fix(chat-engine): instruct LLM not to call document tool for existing memory

Memory context is already injected into system prompt, so add explicit
instruction telling the LLM to use it directly instead of calling the
document tool to re-read memories.

Co-authored-by: CodeFree <codefree@chinatelcom.cn>

server/service/agent/chat-engine.ts

@@ -251,9 +251,11 @@ export async function executeChat(params: ChatEngineParams): Promise<ChatEngineR
   if (user) {
     const memoryContext = await getMemoryContext(user.id);
     if (memoryContext) {
+      const memoryInstruction =
+        "\n\n以下 <memory> 标签内是你的持久记忆，已注入上下文。回答问题时优先参考这些记忆，不要重复调用 document 工具读取已有记忆。";
       systemPrompt = systemPrompt
-        ? `${systemPrompt}\n\n${memoryContext}`
+        ? `${systemPrompt}${memoryInstruction}\n${memoryContext}`
-        : memoryContext;
+        : `${memoryInstruction}\n${memoryContext}`;
     }
   }