修改架构：llm_call后增加观察环节注释

backend/app/main_graph/utils/main_graph_builder.py

@@ -92,9 +92,13 @@ def build_react_main_graph(llm=None, tools=None) -> StateGraph:
       ├─ dictionary_subgraph →────────┤
       ├─ news_analysis_subgraph →─────┤
       ├─ handle_error → (重试或结束) ─┤
-      └─ llm_call → END
-      ↓
-    END
+      └─ llm_call (大模型调用) ←──────┘
+           ↓
+    🔍 观察 (检查 tool_calls)
+           ↓
+    [有工具调用？]
+      ├─ 是 → 执行工具 → 回到 llm_call
+      └─ 否 → END
     """
     # 创建图
     graph = StateGraph(MainGraphState)
@@ -184,8 +188,9 @@ def build_react_main_graph(llm=None, tools=None) -> StateGraph:
     graph.add_edge("dictionary_subgraph", "react_reason")
     graph.add_edge("news_analysis_subgraph", "react_reason")
     graph.add_edge("handle_error", "react_reason")
-
-    # 5. 最终边：llm_call → END
+    
+    # 5. 条件路由：llm_call 后检查是否有工具调用
+    # 注意：这里简化处理，先直接 END，后续再完善工具调用循环
     if llm_node is not None:
         graph.add_edge("llm_call", END)