""" Query parser - main module for query processing. Handles query rewriting, translation, and embedding generation. """ from typing import Dict, List, Optional, Any import numpy as np from config import SearchConfig, QueryConfig from embeddings import BgeEncoder from .language_detector import LanguageDetector from .translator import Translator from .query_rewriter import QueryRewriter, QueryNormalizer class ParsedQuery: """Container for parsed query results.""" def __init__( self, original_query: str, normalized_query: str, rewritten_query: Optional[str] = None, detected_language: str = "unknown", translations: Dict[str, str] = None, query_vector: Optional[np.ndarray] = None, domain: str = "default" ): self.original_query = original_query self.normalized_query = normalized_query self.rewritten_query = rewritten_query or normalized_query self.detected_language = detected_language self.translations = translations or {} self.query_vector = query_vector self.domain = domain def to_dict(self) -> Dict[str, Any]: """Convert to dictionary representation.""" result = { "original_query": self.original_query, "normalized_query": self.normalized_query, "rewritten_query": self.rewritten_query, "detected_language": self.detected_language, "translations": self.translations, "domain": self.domain, "has_vector": self.query_vector is not None } return result class QueryParser: """ Main query parser that processes queries through multiple stages: 1. Normalization 2. Query rewriting (brand/category mappings, synonyms) 3. Language detection 4. Translation to target languages 5. Text embedding generation (for semantic search) """ def __init__( self, config: SearchConfig, text_encoder: Optional[BgeEncoder] = None, translator: Optional[Translator] = None ): """ Initialize query parser. Args: config: Search configuration text_encoder: Text embedding encoder (lazy loaded if not provided) translator: Translator instance (lazy loaded if not provided) """ self.config = config self.query_config = config.query_config self._text_encoder = text_encoder self._translator = translator # Initialize components self.normalizer = QueryNormalizer() self.language_detector = LanguageDetector() self.rewriter = QueryRewriter(self.query_config.rewrite_dictionary) @property def text_encoder(self) -> BgeEncoder: """Lazy load text encoder.""" if self._text_encoder is None and self.query_config.enable_text_embedding: print("[QueryParser] Initializing text encoder...") self._text_encoder = BgeEncoder() return self._text_encoder @property def translator(self) -> Translator: """Lazy load translator.""" if self._translator is None and self.query_config.enable_translation: print("[QueryParser] Initializing translator...") self._translator = Translator( api_key=self.query_config.translation_api_key, use_cache=True, glossary_id=getattr(self.query_config, 'translation_glossary_id', None), translation_context=getattr(self.query_config, 'translation_context', 'e-commerce product search') ) return self._translator def parse(self, query: str, generate_vector: bool = True, context: Optional[Any] = None) -> ParsedQuery: """ Parse query through all processing stages. Args: query: Raw query string generate_vector: Whether to generate query embedding context: Optional request context for tracking and logging Returns: ParsedQuery object with all processing results """ # Initialize logger if context provided logger = context.logger if context else None if logger: logger.info( f"开始查询解析 | 原查询: '{query}' | 生成向量: {generate_vector}", extra={'reqid': context.reqid, 'uid': context.uid} ) # Use print statements for backward compatibility if no context def log_info(msg): if logger: logger.info(msg, extra={'reqid': context.reqid, 'uid': context.uid}) else: print(f"[QueryParser] {msg}") def log_debug(msg): if logger: logger.debug(msg, extra={'reqid': context.reqid, 'uid': context.uid}) else: print(f"[QueryParser] {msg}") # Stage 1: Normalize normalized = self.normalizer.normalize(query) log_debug(f"标准化完成 | '{query}' -> '{normalized}'") if context: context.store_intermediate_result('normalized_query', normalized) # Extract domain if present (e.g., "brand:Nike" -> domain="brand", query="Nike") domain, query_text = self.normalizer.extract_domain_query(normalized) log_debug(f"域提取 | 域: '{domain}', 查询: '{query_text}'") if context: context.store_intermediate_result('extracted_domain', domain) context.store_intermediate_result('domain_query', query_text) # Stage 2: Query rewriting rewritten = None if self.query_config.enable_query_rewrite: rewritten = self.rewriter.rewrite(query_text) if rewritten != query_text: log_info(f"查询重写 | '{query_text}' -> '{rewritten}'") query_text = rewritten if context: context.store_intermediate_result('rewritten_query', rewritten) context.add_warning(f"查询被重写: {query_text}") # Stage 3: Language detection detected_lang = self.language_detector.detect(query_text) log_info(f"语言检测 | 检测到语言: {detected_lang}") if context: context.store_intermediate_result('detected_language', detected_lang) # Stage 4: Translation translations = {} if self.query_config.enable_translation: try: # Determine target languages for translation # If domain has language_field_mapping, only translate to languages in the mapping # Otherwise, use all supported languages target_langs_for_translation = self.query_config.supported_languages # Check if domain has language_field_mapping domain_config = next( (idx for idx in self.config.indexes if idx.name == domain), None ) if domain_config and domain_config.language_field_mapping: # Only translate to languages that exist in the mapping available_languages = set(domain_config.language_field_mapping.keys()) target_langs_for_translation = [ lang for lang in self.query_config.supported_languages if lang in available_languages ] log_debug(f"域 '{domain}' 有语言字段映射，将翻译到: {target_langs_for_translation}") target_langs = self.translator.get_translation_needs( detected_lang, target_langs_for_translation ) if target_langs: log_info(f"开始翻译 | 源语言: {detected_lang} | 目标语言: {target_langs}") # Use e-commerce context for better disambiguation translation_context = getattr(self.query_config, 'translation_context', 'e-commerce product search') translations = self.translator.translate_multi( query_text, target_langs, source_lang=detected_lang, context=translation_context ) log_info(f"翻译完成 | 结果: {translations}") if context: context.store_intermediate_result('translations', translations) for lang, translation in translations.items(): if translation: context.store_intermediate_result(f'translation_{lang}', translation) except Exception as e: error_msg = f"翻译失败 | 错误: {str(e)}" log_info(error_msg) if context: context.add_warning(error_msg) # Stage 5: Text embedding query_vector = None if (generate_vector and self.query_config.enable_text_embedding and domain == "default"): # Only generate vector for default domain try: log_debug("开始生成查询向量") query_vector = self.text_encoder.encode([query_text])[0] log_debug(f"查询向量生成完成 | 形状: {query_vector.shape}") if context: context.store_intermediate_result('query_vector_shape', query_vector.shape) except Exception as e: error_msg = f"查询向量生成失败 | 错误: {str(e)}" log_info(error_msg) if context: context.add_warning(error_msg) # Build result result = ParsedQuery( original_query=query, normalized_query=normalized, rewritten_query=rewritten, detected_language=detected_lang, translations=translations, query_vector=query_vector, domain=domain ) if logger: logger.info( f"查询解析完成 | 原查询: '{query}' | 最终查询: '{rewritten or query_text}' | " f"语言: {detected_lang} | 域: {domain} | " f"翻译数量: {len(translations)} | 向量: {'是' if query_vector is not None else '否'}", extra={'reqid': context.reqid, 'uid': context.uid} ) else: print(f"[QueryParser] Parsing complete") return result def get_search_queries(self, parsed_query: ParsedQuery) -> List[str]: """ Get list of queries to search (original + translations). Args: parsed_query: Parsed query object Returns: List of query strings to search """ queries = [parsed_query.rewritten_query] # Add translations for lang, translation in parsed_query.translations.items(): if translation and translation != parsed_query.rewritten_query: queries.append(translation) return queries def update_rewrite_rules(self, rules: Dict[str, str]) -> None: """ Update query rewrite rules. Args: rules: Dictionary of pattern -> replacement mappings """ for pattern, replacement in rules.items(): self.rewriter.add_rule(pattern, replacement) def get_rewrite_rules(self) -> Dict[str, str]: """Get current rewrite rules.""" return self.rewriter.get_rules()