""" 对话 Agent：基于访谈问题清单，动态选择问题，实时生成回应 支持异步调用和 Redis 会话存储 支持用户基础资料收集和时代背景融入 """ import json import logging from datetime import datetime from typing import List, Optional, Dict, Any from langchain_core.messages import HumanMessage, AIMessage from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder from services.llm_service import llm_service from services.redis_service import redis_service from .prompts import ConversationStage, get_conversation_prompt, get_guided_conversation_prompt, get_opening_prompt from .prompts.profile_prompts import ( get_profile_greeting_prompt, get_profile_extraction_prompt, get_profile_followup_prompt, format_user_profile_context, get_missing_profile_fields, ) from .prompts.conversation_prompts import SLOT_NAME_MAP from .state_schema import MemoirStateSchema logger = logging.getLogger(__name__) class ConversationAgent: """对话 Agent（支持异步和 Redis 存储）""" def __init__(self): # 使用 LLM 服务获取 LLM 实例 self.llm = llm_service.get_llm() async def _get_history_messages(self, conversation_id: str) -> List[Any]: """从 Redis 获取对话历史并转换为 LangChain 消息格式""" history = await redis_service.get_conversation_history(conversation_id) messages = [] for msg in history: if msg["role"] == "human": messages.append(HumanMessage(content=msg["content"])) elif msg["role"] == "ai": messages.append(AIMessage(content=msg["content"])) return messages async def _save_message( self, conversation_id: str, role: str, content: str, message_type: str = "text", voice_session_id: str | None = None, timestamp: datetime | str | int | None = None, ): """保存消息到 Redis""" await redis_service.add_message( conversation_id, role, content, message_type=message_type, voice_session_id=voice_session_id, timestamp=timestamp.isoformat() if isinstance(timestamp, datetime) else timestamp, ) def _format_history_string(self, messages: List[Any]) -> str: """将消息列表格式化为字符串（用于 prompt）""" history_parts = [] for msg in messages: if isinstance(msg, HumanMessage): history_parts.append(f"Human: {msg.content}") elif isinstance(msg, AIMessage): history_parts.append(f"Assistant: {msg.content}") return "\n\n".join(history_parts) async def generate_response( self, conversation_id: str, user_message: str, current_stage: Optional[ConversationStage] = None, covered_topics: Optional[List[str]] = None ) -> str: """ 异步生成 Agent 回应 Args: conversation_id: 对话 ID user_message: 用户消息 current_stage: 当前对话阶段 covered_topics: 已聊过的话题列表 Returns: Agent 回应文本 """ if current_stage is None: current_stage = ConversationStage.CHILDHOOD if covered_topics is None: covered_topics = [] # 如果没有配置 LLM，返回默认回应 if not self.llm: return "抱歉，LLM 服务未配置。请设置 DEEPSEEK_API_KEY 或 LLM_API_KEY 环境变量。" try: # 获取系统提示词 system_prompt = get_conversation_prompt(current_stage, covered_topics, user_message) # 从 Redis 获取对话历史 history_messages = await self._get_history_messages(conversation_id) history_string = self._format_history_string(history_messages) # 构建完整 prompt full_prompt = f"{system_prompt}\n\n{history_string}\n\nHuman: {user_message}\n\nAssistant:" # 异步调用 LLM response = await self.llm.ainvoke(full_prompt) response_text = response.content if hasattr(response, 'content') else str(response) # 保存对话到 Redis await self._save_message(conversation_id, "human", user_message) await self._save_message(conversation_id, "ai", response_text) return response_text except Exception as e: logger.error(f"生成回应失败: {e}") return f"抱歉，生成回应时出现错误: {str(e)}" async def generate_profile_greeting( self, conversation_id: str, missing_fields: List[str], nickname: str = "", ) -> List[str]: """生成资料收集的开场白（首次对话时使用）""" if not self.llm: return ["你好！在开始之前，能告诉我你是哪一年出生的吗？"] try: prompt = get_profile_greeting_prompt(missing_fields, nickname) history_messages = await self._get_history_messages(conversation_id) history_string = self._format_history_string(history_messages) full_prompt = f"{prompt}\n\n{history_string}" if history_string else prompt response = await self.llm.ainvoke(full_prompt) response_text = response.content if hasattr(response, 'content') else str(response) await self._save_message(conversation_id, "ai", response_text) # 最多 2 段，防止 LLM 自问自答 messages = [msg.strip() for msg in response_text.split("[SPLIT]") if msg.strip()] return messages[:2] if messages else [response_text] except Exception as e: logger.error(f"生成资料收集开场白失败: {e}") return ["你好！在我们开始聊人生故事之前，能先简单介绍一下你自己吗？比如你是哪一年出生的？"] async def generate_opening_message( self, conversation_id: str, memoir_state: MemoirStateSchema, user_profile_context: str = "", ) -> List[str]: """ 空对话时 AI 先开口：用户通过「打个招呼」进入，尚未发任何消息。 生成问候 + 一个引导问题，写入 Redis 并返回消息列表。 """ if not self.llm: return ["你好呀～ 有空聊聊你的人生故事吗？你小时候是在哪儿长大的？"] try: empty_slots = memoir_state.empty_slots_for_current_stage() empty_slots_readable = [SLOT_NAME_MAP.get(s, s) for s in empty_slots] if not empty_slots_readable: empty_slots_readable = ["成长的地方", "难忘的事", "重要的人"] prompt = get_opening_prompt( current_stage=memoir_state.current_stage, empty_slots_readable=empty_slots_readable, user_profile_context=user_profile_context, ) full_prompt = f"{prompt}\n\nAssistant:" response = await self.llm.ainvoke(full_prompt) response_text = response.content if hasattr(response, "content") else str(response) await self._save_message(conversation_id, "ai", response_text) # 最多 2 段：问候 + 问题，防止 LLM 自问自答 messages = [msg.strip() for msg in response_text.split("[SPLIT]") if msg.strip()] return messages[:2] if messages else [response_text] except Exception as e: logger.error(f"生成开场白失败: {e}", exc_info=True) return ["你好呀～ 有空聊聊你的人生故事吗？你童年里印象最深的一件事是什么？"] async def extract_profile_from_message( self, user_message: str, missing_fields: List[str], conversation_id: Optional[str] = None, ) -> Dict[str, Any]: """从用户消息中提取基础资料信息；若提供 conversation_id，会结合最近几轮对话一起提取，避免漏提。""" if not self.llm or not missing_fields: return {} recent_dialogue = "" if conversation_id: history_messages = await self._get_history_messages(conversation_id) # 取最近 4 条（2 轮），不包含本轮；本轮由 user_message 单独传入 recent = history_messages[-4:] if len(history_messages) > 4 else history_messages parts = [] for msg in recent: if isinstance(msg, HumanMessage): parts.append(f"用户: {msg.content}") elif isinstance(msg, AIMessage): parts.append(f"助手: {msg.content}") recent_dialogue = "\n".join(parts) if parts else "" try: prompt = get_profile_extraction_prompt( user_message, missing_fields, recent_dialogue=recent_dialogue or None ) response = await self.llm.ainvoke(prompt) content = response.content.strip() parsed = json.loads(content) result = {} if "birth_year" in parsed and parsed["birth_year"] is not None: raw = parsed["birth_year"] if isinstance(raw, int) and 1900 <= raw <= 2100: result["birth_year"] = raw elif isinstance(raw, str) and raw.isdigit(): y = int(raw) if y < 100: # "65" -> 1965 y = 1900 + y if y >= 50 else 2000 + y if 1900 <= y <= 2100: result["birth_year"] = y if "birth_place" in parsed and parsed["birth_place"]: result["birth_place"] = str(parsed["birth_place"]) if "grew_up_place" in parsed and parsed["grew_up_place"]: result["grew_up_place"] = str(parsed["grew_up_place"]) if "occupation" in parsed and parsed["occupation"]: result["occupation"] = str(parsed["occupation"]) return result except (json.JSONDecodeError, Exception) as e: logger.error(f"提取资料信息失败: {e}") return {} async def generate_profile_followup( self, conversation_id: str, user_message: str, missing_fields: List[str], filled_fields: Dict[str, str], nickname: str = "", is_from_voice: bool = False, voice_session_id: str | None = None, user_message_timestamp: datetime | None = None, ) -> List[str]: """在资料收集过程中生成跟进回复""" if not self.llm: return ["谢谢！还能告诉我更多吗？"] try: prompt = get_profile_followup_prompt(missing_fields, filled_fields, user_message, nickname) history_messages = await self._get_history_messages(conversation_id) history_string = self._format_history_string(history_messages) full_prompt = f"{prompt}\n\n{history_string}\n\nHuman: {user_message}\n\nAssistant:" response = await self.llm.ainvoke(full_prompt) response_text = response.content if hasattr(response, 'content') else str(response) human_msg_type = "audio" if is_from_voice else "text" await self._save_message( conversation_id, "human", user_message, message_type=human_msg_type, voice_session_id=voice_session_id, timestamp=user_message_timestamp, ) await self._save_message(conversation_id, "ai", response_text) messages = [msg.strip() for msg in response_text.split("[SPLIT]") if msg.strip()] return messages[:3] if messages else [response_text] except Exception as e: logger.error(f"生成资料跟进回复失败: {e}") return ["谢谢分享！能再告诉我一些吗？"] def _detect_user_stage(self, user_message: str) -> str: """ 通过关键词检测用户当前正在谈论的人生阶段。 返回阶段名称字符串，未检测到返回空字符串。 """ message = user_message.lower() stage_keywords = { "childhood": ["童年", "小时候", "出生", "家乡", "小镇", "爸妈", "父亲", "母亲", "爷爷", "奶奶", "外公", "外婆", "幼儿园"], "education": ["上学", "学校", "老师", "同学", "教育", "大学", "高中", "初中", "小学", "考试", "毕业", "读书", "高考", "课堂"], "career": ["工作", "职业", "事业", "公司", "同事", "创业", "升职", "跳槽", "老板", "行业", "项目", "加班", "薪水", "面试"], "family": ["伴侣", "孩子", "家庭", "家人", "结婚", "爱人", "老婆", "老公", "丈夫", "妻子", "儿子", "女儿", "婚礼", "恋爱"], "belief": ["信念", "价值观", "座右铭", "坚持", "原则", "信仰", "意义", "感悟", "遗憾", "骄傲"], } for stage, keywords in stage_keywords.items(): if any(word in message for word in keywords): return stage return "" async def generate_response_with_state( self, conversation_id: str, user_message: str, memoir_state: MemoirStateSchema, user_profile_context: str = "", is_from_voice: bool = False, voice_session_id: str | None = None, user_message_timestamp: datetime | None = None, ) -> List[str]: """ 基于共享状态异步生成引导式回复 Args: conversation_id: 对话 ID user_message: 用户消息 memoir_state: 共享状态 user_profile_context: 用户基础资料上下文 is_from_voice: 用户消息是否来自语音转写（用于保存正确的 messageType） Returns: Agent 回应文本列表（支持多条消息） """ if not self.llm: return ["抱歉，LLM 服务未配置。请设置 DEEPSEEK_API_KEY 或 LLM_API_KEY 环境变量。"] try: empty_slots = memoir_state.empty_slots_for_current_stage() filled_slots = { key: value.snippet for key, value in memoir_state.slots.get(memoir_state.current_stage, {}).items() if value.snippet } detected_user_stage = self._detect_user_stage(user_message) history_messages = await self._get_history_messages(conversation_id) conversation_turn = len(history_messages) // 2 same_topic_turns = self._estimate_same_topic_turns(history_messages, filled_slots) all_stages_coverage = memoir_state.all_stages_coverage() system_prompt = get_guided_conversation_prompt( current_stage=memoir_state.current_stage, empty_slots=empty_slots, filled_slots=filled_slots, user_message=user_message, conversation_turn=conversation_turn, same_topic_turns=same_topic_turns, all_stages_coverage=all_stages_coverage, detected_user_stage=detected_user_stage, user_profile_context=user_profile_context, ) history_string = self._format_history_string(history_messages) full_prompt = f"{system_prompt}\n\n{history_string}\n\nHuman: {user_message}\n\nAssistant:" response = await self.llm.ainvoke(full_prompt) response_text = response.content if hasattr(response, 'content') else str(response) human_msg_type = "audio" if is_from_voice else "text" await self._save_message( conversation_id, "human", user_message, message_type=human_msg_type, voice_session_id=voice_session_id, timestamp=user_message_timestamp, ) await self._save_message(conversation_id, "ai", response_text) messages = [msg.strip() for msg in response_text.split("[SPLIT]") if msg.strip()] return messages[:3] if messages else [response_text] except Exception as e: logger.error(f"生成回应失败: {e}") return [f"抱歉，生成回应时出现错误: {str(e)}"] def _estimate_same_topic_turns(self, history_messages: List[Any], current_filled_slots: dict) -> int: """ 估算同一话题的轮数 通过分析最近几轮对话来判断是否一直在同一个话题上 """ if len(history_messages) < 4: return len(history_messages) // 2 # 简单策略：检查最近的对话是否有重复关键词 recent_messages = history_messages[-6:] # 最近3轮 # 提取关键词（简单实现） keywords_per_turn = [] for i in range(0, len(recent_messages), 2): if i + 1 < len(recent_messages): human_msg = recent_messages[i].content if hasattr(recent_messages[i], 'content') else str(recent_messages[i]) ai_msg = recent_messages[i+1].content if hasattr(recent_messages[i+1], 'content') else str(recent_messages[i+1]) combined = human_msg + ai_msg keywords_per_turn.append(combined[:100]) # 取前100字作为特征 # 如果连续3轮都在讨论相似内容，认为同一话题 if len(keywords_per_turn) >= 3: return 3 return len(keywords_per_turn) def detect_stage(self, conversation_id: str, user_message: str) -> ConversationStage: """ 检测对话阶段 Args: conversation_id: 对话 ID user_message: 用户消息 Returns: 检测到的对话阶段 """ # 简单的关键词检测（实际应该使用更智能的方法） message_lower = user_message.lower() if any(word in message_lower for word in ["童年", "小时候", "出生", "家庭背景"]): return ConversationStage.CHILDHOOD elif any(word in message_lower for word in ["上学", "学校", "老师", "同学", "教育"]): return ConversationStage.EDUCATION elif any(word in message_lower for word in ["工作", "职业", "事业", "公司", "同事"]): return ConversationStage.CAREER elif any(word in message_lower for word in ["伴侣", "孩子", "家庭", "家人", "结婚"]): return ConversationStage.FAMILY elif any(word in message_lower for word in ["信念", "价值观", "座右铭", "坚持", "原则"]): return ConversationStage.BELIEFS elif any(word in message_lower for word in ["总结", "回顾", "感激", "希望", "未来"]): return ConversationStage.SUMMARY else: # 默认返回当前阶段或童年阶段 return ConversationStage.CHILDHOOD async def clear_memory(self, conversation_id: str): """清除对话记忆（从 Redis）""" await redis_service.clear_conversation_history(conversation_id)