
        news_summary = ""
        for i, news in enumerate(results.get('news_data', []), 1):
            sentiment_emoji = {'Positive': '📈', 'Negative': '📉', 'Neutral': '📊'}.get(news.get('sentiment', 'Neutral'), '📊')
            news_summary += f"{i}. {sentiment_emoji} **{news.get('title', 'N/A')}**\n"
            news_summary += f"   _{news.get('publisher', 'Unknown')} - {news.get('sentiment', 'Neutral')} sentiment_\n\n"
        
        # Format analysis summary
        analysis = results.get('analysis_data', {})
        analysis_summary = f"""
**📊 Portfolio Overview**
• Total Stocks Analyzed: {analysis.get('total_stocks', 0)}
• Risk Level: {analysis.get('risk_color', '🟡')} {analysis.get('risk_level', 'Medium')}
• Average Change: {analysis.get('avg_change_percent', 0):+.2f}%
• Volatility: {analysis.get('volatility', 0):.2f}%

**📈 Market Movers**
• Positive: {analysis.get('positive_movers', 0)} stocks
• Negative: {analysis.get('negative_movers', 0)} stocks
• Neutral: {analysis.get('neutral_movers', 0)} stocks

**⏰ Last Updated: {analysis.get('timestamp', 'N/A')}**
**🔍 Data Quality: {analysis.get('data_quality', 'Unknown')}**
        """
        
        # Combine market brief and risk assessment
        comprehensive_brief = f"""
{results.get('market_brief', 'No brief available')}

---

**🎯 Risk Assessment**
{results.get('risk_assessment', 'No risk assessment available')}

---

**⚡ Processing Info**
• Processing Time: {results.get('processing_time', 0)} seconds
• Symbols: {', '.join(results.get('symbols_analyzed', []))}
• Voice Input: {'✅' if results.get('voice_input') else '❌'}
• Voice Output: {'✅' if results.get('voice_output') else '❌'}
        """
        
        return stocks_df, news_summary, analysis_summary, comprehensive_brief

# Initialize the orchestrator
orchestrator = MultiAgentOrchestrator()

def process_query(query, include_voice, custom_stocks, voice_input_file):
    """Main processing function for Gradio interface"""
    try:
        results = orchestrator.process_market_query(
            query=query,
            include_voice=include_voice,
            custom_stocks=custom_stocks,
            voice_input_file=voice_input_file
        )
        
        stocks_df, news_summary, analysis_summary, comprehensive_brief = orchestrator.format_display_data(results)
        
        # Handle voice output
        voice_output_file = None
        if results.get('voice_file_path'):
            voice_output_file = results['voice_file_path']
        
        return stocks_df, news_summary, analysis_summary, comprehensive_brief, voice_output_file
        
    except Exception as e:
        error_msg = f"Error processing query: {str(e)}"
        return error_msg, "", "", "", None

def get_live_update(custom_stocks):
    """Get live market updates"""
    try:
        symbols = [s.strip().upper() for s in custom_stocks.split(',') if s.strip()] if custom_stocks.strip() else None
        update_data = orchestrator.get_real_time_update(symbols)
        
        if update_data.get('cache_valid'):
            return "📱 Using cached data (updated within last 30 seconds)", "", ""
        
        # Format the update
        stocks_data = update_data.get('stocks_data', [])
        analysis_data = update_data.get('analysis_data', {})
        
        # Quick summary
        avg_change = analysis_data.get('avg_change_percent', 0)
        risk_level = analysis_data.get('risk_level', 'Medium')
        timestamp = update_data.get('timestamp', 'N/A')
        
        summary = f"""
🔄 **Live Market Update - {timestamp}**

📊 Portfolio Status: {avg_change:+.2f}% average change
🎯 Risk Level: {risk_level}
📈 Positive Movers: {analysis_data.get('positive_movers', 0)}
📉 Negative Movers: {analysis_data.get('negative_movers', 0)}
        """
        
        # Top movers
        top_movers = sorted(stocks_data, key=lambda x: abs(x.get('change_percent', 0)), reverse=True)[:3]
        movers_text = "**🚀 Top Movers:**\n"
        for stock in top_movers:
            direction = "📈" if stock.get('change_percent', 0) > 0 else "📉"
            movers_text += f"• {direction} {stock.get('symbol', 'N/A')}: {stock.get('change_percent', 0):+.2f}%\n"
        
        return summary, movers_text, f"Updated: {timestamp}"
        
    except Exception as e:
        return f"Update failed: {str(e)}", "", ""

# Create Gradio Interface
def create_interface():
    """Create the main Gradio interface"""
    
    with gr.Blocks(
        title="🚀 Multi-Agent Market Analysis System",
        theme=gr.themes.Soft(),
        css="""
        .gradio-container {
            max-width: 1200px !important;
        }
        .main-header {
            text-align: center;
            background: linear-gradient(90deg, #667eea 0%, #764ba2 100%);
            color: white;
            padding: 20px;
            border-radius: 10px;
            margin-bottom: 20px;
        }
        """
    ) as demo:
        
        # Header
        gr.HTML("""
        <div class="main-header">
            <h1>🚀 Multi-Agent Market Analysis System</h1>
            <p>Real-time market analysis with AI-powered insights, news sentiment, and voice capabilities</p>
        </div>
        """)
        
        with gr.Tab("📊 Market Analysis"):
            with gr.Row():
                with gr.Column(scale=1):
                    query_input = gr.Textbox(
                        label="🔍 Market Query",
                        placeholder="Ask about market trends, specific stocks, or analysis...",
                        value="What's the current market sentiment for tech stocks?",
                        lines=2
                    )
                    
                    custom_stocks_input = gr.Textbox(
                        label="📈 Custom Stock Symbols (comma-separated)",
                        placeholder="AAPL,GOOGL,MSFT,NVDA... (leave empty for default portfolio)",
                        value=""
                    )
                    
                    with gr.Row():
                        include_voice_checkbox = gr.Checkbox(
                            label="🔊 Generate Voice Response",
                            value=False
                        )
                        
                        voice_input_file = gr.Audio(
                            label="🎤 Voice Input (optional)",
                            type="filepath"
                        )
                    
                    analyze_button = gr.Button("🚀 Analyze Market", variant="primary", size="lg")
                
                with gr.Column(scale=2):
                    with gr.Tab("📊 Stock Data"):
                        stocks_output = gr.Dataframe(
                            label="Real-time Stock Data",
                            headers=["Symbol", "Price", "Change %", "Volume", "Source"],
                            interactive=False
                        )
                    
                    with gr.Tab("📰 Market News"):
                        news_output = gr.Markdown(label="Latest Market News & Sentiment")
                    
                    with gr.Tab("📈 Analysis"):
                        analysis_output = gr.Markdown(label="Portfolio Analysis")
                    
                    with gr.Tab("🎯 AI Brief"):
                        brief_output = gr.Markdown(label="Comprehensive Market Brief")
                    
                    # Voice output
                    voice_output = gr.Audio(label="🔊 Voice Response", visible=False)
        
        with gr.Tab("📱 Live Updates"):
            gr.Markdown("### 🔄 Real-time Market Monitor")
            
            with gr.Row():
                live_stocks_input = gr.Textbox(
                    label="Stock Symbols for Live Updates",
                    placeholder="Leave empty for default portfolio",
                    value=""
                )
                update_button = gr.Button("🔄 Get Live Update", variant="secondary")
            
            with gr.Row():
                with gr.Column():
                    live_summary = gr.Markdown(label="Market Summary")
                with gr.Column():
                    live_movers = gr.Markdown(label="Top Movers")
                with gr.Column():
                    live_timestamp = gr.Markdown(label="Last Update")
        
        with gr.Tab("ℹ️ About"):
            gr.Markdown("""
            ### 🤖 Multi-Agent System Architecture
            
            This system uses multiple specialized AI agents working together:
            
            **🔗 API Agent**: Fetches real-time market data from multiple sources with fallback mechanisms
            
            **📰 Scraping Agent**: Gathers market news and performs sentiment analysis
            
            **🗃️ Retriever Agent**: Indexes and retrieves relevant market information
            
            **📊 Analysis Agent**: Performs quantitative analysis and risk assessment
            
            **🤖 Language Agent**: Synthesizes insights using Google's Gemini AI
            
            **🎤 Voice Agent**: Handles speech-to-text and text-to-speech functionality
            
            **🎛️ Orchestrator**: Coordinates all agents for comprehensive market analysis
            
            ### 🎯 Key Features
            - Real-time stock data with multiple fallback sources
            - AI-powered market sentiment analysis
            - Voice input and output capabilities
            - Risk assessment and portfolio metrics
            - Live market updates with caching
            - Comprehensive market briefs
            
            ### 📝 Usage Tips
            1. Use natural language queries like "How are tech stocks performing?"
            2. Specify custom stocks or use the default tech portfolio
            3. Enable voice output for audio briefings
            4. Use voice input to ask questions hands-free
            5. Check live updates for real-time monitoring
            
            **Note**: This system uses both live market data (when available) and demo data for demonstration purposes.
            """)
        
        # Event handlers
        analyze_button.click(
            process_query,
            inputs=[query_input, include_voice_checkbox, custom_stocks_input, voice_input_file],
            outputs=[stocks_output, news_output, analysis_output, brief_output, voice_output]
        ).then(
            lambda: gr.update(visible=True),
            outputs=[voice_output]
        )
        
        update_button.click(
            get_live_update,
            inputs=[live_stocks_input],
            outputs=[live_summary, live_movers, live_timestamp]
        )
        
        # Auto-refresh live updates every 60 seconds
        demo.load(
            get_live_update,
            inputs=[gr.Textbox(value="", visible=False)],
            outputs=[live_summary, live_movers, live_timestamp],
            every=60
        )
    
    return demo

# Launch the application
if __name__ == "__main__":
    print("🚀 Starting Multi-Agent Market Analysis System...")
    
    # Check for required API keys
    if not GEMINI_API_KEY:
        print("⚠️  Warning: GEMINI_API_KEY not found. Using fallback text generation.")
    
    print("✅ System initialized successfully!")
    print("📊 Loading market data sources...")
    print("🎤 Voice capabilities enabled")
    print("🔄 Real-time updates configured")
    
    # Create and launch the interface
    demo = create_interface()
    demo.launch(
        server_name="0.0.0.0",
        server_port=7860,
        share=True,
        debug=True,
        show_error=True
    )