feat 增加推荐对话可控参数

src/routers/chat.py

@@ -25,6 +25,8 @@ async def chat_stream(request: ChatRequest):
     * `message`: 用户的设计意图（如：'我想设计一个极简风格的橡木办公桌'）。
     * `thread_id`: (可选) 现有项目的唯一标识。若不传，系统将自动分配并返回。
     * `checkpoint_id`: (可选) 历史快照 ID。
+    * `config_params`: (可选) 对话配置参数
+    * `require_suggestion`: (可选) 是否需要建议按钮
 
     #### 3. 响应流说明 (Data Format)
     响应以 `data: ` 开头的 JSON 字符串流形式发送：
@@ -122,7 +124,10 @@ async def chat_stream(request: ChatRequest):
         #     )
         #     new_messages.append(force_instruction)
 
-        input_data = {"messages": new_messages}
+        input_data = {
+            "messages": new_messages,
+            "require_suggestion": request.need_suggestion  # 初始由前端决定
+        }
 
         async for event in app.astream(
                 input_data,

src/schemas/chat.py

@@ -14,7 +14,7 @@ class ChatRequest(BaseModel):
     thread_id: Optional[str] = Field(None, description="会话线程ID，不传则开启新会话")
     checkpoint_id: Optional[str] = Field(None, description="回溯点的ID，用于从历史点开启新对话")
     config_params: Optional[AgentConfig] = None
-    # force_sketch: bool = False  # 新增：是否强制绘图
+    need_suggestion: bool = False
 
 
 class HistoryItem(BaseModel):

src/server/agent/agents.py

@@ -42,9 +42,10 @@ async def designer_node(state: AgentState, config: RunnableConfig):
     system_text = get_agent_prompt("designer")
 
     system_prompt = SystemMessage(content=system_text)
+    should_suggest = len(state["messages"]) % 5 == 0
     # 改为异步调用 ainvoke
     response = await model.ainvoke([system_prompt] + messages)
-    return {"messages": [response]}
+    return {"messages": [response], "require_suggestion": should_suggest}
 
 
 # --- 2. Researcher Agent (情报专家) ---

src/server/agent/graph.py

@@ -18,7 +18,7 @@ from langgraph.checkpoint.mongodb import MongoDBSaver
 # 定义路由的输出结构，强制 LLM 选择一个
 class RouteResponse(BaseModel):
     # 将 FINISH 替换或增加 Suggester
-    next: Literal["Designer", "Researcher", "Visualizer", "Suggester"]
+    next: Literal["Designer", "Researcher", "Visualizer", "Suggester", "FINISH"]
 
 
 creds = service_account.Credentials.from_service_account_file(
@@ -42,7 +42,13 @@ def supervisor_node(state: AgentState):
     # --- 拦截逻辑修改 ---
     # 如果专家已经回复完了（AIMessage 且无工具调用），则交给 Suggester 生成按钮
     if isinstance(last_message, AIMessage) and not last_message.tool_calls:
+        should_go_to_suggester = state.get("require_suggestion", False)
+
+        # 如果符合建议条件
+        if should_go_to_suggester:
             return {"next": "Suggester"}
+        else:
+            return {"next": "FINISH"}
 
     system_prompt = """你是家具设计主管。分配任务给专家：
     - Designer: 设计建议、参数细化。
@@ -74,7 +80,8 @@ workflow.add_conditional_edges(
         "Designer": "Designer",
         "Researcher": "Researcher",
         "Visualizer": "Visualizer",
-        "Suggester": "Suggester"  # 原本的 FINISH 现在指向 Suggester
+        "Suggester": "Suggester",  # 原本的 FINISH 现在指向 Suggester
+        "FINISH": END  # 直接结束，不给建议
     }
 )
 
@@ -82,8 +89,7 @@ workflow.add_conditional_edges(
 workflow.add_edge("Designer", "Supervisor")
 workflow.add_edge("Researcher", "Supervisor")
 workflow.add_edge("Visualizer", "Supervisor")
-
-# 重点：Suggester 是整个流程的终点
+# 重点：Suggester 可以是整个流程的终点
 workflow.add_edge("Suggester", END)
 
 client = MongoClient(MONGO_URI)

src/server/agent/state.py

@@ -7,3 +7,4 @@ class AgentState(TypedDict):
     messages: Annotated[Sequence[BaseMessage], operator.add]
     # next 存储 Supervisor 决定的下一步是谁
     next: str
+    require_suggestion: bool  # 是否需要建议按钮