> ## Documentation Index
> Fetch the complete documentation index at: https://ppio.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Browser Use 集成
[BrowserUse](https://github.com/browser-use/browser-use) 是一个强大的 AI 浏览器 agent。结合 PPIO Sandbox 提供的隔离沙箱环境,你可以构建高并发、多任务的浏览器 AI agent。
本文档详细介绍如何基于 PPIO Sandbox 服务运行 BrowserUse 项目。
本文档使用 PPIO 发布的 `browser-chromium` Sandbox Template。
### 1. 获取 PPIO API Key
### 2. 安装依赖
安装所需的 Python 包:
```bash Python theme={null}
pip install browser-use
```
### 3. 示例代码
开始之前,你需要先配置必要的环境变量:
```bash Bash theme={null}
export PPIO_API_KEY=""
export LLM_API_KEY=""
export LLM_BASE_URL=https://api.ppinfra.com/v3/openai
export LLM_MODEL=""
```
保存以下代码:
**agent.py**
```python theme={null}
import asyncio
import base64
import os
import time
import dotenv
dotenv.load_dotenv(override=True)
from browser_use import Agent, BrowserSession
from browser_use.llm import ChatOpenAI
from ppio_sandbox import PPIO
ppio = PPIO()
async def screenshot(agent: Agent):
# 截图函数
print("Taking screenshot...")
page = await agent.browser_session.get_current_page()
screenshot_bytes = await page.screenshot(format='png')
# screenshot 方法返回图片的二进制数据，需将其保存为 PNG 文件
screenshots_dir = os.path.join(".", "screenshots")
os.makedirs(screenshots_dir, exist_ok=True)
screenshot_path = os.path.join(screenshots_dir, f"{time.time()}.png")
if isinstance(screenshot_bytes, str):
screenshot_data = base64.b64decode(screenshot_bytes)
else:
screenshot_data = screenshot_bytes
with open(screenshot_path, "wb") as f:
f.write(screenshot_data)
print(f"Screenshot saved to {screenshot_path}")
async def main():
# 创建 PPIO Sandbox 实例
sandbox = ppio.sandbox.create(
"browser-chromium", # This template contains chromium browser and exposes port 9223 for remote connection
timeout=600, # Timeout in seconds
)
try:
# 从 Sandbox 获取 Chrome 调试端口地址
host = sandbox.get_host(9223) # Get sandbox port 9223 address
cdp_url = f"https://{host}"
print(f"Chrome Debug Protocol URL: {cdp_url}")
# 等待 Chrome 和反向代理就绪
print("Waiting for Chrome and reverse proxy to be ready...")
import requests
max_retries = 20
for i in range(max_retries):
try:
response = requests.get(f"{cdp_url}/json/version", verify=False, timeout=5)
if response.status_code == 200:
data = response.json()
if 'webSocketDebuggerUrl' in data:
print(f"✅ CDP endpoint ready: {data.get('Browser', 'Unknown')}")
print(f"WebSocket URL: {data['webSocketDebuggerUrl']}")
break
else:
print(f"Attempt {i+1}/{max_retries}: Response missing webSocketDebuggerUrl")
print(f"Response data: {data}")
else:
print(f"Attempt {i+1}/{max_retries}: HTTP {response.status_code}")
except Exception as e:
print(f"Attempt {i+1}/{max_retries}: {type(e).__name__}: {e}")
if i == max_retries - 1:
print("❌ Failed to connect to CDP endpoint after all retries")
print("Checking startup script logs...")
logs = sandbox.commands.run("tail -50 /tmp/*.log 2>/dev/null || echo 'No logs found'")
print(f"Logs: {logs}")
raise RuntimeError("CDP endpoint not ready")
time.sleep(2)
# 创建 BrowserUse 会话
browser_session = BrowserSession(cdp_url=cdp_url)
await browser_session.start()
print("BrowserUse session created successfully")
# 创建 AI Agent
agent = Agent(
task="Go to ppio.com and tell me what you see",
llm=ChatOpenAI(
api_key=os.getenv("LLM_API_KEY"),
base_url=os.getenv("LLM_BASE_URL"),
model=os.getenv("LLM_MODEL"),
temperature=1
),
browser_session=browser_session,
)
# 运行 Agent 任务
print("Starting Agent task execution...")
await agent.run(
on_step_end=screenshot, # Take screenshot after each step
)
# 关闭浏览器会话
await browser_session.kill()
print("Task execution completed")
finally:
# 清理 Sandbox 资源
sandbox.kill()
print("Sandbox resources cleaned up")
if __name__ == "__main__":
asyncio.run(main())
```
### 4. 运行 Agent
安装完依赖并设置好环境变量后,即可运行示例代码。如果一切顺利,你会在终端中看到如下输出。Browser-use 正在你的 Sandbox 内的远程浏览器中执行任务。
它会生成如下的截图:
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.