""" 智能新闻检索工具 提供模糊搜索、链接查询、历史相关新闻检索等高级搜索功能。 """ import re from collections import Counter from datetime import datetime, timedelta from difflib import SequenceMatcher from typing import Dict, List, Optional, Tuple from ..services.data_service import DataService from ..utils.validators import validate_keyword, validate_limit from ..utils.errors import MCPError, InvalidParameterError, DataNotFoundError class SearchTools: """智能新闻检索工具类""" def __init__(self, project_root: str = None): """ 初始化智能检索工具 Args: project_root: 项目根目录 """ self.data_service = DataService(project_root) # 中文停用词列表 self.stopwords = { '的', '了', '在', '是', '我', '有', '和', '就', '不', '人', '都', '一', '一个', '上', '也', '很', '到', '说', '要', '去', '你', '会', '着', '没有', '看', '好', '自己', '这', '那', '来', '被', '与', '为', '对', '将', '从', '以', '及', '等', '但', '或', '而', '于', '中', '由', '可', '可以', '已', '已经', '还', '更', '最', '再', '因为', '所以', '如果', '虽然', '然而' } def search_news_unified( self, query: str, search_mode: str = "keyword", date_range: Optional[Dict[str, str]] = None, platforms: Optional[List[str]] = None, limit: int = 50, sort_by: str = "relevance", threshold: float = 0.6, include_url: bool = False ) -> Dict: """ 统一新闻搜索工具 - 整合多种搜索模式 Args: query: 查询内容（必需）- 关键词、内容片段或实体名称 search_mode: 搜索模式，可选值： - "keyword": 精确关键词匹配（默认） - "fuzzy": 模糊内容匹配（使用相似度算法） - "entity": 实体名称搜索（自动按权重排序） date_range: 日期范围，格式: {"start": "YYYY-MM-DD", "end": "YYYY-MM-DD"} 不指定则默认查询今天 platforms: 平台过滤列表，如 ['zhihu', 'weibo'] limit: 返回条数限制，默认50 sort_by: 排序方式，可选值： - "relevance": 按相关度排序（默认） - "weight": 按新闻权重排序 - "date": 按日期排序 threshold: 相似度阈值（仅fuzzy模式有效），0-1之间，默认0.6 include_url: 是否包含URL链接，默认False（节省token） Returns: 搜索结果字典，包含匹配的新闻列表 Examples: - search_news_unified(query="人工智能", search_mode="keyword") - search_news_unified(query="特斯拉降价", search_mode="fuzzy", threshold=0.4) - search_news_unified(query="马斯克", search_mode="entity", limit=20) - search_news_unified(query="iPhone 16发布", search_mode="keyword") """ try: # 参数验证 query = validate_keyword(query) if search_mode not in ["keyword", "fuzzy", "entity"]: raise InvalidParameterError( f"无效的搜索模式: {search_mode}", suggestion="支持的模式: keyword, fuzzy, entity" ) if sort_by not in ["relevance", "weight", "date"]: raise InvalidParameterError( f"无效的排序方式: {sort_by}", suggestion="支持的排序: relevance, weight, date" ) limit = validate_limit(limit, default=50) threshold = max(0.0, min(1.0, threshold)) # 处理日期范围 if date_range: from ..utils.validators import validate_date_range date_range_tuple = validate_date_range(date_range) start_date, end_date = date_range_tuple else: # 默认今天 start_date = end_date = datetime.now() # 收集所有匹配的新闻 all_matches = [] current_date = start_date while current_date <= end_date: try: all_titles, id_to_name, timestamps = self.data_service.parser.read_all_titles_for_date( date=current_date, platform_ids=platforms ) # 根据搜索模式执行不同的搜索逻辑 if search_mode == "keyword": matches = self._search_by_keyword_mode( query, all_titles, id_to_name, current_date, include_url ) elif search_mode == "fuzzy": matches = self._search_by_fuzzy_mode( query, all_titles, id_to_name, current_date, threshold, include_url ) else: # entity matches = self._search_by_entity_mode( query, all_titles, id_to_name, current_date, include_url ) all_matches.extend(matches) except DataNotFoundError: # 该日期没有数据，继续下一天 pass current_date += timedelta(days=1) if not all_matches: time_desc = "今天" if start_date == end_date else f"{start_date.strftime('%Y-%m-%d')} 至 {end_date.strftime('%Y-%m-%d')}" return { "success": True, "results": [], "total": 0, "query": query, "search_mode": search_mode, "time_range": time_desc, "message": f"未找到匹配的新闻（{time_desc}）" } # 统一排序逻辑 if sort_by == "relevance": all_matches.sort(key=lambda x: x.get("similarity_score", 1.0), reverse=True) elif sort_by == "weight": from .analytics import calculate_news_weight all_matches.sort(key=lambda x: calculate_news_weight(x), reverse=True) elif sort_by == "date": all_matches.sort(key=lambda x: x.get("date", ""), reverse=True) # 限制返回数量 results = all_matches[:limit] # 构建时间范围描述 if start_date == end_date: time_range_desc = start_date.strftime("%Y-%m-%d") else: time_range_desc = f"{start_date.strftime('%Y-%m-%d')} 至 {end_date.strftime('%Y-%m-%d')}" result = { "success": True, "summary": { "total_found": len(all_matches), "returned_count": len(results), "requested_limit": limit, "search_mode": search_mode, "query": query, "platforms": platforms or "所有平台", "time_range": time_range_desc, "sort_by": sort_by }, "results": results } if search_mode == "fuzzy": result["summary"]["threshold"] = threshold if len(all_matches) < limit: result["note"] = f"模糊搜索模式下，相似度阈值 {threshold} 仅匹配到 {len(all_matches)} 条结果" return result except MCPError as e: return { "success": False, "error": e.to_dict() } except Exception as e: return { "success": False, "error": { "code": "INTERNAL_ERROR", "message": str(e) } } def _search_by_keyword_mode( self, query: str, all_titles: Dict, id_to_name: Dict, current_date: datetime, include_url: bool ) -> List[Dict]: """ 关键词搜索模式（精确匹配） Args: query: 搜索关键词 all_titles: 所有标题字典 id_to_name: 平台ID到名称映射 current_date: 当前日期 Returns: 匹配的新闻列表 """ matches = [] query_lower = query.lower() for platform_id, titles in all_titles.items(): platform_name = id_to_name.get(platform_id, platform_id) for title, info in titles.items(): # 精确包含判断 if query_lower in title.lower(): news_item = { "title": title, "platform": platform_id, "platform_name": platform_name, "date": current_date.strftime("%Y-%m-%d"), "similarity_score": 1.0, # 精确匹配，相似度为1 "ranks": info.get("ranks", []), "count": len(info.get("ranks", [])), "rank": info["ranks"][0] if info["ranks"] else 999 } # 条件性添加 URL 字段 if include_url: news_item["url"] = info.get("url", "") news_item["mobileUrl"] = info.get("mobileUrl", "") matches.append(news_item) return matches def _search_by_fuzzy_mode( self, query: str, all_titles: Dict, id_to_name: Dict, current_date: datetime, threshold: float, include_url: bool ) -> List[Dict]: """ 模糊搜索模式（使用相似度算法） Args: query: 搜索内容 all_titles: 所有标题字典 id_to_name: 平台ID到名称映射 current_date: 当前日期 threshold: 相似度阈值 Returns: 匹配的新闻列表 """ matches = [] for platform_id, titles in all_titles.items(): platform_name = id_to_name.get(platform_id, platform_id) for title, info in titles.items(): # 模糊匹配 is_match, similarity = self._fuzzy_match(query, title, threshold) if is_match: news_item = { "title": title, "platform": platform_id, "platform_name": platform_name, "date": current_date.strftime("%Y-%m-%d"), "similarity_score": round(similarity, 4), "ranks": info.get("ranks", []), "count": len(info.get("ranks", [])), "rank": info["ranks"][0] if info["ranks"] else 999 } # 条件性添加 URL 字段 if include_url: news_item["url"] = info.get("url", "") news_item["mobileUrl"] = info.get("mobileUrl", "") matches.append(news_item) return matches def _search_by_entity_mode( self, query: str, all_titles: Dict, id_to_name: Dict, current_date: datetime, include_url: bool ) -> List[Dict]: """ 实体搜索模式（自动按权重排序） Args: query: 实体名称 all_titles: 所有标题字典 id_to_name: 平台ID到名称映射 current_date: 当前日期 Returns: 匹配的新闻列表 """ matches = [] for platform_id, titles in all_titles.items(): platform_name = id_to_name.get(platform_id, platform_id) for title, info in titles.items(): # 实体搜索：精确包含实体名称 if query in title: news_item = { "title": title, "platform": platform_id, "platform_name": platform_name, "date": current_date.strftime("%Y-%m-%d"), "similarity_score": 1.0, "ranks": info.get("ranks", []), "count": len(info.get("ranks", [])), "rank": info["ranks"][0] if info["ranks"] else 999 } # 条件性添加 URL 字段 if include_url: news_item["url"] = info.get("url", "") news_item["mobileUrl"] = info.get("mobileUrl", "") matches.append(news_item) return matches def _calculate_similarity(self, text1: str, text2: str) -> float: """ 计算两个文本的相似度 Args: text1: 文本1 text2: 文本2 Returns: 相似度分数 (0-1之间) """ # 使用 difflib.SequenceMatcher 计算序列相似度 return SequenceMatcher(None, text1.lower(), text2.lower()).ratio() def _fuzzy_match(self, query: str, text: str, threshold: float = 0.3) -> Tuple[bool, float]: """ 模糊匹配函数 Args: query: 查询文本 text: 待匹配文本 threshold: 匹配阈值 Returns: (是否匹配, 相似度分数) """ # 直接包含判断 if query.lower() in text.lower(): return True, 1.0 # 计算整体相似度 similarity = self._calculate_similarity(query, text) if similarity >= threshold: return True, similarity # 分词后的部分匹配 query_words = set(self._extract_keywords(query)) text_words = set(self._extract_keywords(text)) if not query_words or not text_words: return False, 0.0 # 计算关键词重合度 common_words = query_words & text_words keyword_overlap = len(common_words) / len(query_words) if keyword_overlap >= 0.5: # 50%的关键词重合 return True, keyword_overlap return False, similarity def _extract_keywords(self, text: str, min_length: int = 2) -> List[str]: """ 从文本中提取关键词 Args: text: 输入文本 min_length: 最小词长 Returns: 关键词列表 """ # 移除URL和特殊字符 text = re.sub(r'http[s]?://\S+', '', text) text = re.sub(r'\[.*?\]', '', text) # 移除方括号内容 # 使用正则表达式分词（中文和英文） words = re.findall(r'[\w]+', text) # 过滤停用词和短词 keywords = [ word for word in words if word and len(word) >= min_length and word not in self.stopwords ] return keywords def _calculate_keyword_overlap(self, keywords1: List[str], keywords2: List[str]) -> float: """ 计算两个关键词列表的重合度 Args: keywords1: 关键词列表1 keywords2: 关键词列表2 Returns: 重合度分数 (0-1之间) """ if not keywords1 or not keywords2: return 0.0 set1 = set(keywords1) set2 = set(keywords2) # Jaccard 相似度 intersection = len(set1 & set2) union = len(set1 | set2) if union == 0: return 0.0 return intersection / union def search_related_news_history( self, reference_text: str, time_range: str = "yesterday", start_date: Optional[datetime] = None, end_date: Optional[datetime] = None, threshold: float = 0.4, limit: int = 50, include_url: bool = False ) -> Dict: """ 在历史数据中搜索与给定新闻相关的新闻 Args: reference_text: 参考新闻标题或内容 time_range: 时间范围预设值，可选： - "yesterday": 昨天 - "last_week": 上周 (7天) - "last_month": 上个月 (30天) - "custom": 自定义日期范围（需要提供 start_date 和 end_date） start_date: 自定义开始日期（仅当 time_range="custom" 时有效） end_date: 自定义结束日期（仅当 time_range="custom" 时有效） threshold: 相似度阈值 (0-1之间)，默认0.4 limit: 返回条数限制，默认50 include_url: 是否包含URL链接，默认False（节省token） Returns: 搜索结果字典，包含相关新闻列表 Example: >>> tools = SearchTools() >>> result = tools.search_related_news_history( ... reference_text="人工智能技术突破", ... time_range="last_week", ... threshold=0.4, ... limit=50 ... ) >>> for news in result['results']: ... print(f"{news['date']}: {news['title']} (相似度: {news['similarity_score']})") """ try: # 参数验证 reference_text = validate_keyword(reference_text) threshold = max(0.0, min(1.0, threshold)) limit = validate_limit(limit, default=50) # 确定查询日期范围 today = datetime.now() if time_range == "yesterday": search_start = today - timedelta(days=1) search_end = today - timedelta(days=1) elif time_range == "last_week": search_start = today - timedelta(days=7) search_end = today - timedelta(days=1) elif time_range == "last_month": search_start = today - timedelta(days=30) search_end = today - timedelta(days=1) elif time_range == "custom": if not start_date or not end_date: raise InvalidParameterError( "自定义时间范围需要提供 start_date 和 end_date", suggestion="请提供 start_date 和 end_date 参数" ) search_start = start_date search_end = end_date else: raise InvalidParameterError( f"不支持的时间范围: {time_range}", suggestion="请使用 'yesterday', 'last_week', 'last_month' 或 'custom'" ) # 提取参考文本的关键词 reference_keywords = self._extract_keywords(reference_text) if not reference_keywords: raise InvalidParameterError( "无法从参考文本中提取关键词", suggestion="请提供更详细的文本内容" ) # 收集所有相关新闻 all_related_news = [] current_date = search_start while current_date <= search_end: try: # 读取该日期的数据 all_titles, id_to_name, _ = self.data_service.parser.read_all_titles_for_date(current_date) # 搜索相关新闻 for platform_id, titles in all_titles.items(): platform_name = id_to_name.get(platform_id, platform_id) for title, info in titles.items(): # 计算标题相似度 title_similarity = self._calculate_similarity(reference_text, title) # 提取标题关键词 title_keywords = self._extract_keywords(title) # 计算关键词重合度 keyword_overlap = self._calculate_keyword_overlap( reference_keywords, title_keywords ) # 综合相似度 (70% 关键词重合 + 30% 文本相似度) combined_score = keyword_overlap * 0.7 + title_similarity * 0.3 if combined_score >= threshold: news_item = { "title": title, "platform": platform_id, "platform_name": platform_name, "date": current_date.strftime("%Y-%m-%d"), "similarity_score": round(combined_score, 4), "keyword_overlap": round(keyword_overlap, 4), "text_similarity": round(title_similarity, 4), "common_keywords": list(set(reference_keywords) & set(title_keywords)), "rank": info["ranks"][0] if info["ranks"] else 0 } # 条件性添加 URL 字段 if include_url: news_item["url"] = info.get("url", "") news_item["mobileUrl"] = info.get("mobileUrl", "") all_related_news.append(news_item) except DataNotFoundError: # 该日期没有数据，继续下一天 pass except Exception as e: # 记录错误但继续处理其他日期 print(f"Warning: 处理日期 {current_date.strftime('%Y-%m-%d')} 时出错: {e}") # 移动到下一天 current_date += timedelta(days=1) if not all_related_news: return { "success": True, "results": [], "total": 0, "query": reference_text, "time_range": time_range, "date_range": { "start": search_start.strftime("%Y-%m-%d"), "end": search_end.strftime("%Y-%m-%d") }, "message": "未找到相关新闻" } # 按相似度排序 all_related_news.sort(key=lambda x: x["similarity_score"], reverse=True) # 限制返回数量 results = all_related_news[:limit] # 统计信息 platform_distribution = Counter([news["platform"] for news in all_related_news]) date_distribution = Counter([news["date"] for news in all_related_news]) result = { "success": True, "summary": { "total_found": len(all_related_news), "returned_count": len(results), "requested_limit": limit, "threshold": threshold, "reference_text": reference_text, "reference_keywords": reference_keywords, "time_range": time_range, "date_range": { "start": search_start.strftime("%Y-%m-%d"), "end": search_end.strftime("%Y-%m-%d") } }, "results": results, "statistics": { "platform_distribution": dict(platform_distribution), "date_distribution": dict(date_distribution), "avg_similarity": round( sum([news["similarity_score"] for news in all_related_news]) / len(all_related_news), 4 ) if all_related_news else 0.0 } } if len(all_related_news) < limit: result["note"] = f"相关性阈值 {threshold} 下仅找到 {len(all_related_news)} 条相关新闻" return result except MCPError as e: return { "success": False, "error": e.to_dict() } except Exception as e: return { "success": False, "error": { "code": "INTERNAL_ERROR", "message": str(e) } }