test-description.js•2.5 kB
#!/usr/bin/env node
const { spawn } = require('child_process');
// 启动 MCP 服务器进程，测试新的 DESCRIPTION 变量
const mcpServer = spawn('node', ['build/index.js'], {
stdio: ['pipe', 'pipe', 'pipe'],
env: {
...process.env,
API_FORMAT: 'openai',
API_ENDPOINT: 'https://api.tu-zi.com/v1',
API_KEY: 'sk-WJqJGTf09kvMb7XklNQTxDZhDp0nMI4lZbfHGT1POw2CLtgK',
DEFAULT_MODEL: 'gemini-2.5-pro-preview-06-05',
DESCRIPTION: 'Supported models:\n- claude-sonnet-4-20250514: Claude Sonnet 4 model with extended thinking\n- gemini-2.5-pro-preview-06-05: Google Gemini 2.5 Pro preview model\n\nNote: This gateway supports custom endpoints and multiple AI providers.'
}
});
console.log('🚀 Testing MCP AI Gateway with DESCRIPTION variable...');
let responseBuffer = '';
mcpServer.stdout.on('data', (data) => {
const chunk = data.toString();
responseBuffer += chunk;
// 检查是否收到完整的 JSON-RPC 响应
const lines = responseBuffer.split('\n');
for (const line of lines) {
if (line.trim()) {
try {
const response = JSON.parse(line.trim());
if (response.id === 1) {
console.log('✅ 工具描述响应:');
console.log('工具名称:', response.result.tools[0].name);
console.log('工具描述:\n', response.result.tools[0].description);
mcpServer.kill();
process.exit(0);
}
} catch (e) {
// 忽略不完整的 JSON
}
}
}
});
mcpServer.stderr.on('data', (data) => {
console.error('❌ 错误:', data.toString());
});
mcpServer.on('close', (code) => {
console.log(`📋 服务器退出，代码: ${code}`);
});
// 等待服务器启动，然后发送测试请求
setTimeout(() => {
console.log('📤 发送工具列表请求...');
// 发送 MCP 协议初始化
mcpServer.stdin.write(JSON.stringify({
jsonrpc: "2.0",
id: 0,
method: "initialize",
params: {
protocolVersion: "2024-11-05",
capabilities: {},
clientInfo: {
name: "test-client",
version: "1.0.0"
}
}
}) + '\n');
// 发送工具列表请求
mcpServer.stdin.write(JSON.stringify({
jsonrpc: "2.0",
id: 1,
method: "tools/list"
}) + '\n');
}, 1000);