概述

在 AI Agent 开发中,执行 LLM 生成的代码是常见需求,但直接运行存在严重的安全风险,包括恶意代码注入、环境污染、资源滥用等。沙箱服务通过云端隔离的执行环境来解决这些安全问题。

为什么需要代码隔离

直接在主机环境中执行不受信任的代码存在以下风险:

• 恶意代码注入 - LLM 可能生成包含恶意逻辑的代码
• 环境污染 - 代码执行可能修改系统配置或安装冲突的依赖
• 资源滥用 - 无限循环或大量内存分配可能导致系统崩溃
• 多租户安全 - 一个用户的代码可能影响其他用户

沙箱服务提供完全隔离的云端执行环境,确保安全性和可靠性。

支持的语言

沙箱支持多种主流编程语言:

• Python
• JavaScript & TypeScript
• R
• Java
• Bash

需要其他语言支持?可以通过自定义模板实现。

Python

import { Sandbox } from '@e2b/code-interpreter'

async function main() {
  const sandbox = await Sandbox.create()

  const execution = await sandbox.runCode(
    `
import requests

response = requests.get("https://httpbin.org/json")
print(response.json())
    `,
    { language: 'python' }
  )

  console.log(execution.logs.stdout)

  await sandbox.kill()
}

main()

JavaScript & TypeScript

import { Sandbox } from '@e2b/code-interpreter'

async function main() {
  const sandbox = await Sandbox.create()

  const execution = await sandbox.runCode(
    `
const https = require('https');

https.get('https://httpbin.org/json', (res) => {
  let data = '';
  res.on('data', (chunk) => { data += chunk; });
  res.on('end', () => { console.log(JSON.parse(data)); });
});
    `,
    { language: 'js' }
  )

  console.log(execution.logs.stdout)

  await sandbox.kill()
}

main()

R

import { Sandbox } from '@e2b/code-interpreter'

async function main() {
  const sandbox = await Sandbox.create()

  const execution = await sandbox.runCode(
    `
library(httr)

response <- GET("https://httpbin.org/json")
content <- content(response, "parsed")
print(content)
    `,
    { language: 'r' }
  )

  console.log(execution.logs.stdout)

  await sandbox.kill()
}

main()

Java

import { Sandbox } from '@e2b/code-interpreter'

async function main() {
  const sandbox = await Sandbox.create()

  const execution = await sandbox.runCode(
    `
import java.net.*;
import java.io.*;

public class Main {
    public static void main(String[] args) throws Exception {
        URL url = new URL("https://httpbin.org/json");
        HttpURLConnection conn = (HttpURLConnection) url.openConnection();
        BufferedReader in = new BufferedReader(new InputStreamReader(conn.getInputStream()));
        String inputLine;
        while ((inputLine = in.readLine()) != null) {
            System.out.println(inputLine);
        }
        in.close();
    }
}
    `,
    { language: 'java' }
  )

  console.log(execution.logs.stdout)

  await sandbox.kill()
}

main()

Bash

import { Sandbox } from '@e2b/code-interpreter'

async function main() {
  const sandbox = await Sandbox.create()

  const execution = await sandbox.runCode(
    `
curl -s https://httpbin.org/json
    `,
    { language: 'bash' }
  )

  console.log(execution.logs.stdout)

  await sandbox.kill()
}

main()

流式输出

Code Interpreter 支持在代码执行期间实时流式获取标准输出、错误输出和执行结果。

流式获取 stdout/stderr

import { Sandbox } from '@e2b/code-interpreter'

async function main() {
  const sandbox = await Sandbox.create()

  await sandbox.runCode('print("hello"); import time; time.sleep(1); print("world")', {
    onStdout: (data) => {
      console.log(data)
    },
    onStderr: (data) => {
      console.log(data)
    },
  })

  await sandbox.kill()
}

main()

流式获取执行结果

import { Sandbox } from '@e2b/code-interpreter'

async function main() {
  const sandbox = await Sandbox.create()

  await sandbox.runCode('x = 1', {
    onResult: (result) => {
      console.log(result)
    },
  })

  await sandbox.kill()
}

main()