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/adapters/llm/deepseek.py

@@ -52,8 +52,9 @@ class DeepSeekLLMProvider:
         *,
         temperature: float | None = None,
         model: str | None = None,
+        max_tokens: int | None = None,
     ) -> str:
-        llm = self._get_llm(temperature, model)
+        llm = self._get_llm(temperature, model, max_tokens)
         lc_messages = _to_langchain_messages(messages)
         result = await llm.ainvoke(lc_messages)
         return str(result.content)
@@ -64,21 +65,29 @@ class DeepSeekLLMProvider:
         *,
         temperature: float | None = None,
         model: str | None = None,
+        max_tokens: int | None = None,
     ) -> AsyncIterator[str]:
-        llm = self._get_llm(temperature, model)
+        llm = self._get_llm(temperature, model, max_tokens)
         lc_messages = _to_langchain_messages(messages)
         async for chunk in llm.astream(lc_messages):
             if chunk.content:
                 yield str(chunk.content)
 
-    def _get_llm(self, temperature: float | None, model: str | None):
-        if temperature is None and model is None:
+    def _get_llm(
+        self,
+        temperature: float | None,
+        model: str | None,
+        max_tokens: int | None = None,
+    ):
+        if temperature is None and model is None and max_tokens is None:
             return self._llm
         kwargs: dict = {}
         if temperature is not None:
             kwargs["temperature"] = temperature
         if model is not None:
             kwargs["model"] = model
+        if max_tokens is not None:
+            kwargs["max_tokens"] = max_tokens
         return self._llm.bind(**kwargs) if kwargs else self._llm