diff --git a/server/remote_test_server.py b/server/remote_test_server.py
new file mode 100644
index 0000000..4b11bd3
--- /dev/null
+++ b/server/remote_test_server.py
@@ -0,0 +1,93 @@
+
+import sys
+import os
+from fastapi import FastAPI, HTTPException
+from pydantic import BaseModel, Field
+from typing import Dict, Any
+import uvicorn
+
+# 确保 pytrace 在 Python 路径中
+sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), '..')))
+
+from pytrace.model.graph import WorkflowGraph
+from pytrace.runtime.context import TraceContext
+from pytrace.runtime.executor import WorkflowExecutor
+from pytrace.runtime.repository import NodeRepository
+# 导入项目中已有的 TraceServer
+from app.services import trace_server
+# 假定数学节点套件已在某处定义并可以导入，以便在图中本地使用
+from pytrace.nodes.test_math_suite import MathSuite
+
+
+# --- 1. 数据模型 ---
+
+class GraphExecutionRequest(BaseModel):
+    graph_data: Dict[str, Any]
+    initial_inputs: Dict[str, Any] = Field(default_factory=dict)
+
+
+# --- 2. FastAPI 应用定义 ---
+
+app = FastAPI(
+    title="TraceStudio - Remote Test Server",
+    description="一个用于测试远程节点注册和执行的服务器。它同时开启HTTP和TCP服务。",
+    version="2.0.0",
+)
+
+@app.on_event("startup")
+async def startup_event():
+    # 1. 在后台启动TCP服务器，用于监听客户端的连接和注册
+    trace_server.start_background()
+    
+    # 2. 将TCP服务器的远程节点注册表配置给全局的NodeRepository
+    NodeRepository.set_remote_registry(trace_server.remote_registry)
+    
+    print("--- Remote Test Server ---")
+    print(f"TCP RPC Server started on port {trace_server.port}. Waiting for clients to connect and register.")
+    print("HTTP Server ready to execute graphs at /execute_graph")
+
+
+@app.post("/execute_graph", tags=["Execution"])
+async def execute_graph(request: GraphExecutionRequest):
+    """
+    接收图的JSON并执行，可能会调用已通过TCP连接注册的远程节点。
+    """
+    try:
+        print("\n--- New Graph Execution Request ---")
+        # 1. 从请求数据构建 WorkflowGraph
+        graph = WorkflowGraph.from_dict(request.graph_data)
+        print(f"Executing graph: '{graph.name}'")
+        
+        # 2. 准备执行上下文，注入 trace_server 实例作为RPC服务
+        # RemoteProxyNode 将通过这个服务来回调客户端
+        context = TraceContext(services={"rpc": trace_server})
+        
+        # 3. 实例化执行器并运行
+        executor = WorkflowExecutor(graph, context)
+        executor.execute_batch(request.initial_inputs)
+        
+        # 4. 获取并返回最终输出
+        final_outputs = context.get_outputs()
+        print(f"Graph execution finished. Final outputs: {final_outputs}")
+        return {
+            "success": True,
+            "outputs": final_outputs
+        }
+    except Exception as e:
+        import traceback
+        print(f"ERROR: Graph execution failed: {e}\n{traceback.format_exc()}")
+        raise HTTPException(
+            status_code=500,
+            detail={
+                "error": str(e),
+                "type": type(e).__name__,
+                "traceback": traceback.format_exc()
+            }
+        )
+
+# --- 3. 启动服务器 ---
+
+if __name__ == "__main__":
+    # 使用 uvicorn 启动HTTP服务器。TCP服务器会在startup事件中被启动。
+    # 您可以通过命令行启动: uvicorn server.remote_test_server:app --reload --port 8000
+    uvicorn.run(app, host="0.0.0.0", port=8000)
diff --git a/tests/assets/complex_remote_graph.json b/tests/assets/complex_remote_graph.json
new file mode 100644
index 0000000..f1c4420
--- /dev/null
+++ b/tests/assets/complex_remote_graph.json
@@ -0,0 +1,103 @@
+
+{
+  "name": "Complex Remote and Subgraph Test",
+  "nodes": [
+    {
+      "id": "main_input",
+      "class_id": "InputTraceNode",
+      "params": {
+        "main_text": { "mode": "context", "value": "inputs.main_text" },
+        "main_num": { "mode": "context", "value": "inputs.main_num" }
+      }
+    },
+    {
+      "id": "remote_text_processor",
+      "class_id": "remote.process_text"
+    },
+    {
+      "id": "sub_graph_node",
+      "class_id": "core.Graph",
+      "data": {
+        "name": "My SubGraph",
+        "nodes": [
+          {
+            "id": "sub_input",
+            "class_id": "InputTraceNode",
+            "params": {
+              "sub_in_1": { "mode": "context", "value": "inputs.sub_in_1" },
+              "sub_in_2": { "mode": "context", "value": "inputs.sub_in_2" }
+            }
+          },
+          {
+            "id": "local_add",
+            "class_id": "math.add"
+          },
+          {
+            "id": "remote_ratio_calculator",
+            "class_id": "remote.calculate_ratio",
+            "params": {
+                "denominator": { "mode": "static", "value": 10.0 }
+            }
+          },
+          {
+            "id": "sub_output",
+            "class_id": "OutputTraceNode"
+          }
+        ],
+        "edges": [
+          {
+            "source_id": "sub_input", "source_port": "sub_in_1",
+            "target_id": "local_add", "target_port": "a"
+          },
+          {
+            "source_id": "sub_input", "source_port": "sub_in_2",
+            "target_id": "local_add", "target_port": "b"
+          },
+          {
+            "source_id": "local_add", "source_port": "c",
+            "target_id": "remote_ratio_calculator", "target_port": "numerator"
+          },
+          {
+            "source_id": "remote_ratio_calculator", "source_port": "ratio",
+            "target_id": "sub_output", "target_port": "sub_ratio"
+          }
+        ],
+        "interface": {
+          "inputs": { "sub_in_1": "integer", "sub_in_2": "integer" },
+          "outputs": { "sub_ratio": "float" }
+        }
+      }
+    },
+    {
+      "id": "main_output",
+      "class_id": "OutputTraceNode"
+    }
+  ],
+  "edges": [
+    {
+      "source_id": "main_input", "source_port": "main_text",
+      "target_id": "remote_text_processor", "target_port": "text_in"
+    },
+    {
+      "source_id": "remote_text_processor", "source_port": "length",
+      "target_id": "sub_graph_node", "target_port": "sub_in_1"
+    },
+    {
+      "source_id": "main_input", "source_port": "main_num",
+      "target_id": "sub_graph_node", "target_port": "sub_in_2"
+    },
+    {
+      "source_id": "sub_graph_node", "source_port": "sub_ratio",
+      "target_id": "main_output", "target_port": "final_ratio"
+    }
+  ],
+  "interface": {
+    "inputs": {
+      "main_text": "string",
+      "main_num": "integer"
+    },
+    "outputs": {
+      "final_ratio": "float"
+    }
+  }
+}
diff --git a/tests/demo_client.py b/tests/demo_client.py
deleted file mode 100644
index f9476ca..0000000
--- a/tests/demo_client.py
+++ /dev/null
@@ -1,73 +0,0 @@
-import sys
-import os
-import time
-
-# Ensure pytrace is in python path
-sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), '..')))
-
-from pytrace.api.node import TraceNode
-from pytrace.api.decorators import register_class, register_node, input_port, output_port
-from pytrace.runtime.context import TraceContext
-from pytrace.remote.worker import TraceWorker, expose_node
-
-# =============================================================================
-# 1. Define Your Custom Nodes
-# =============================================================================
-# Use @expose_node to make the class available to the Remote Server.
-# Use @register_class to define the category and display name in the UI.
-
-@expose_node
-@register_class(category="Tutorial", display_name="Client Demo")
-class ClientDemoNode(TraceNode):
-    """
-    A collection of demo nodes running on the client side.
-    Developers can add their own methods here.
-    """
-    
-    @register_node(display_name="Remote Add", description="Performs addition on the client.")
-    @input_port(name="a", type="float")
-    @input_port(name="b", type="float")
-    @output_port(name="result", type="float")
-    def add(self, ctx: TraceContext):
-        # 1. Get Inputs
-        a = ctx.get_input("a", 0.0)
-        b = ctx.get_input("b", 0.0)
-        
-        # 2. Execute Logic
-        result = a + b
-        
-        # 3. Log execution (This log will appear on the Server console!)
-        ctx.log(f"[Client] Executing Add: {a} + {b} = {result}")
-        
-        # 4. Set Output
-        ctx.set_output("result", result)
-
-    @register_node(display_name="Remote Echo", description="Echoes a message.")
-    @input_port(name="message", type="string")
-    @output_port(name="echo", type="string")
-    def echo(self, ctx: TraceContext):
-        msg = ctx.get_input("message", "")
-        ctx.log(f"[Client] Echoing: {msg}")
-        ctx.set_output("echo", f"Client says: {msg}")
-
-# =============================================================================
-# 2. Start the Worker
-# =============================================================================
-
-if __name__ == "__main__":
-    SERVER_HOST = 'localhost'
-    SERVER_PORT = 9999
-    
-    print(f"--- TraceStudio Client Demo ---")
-    print(f"Connecting to Server at {SERVER_HOST}:{SERVER_PORT}...")
-    
-    # Initialize the Worker
-    worker = TraceWorker(host=SERVER_HOST, port=SERVER_PORT)
-    
-    try:
-        # Start the worker (This blocks until stopped)
-        worker.start()
-    except KeyboardInterrupt:
-        print("\n[Client] Worker stopped.")
-    except Exception as e:
-        print(f"\n[Client] Error: {e}")
\ No newline at end of file
diff --git a/tests/demo_server.py b/tests/demo_server.py
deleted file mode 100644
index ec027e7..0000000
--- a/tests/demo_server.py
+++ /dev/null
@@ -1,220 +0,0 @@
-import socket
-import json
-import threading
-import sys
-import os
-import time
-
-# Add project root
-sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), '..')))
-
-from pytrace.api.decorators import output_port, register_class
-from pytrace.remote.manager import RPCManager
-from pytrace.remote.protocol import MessageType
-from pytrace.model.graph import WorkflowGraph, Node, Edge
-from pytrace.model.enums import DimensionMode
-from pytrace.runtime.context import TraceContext
-from pytrace.runtime.executor import WorkflowExecutor
-from pytrace.remote.registry import RemoteRegistry
-from pytrace.runtime.repository import NodeRepository
-from pytrace.model.data import TraceParam, ParamMode
-from pytrace.api.node import GraphTraceNode, TraceNode
-from pytrace.runtime.registry import NodeRegistry as LocalNodeRegistry
-
-class ClientSession:
-    """
-    Represents a connected Worker Client on the Server.
-    Wraps the Socket connection and the RPCManager.
-    """
-    def __init__(self, conn, addr, server: 'TraceServer'):
-        self.conn = conn
-        self.addr = addr
-        self.server = server
-        self.client_id = None
-        self.rpc = RPCManager(send_callback=self._send_json)
-        
-        # Register internal handlers
-        self.rpc.register_handler(MessageType.REGISTER_NODES, self._handle_register)
-
-    def _send_json(self, data):
-        try:
-            msg = json.dumps(data) + "\n"
-            self.conn.sendall(msg.encode('utf-8'))
-        except Exception as e:
-            print(f"[Session] Send error: {e}")
-            self.close()
-
-    def _handle_register(self, payload, msg_id):
-        self.client_id = payload.get("client_id")
-        nodes = payload.get("nodes", [])
-        print(f"[Server] Client {self.client_id} registered {len(nodes)} nodes.")
-        
-        # Update Server Registry
-        self.server.remote_registry.register_client(self.client_id, nodes)
-        self.server.register_session(self.client_id, self)
-
-    def start_loop(self):
-        """Blocking read loop."""
-        f = self.conn.makefile('r', encoding='utf-8')
-        try:
-            for line in f:
-                if not line: break
-                try:
-                    data = json.loads(line)
-                    self.rpc.process_message(data)
-                except Exception as e:
-                    print(f"[Session] Msg error: {e}")
-        finally:
-            self.close()
-
-    def close(self):
-        print(f"[Server] Session {self.client_id} disconnected")
-        self.rpc.close()
-        try:
-            self.conn.close()
-        except:
-            pass
-        if self.client_id:
-            self.server.unregister_session(self.client_id)
-
-class TraceServer:
-    def __init__(self, socket_port=9999):
-        self.socket_port = socket_port
-        self.sessions = {} # client_id -> ClientSession
-        self.lock = threading.Lock()
-        self.remote_registry = RemoteRegistry()
-
-    def start(self):
-        # 1. Start Socket Server (For Workers)
-        threading.Thread(target=self._socket_server_loop, daemon=True).start()
-        
-        # 2. Start "HTTP" Server (For Users - Simulated)
-        threading.Thread(target=self._http_server_simulation, daemon=True).start()
-        
-        print(f"[Server] Running. Socket Port: {self.socket_port}")
-        
-        # Keep main thread alive
-        while True:
-            time.sleep(1)
-
-    def _socket_server_loop(self):
-        server_socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
-        server_socket.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
-        server_socket.bind(('localhost', self.socket_port))
-        server_socket.listen(5)
-        
-        while True:
-            conn, addr = server_socket.accept()
-            session = ClientSession(conn, addr, self)
-            threading.Thread(target=session.start_loop, daemon=True).start()
-
-    def register_session(self, client_id, session):
-        with self.lock:
-            self.sessions[client_id] = session
-
-    def unregister_session(self, client_id):
-        with self.lock:
-            self.sessions.pop(client_id, None)
-
-    # --- RPC Service Interface for Workflow ---
-    def call(self, target_id: str, method: str, params: dict, timeout=3600):
-        # 1. Routing Logic
-        if not target_id:
-            providers = self.remote_registry.find_providers(method)
-            if not providers:
-                raise RuntimeError(f"No active client provides node '{method}'")
-            target_id = providers[0]
-            print(f"[Server] Auto-routing '{method}' to client '{target_id}'")
-
-        # 2. Get Session
-        session = self.sessions.get(target_id)
-        if not session:
-            raise RuntimeError(f"Client {target_id} not connected")
-        
-        # 3. Execute via Session's RPC
-        return session.rpc.call(target_id, method, params, timeout)
-
-    # --- HTTP API Simulation ---
-    def _http_server_simulation(self):
-        """
-        Simulates an HTTP endpoint that receives workflow execution requests.
-        In a real app, this would be Flask/FastAPI.
-        """
-        print("[Server] HTTP API ready. Waiting for requests...")
-        
-        # Simulate a user request coming in after 5 seconds
-        time.sleep(5)
-        
-        # Mock Request Payload
-        # User explicitly wants 'ClientMathNode.add' to run on a specific worker if available
-        # Or they might leave it empty for auto-routing.
-        
-        # Let's find a connected client to simulate the user picking one
-        target_worker_id = None
-        while not target_worker_id:
-            with self.lock:
-                if self.sessions:
-                    target_worker_id = list(self.sessions.keys())[0]
-            if not target_worker_id:
-                print("[Server] HTTP Waiting for workers to connect...")
-                time.sleep(2)
-
-        print(f"\n[Server] >>> HTTP Request Received: Execute Workflow (Target: {target_worker_id}) <<<")
-        
-        # The User Request Data
-        request_payload = {
-            "inputs": {"x": 10, "y": 20},
-            "node_targets": {
-                # The user specifies that the node with ID 'node_remote_1' in the graph
-                # MUST be executed by 'target_worker_id'
-                "node_remote_1": target_worker_id 
-            }
-        }
-        
-        self.run_workflow(request_payload)
-        
-    def run_workflow(self, payload):
-        print("[Server] Loading test workflow: tests/assets/main.flow")
-        flow_path = os.path.abspath("tests/assets/main.flow")
-        with open(flow_path, 'r', encoding='utf-8') as f:
-            graph_data = json.load(f)
-        
-        # Inject a node ID into the graph data for demonstration if needed, 
-        # or assume the graph already has 'node_remote_1'. 
-        # For this demo, let's assume the graph's first node is the one we want to target.
-        # (In reality, graph_data comes from the frontend with IDs)
-        
-        graph = WorkflowGraph.from_dict(graph_data)
-        
-        # Prepare Repository
-        repo = NodeRepository(remote_registry=self.remote_registry)
-        
-        # Prepare Context with User Preferences (node_targets)
-        context = TraceContext(
-            contexts={}, 
-            services={"rpc": self, "node_repository": repo},
-            node_targets=payload.get("node_targets")
-        )
-        
-        # Execute
-        print("[Server] Executing workflow...")
-        executor = WorkflowExecutor(graph, context, node_repository=repo)
-        
-        inputs = payload.get("inputs")
-        executor.execute_batch(inputs)
-        
-        # 7. Verify Results
-        outputs = context.context.get_outputs()
-        print(f"[Server] Workflow finished. Outputs: {outputs}")
-        
-        res = outputs.get("final_result")
-        val = res.value if hasattr(res, 'value') else res
-        
-        if val == 15.0: # main.flow logic: x(10) -> sub_flow(a=10, b=5) -> 10+5=15
-            print(f"[Server] ✅ Remote Execution SUCCESS! Result: {val}")
-        else:
-            print(f"[Server] ❌ Remote Execution FAILED. Expected 15.0, got {val}")
-
-if __name__ == "__main__":
-    server = TraceServer()
-    server.start()
\ No newline at end of file
diff --git a/tests/run_remote_test.py b/tests/run_remote_test.py
new file mode 100644
index 0000000..ea13595
--- /dev/null
+++ b/tests/run_remote_test.py
@@ -0,0 +1,93 @@
+
+import json
+import httpx
+import os
+
+# --- 配置 ---
+SERVER_URL = "http://localhost:8000"
+GRAPH_EXECUTE_ENDPOINT = f"{SERVER_URL}/execute_graph"
+GRAPH_FILE_PATH = os.path.join(os.path.dirname(__file__), "assets", "complex_remote_graph.json")
+
+def run_test():
+    """
+    执行端到端远程调用测试。
+    
+    使用方法:
+    1. 确保 remote_test_server.py 正在运行 (uvicorn server.remote_test_server:app --reload --port 8000)
+    2. 确保 test_complex_remote_client.py 正在运行 (python tests/test_complex_remote_client.py)
+    3. 运行此脚本 (python tests/run_remote_test.py)
+    """
+    
+    print("--- Running End-to-End Remote Execution Test ---")
+    
+    # 1. 加载图JSON文件
+    try:
+        with open(GRAPH_FILE_PATH, 'r', encoding='utf-8') as f:
+            graph_data = json.load(f)
+        print(f"Successfully loaded graph from: {GRAPH_FILE_PATH}")
+    except FileNotFoundError:
+        print(f"ERROR: Graph file not found at {GRAPH_FILE_PATH}")
+        return
+        
+    # 2. 准备初始输入
+    # 根据 complex_remote_graph.json 的 "interface" 定义
+    initial_inputs = {
+        "main_text": "hello_remote_world",
+        "main_num": 5
+    }
+    print(f"Initial inputs: {initial_inputs}")
+
+    # 3. 构建请求体
+    request_payload = {
+        "graph_data": graph_data,
+        "initial_inputs": initial_inputs
+    }
+
+    # 4. 发送HTTP请求到服务器
+    print(f"Sending POST request to {GRAPH_EXECUTE_ENDPOINT}...")
+    try:
+        with httpx.Client(timeout=60.0) as client:
+            response = client.post(GRAPH_EXECUTE_ENDPOINT, json=request_payload)
+            response.raise_for_status() # 如果状态码不是 2xx，则会引发异常
+            
+            result = response.json()
+            
+            print("\n--- Test Result ---")
+            print("Request successful!")
+            print("Server response:")
+            # 使用json.dumps美化输出
+            print(json.dumps(result, indent=2, ensure_ascii=False))
+
+            # 预期结果计算：
+            # 1. main_text "hello_remote_world" 长度为 18
+            # 2. 子图接收 sub_in_1=18, sub_in_2=5
+            # 3. 子图内本地相加: 18 + 5 = 23
+            # 4. 子图内远程计算比率: 23 / 10.0 = 2.3
+            # 5. 主图接收到最终结果 2.3
+            expected_ratio = 2.3
+            final_ratio = result.get("outputs", {}).get("final_ratio", {}).get("value")
+            
+            print("\n--- Verification ---")
+            print(f"Expected final_ratio: {expected_ratio}")
+            print(f"Actual final_ratio:   {final_ratio}")
+            assert final_ratio == expected_ratio, "Test assertion failed!"
+            print("✅ Test Passed Successfully!")
+
+            
+    except httpx.RequestError as e:
+        print("\n--- TEST FAILED ---")
+        print(f"ERROR: Could not connect to the server at {SERVER_URL}.")
+        print("Please ensure the 'remote_test_server.py' is running.")
+        print(f"Details: {e}")
+    except httpx.HTTPStatusError as e:
+        print("\n--- TEST FAILED ---")
+        print(f"ERROR: Server returned a non-200 status code: {e.response.status_code}")
+        print("Server response:")
+        try:
+            # 尝试打印JSON错误详情
+            print(json.dumps(e.response.json(), indent=2, ensure_ascii=False))
+        except json.JSONDecodeError:
+            print(e.response.text)
+            
+if __name__ == "__main__":
+    run_test()
diff --git a/tests/test_complex_remote_client.py b/tests/test_complex_remote_client.py
new file mode 100644
index 0000000..139abfc
--- /dev/null
+++ b/tests/test_complex_remote_client.py
@@ -0,0 +1,104 @@
+
+import sys
+import os
+import time
+
+# 确保 pytrace 在 Python 路径中
+sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), '..')))
+
+from pytrace.api.node import TraceNode
+from pytrace.api.decorators import register_class, register_node, input_port, output_port
+from pytrace.runtime.context import TraceContext
+from pytrace.remote.worker import TraceWorker, expose_node
+
+# --- 1. 定义可被远程调用的节点 ---
+# 使用 @expose_node 将此类暴露给服务器
+@expose_node
+@register_class(category="ComplexRemote", display_name="Complex Remote Node")
+class ComplexRemoteNode(TraceNode):
+    """
+    这个类包含一组用于远程调用的方法（节点）。
+    每个方法都添加了详细的打印语句，方便您在客户端控制台进行调试。
+    """
+    
+    @register_node(
+        class_id="remote.process_text", 
+        display_name="Process Text (Remote)", 
+        description="在客户端处理文本，计算长度并添加前缀。"
+    )
+    @input_port(name="text_in", type="string")
+    @output_port(name="text_out", type="string")
+    @output_port(name="length", type="integer")
+    def process_text(self, ctx: TraceContext):
+        text = ctx.get_input("text_in", "")
+        
+        print("\n--- [REMOTE] ---")
+        print(f"Node 'remote.process_text' is executing.")
+        print(f"  - Received input 'text_in': '{text}'")
+        
+        # 模拟一些耗时操作
+        time.sleep(0.5)
+        
+        processed_text = f"processed_by_client:_{text}"
+        text_length = len(text)
+        
+        ctx.set_output("text_out", processed_text)
+        ctx.set_output("length", text_length)
+        
+        print(f"  - Sent output 'text_out': '{processed_text}'")
+        print(f"  - Sent output 'length': {text_length}")
+        print("--- [REMOTE] ---\
+")
+
+
+    @register_node(
+        class_id="remote.calculate_ratio", 
+        display_name="Calculate Ratio (Remote)", 
+        description="在客户端根据两个数字计算比率。"
+    )
+    @input_port(name="numerator", type="float", label="分子")
+    @input_port(name="denominator", type="float", label="分母")
+    @output_port(name="ratio", type="float", label="比率")
+    def calculate_ratio(self, ctx: TraceContext):
+        numerator = ctx.get_input("numerator", 0.0)
+        denominator = ctx.get_input("denominator", 1.0)
+        
+        print("\n--- [REMOTE] ---")
+        print(f"Node 'remote.calculate_ratio' is executing.")
+        print(f"  - Received input 'numerator': {numerator}")
+        print(f"  - Received input 'denominator': {denominator}")
+        
+        if denominator == 0:
+            print("  - Error: Denominator is zero.")
+            # 在实际应用中可能需要更复杂的错误处理
+            ratio = 0.0
+        else:
+            ratio = numerator / denominator
+
+        ctx.set_output("ratio", ratio)
+        print(f"  - Sent output 'ratio': {ratio}")
+        print("--- [REMOTE] ---\
+")
+
+
+# --- 2. 启动客户端 Worker ---
+if __name__ == "__main__":
+    # 服务器的TCP RPC地址和端口
+    SERVER_HOST = 'localhost'
+    SERVER_PORT = 9999 # 这应该与 remote_test_server.py 中的端口一致
+    
+    print(f"--- Complex Remote Test Client ---")
+    print(f"Attempting to connect to RPC Server at {SERVER_HOST}:{SERVER_PORT}...")
+    
+    # 初始化并启动 Worker
+    # 它会自动查找所有被 @expose_node 装饰的类，并向服务器注册
+    worker = TraceWorker(host=SERVER_HOST, port=SERVER_PORT)
+    
+    try:
+        # worker.start() 是一个阻塞操作，它会保持运行以监听服务器的调用
+        worker.start()
+    except KeyboardInterrupt:
+        print("\n[Client] Worker stopped by user.")
+    except Exception as e:
+        print(f"\n[Client] An error occurred: {e}")
+        print("[Client] Is the test server running?")