""" 新闻智能体 - 主控制器 实时抓取、分析、通知重要新闻 """ import asyncio from typing import Dict, Any, List, Optional from datetime import datetime, timedelta from app.utils.logger import logger from app.config import get_settings from app.news_agent.sources import get_enabled_sources from app.news_agent.fetcher import NewsFetcher, NewsItem from app.news_agent.filter import NewsDeduplicator, NewsFilter from app.news_agent.analyzer import NewsAnalyzer, NewsAnalyzerSimple from app.news_agent.news_db_service import get_news_db_service from app.news_agent.notifier import get_news_notifier class NewsAgent: """新闻智能体 - 主控制器""" _instance = None _initialized = False def __new__(cls, *args, **kwargs): """单例模式""" if cls._instance is None: cls._instance = super().__new__(cls) return cls._instance def __init__(self): """初始化新闻智能体""" if NewsAgent._initialized: return NewsAgent._initialized = True self.settings = get_settings() # 核心组件 self.fetcher = NewsFetcher() self.deduplicator = NewsDeduplicator() self.filter = NewsFilter() self.analyzer = NewsAnalyzer() # LLM 分析器 self.simple_analyzer = NewsAnalyzerSimple() # 规则分析器（备用） self.db_service = get_news_db_service() self.notifier = get_news_notifier() # 配置 self.fetch_interval = 300 # 抓取间隔（秒）= 5分钟 self.min_priority = 40.0 # 最低通知优先级 self.use_llm = True # 使用 LLM 批量分析 # 统计数据 self.stats = { 'total_fetched': 0, 'total_saved': 0, 'total_analyzed': 0, 'total_notified': 0, 'last_fetch_time': None, 'last_notify_time': None } # 运行状态 self.running = False self._task = None logger.info("新闻智能体初始化完成") async def start(self): """启动新闻智能体""" if self.running: logger.warning("新闻智能体已在运行") return self.running = True # 发送启动通知 sources = get_enabled_sources() crypto_count = sum(1 for s in sources if s['category'] == 'crypto') stock_count = sum(1 for s in sources if s['category'] == 'stock') await self.notifier.notify_startup({ 'crypto_sources': crypto_count, 'stock_sources': stock_count, 'fetch_interval': self.fetch_interval }) # 启动后台任务 self._task = asyncio.create_task(self._run_loop()) logger.info("新闻智能体已启动") async def stop(self): """停止新闻智能体""" if not self.running: return self.running = False if self._task: self._task.cancel() try: await self._task except asyncio.CancelledError: pass await self.fetcher.close() logger.info("新闻智能体已停止") async def _run_loop(self): """主循环""" while self.running: try: await self._fetch_and_process_news() except Exception as e: logger.error(f"新闻处理循环出错: {e}") await self.notifier.notify_error(str(e)) # 等待下一次抓取 await asyncio.sleep(self.fetch_interval) async def _fetch_and_process_news(self): """抓取并处理新闻""" logger.info("=" * 60) logger.info("开始新闻处理周期") # 1. 抓取新闻 items = await self.fetcher.fetch_all_news() self.stats['total_fetched'] += len(items) self.stats['last_fetch_time'] = datetime.utcnow().isoformat() if not items: logger.info("没有获取到新新闻") return logger.info(f"获取到 {len(items)} 条新闻") # 2. 去重 items = self.deduplicator.deduplicate_list(items) logger.info(f"去重后剩余 {len(items)} 条") # 3. 过滤 filtered_items = self.filter.filter_news(items) logger.info(f"过滤后剩余 {len(filtered_items)} 条") if not filtered_items: logger.info("没有符合条件的新闻") return # 4. 保存到数据库 saved_articles = [] for item in filtered_items: # 检查数据库中是否已存在 if self.db_service.check_duplicate_by_hash(item.content_hash): continue # 保存 article_data = { 'title': item.title, 'content': item.content, 'url': item.url, 'source': item.source, 'author': item.author, 'category': item.category, 'tags': item.tags, 'published_at': item.published_at, 'crawled_at': item.crawled_at, 'content_hash': item.content_hash, 'quality_score': getattr(item, 'quality_score', 0.5), } article = self.db_service.save_article(article_data) if article: saved_articles.append((article, item)) self.stats['total_saved'] += len(saved_articles) logger.info(f"保存了 {len(saved_articles)} 条新文章") if not saved_articles: return # 5. LLM 分析（仅批量分析） analyzed_count = 0 high_priority_articles = [] if self.use_llm: # 只使用批量分析 (异步) items_to_analyze = [item for _, item in saved_articles] results = await self.analyzer.analyze_batch(items_to_analyze) for (article, _), result in zip(saved_articles, results): if result: priority = self.analyzer.calculate_priority( result, getattr(article, 'quality_score', 0.5) ) self.db_service.mark_as_analyzed(article.id, result, priority) analyzed_count += 1 # 只发送重大影响（high）的新闻 if result.get('market_impact') == 'high': article_dict = article.to_dict() article_dict.update({ 'llm_analyzed': True, 'market_impact': result.get('market_impact'), 'impact_type': result.get('impact_type'), 'sentiment': result.get('sentiment'), 'summary': result.get('summary'), 'key_points': result.get('key_points'), 'trading_advice': result.get('trading_advice'), 'relevant_symbols': result.get('relevant_symbols'), 'priority': priority, }) high_priority_articles.append(article_dict) else: # 使用规则分析 for article, item in saved_articles: result = self.simple_analyzer.analyze_single(item) priority = result.get('confidence', 50) self.db_service.mark_as_analyzed(article.id, result, priority) analyzed_count += 1 # 只发送重大影响（high）的新闻 if result.get('market_impact') == 'high': article_dict = article.to_dict() article_dict.update({ 'llm_analyzed': True, 'market_impact': result.get('market_impact'), 'impact_type': result.get('impact_type'), 'sentiment': result.get('sentiment'), 'summary': result.get('summary'), 'key_points': result.get('key_points'), 'trading_advice': result.get('trading_advice'), 'relevant_symbols': result.get('relevant_symbols'), 'priority': priority, }) high_priority_articles.append(article_dict) self.stats['total_analyzed'] += analyzed_count logger.info(f"分析了 {analyzed_count} 条文章") # 6. 发送通知（仅批量发送）- 增加过滤条件 if high_priority_articles: # 按优先级排序 high_priority_articles.sort( key=lambda x: x.get('priority', 0), reverse=True ) # 额外过滤：只推送真正重要的新闻 truly_important_articles = [] for article in high_priority_articles: impact = article.get('market_impact', 'low') priority = article.get('priority', 0) confidence = article.get('llm_analyzed', False) and article.get('relevant_symbols') # 推送条件（满足其一即可）： # 1. high 影响 + 优先级 >= 55 # 2. high 影响 + 有明确相关代码 # 3. 优先级 >= 60（特别重要） should_notify = ( (impact == 'high' and priority >= 55) or (impact == 'high' and confidence) or (priority >= 60) ) if should_notify: truly_important_articles.append(article) # 批量发送最多10条 if truly_important_articles: await self.notifier.notify_news_batch(trulyy_important_articles[:10]) for article in truly_important_articles[:10]: self.db_service.mark_as_notified(article['id']) self.stats['total_notified'] += 1 logger.info(f"推送了 {len(trulyy_important_articles)} 条真正重要的新闻（从 {len(high_priority_articles)} 条 high 中筛选）") else: logger.info(f"没有达到推送标准的新闻（{len(high_priority_articles)} 条 high 但不够重要）") self.stats['last_notify_time'] = datetime.utcnow().isoformat() logger.info("=" * 60) def get_stats(self) -> Dict[str, Any]: """获取统计数据""" stats = self.stats.copy() stats['running'] = self.running stats['fetch_interval'] = self.fetch_interval stats['use_llm'] = self.use_llm # 从数据库获取更多统计 db_stats = self.db_service.get_stats(hours=24) stats['db_stats'] = db_stats return stats async def manual_fetch(self, category: str = None) -> Dict[str, Any]: """ 手动触发新闻抓取 Args: category: 分类过滤 Returns: 处理结果 """ logger.info(f"手动触发新闻抓取: category={category}") items = await self.fetcher.fetch_all_news(category) result = { 'fetched': len(items), 'timestamp': datetime.utcnow().isoformat() } if items: # 这里可以触发处理流程 # 为简化，只返回抓取结果 result['items'] = [item.to_dict() for item in items[:5]] return result # 全局实例 _news_agent = None def get_news_agent() -> NewsAgent: """获取新闻智能体单例""" global _news_agent if _news_agent is None: _news_agent = NewsAgent() return _news_agent