from typing import Dict, List, Optional from landmark_prompts import LandmarkPrompts from brand_prompts import BrandPrompts from scene_prompts import ScenePrompts from universal_object_prompts import UniversalObjectPrompts class PromptLibraryManager: """ Facade 模式：統一管理所有 Prompt 子模組 提供單一介面存取品牌、地標、場景、通用物品等 prompts """ def __init__(self): """初始化所有 Prompt 子模組""" print("Initializing Prompt Library Manager (Facade)...") # 載入所有子模組 self.brand_prompts = BrandPrompts() self.landmark_prompts = LandmarkPrompts() self.scene_prompts = ScenePrompts() self.object_prompts = UniversalObjectPrompts() # 統計資訊 total_brands = self._count_brands() total_landmarks = len(self.landmark_prompts.landmarks) total_scenes = len(self.scene_prompts.scene_vocabularies) total_objects = len(self.object_prompts.object_vocabularies) print(f"✓ Prompt Library Manager initialized:") print(f" - {total_brands} brands across {len(self.brand_prompts.brand_prompts)} categories") print(f" - {total_landmarks} world landmarks") print(f" - {total_scenes} scene categories") print(f" - {total_objects} universal object categories") def _count_brands(self) -> int: """計算總品牌數量""" total = 0 for category in self.brand_prompts.brand_prompts.values(): total += len(category) return total # ===== 品牌相關方法 Brand Methods ===== def get_brand_prompts(self, brand_name: str) -> Optional[Dict]: """ 取得特定品牌的完整 prompt 資料 Args: brand_name: 品牌名稱 Returns: 品牌資料字典 """ return self.brand_prompts.get_prompts(brand_name) def get_brand_category(self, brand_name: str) -> str: """取得品牌類別""" return self.brand_prompts.get_brand_category(brand_name) def get_all_brands(self) -> Dict: """取得所有品牌的扁平化字典""" return self.brand_prompts.get_all_brands() def get_brands_by_category(self, category: str) -> Dict: """取得特定類別的所有品牌""" return self.brand_prompts.get_brands_by_category(category) def search_brand_by_alias(self, alias: str) -> Optional[str]: """根據別名搜尋品牌名稱""" return self.brand_prompts.search_brand_by_alias(alias) # ===== 地標相關方法 Landmark Methods ===== def get_landmark_prompts(self, landmark_name: str) -> Optional[Dict]: """ 取得特定地標的完整 prompt 資料 Args: landmark_name: 地標名稱 Returns: 地標資料字典 """ return self.landmark_prompts.get_prompts(landmark_name) def get_all_landmarks(self) -> Dict: """取得所有地標資料""" return self.landmark_prompts.get_all_landmarks() def search_landmark_by_location(self, city: str = None, country: str = None) -> List[str]: """ 根據地理位置搜尋地標 Args: city: 城市名稱 country: 國家名稱 Returns: 符合條件的地標名稱列表 """ return self.landmark_prompts.search_by_location(city, country) def get_landmark_visual_prompts(self, landmark_name: str, context: str = 'iconic_view') -> List[str]: """ 取得地標的視覺描述 prompts Args: landmark_name: 地標名稱 context: 情境類型 Returns: 視覺描述列表 """ return self.landmark_prompts.get_visual_prompts(landmark_name, context) # Scene Methods def get_scene_prompts(self, scene_category: str, subcategory: str = None) -> List[str]: """ 取得場景 prompts Args: scene_category: 場景類別 subcategory: 子類別（可選） Returns: Prompt 列表 """ return self.scene_prompts.get_prompts(scene_category, subcategory) def get_all_scene_categories(self) -> List[str]: """取得所有場景類別""" return self.scene_prompts.get_all_categories() def get_scene_subcategories(self, scene_category: str) -> List[str]: """取得場景的子類別""" return self.scene_prompts.get_subcategories(scene_category) # Universal Object Methods def get_object_prompts(self, category: str, subcategory: str = None) -> List[str]: """ 取得通用物品 prompts Args: category: 物品類別 (如 'animals', 'vehicles') subcategory: 子類別 (如 'dogs', 'cats') Returns: Prompt 列表 """ return self.object_prompts.get_prompts(category, subcategory) def get_all_object_categories(self) -> List[str]: """取得所有通用物品類別""" return self.object_prompts.get_all_categories() def get_object_subcategories(self, category: str) -> List[str]: """取得物品的子類別""" return self.object_prompts.get_subcategories(category) def detect_object_category(self, detected_objects: List[str]) -> Optional[str]: """根據檢測到的物體推測主要類別""" return self.object_prompts.detect_object_category(detected_objects) # Smart Hashtag Generation def get_hashtags_for_content(self, detected_items: Dict, language: str = 'zh') -> List[str]: """ 智能標籤生成：整合品牌、地標、場景的標籤 Args: detected_items: 檢測到的內容字典 { 'landmarks': ['Big Ben', ...], 'brands': ['Apple', ...], 'scene_category': 'urban', 'scene_subcategory': 'city_canyon' } language: 語言 ('zh', 'en', 或 'zh-en') Returns: Hashtag 列表（去重並排序） """ hashtags = [] # 1. 地標標籤（最高優先級） landmarks = detected_items.get('landmarks', []) for landmark in landmarks: landmark_tags = self.landmark_prompts.get_hashtags(landmark, language) hashtags.extend(landmark_tags) # 2. 品牌標籤（高優先級） brands = detected_items.get('brands', []) for brand in brands: brand_tags = self.brand_prompts.get_hashtags(brand, language) hashtags.extend(brand_tags) # 3. 場景標籤（中優先級） scene_category = detected_items.get('scene_category') if scene_category: scene_tags = self.scene_prompts.get_hashtags(scene_category, language) hashtags.extend(scene_tags) # 去重並保持順序（地標 > 品牌 > 場景） seen = set() unique_hashtags = [] for tag in hashtags: if tag not in seen: seen.add(tag) unique_hashtags.append(tag) # 返回前 10 個 return unique_hashtags[:10] # Search Functions def search_by_location(self, city: str = None, country: str = None) -> Dict: """ 根據地點搜尋所有相關內容（地標、品牌） Args: city: 城市名稱 country: 國家名稱 Returns: 搜尋結果字典 """ results = { 'landmarks': [], 'brands': [] } # 搜尋地標 landmarks = self.landmark_prompts.search_by_location(city, country) results['landmarks'] = landmarks # 品牌通常不按地理位置分類，但可以擴展此功能 return results def detect_landmark_from_image_context(self, detected_objects: List[str], scene_analysis: Dict) -> Optional[str]: """ 根據檢測到的物體和場景分析推測可能的地標 Args: detected_objects: 檢測到的物體列表 scene_analysis: 場景分析結果 Returns: 推測的地標名稱，若無法推測則返回 None """ # 關鍵字映射到地標 landmark_keywords = { 'Big Ben': ['clock tower', 'tower', 'bridge', 'river'], 'Eiffel Tower': ['tower', 'iron structure', 'landmark'], 'Statue of Liberty': ['statue', 'monument', 'island', 'harbor'], 'Sydney Opera House': ['building', 'harbor', 'architecture'], 'Taj Mahal': ['building', 'monument', 'dome'], 'Pyramids of Giza': ['pyramid', 'desert', 'monument'], # 可以擴展更多 } # 簡單的關鍵字匹配 for landmark, keywords in landmark_keywords.items(): match_count = sum(1 for obj in detected_objects if any(kw in obj.lower() for kw in keywords)) if match_count >= 2: # 至少匹配 2 個關鍵字 return landmark return None print("✓ PromptLibraryManager (Facade) defined")