refactor(agents): 抽取阶段常量与对话上下文；快档 LLM；图片 prompt 可禁止回退

访谈与阶段
- 新增 app/agents/stage_constants.py：集中 CHAT_STAGES、章节分类/顺序、阶段到默认 memoir 类别等，与 MemoirState 默认槽位顺序对齐；减少散落在 prompts 内的重复常量。
- 新增 app/agents/chat/prompt_context.py：以 ChatPromptContext 汇总 guided 系统提示所需字段（阶段、槽位、轮次、人设、记忆证据、回复长度模式、背景声线、职业等），统一走 get_guided_conversation_prompt。
- 大幅收敛 app/agents/chat/prompts_conversation.py；调整 prompts.py、stage_prompts.py、stage_detection.py；同步 interview_agent、profile_agent、helpers 与 state_schema，使对话侧构造提示的方式一致、可测。

回忆录流水线
- memoir/prompts.py 删除已迁至 stage_constants / 独立模板的大段常量与图片占位相关逻辑；classification / extraction / fidelity / narrative agents 与 orchest（全量历史仍可用于计数，注入模型时按轮次与字符上限截断）、image_prompt_fallback_disabled。
- dependencies 增加 get_llm_provider_fast（LRU 缓存，可与默认共用密钥与 base_url）。

任务与编排
- memoir_tasks：prepare_batches 注入 llm_fast；开启独立快档模型时打结构化日志。
- chapter_cover_tasks、story_image_tasks：与图片 prompt / JSON 工具路径或策略变更对齐（import 与行为一致）。
- story_pipeline_sync 等小处同步。

其它核心
- langchain_llm、text_normalize 随上述调用链微调。

开发者体验
- .cursor/settings.json：启用 redis-development、postman 插件。

测试
- 新增 test_image_prompt_policy：覆盖「禁止回退」等图片 prompt 策略。
- 更新 test_interview_prompts、test_interview_reply_length、test_experience_regressions、test_json_and_memory_utils，匹配新常量位置、json_utils 与对话/长度行为。

api/app/agents/chat/profile_agent.py

@@ -4,27 +4,28 @@ ProfileAgent：用户资料收集 Specialist
 """
 
 import json
+import time
 from typing import Any, Dict, List, Optional
 
-from langchain_core.messages import AIMessage, HumanMessage
+from langchain_core.messages import AIMessage, HumanMessage, SystemMessage
 
-from app.agents.chat.helpers import format_history_string, get_history_messages
+from app.agents.chat.helpers import format_history_string, get_history_with_window
 from app.agents.chat.prompts_profile import (
     get_profile_extraction_prompt,
     get_profile_followup_prompt,
     get_profile_greeting_prompt,
 )
-from app.core.dependencies import get_llm_provider
-from app.core.langchain_llm import ainvoke_json_object
 from app.core.agent_logging import agent_span, log_agent_payload, log_agent_summary
 from app.core.config import settings
+from app.core.dependencies import get_llm_provider
+from app.core.json_utils import extract_json_payload
+from app.core.langchain_llm import ainvoke_json_object
 from app.core.logging import get_logger
 from app.agents.chat.reply_limits import (
     nonempty_segments_or_fallback,
     segments_from_llm_response,
     truncate_chat_segments,
 )
-from app.features.memoir.memoir_images.json_payload import extract_json_payload
 
 logger = get_logger(__name__)
 
@@ -37,6 +38,15 @@ def _get_langchain_llm():
         return None
 
 
+def _message_contents_char_count(messages: List[Any]) -> int:
+    n = 0
+    for m in messages:
+        c = getattr(m, "content", None)
+        if isinstance(c, str):
+            n += len(c)
+    return n
+
+
 class ProfileAgent:
     """用户资料收集 Specialist Agent"""
 
@@ -54,10 +64,12 @@ class ProfileAgent:
             return {}
         recent_dialogue = ""
         if conversation_id:
-            history_messages = await get_history_messages(conversation_id)
-            recent = (
-                history_messages[-4:] if len(history_messages) > 4 else history_messages
+            hw = await get_history_with_window(
+                conversation_id,
+                max_pairs=settings.chat_history_max_pairs,
+                max_chars=settings.chat_history_max_chars,
             )
+            recent = hw.window[-4:] if len(hw.window) > 4 else hw.window
             parts = []
             for msg in recent:
                 if isinstance(msg, HumanMessage):
@@ -118,21 +130,41 @@ class ProfileAgent:
                 nickname,
                 interview_stage_hint=interview_stage_hint,
             )
-            history_messages = await get_history_messages(conversation_id)
-            history_string = format_history_string(history_messages)
-            full_prompt = (
-                f"{prompt}\n\n{history_string}\n\nHuman: {user_message}\n\nAssistant:"
+            hw = await get_history_with_window(
+                conversation_id,
+                max_pairs=settings.chat_history_max_pairs,
+                max_chars=settings.chat_history_max_chars,
+            )
+            messages: List[Any] = [SystemMessage(content=prompt)]
+            messages.extend(hw.window)
+            messages.append(HumanMessage(content=user_message))
+            log_agent_payload(
+                logger,
+                "ProfileAgent.followup.prompt",
+                format_history_string(messages),
             )
-            log_agent_payload(logger, "ProfileAgent.followup.prompt", full_prompt)
             chat_llm = self.llm.bind(
                 max_tokens=settings.chat_profile_followup_max_tokens
             )
+            llm_t0 = time.perf_counter()
             with agent_span(
                 logger,
                 "ProfileAgent.followup.llm",
                 conversation_id=conversation_id,
             ):
-                response = await chat_llm.ainvoke(full_prompt)
+                logger.info(
+                    "event=chat_prompt_built agent=ProfileAgent.generate_profile_followup "
+                    "prompt_chars={} history_pairs_total={} history_pairs_windowed={}",
+                    _message_contents_char_count(messages),
+                    hw.turn_total,
+                    len(hw.window) // 2,
+                )
+                response = await chat_llm.ainvoke(messages)
+            logger.info(
+                "event=chat_llm_done agent=ProfileAgent.generate_profile_followup "
+                "response_latency_ms={:.2f}",
+                (time.perf_counter() - llm_t0) * 1000,
+            )
             response_text = (
                 response.content if hasattr(response, "content") else str(response)
             )
@@ -181,19 +213,44 @@ class ProfileAgent:
             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 get_history_messages(conversation_id)
-            history_string = format_history_string(history_messages)
-            full_prompt = f"{prompt}\n\n{history_string}" if history_string else prompt
-            log_agent_payload(logger, "ProfileAgent.greeting.prompt", full_prompt)
+            hw = await get_history_with_window(
+                conversation_id,
+                max_pairs=settings.chat_history_max_pairs,
+                max_chars=settings.chat_history_max_chars,
+            )
+            messages: List[Any] = [SystemMessage(content=prompt)]
+            messages.extend(hw.window)
+            if hw.window:
+                messages.append(
+                    HumanMessage(content="（请根据上文自然接话，继续资料收集开场。）")
+                )
+            else:
+                messages.append(HumanMessage(content="（请说出资料收集开场白。）"))
+            log_agent_payload(
+                logger, "ProfileAgent.greeting.prompt", format_history_string(messages)
+            )
             chat_llm = self.llm.bind(
                 max_tokens=settings.chat_profile_followup_max_tokens
             )
+            llm_t0 = time.perf_counter()
             with agent_span(
                 logger,
                 "ProfileAgent.greeting.llm",
                 conversation_id=conversation_id,
             ):
-                response = await chat_llm.ainvoke(full_prompt)
+                logger.info(
+                    "event=chat_prompt_built agent=ProfileAgent.generate_profile_greeting "
+                    "prompt_chars={} history_pairs_total={} history_pairs_windowed={}",
+                    _message_contents_char_count(messages),
+                    hw.turn_total,
+                    len(hw.window) // 2,
+                )
+                response = await chat_llm.ainvoke(messages)
+            logger.info(
+                "event=chat_llm_done agent=ProfileAgent.generate_profile_greeting "
+                "response_latency_ms={:.2f}",
+                (time.perf_counter() - llm_t0) * 1000,
+            )
             response_text = (
                 response.content if hasattr(response, "content") else str(response)
             )