feat(i18n): persist language preference and thread through chat, memoir, TTS

- Add users.language_preference (Alembic 0018, default zh); capture at signup/SMS
  only; expose on auth and profile APIs
- Lite English prompts for chat and memoir; localized stage labels and agent
  names (Life Echo / 岁月知己)
- Tencent TTS: language-aware synthesis, ModelType=1 for 501004, English chunking
- WebSocket pipeline: emit all AGENT_RESPONSE segments when TTS cancels; INFO logs
  for tts_this_turn and TTS decisions; on-demand TTS logging
- Expo: device language on auth, i18n tiers/agent name, [SPLIT] streaming UX fixes
- Tests for migration, prompts, pipeline, router tts_this_turn, reply segments

Co-authored-by: Cursor <cursoragent@cursor.com>

api/app/agents/chat/profile_agent.py

@@ -29,6 +29,31 @@ from app.ports.llm import LLMProvider
 
 logger = get_logger(__name__)
 
+_FOLLOWUP_FALLBACK_ZH = "谢谢分享！能再告诉我一些吗？"
+_FOLLOWUP_FALLBACK_EN = "Thanks for sharing — could you tell me a bit more?"
+_GREETING_FALLBACK_ZH = "你好！在开始之前，能告诉我你是哪一年出生的吗？"
+_GREETING_FALLBACK_EN = (
+    "Hi! Before we get started, could you tell me what year you were born?"
+)
+_GREETING_FALLBACK_FULL_ZH = (
+    "你好！在我们开始聊人生故事之前，能先简单介绍一下你自己吗？比如你是哪一年出生的？"
+)
+_GREETING_FALLBACK_FULL_EN = (
+    "Hi! Before we dive into life stories, could you introduce yourself a little — for example, what year were you born?"
+)
+
+
+def _profile_followup_fallback(language: str) -> str:
+    return _FOLLOWUP_FALLBACK_EN if language == "en" else _FOLLOWUP_FALLBACK_ZH
+
+
+def _profile_greeting_fallback(language: str) -> str:
+    return _GREETING_FALLBACK_EN if language == "en" else _GREETING_FALLBACK_ZH
+
+
+def _profile_greeting_fallback_full(language: str) -> str:
+    return _GREETING_FALLBACK_FULL_EN if language == "en" else _GREETING_FALLBACK_FULL_ZH
+
 
 class _ProviderBackedProfileGateway:
     def __init__(self, provider: LLMProvider) -> None:
@@ -173,6 +198,7 @@ class ProfileAgent:
         user_message: str,
         missing_fields: List[str],
         conversation_id: Optional[str] = None,
+        language: str = "zh",
     ) -> Dict[str, Any]:
         """从用户消息中提取资料字段，不持久化"""
         if not missing_fields:
@@ -186,15 +212,20 @@ class ProfileAgent:
             )
             recent = hw.window[-4:] if len(hw.window) > 4 else hw.window
             parts = []
+            user_label = "User" if language == "en" else "用户"
+            asst_label = "Assistant" if language == "en" else "助手"
             for msg in recent:
                 if isinstance(msg, HumanMessage):
-                    parts.append(f"用户: {msg.content}")
+                    parts.append(f"{user_label}: {msg.content}")
                 elif isinstance(msg, AIMessage):
-                    parts.append(f"助手: {msg.content}")
+                    parts.append(f"{asst_label}: {msg.content}")
             recent_dialogue = "\n".join(parts) if parts else ""
         try:
             prompt = get_profile_extraction_prompt(
-                user_message, missing_fields, recent_dialogue=recent_dialogue or None
+                user_message,
+                missing_fields,
+                recent_dialogue=recent_dialogue or None,
+                language=language,
             )
             parsed = await self._llm_gateway.json_object(
                 prompt,
@@ -241,6 +272,7 @@ class ProfileAgent:
         filled_fields: Dict[str, str],
         nickname: str = "",
         interview_stage_hint: str = "",
+        language: str = "zh",
     ) -> List[str]:
         """生成资料追问回复，不持久化（由 Orchestrator 负责）"""
         try:
@@ -249,6 +281,7 @@ class ProfileAgent:
                 filled_fields,
                 nickname,
                 interview_stage_hint=interview_stage_hint,
+                language=language,
             )
             hw = await get_history_with_window(
                 conversation_id,
@@ -284,7 +317,7 @@ class ProfileAgent:
                 response_text,
                 max_segments=3,
                 max_chars_per_segment=settings.chat_interview_max_chars_per_segment,
-                fallback="谢谢分享！能再告诉我一些吗？",
+                fallback=_profile_followup_fallback(language),
             )
             log_agent_summary(
                 logger,
@@ -295,17 +328,20 @@ class ProfileAgent:
             return segments
         except Exception as e:
             logger.error("生成资料跟进回复失败: {}", e)
-            return ["谢谢分享！能再告诉我一些吗？"]
+            return [_profile_followup_fallback(language)]
 
     async def generate_profile_greeting(
         self,
         conversation_id: str,
         missing_fields: List[str],
         nickname: str = "",
+        language: str = "zh",
     ) -> List[str]:
         """生成资料收集开场白，不持久化（由 Orchestrator 负责）"""
         try:
-            prompt = get_profile_greeting_prompt(missing_fields, nickname)
+            prompt = get_profile_greeting_prompt(
+                missing_fields, nickname, language=language
+            )
             hw = await get_history_with_window(
                 conversation_id,
                 max_pairs=settings.chat_history_max_pairs,
@@ -313,12 +349,19 @@ class ProfileAgent:
             )
             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。）")
+            if language == "en":
+                kickoff = (
+                    "(Continue from the context above and warmly carry on the profile-gathering opener.)"
+                    if hw.window
+                    else "(Please deliver your profile-gathering opener.)"
                 )
             else:
-                messages.append(HumanMessage(content="（请说出资料收集开场白。）"))
+                kickoff = (
+                    "（请根据上文自然接话，继续资料收集开场。）"
+                    if hw.window
+                    else "（请说出资料收集开场白。）"
+                )
+            messages.append(HumanMessage(content=kickoff))
             log_agent_payload(
                 logger,
                 "ProfileAgent.greeting.prompt",
@@ -345,7 +388,7 @@ class ProfileAgent:
                 response_text,
                 max_segments=2,
                 max_chars_per_segment=settings.chat_interview_max_chars_per_segment,
-                fallback="你好！在开始之前，能告诉我你是哪一年出生的吗？",
+                fallback=_profile_greeting_fallback(language),
             )
             log_agent_summary(
                 logger,
@@ -356,6 +399,4 @@ class ProfileAgent:
             return segments
         except Exception as e:
             logger.error("生成资料收集开场白失败: {}", e)
-            return [
-                "你好！在我们开始聊人生故事之前，能先简单介绍一下你自己吗？比如你是哪一年出生的？"
-            ]
+            return [_profile_greeting_fallback_full(language)]