feat(api): 叙事 prompt、职业上下文、读路径章节、WS 解耦与错误脱敏

- 回忆录：事实边界补充允许清单；传记文体示例与 JSON 叙事要求对齐
- default 职业提示 occupation_context；cadre/military 退休语境
- GET 章节读路径零写入，prepare_chapter_read_view + markdown_for_response
- 文本归一抽到 core/text_normalize；移除弃用 reply 策略与 recompose_chapters_for_story
- ConversationService：WS 连接/用户段落/结束对话；对外错误固定文案
- 测试：HTTP 脱敏契约、章节读视图、occupation 与 background_voice

api/tests/test_interview_reply_length.py

@@ -7,8 +7,7 @@ import pytest
 
 from app.agents.chat.interview_reply_length import (
     ReplyLengthMode,
-    bump_reply_length_strategy_for_background_voice,
-    compute_reply_length_strategy,
+    bump_reply_plan_for_background_voice,
     compute_reply_plan,
 )
 from app.agents.state_schema import MemoirStateSchema
@@ -29,10 +28,9 @@ def _fake_settings(**overrides: object) -> SimpleNamespace:
 
 
 def test_strategy_brief_when_very_short() -> None:
-    s = compute_reply_length_strategy(
-        5,
-        likely_new_detail=False,
-        likely_chit_chat=False,
+    s = compute_reply_plan(
+        "x" * 5,
+        background_voice=None,
         settings=_fake_settings(),
     )
     assert s.mode == ReplyLengthMode.brief
@@ -41,10 +39,9 @@ def test_strategy_brief_when_very_short() -> None:
 
 
 def test_strategy_standard_mid_length() -> None:
-    s = compute_reply_length_strategy(
-        50,
-        likely_new_detail=True,
-        likely_chit_chat=False,
+    s = compute_reply_plan(
+        "x" * 50,
+        background_voice=None,
         settings=_fake_settings(),
     )
     assert s.mode == ReplyLengthMode.standard
@@ -53,10 +50,11 @@ def test_strategy_standard_mid_length() -> None:
 
 
 def test_strategy_long_chit_stays_standard() -> None:
-    s = compute_reply_length_strategy(
-        120,
-        likely_new_detail=False,
-        likely_chit_chat=True,
+    msg = "今天天气真好哈哈" * 11
+    assert len(msg) >= 80
+    s = compute_reply_plan(
+        msg,
+        background_voice=None,
         settings=_fake_settings(),
     )
     assert s.mode == ReplyLengthMode.standard
@@ -64,10 +62,12 @@ def test_strategy_long_chit_stays_standard() -> None:
 
 
 def test_strategy_long_with_new_detail_expanded() -> None:
-    s = compute_reply_length_strategy(
-        120,
-        likely_new_detail=True,
-        likely_chit_chat=False,
+    base = "第一次认识他"
+    msg = (base + "x" * 200)[:120]
+    assert len(msg) == 120
+    s = compute_reply_plan(
+        msg,
+        background_voice=None,
         settings=_fake_settings(),
     )
     assert s.mode == ReplyLengthMode.expanded
@@ -76,24 +76,27 @@ def test_strategy_long_with_new_detail_expanded() -> None:
 
 
 def test_strategy_boundary_len_20_brief_len_21_standard() -> None:
-    a = compute_reply_length_strategy(
-        20, likely_new_detail=False, likely_chit_chat=False, settings=_fake_settings()
+    a = compute_reply_plan(
+        "x" * 20,
+        background_voice=None,
+        settings=_fake_settings(),
     )
-    b = compute_reply_length_strategy(
-        21, likely_new_detail=False, likely_chit_chat=False, settings=_fake_settings()
+    b = compute_reply_plan(
+        "x" * 21,
+        background_voice=None,
+        settings=_fake_settings(),
     )
     assert a.mode == ReplyLengthMode.brief
     assert b.mode == ReplyLengthMode.standard
 
 
 def test_bump_standard_only_for_cadre_military() -> None:
-    s0 = compute_reply_length_strategy(
-        50,
-        likely_new_detail=False,
-        likely_chit_chat=False,
+    s0 = compute_reply_plan(
+        "x" * 50,
+        background_voice=None,
         settings=_fake_settings(),
     )
-    bumped = bump_reply_length_strategy_for_background_voice(
+    bumped = bump_reply_plan_for_background_voice(
         s0,
         background_voice="cadre",
         settings=_fake_settings(
@@ -104,16 +107,15 @@ def test_bump_standard_only_for_cadre_military() -> None:
     assert bumped.max_tokens == s0.max_tokens + 40
     assert bumped.max_chars_per_segment == s0.max_chars_per_segment + 40
 
-    brief = compute_reply_length_strategy(
-        5,
-        likely_new_detail=False,
-        likely_chit_chat=False,
+    brief = compute_reply_plan(
+        "x" * 5,
+        background_voice=None,
         settings=_fake_settings(
             chat_interview_cadre_military_standard_extra_tokens=40,
             chat_interview_cadre_military_standard_extra_chars=40,
         ),
     )
-    same = bump_reply_length_strategy_for_background_voice(
+    same = bump_reply_plan_for_background_voice(
         brief,
         background_voice="military",
         settings=_fake_settings(
@@ -149,13 +151,14 @@ def test_plan_long_chit_stays_standard_not_expanded() -> None:
 
 
 def test_strategy_boundary_len_79_standard_len_80_long_branch() -> None:
-    a = compute_reply_length_strategy(
-        79, likely_new_detail=False, likely_chit_chat=False, settings=_fake_settings()
+    a = compute_reply_plan(
+        "x" * 79,
+        background_voice=None,
+        settings=_fake_settings(),
     )
-    b = compute_reply_length_strategy(
-        80,
-        likely_new_detail=False,
-        likely_chit_chat=False,
+    b = compute_reply_plan(
+        "x" * 80,
+        background_voice=None,
         settings=_fake_settings(),
     )
     assert a.mode == ReplyLengthMode.standard