feat(api): 收敛对话与记忆流程边界，引入 LLM 网关与专用服务

- MemoryService 异步路径委托 MemoryIngestService / MemoryRetrievalService；富化派发经 MemoryEnrichmentScheduler
- WebSocket pipeline 经 ChatTurnService 与显式 DTO 编排单轮对话；回忆录片段入队由 MemoirIngestScheduler 封装
- 新增 LlmGateway（LlmUseCase），各 agent、任务与适配器对齐 ports
- 补充 memory 提示适配、runtime 类型、memory-retrieval 文档、ai-touchpoints 说明与扫描脚本及配套测试

Made-with: Cursor

api/app/agents/chat/profile_agent.py

@@ -24,19 +24,36 @@ from app.core.agent_logging import agent_span, log_agent_payload, log_agent_summ
 from app.core.config import settings
 from app.core.dependencies import get_llm_provider
 from app.core.llm_call import allm_json_call
+from app.core.llm_gateway import LlmGateway, LlmUseCase
 from app.core.logging import get_logger
+from app.ports.llm import LLMProvider
 
 logger = get_logger(__name__)
 
 
 def _get_langchain_llm():
     try:
-        provider = get_llm_provider()
-        return getattr(provider, "langchain_llm", None)
+        return LlmGateway().langchain_llm_for(LlmUseCase("chat.profile"))
     except Exception:
         return None
 
 
+def _langchain_messages_to_port(messages: List[Any]) -> list[dict]:
+    """LangChain message 列表 → ``LLMProvider.complete`` 的 ``role/content`` 结构。"""
+    out: list[dict] = []
+    for m in messages:
+        if isinstance(m, SystemMessage):
+            out.append({"role": "system", "content": str(m.content)})
+        elif isinstance(m, HumanMessage):
+            out.append({"role": "user", "content": str(m.content)})
+        elif isinstance(m, AIMessage):
+            out.append({"role": "assistant", "content": str(m.content)})
+        else:
+            c = getattr(m, "content", None)
+            out.append({"role": "user", "content": str(c) if c is not None else ""})
+    return out
+
+
 def _message_contents_char_count(messages: List[Any]) -> int:
     n = 0
     for m in messages:
@@ -49,9 +66,15 @@ def _message_contents_char_count(messages: List[Any]) -> int:
 class ProfileAgent:
     """用户资料收集 Specialist Agent"""
 
-    def __init__(self):
+    def __init__(self, llm_provider: LLMProvider | None = None):
+        self._llm_provider = llm_provider
         self.llm = _get_langchain_llm()
 
+    def _provider(self) -> LLMProvider:
+        if self._llm_provider is not None:
+            return self._llm_provider
+        return get_llm_provider()
+
     async def _invoke_chat(
         self,
         messages: List[Any],
@@ -60,20 +83,21 @@ class ProfileAgent:
         conversation_id: Optional[str],
         agent_name: str,
     ) -> str:
-        chat_llm = self.llm.bind(max_tokens=max_tokens)
+        port_messages = _langchain_messages_to_port(messages)
         llm_t0 = time.perf_counter()
         with agent_span(
             logger, f"{agent_name}.llm", conversation_id=conversation_id or ""
         ):
-            response = await chat_llm.ainvoke(messages)
+            response_text = await self._provider().complete(
+                port_messages,
+                max_tokens=max_tokens,
+            )
         logger.info(
             "event=chat_llm_done agent={} response_latency_ms={:.2f}",
             agent_name,
             (time.perf_counter() - llm_t0) * 1000,
         )
-        return (
-            response.content if hasattr(response, "content") else str(response)
-        ) or ""
+        return response_text or ""
 
     async def _segments_from_response(
         self,