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/tests/test_interview_prompts.py

@@ -1,5 +1,8 @@
 """访谈提示词：追问触发与性格（Persona）拼接回归。"""
 
+from langchain_core.messages import AIMessage, HumanMessage, SystemMessage
+
+from app.agents.chat.helpers import format_history_string
 from app.agents.chat.personas import normalize_interview_persona
 from app.agents.chat.prompts_conversation import (
     get_guided_conversation_prompt,
@@ -13,24 +16,40 @@ def test_guided_prompt_contains_mandatory_followup_when_heuristic_matches():
         empty_slots=["place", "people"],
         filled_slots={},
         user_message="厉害吧 那个女生叫娟娟",
-        conversation_turn=1,
+        conversation_turn_total=1,
         same_topic_turns=1,
         all_stages_coverage=None,
         detected_user_stage="childhood",
         user_profile_context="",
         persona="default",
     )
-    assert "什么时候追问" in p
+    assert "本轮追问判定" in p
     assert "本轮判定" in p
 
 
+def test_guided_prompt_does_not_embed_raw_user_message_in_system_text():
+    p = get_guided_conversation_prompt(
+        current_stage="childhood",
+        empty_slots=["place"],
+        filled_slots={},
+        user_message="__USER_SECRET_PHRASE_XYZ__",
+        conversation_turn_total=0,
+        same_topic_turns=0,
+        all_stages_coverage=None,
+        detected_user_stage="childhood",
+        user_profile_context="",
+        persona="default",
+    )
+    assert "__USER_SECRET_PHRASE_XYZ__" not in p
+
+
 def test_guided_prompt_persona_curious_guide():
     p = get_guided_conversation_prompt(
         current_stage="education",
         empty_slots=["school"],
         filled_slots={},
         user_message="还行吧",
-        conversation_turn=0,
+        conversation_turn_total=0,
         same_topic_turns=0,
         all_stages_coverage=None,
         detected_user_stage="education",
@@ -51,7 +70,7 @@ def test_guided_prompt_contains_memoir_orientation():
         empty_slots=["place"],
         filled_slots={},
         user_message="后来我就去上班了",
-        conversation_turn=0,
+        conversation_turn_total=0,
         same_topic_turns=0,
         all_stages_coverage=None,
         detected_user_stage="childhood",
@@ -68,7 +87,7 @@ def test_guided_prompt_contains_memory_section_when_evidence():
         empty_slots=["place"],
         filled_slots={},
         user_message="后来我就去上班了",
-        conversation_turn=0,
+        conversation_turn_total=0,
         same_topic_turns=0,
         all_stages_coverage=None,
         detected_user_stage="childhood",
@@ -87,7 +106,7 @@ def test_guided_prompt_chit_chat_hint():
         empty_slots=["place"],
         filled_slots={},
         user_message="今天天气真好哈哈",
-        conversation_turn=0,
+        conversation_turn_total=0,
         same_topic_turns=0,
         all_stages_coverage=None,
         detected_user_stage="childhood",
@@ -103,7 +122,7 @@ def test_guided_prompt_reply_length_section_explicit_expanded():
         empty_slots=["place"],
         filled_slots={},
         user_message="还行吧",
-        conversation_turn=0,
+        conversation_turn_total=0,
         same_topic_turns=0,
         all_stages_coverage=None,
         detected_user_stage="childhood",
@@ -123,7 +142,7 @@ def test_guided_prompt_reply_length_explicit_brief():
         empty_slots=["place"],
         filled_slots={},
         user_message="嗯",
-        conversation_turn=0,
+        conversation_turn_total=0,
         same_topic_turns=0,
         all_stages_coverage=None,
         detected_user_stage="childhood",
@@ -140,7 +159,7 @@ def test_guided_prompt_background_voice_military() -> None:
         empty_slots=["place"],
         filled_slots={},
         user_message="后来我就去上班了",
-        conversation_turn=0,
+        conversation_turn_total=0,
         same_topic_turns=0,
         all_stages_coverage=None,
         detected_user_stage="childhood",
@@ -149,7 +168,7 @@ def test_guided_prompt_background_voice_military() -> None:
         background_voice="military",
     )
     assert "背景语气：军队语境" in p
-    assert "真诚承接" in p
+    assert "先接住对方" in p
 
 
 def test_opening_prompt_military_has_examples_note() -> None:
@@ -162,3 +181,17 @@ def test_opening_prompt_military_has_examples_note() -> None:
     )
     assert "军队语境" in p
     assert "（军队语境：简洁" in p or "军队语境" in p
+
+
+def test_format_history_string_includes_system_for_debug_logs() -> None:
+    """log_agent_payload 依赖本函数时需包含 System，避免生产上丢失主 system prompt。"""
+    s = format_history_string(
+        [
+            SystemMessage(content="SYS_INSTRUCTIONS"),
+            HumanMessage(content="hi"),
+            AIMessage(content="hello"),
+        ]
+    )
+    assert "System: SYS_INSTRUCTIONS" in s
+    assert "Human: hi" in s
+    assert "Assistant: hello" in s