feat(api): 访谈路径轻量门控、Memoir Phase1 批处理与叙事/记忆管线加固

- 新增 utterance_substance：短时/应答/元话语可跳过记忆检索、阶段 LLM 与资料抽取 LLM；可配置
- 输入归一化：LLM 模式默认仅语音/ASR；配置项写入 .env.example
- Memoir Phase1：可选 batch LLM 一次性抽取+分类（失败回退逐段）；Extraction 空槽位时阶段与 current_stage 对齐，prompt 约束收紧
- 叙事与忠实度：narrative_safety、证据重叠/场合锚点、标题 slots 与履历短语 grounded；fidelity 解析失败 fail-open 可配置
- 章节管线：锁 TTL 上调、锁竞争 Celery 重试、Phase2 immediate singleflight 等；story_pipeline_sync / chapter_compose / memoir_tasks 联动
- Memory：compaction / repo / summarizer / evidence 小修；事实 FTS 未命中是否回退最近事实可配置
- 新增 memoir_pipeline_trace；补充 memoir_reliability 文档与多项回归/门控测试

api/tests/test_chat_input_normalize.py

@@ -1,6 +1,6 @@
 """聊天输入归一：与 memoir 规则共用，配置独立。"""
 
-from unittest.mock import patch
+from unittest.mock import MagicMock, patch
 
 from app.features.conversation.input_normalize import (
     apply_conversation_input_rules,
@@ -37,3 +37,22 @@ def test_normalize_chat_off_mode() -> None:
         m.chat_input_normalize_enabled = True
         m.chat_input_normalize_mode = "off"
         assert normalize_chat_input_for_agent(raw, llm=None) == raw
+
+
+def test_normalize_llm_mode_voice_only_passes_no_llm_for_typing() -> None:
+    raw = "美看上我"
+    fake = MagicMock()
+    with patch("app.features.conversation.input_normalize.settings") as m:
+        m.chat_input_normalize_enabled = True
+        m.chat_input_normalize_mode = "llm"
+        m.chat_input_normalize_llm_voice_only = True
+        m.chat_input_normalize_llm_max_tokens = 512
+        m.chat_input_normalize_llm_max_input_chars = 8000
+        with patch(
+            "app.features.conversation.input_normalize._llm_normalize_chat_input"
+        ) as llm_norm:
+            llm_norm.return_value = None
+            out = normalize_chat_input_for_agent(raw, llm=fake, is_from_voice=False)
+            llm_norm.assert_called_once()
+            assert llm_norm.call_args[0][1] is None
+            assert out == "没看上我"