app/utils/breaking_news_scorer.py

"""
Breaking News Scoring System
Identifies highest-impact financial news using multi-factor weighted scoring
"""

import re
from datetime import datetime, timedelta
from typing import Dict, List
import logging

logger = logging.getLogger(__name__)


class BreakingNewsScorer:
    """
    Sophisticated scoring system for breaking financial news
    Uses weighted factors to identify market-moving events
    """

    # Critical keywords with high market impact (weight: 3.0)
    CRITICAL_KEYWORDS = [
        # Central Bank Actions
        'rate hike', 'rate cut', 'interest rate', 'fed raises', 'fed cuts',
        'fomc decision', 'monetary policy', 'quantitative easing', 'qe',
        'emergency meeting', 'powell', 'lagarde', 'yellen',

        # Market Events
        'market crash', 'flash crash', 'circuit breaker', 'trading halt',
        'all-time high', 'all time high', 'record high', 'record low',
        'biggest drop', 'biggest gain', 'historic', 'unprecedented',

        # Economic Data
        'gdp', 'jobs report', 'unemployment', 'inflation',
        'cpi', 'ppi', 'nonfarm payroll', 'nfp',

        # Corporate Events
        'earnings beat', 'earnings miss', 'profit warning',
        'bankruptcy', 'chapter 11', 'delisted',
        'merger', 'acquisition', 'takeover', 'buyout',

        # Geopolitical
        'war', 'invasion', 'sanctions', 'trade war',
        'embargo', 'default', 'debt ceiling', 'shutdown',
        'impeachment', 'coup', 'terrorist attack'
    ]

    # High-impact keywords (weight: 2.0)
    HIGH_IMPACT_KEYWORDS = [
        # Market Movement
        'surge', 'plunge', 'soar', 'tumble', 'rally', 'selloff',
        'volatility', 'whipsaw', 'correction', 'bear market', 'bull market',

        # Economic Indicators
        'retail sales', 'housing starts', 'consumer confidence',
        'manufacturing index', 'pmi', 'trade deficit',

        # Corporate
        'revenue beat', 'guidance', 'dividend', 'stock split',
        'ipo', 'listing', 'secondary offering',

        # Crypto/Tech
        'bitcoin', 'crypto crash', 'hack', 'breach',
        'antitrust', 'regulation', 'sec investigation',

        # Commodities
        'oil', 'gold', 'crude', 'opec', 'energy crisis',
        'supply chain', 'shortage', 'surplus'
    ]

    # Medium-impact keywords (weight: 1.5)
    MEDIUM_IMPACT_KEYWORDS = [
        'analyst', 'upgrade', 'downgrade', 'target price',
        'forecast', 'outlook', 'projection', 'estimate',
        'conference call', 'ceo', 'cfo', 'executive',
        'lawsuit', 'settlement', 'fine', 'penalty',
        'product launch', 'partnership', 'deal', 'contract'
    ]

    # Premium source weights (multipliers)
    SOURCE_WEIGHTS = {
        # Tier 1: Breaking News Specialists (2.0x)
        'walter_bloomberg': 2.0,
        'fxhedge': 2.0,
        'deitaone': 2.0,
        'firstsquawk': 1.9,
        'livesquawk': 1.9,

        # Tier 2: Major Financial Media (1.8x)
        'reuters': 1.8,
        'bloomberg': 1.8,
        'ft': 1.7,
        'wsj': 1.7,

        # Tier 3: Mainstream Media (1.5x)
        'cnbc': 1.5,
        'bbc': 1.5,
        'marketwatch': 1.5,

        # Tier 4: Alternative/Community (1.2x)
        'zerohedge': 1.2,
        'wallstreetbets': 1.2,
        'reddit': 1.2,

        # Default
        'default': 1.0
    }

    # Ticker mention bonus (companies that move markets)
    MAJOR_TICKERS = [
        'SPY', 'QQQ', 'DIA', 'IWM',  # Market indices
        'AAPL', 'MSFT', 'GOOGL', 'AMZN', 'NVDA', 'TSLA', 'META',  # Mega caps
        'JPM', 'BAC', 'GS', 'MS', 'WFC',  # Banks
        'XOM', 'CVX', 'COP',  # Energy
        'BTC', 'ETH', 'BTCUSD', 'ETHUSD'  # Crypto
    ]

    def __init__(self):
        """Initialize the breaking news scorer"""
        logger.info("BreakingNewsScorer initialized")

    def calculate_impact_score(self, news_item: Dict) -> float:
        """
        Calculate comprehensive impact score for a news item

        Args:
            news_item: Dictionary containing news metadata

        Returns:
            Impact score (0-100, higher = more impactful)
        """
        score = 0.0

        # Extract key fields
        title = news_item.get('title', '').lower()
        summary = news_item.get('summary', '').lower()
        source = news_item.get('source', '').lower()
        timestamp = news_item.get('timestamp', datetime.now())
        sentiment = news_item.get('sentiment', 'neutral')
        impact_level = news_item.get('impact', 'low')
        category = news_item.get('category', 'markets')

        # Combine title and summary for keyword analysis
        text = f"{title} {summary}"

        # 1. KEYWORD SCORING (30 points max)
        keyword_score = self._score_keywords(text)
        score += keyword_score

        # 2. RECENCY SCORING (20 points max)
        recency_score = self._score_recency(timestamp)
        score += recency_score

        # 3. SOURCE CREDIBILITY (20 points max)
        source_score = self._score_source(source)
        score += source_score

        # 4. ENGAGEMENT SCORING (15 points max)
        engagement_score = self._score_engagement(news_item)
        score += engagement_score

        # 5. SENTIMENT EXTREMITY (10 points max)
        sentiment_score = self._score_sentiment(sentiment)
        score += sentiment_score

        # 6. CATEGORY RELEVANCE (5 points max)
        category_score = self._score_category(category)
        score += category_score

        # 7. TICKER MENTIONS (bonus up to 10 points)
        ticker_score = self._score_tickers(text)
        score += ticker_score

        # 8. URGENCY INDICATORS (bonus up to 10 points)
        urgency_score = self._score_urgency(text)
        score += urgency_score

        # 9. EXISTING IMPACT LEVEL (weight existing classification)
        if impact_level == 'high':
            score *= 1.2
        elif impact_level == 'medium':
            score *= 1.1

        # Cap at 100
        score = min(score, 100.0)

        logger.debug(f"News '{title[:50]}...' scored: {score:.2f}")

        return score

    def _score_keywords(self, text: str) -> float:
        """Score based on keyword presence and frequency"""
        score = 0.0

        # Critical keywords (3.0 points each, max 18)
        critical_matches = sum(1 for kw in self.CRITICAL_KEYWORDS if kw in text)
        score += min(critical_matches * 3.0, 18.0)

        # High-impact keywords (2.0 points each, max 8)
        high_matches = sum(1 for kw in self.HIGH_IMPACT_KEYWORDS if kw in text)
        score += min(high_matches * 2.0, 8.0)

        # Medium-impact keywords (1.0 points each, max 4)
        medium_matches = sum(1 for kw in self.MEDIUM_IMPACT_KEYWORDS if kw in text)
        score += min(medium_matches * 1.0, 4.0)

        return min(score, 30.0)

    def _score_recency(self, timestamp: datetime) -> float:
        """Score based on how recent the news is"""
        try:
            if isinstance(timestamp, str):
                timestamp = datetime.fromisoformat(timestamp.replace('Z', '+00:00'))

            age_seconds = (datetime.now() - timestamp.replace(tzinfo=None)).total_seconds()
            age_minutes = age_seconds / 60

            # Exponential decay: most recent = highest score
            if age_minutes < 5:
                return 20.0  # Within 5 minutes: full score
            elif age_minutes < 15:
                return 18.0  # 5-15 minutes
            elif age_minutes < 30:
                return 15.0  # 15-30 minutes
            elif age_minutes < 60:
                return 10.0  # 30-60 minutes
            elif age_minutes < 180:
                return 5.0   # 1-3 hours
            else:
                return 1.0   # Older than 3 hours
        except:
            return 5.0  # Default if timestamp parsing fails

    def _score_source(self, source: str) -> float:
        """Score based on source credibility"""
        source = source.lower().replace(' ', '_').replace('/', '').replace('@', '')

        # Check for known sources
        for source_key, weight in self.SOURCE_WEIGHTS.items():
            if source_key in source:
                return weight * 10.0  # Scale to max 20 points

        return self.SOURCE_WEIGHTS['default'] * 10.0

    def _score_engagement(self, news_item: Dict) -> float:
        """Score based on social engagement metrics"""
        score = 0.0
        has_engagement = False

        # Twitter engagement (top-level fields)
        likes = news_item.get('likes', 0)
        if likes > 0:
            has_engagement = True
            score += min(likes / 1000, 5.0)  # Max 5 points for likes

        retweets = news_item.get('retweets', 0)
        if retweets > 0:
            has_engagement = True
            score += min(retweets / 500, 5.0)  # Max 5 points for retweets

        # Reddit engagement (top-level fields)
        reddit_score = news_item.get('reddit_score', 0)
        if reddit_score > 0:
            has_engagement = True
            score += min(reddit_score / 1000, 5.0)  # Max 5 points for score

        comments = news_item.get('reddit_comments', 0)
        if comments > 0:
            has_engagement = True
            score += min(comments / 200, 5.0)  # Max 5 points for comments

        # If no engagement data, return default score
        if not has_engagement:
            return 5.0

        return min(score, 15.0)

    def _score_sentiment(self, sentiment: str) -> float:
        """Score based on sentiment extremity (extreme = more impactful)"""
        if sentiment == 'positive':
            return 8.0  # Strong positive news moves markets
        elif sentiment == 'negative':
            return 10.0  # Negative news tends to have more impact
        else:
            return 3.0  # Neutral news less impactful

    def _score_category(self, category: str) -> float:
        """Score based on category relevance"""
        if category == 'macro':
            return 5.0  # Macro news affects entire market
        elif category == 'markets':
            return 4.0  # Direct market news
        elif category == 'geopolitical':
            return 3.0  # Geopolitical can be high impact
        else:
            return 2.0  # Other categories

    def _score_tickers(self, text: str) -> float:
        """Bonus score for mentioning major market-moving tickers"""
        text_upper = text.upper()

        # Count major ticker mentions
        ticker_mentions = sum(1 for ticker in self.MAJOR_TICKERS if ticker in text_upper)

        # 2 points per ticker, max 10 points
        return min(ticker_mentions * 2.0, 10.0)

    def _score_urgency(self, text: str) -> float:
        """Bonus score for urgency indicators"""
        urgency_patterns = [
            r'\bbreaking\b', r'\balert\b', r'\burgent\b', r'\bjust in\b',
            r'\bemergency\b', r'\bimmediate\b', r'\bnow\b', r'\btoday\b',
            r'‼️', r'🚨', r'⚠️', r'🔴', r'❗'
        ]

        score = 0.0
        for pattern in urgency_patterns:
            if re.search(pattern, text, re.IGNORECASE):
                score += 2.0

        return min(score, 10.0)

    def get_breaking_news(self, news_items: List[Dict], top_n: int = 1) -> List[Dict]:
        """
        Identify top breaking news from a list

        Args:
            news_items: List of news item dictionaries
            top_n: Number of top items to return

        Returns:
            List of top breaking news items with scores
        """
        if not news_items:
            return []

        # Calculate scores for all items
        scored_items = []
        for item in news_items:
            score = self.calculate_impact_score(item)
            scored_items.append({
                **item,
                'breaking_score': score
            })

        # Sort by score (descending)
        scored_items.sort(key=lambda x: x['breaking_score'], reverse=True)

        # Log top items
        logger.info(f"Top {top_n} breaking news:")
        for i, item in enumerate(scored_items[:top_n], 1):
            logger.info(f"  {i}. [{item['breaking_score']:.1f}] {item['title'][:60]}...")

        return scored_items[:top_n]

    def get_breaking_threshold(self) -> float:
        """Get minimum score threshold for breaking news display"""
        return 40.0  # Only show news with score >= 40 (out of 100)


# Singleton instance
_scorer_instance = None

def get_breaking_news_scorer() -> BreakingNewsScorer:
    """Get singleton instance of BreakingNewsScorer"""
    global _scorer_instance
    if _scorer_instance is None:
        _scorer_instance = BreakingNewsScorer()
    return _scorer_instance