is_running = mag.is_running()
print(f"服务运行状态: {is_running}")

new_model = {
    "name": "deepseek-chat",
    "base_url": "https://api.deepseek.com",
    "api_key": "",
    "model": "deepseek-chat"
}
mag.add_model(new_model)

models = mag.list_models()
print(f"找到 {len(models)} 个模型")
for model in models:
    print(f"\n模型名称：{model['name']} \nBase_url：{model['base_url']}\n")

updated_model = {
    "name": "deepseek-chat",
    "base_url": "https://api.deepseek.com",
    "api_key": "",
    "model": "deepseek-chat"
}
result = mag.update_model("deepseek-chat", updated_model)
print(result)

mag.delete_model("deepseek-chat")

result = mag.remove_server(["fetch","memory"])
print(json.dumps(result, indent=4, ensure_ascii=False))

mcp_config = mag.get_mcp_config()
print(json.dumps(mcp_config, indent=8, ensure_ascii=False))

mcp_status = mag.get_mcp_status()
for server, status in mcp_status.items():
    print(f"\n{server}: {'已连接' if status.get('connected') else '未连接'}")
    print(f"{server}工具：{status.get('tools')}")
print(json.dumps(mcp_status, indent=4, ensure_ascii=False))

mcp_tools = mag.get_tools()
print(json.dumps(mcp_tools, indent=2, ensure_ascii=False))

script_files = {
    "main.py": """#!/usr/bin/env python3
import asyncio
from mcp.server.fastmcp import FastMCP

mcp = FastMCP("My Custom Tool")

@mcp.tool()
def hello_world(name: str) -> str:
    \"\"\"Say hello to someone\"\"\"
    return f"Hello, {name}!"

if __name__ == "__main__":
    mcp.run()
""",
    "requirements.txt": "fastmcp>=0.1.0"
}

result = mag.register_mcp(
    folder_name="my_custom_tool",
    script_files=script_files,
    readme="# My Custom Tool\n\nA simple hello world MCP tool.",
    dependencies="fastmcp>=0.1.0",
    port=9001
)

print(f"注册结果: {result}")

# 测试fetch工具
result = mag.test_mcp(
    server_name="fetch",
    tool_name="fetch",
    params={"url": "https://httpbin.org/json"}
)

print(f"测试状态: {result['status']}")
print(f"执行时间: {result.get('execution_time', 'N/A')}秒")

if result['status'] == 'success':
    print(f"工具结果: {result['result']}")
else:
    print(f"错误信息: {result['error']}")

# 测试不带参数的工具
simple_result = mag.test_mcp("memory", "list_memories")
print(f"简单测试结果: {simple_result}")

# 1. 生成AI MCP工具
print("=== 生成AI MCP工具 ===")
result = mag.generate_mcp(
    requirement="创建一个时间管理工具，提供获取当前时间、时区转换、倒计时等功能",
    model_name="deepseek-chat"
)

if result["status"] == "success":
    tool_name = result["tool_name"]
    print(f"成功生成工具: {tool_name}")
    
    # 2. 连接新生成的工具
    print(f"\n=== 连接工具 {tool_name} ===")
    connect_result = mag.connect_mcp(tool_name)
    print(f"连接结果: {connect_result['status']}")
    
    # 3. 获取工具列表并测试
    if connect_result['status'] == 'connected':
        tools = connect_result.get('tools', [])
        print(f"可用工具: {tools}")
        
        # 4. 测试第一个工具（如果存在）
        if tools:
            first_tool = tools[0]
            print(f"\n=== 测试工具 {first_tool} ===")
            test_result = mag.test_mcp(tool_name, first_tool, {})
            print(f"测试结果: {test_result}")
    
    # 5. 列出所有AI工具
    print(f"\n=== 当前AI工具列表 ===")
    ai_tools = mag.list_ai_mcp()
    for tool in ai_tools:
        print(f"- {tool}")
        
else:
    print(f"生成失败: {result['error']}")

# 6. 获取当前MCP状态
print(f"\n=== MCP服务器状态 ===")
status = mag.get_mcp_status()
for server, info in status.items():
    connected = "✓" if info.get("connected") else "✗"
    tools_count = len(info.get("tools", []))
    print(f"{connected} {server}: {tools_count} 个工具")

graph = mag.export_graph("news_processing_workflow")
print(json.dumps(graph, indent=2, ensure_ascii=False))

graph_detail = mag.get_graph_detail("deepresearch")

# 渲染README
if graph_detail['readme']:
    display(Markdown(graph_detail['readme']))
else:
    print("未找到README文件")

result = mag.run("deepresearch",input_text="深度探究mcp-agent-graph项目",parallel=False)
print(result)

# 首先运行一个图
initial_result = mag.run(
    "deepresearch", 
    "深度探究mcp-agent-graph项目",
    parallel=True
)
conversation_id = initial_result['conversation_id']

# 从断点继续（不添加新输入）
checkpoint_result = mag.continue_run(
    conversation_id=conversation_id,
    continue_from_checkpoint=True
)

print(f"断点继续结果: {checkpoint_result}")

# 重新开启会话，使用新输入
continue_result = mag.continue_run(
    conversation_id=conversation_id,
    input_text="深度探究mcp-agent-graph项目",
    parallel=False
)

print(f"继续执行结果: {continue_result}")

mcp = mag.generate_mcp_script("news_processing_workflow")
# 保存并行执行脚本
with open("news_workflow_parallel.py", "w", encoding="utf-8") as f:
    f.write(mcp["parallel_script"])

# 保存顺序执行脚本
with open("news_workflow_sequential.py", "w", encoding="utf-8") as f:
    f.write(mcp["sequential_script"])

result = mag.list_conversations()
result

result = mag.get_conversation('')

# 获取附件
attactments = result.get("attachments")
print(json.dumps(attactments, indent=2, ensure_ascii=False))

def display_attachments(attachments):
    """
    显示附件内容，支持md、html文件,这里只提供demo，运行时产生其他扩展文件请自行设计。或直接前往目录查看。
    使用notebook的markdown进行展示
    
    参数:
    attachments - 附件列表
    """
    if not attachments:
        print("没有可显示的附件")
        return
    
    for attachment in attachments:
        file_type = attachment["type"].lower()
        file_path = attachment["path"]
        file_name = attachment["name"]
        
        # 跳过不支持的文件类型
        if file_type not in ["md", "html", "docx", "pdf"]:
            continue
            
        print(f"\n📄 {file_name} ({file_type.upper()}):")
        
        try:
            # Markdown文件 - 使用IPython的Markdown显示
            if file_type == "md":
                with open(file_path, 'r', encoding='utf-8') as f:
                    md_content = f.read()
                display(Markdown(md_content))
                
            # HTML文件 - 使用IPython的HTML显示，或在浏览器中打开
            elif file_type == "html":
                try:
                    webbrowser.open(file_path)
                except:
                    print(f"已在浏览器中打开 HTML 文件")
                
        except Exception as e:
            print(f"无法显示文件 {file_name}: {str(e)}")

display_attachments(attactments)