feat(api+app): 对话阶段化、回忆录流水线与客户端会话体验

- DB: segments 用户输入文本（Alembic 0002）
- Chat: 阶段检测/阶段提示/回复限制，编排与访谈/画像 prompts 调整
- Memoir: 忠实度检查 agent，叙事与分类等链路更新
- Core: agent 日志、Alembic 启动、LangChain/日志/配置等
- Story: time_hints；Memory 检索与相关测试
- Expo: 助手头像、会话页与消息拆分、实时会话与文案/i18n
- Docs/scripts/tests: 迁移脚本、LLM JSON/记忆检索文档、新增单测

api/app/agents/chat/interview_agent.py

@@ -3,9 +3,10 @@ InterviewAgent：正式访谈 Specialist
 负责状态感知回复、开场白，不负责 Redis 持久化（由 Orchestrator 统一处理）
 """
 
-from typing import Any, List
+from typing import Any, List, Optional
 
 from app.agents.chat.agent_turn import AgentChatTurn
+from app.agents.chat.stage_detection import keyword_fallback_primary_stage
 from app.core.dependencies import get_llm_provider
 from app.core.logging import get_logger
 
@@ -16,6 +17,13 @@ from app.agents.chat.prompts_conversation import (
     get_opening_prompt,
 )
 from app.agents.state_schema import MemoirStateSchema
+from app.agents.chat.reply_limits import truncate_chat_segments
+from app.core.agent_logging import (
+    agent_span,
+    log_agent_payload,
+    log_agent_summary,
+)
+from app.core.config import settings
 
 logger = get_logger(__name__)
 
@@ -38,89 +46,8 @@ class InterviewAgent:
         self.llm = _get_langchain_llm()
 
     def _detect_user_stage(self, user_message: str) -> str:
-        """根据关键词检测用户正在谈论的人生阶段"""
-        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 ""
+        """关键词回退：与 stage_detection 一致（多阶段打分）。"""
+        return keyword_fallback_primary_stage(user_message)
 
     def _estimate_same_topic_turns(
         self, history_messages: List[Any], current_filled_slots: dict
@@ -153,6 +80,7 @@ class InterviewAgent:
         user_message: str,
         memoir_state: MemoirStateSchema,
         user_profile_context: str = "",
+        detected_user_stage: Optional[str] = None,
     ) -> AgentChatTurn:
         """生成状态感知的访谈回复，不持久化（由 Orchestrator 负责）"""
         if not self.llm:
@@ -167,7 +95,10 @@ class InterviewAgent:
                 ).items()
                 if value.snippet
             }
-            detected_user_stage = self._detect_user_stage(user_message)
+            if detected_user_stage is not None:
+                du = detected_user_stage
+            else:
+                du = self._detect_user_stage(user_message)
             history_messages = await get_history_messages(conversation_id)
             conversation_turn = len(history_messages) // 2
             same_topic_turns = self._estimate_same_topic_turns(
@@ -182,22 +113,52 @@ class InterviewAgent:
                 conversation_turn=conversation_turn,
                 same_topic_turns=same_topic_turns,
                 all_stages_coverage=all_stages_coverage,
-                detected_user_stage=detected_user_stage,
+                detected_user_stage=du,
                 user_profile_context=user_profile_context,
             )
             history_string = 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)
+            log_agent_payload(
+                logger, "InterviewAgent.generate_response.prompt", full_prompt
+            )
+            chat_llm = self.llm.bind(max_tokens=settings.chat_interview_max_tokens)
+            with agent_span(
+                logger,
+                "InterviewAgent.generate_response.llm",
+                conversation_id=conversation_id,
+                stage=memoir_state.current_stage,
+            ):
+                response = await chat_llm.ainvoke(full_prompt)
             response_text = (
                 response.content if hasattr(response, "content") else str(response)
             )
+            log_agent_payload(
+                logger, "InterviewAgent.generate_response.raw_response", response_text
+            )
             messages = [
                 msg.strip() for msg in response_text.split("[SPLIT]") if msg.strip()
             ]
-            out = messages[:3] if messages else [response_text]
+            raw_list = messages if messages else [response_text.strip()]
+            out = truncate_chat_segments(
+                raw_list,
+                max_segments=settings.chat_interview_max_segments,
+                max_chars_per_segment=settings.chat_interview_max_chars_per_segment,
+            )
+            if not out:
+                out = [
+                    response_text.strip()[
+                        : settings.chat_interview_max_chars_per_segment
+                    ]
+                ]
+            log_agent_summary(
+                logger,
+                "InterviewAgent.generate_response segments={} conversation_id={}",
+                len(out),
+                conversation_id,
+            )
             return AgentChatTurn(messages=out, skip_tts=False)
         except Exception as e:
-            logger.error("生成回应失败: %s", e, exc_info=True)
+            logger.error("生成回应失败: {}", e, exc_info=True)
             return AgentChatTurn(messages=[_FALLBACK_REPLY], skip_tts=True)
 
     async def generate_opening_message(
@@ -218,14 +179,44 @@ class InterviewAgent:
                 user_profile_context=user_profile_context,
             )
             full_prompt = f"{prompt}\n\nAssistant:"
-            response = await self.llm.ainvoke(full_prompt)
+            log_agent_payload(logger, "InterviewAgent.opening.prompt", full_prompt)
+            opening_llm = self.llm.bind(max_tokens=settings.chat_opening_max_tokens)
+            with agent_span(
+                logger,
+                "InterviewAgent.opening.llm",
+                conversation_id=conversation_id,
+            ):
+                response = await opening_llm.ainvoke(full_prompt)
             response_text = (
                 response.content if hasattr(response, "content") else str(response)
             )
+            log_agent_payload(
+                logger, "InterviewAgent.opening.raw_response", response_text
+            )
             messages = [
                 msg.strip() for msg in response_text.split("[SPLIT]") if msg.strip()
             ]
-            return messages[:2] if messages else [response_text]
+            raw_list = messages if messages else [response_text.strip()]
+            out = truncate_chat_segments(
+                raw_list,
+                max_segments=2,
+                max_chars_per_segment=settings.chat_interview_max_chars_per_segment,
+            )
+            log_agent_summary(
+                logger,
+                "InterviewAgent.opening segments={} conversation_id={}",
+                len(out),
+                conversation_id,
+            )
+            return (
+                out
+                if out
+                else [
+                    response_text.strip()[
+                        : settings.chat_interview_max_chars_per_segment
+                    ]
+                ]
+            )
         except Exception as e:
-            logger.error("生成开场白失败: %s", e, exc_info=True)
+            logger.error("生成开场白失败: {}", e, exc_info=True)
             return ["你好呀～ 又见面了，最近有没有什么事想跟我说说？"]