# 自动学习群组特定用语
"发财了" → "表示惊喜或获得好处"
"下次一定" → "委婉拒绝的表达"
"🦌" → "xxxxx"

# 自动检测对话目标并动态规划阶段
用户: "我最近工作压力好大..."
Bot: 检测目标 → emotional_support (情感支持)
     规划阶段 → ["倾听诉说", "识别核心问题", "表达理解", "提供建议", "给予鼓励"]
     当前阶段 → "倾听诉说"

# 统一的服务创建和管理
factory_manager = FactoryManager()
factory_manager.initialize_factories(config, context)

# 服务工厂
service_factory = factory_manager.get_service_factory()
db_manager = service_factory.create_database_manager()

# 组件工厂
component_factory = factory_manager.get_component_factory()
expression_learner = component_factory.create_expression_pattern_learner()

# 灵活的学习策略
learning_strategy = StrategyFactory.create_strategy(
    LearningStrategyType.BATCH,  # INCREMENTAL / REINFORCEMENT
    config={'batch_size': 100}
)

# 服务间松耦合
class MultidimensionalAnalyzer:
    def __init__(self, config, db_manager, llm_adapter, ...):
        self.config = config
        self.db = db_manager
        self.llm = llm_adapter

# 数据访问层抽象
affection_repo = AffectionRepository(session)
user_affection = await affection_repo.get_user_affection(group_id, user_id)

class FactoryManager:
    """全局工厂管理器 - 单例模式"""

    def initialize_factories(self, config, context):
        self._service_factory = ServiceFactory(config, context)
        self._component_factory = ComponentFactory(config, self._service_factory)

    def get_service_factory(self) -> ServiceFactory:
        # 服务工厂：创建业务服务（数据库、学习、分析等）
        return self._service_factory

    def get_component_factory(self) -> ComponentFactory:
        # 组件工厂：创建轻量级组件（过滤器、调度器等）
        return self._component_factory

class StrategyFactory:
    @staticmethod
    def create_strategy(strategy_type: LearningStrategyType, config: dict):
        strategies = {
            LearningStrategyType.BATCH: BatchLearningStrategy,
            LearningStrategyType.INCREMENTAL: IncrementalLearningStrategy,
            LearningStrategyType.REINFORCEMENT: ReinforcementLearningStrategy
        }
        return strategies[strategy_type](config)

class BaseRepository:
    """基础仓储 - 提供通用 CRUD 操作"""

    async def get(self, id: int):
        async with self.session() as session:
            return await session.get(self.model, id)

    async def save(self, entity):
        async with self.session() as session:
            session.add(entity)
            await session.commit()

class ExpressionRepository(BaseRepository):
    """表达模式仓储 - 专门处理表达模式数据"""

    async def get_patterns_by_group(self, group_id: str, limit: int = 10):
        # 特定业务逻辑
        ...

class EventBus:
    """事件总线 - 发布/订阅模式"""

    def subscribe(self, event_type: str, handler: Callable):
        self._handlers[event_type].append(handler)

    async def publish(self, event_type: str, data: Any):
        for handler in self._handlers[event_type]:
            await handler(data)

# 使用示例
event_bus.subscribe("learning_completed", on_learning_completed)
await event_bus.publish("learning_completed", learning_result)

# 1. 候选词提取
candidates = await jargon_miner.extract_candidates(recent_messages)
# 输出: ["发财了", "下次一定", "真香"]

# 2. LLM 推断含义
meanings = await jargon_miner.infer_meanings(candidates, context)
# 输出: {
#   "发财了": "表示惊喜或获得好处",
#   "下次一定": "委婉拒绝的表达"
# }

# 3. 保存到数据库
await db.save_jargon_batch(chat_id, meanings)

# 4. LLM 请求时注入理解
jargon_explanation = await jargon_query.check_and_explain_jargon(
    text="今天真是发财了！",
    chat_id=group_id
)
# 注入到 LLM prompt: "文本中包含的黑话: 「发财了」: 表示惊喜或获得好处"

# 情感支持类 (5种)
comfort           # 安慰用户
emotional_support # 情感支持
empathy          # 深度共情
encouragement    # 鼓励打气
calm_down        # 情绪安抚

# 信息交流类 (5种)
qa               # 解答疑问
guide_share      # 引导分享
teach            # 教学指导
discuss          # 深度讨论
storytelling     # 讲故事

# 娱乐互动类 (6种)
casual_chat      # 闲聊互动
tease            # 友好调侃
flirt            # 俏皮调戏
joke             # 幽默搞笑
meme             # 梗文化互动
roleplay         # 角色扮演

# 社交互动类 (5种)
greeting         # 问候寒暄
compliment       # 赞美夸奖
celebrate        # 庆祝祝贺
apologize        # 道歉和解
gossip           # 八卦闲聊

# 建议指导类 (4种)
advise           # 提供建议
brainstorm       # 头脑风暴
plan             # 制定计划
analyze          # 分析问题

# 情绪调节类 (3种)
vent             # 倾听发泄
motivate         # 激励鼓舞
complaint        # 抱怨吐槽

# 兴趣分享类 (3种)
recommend        # 推荐分享
review           # 评价点评
hobby_chat       # 爱好交流

# 特殊场景类 (3种)
debate           # 友好辩论
confess          # 倾诉秘密
nostalgia        # 怀旧回忆

# 冲突场景类 (4种)
argument         # 激烈争论
quarrel          # 吵架互怼
insult_exchange  # 互骂对喷
provoke          # 挑衅应对

# 会话ID生成 (24小时不变)
session_id = MD5(group_id + user_id + date)

# 会话数据结构
{
  "session_id": "sess_a1b2c3d4e5f6",
  "final_goal": {
    "type": "emotional_support",
    "name": "情感支持",
    "topic": "工作压力",
    "topic_status": "active"
  },
  "current_stage": {
    "index": 2,
    "task": "识别核心问题",
    "strategy": "顺序推进",
    "adjustment_reason": "上一阶段已完成"
  },
  "planned_stages": [
    "倾听诉说", "识别核心问题", "表达理解",
    "提供建议", "给予鼓励"
  ],
  "metrics": {
    "rounds": 5,
    "user_engagement": 0.8,
    "goal_progress": 0.4
  },
  "conversation_history": [...]
}

# 自动记录用户互动
await social_relation_manager.record_interaction(
    group_id="123456",
    user_a="10001",
    user_b="10002",
    interaction_type="mention"  # at / reply / topic_discussion
)

# 分析关系强度
strength = await social_relation_manager.calculate_relationship_strength(
    group_id, user_a, user_b
)
# 输出: 0.75 (基于互动频率、类型、时间衰减)

# 处理用户消息互动
result = await affection_manager.process_message_interaction(
    group_id, user_id, message_text
)

# 自动识别互动类型
interaction_types = {
    "praise": +10,      # 称赞
    "encourage": +5,    # 鼓励
    "insult": -15,      # 侮辱
    "harass": -20       # 骚扰
}

# 好感度限制
- 单用户上限: 100分
- 群组总分上限: 250分
- 超额时自动衰减旧好感度

# 直接在原人格末尾增量更新
await persona_manager_updater.apply_incremental_update(
    group_id=group_id,
    update_content="【学习成果】\n新增表达模式: ..."
)

# 优势:
# ✅ 自动创建备份人格
# ✅ 无需手动执行应用命令
# ✅ 更好的版本管理

# 临时存储到 persona_updates.txt
await temporary_persona_updater.write_to_updates_file(content)

# 手动审查后应用
/apply_persona_updates

# 优势:
# ✅ 人工审核把关
# ✅ 批量应用更新