4天开发，1700万美元融资：开源的Browser Use为啥这么火？

import asyncio
from playwright.async_api import async_playwright
from openai import OpenAI
import json

client = OpenAI()  # 替换为你的AI模型客户端

async def main():
    # 初始化浏览器
    async with async_playwright() as p:
        browser = await p.chromium.launch(headless=False)
        page = await browser.new_page()
        
        # 导航到目标网站
        await page.goto("https://example.com")
        
        # 主循环
        for step in range(5):  # 最多执行5步
            # 1. 观察：获取页面状态
            state = await get_page_state(page)
            
            # 2. 思考：让AI决策下一步
            next_action = await get_ai_decision(state)
            print(f"AI决定: {next_action}")
            
            # 如果AI决定完成任务，退出循环
            if next_action["action"] == "done":
                print(f"任务完成: {next_action.get('reason', '')}")
                break
                
            # 3. 行动：执行AI决策
            await execute_action(page, next_action)
            await page.wait_for_load_state("networkidle")
        
        await browser.close()

async def get_page_state(page):
    """获取当前页面状态"""
    # 获取页面基本信息
    url = page.url
    title = await page.title()
    
    # 获取可交互元素
    elements = await page.evaluate("""() => {
        const results = [];
        const all = document.querySelectorAll('a, button, input, select, textarea');
        
        all.forEach((el, index) => {
            const rect = el.getBoundingClientRect();
            // 只收集可见元素
            if (rect.width > 0 && rect.height > 0) {
                results.push({
                    index: index,
                    tag: el.tagName.toLowerCase(),
                    text: el.innerText || el.value || '',
                    type: el.type || '',
                    id: el.id || '',
                    name: el.name || '',
                    className: el.className || ''
                });
            }
        });
        
        return results;
    }""")
    
    # 将元素列表格式化为易读形式
    elements_text = ""
    for el in elements:
        elements_text += f"[{el['index']}] <{el['tag']}> {el['text']} {el['id']}\n"
    
    return {
        "url": url,
        "title": title,
        "elements": elements,
        "elements_text": elements_text
    }

async def get_ai_decision(state):
    """使用AI决定下一步行动"""
    prompt = f"""
    你是一个浏览器自动化助手。根据当前页面状态，决定下一步操作。
    
    当前URL: {state['url']}
    页面标题: {state['title']}
    
    可交互元素:
    {state['elements_text']}
    
    请从以下选择一个动作:
    1. click - 点击元素，参数: index (整数)
    2. input - 输入文本，参数: index (整数), text (字符串)
    3. goto - 导航到URL，参数: url (字符串)
    4. done - 完成任务，参数: reason (字符串)
    
    以JSON格式返回决策，例如: {"action": "click", "index": 2}
    """
    
    response = client.chat.completions.create(
        model="gpt-4",
        messages=[{"role": "user", "content": prompt}],
        response_format={"type": "json_object"}
    )
    
    # 解析AI响应
    try:
        return json.loads(response.choices[0].message.content)
    except:
        # 解析失败时的后备策略
        return {"action": "done", "reason": "无法解析AI响应"}

async def execute_action(page, action):
    """执行AI决定的动作"""
    if action["action"] == "click":
        # 获取元素并点击
        elements = await page.evaluate("""() => {
            const results = [];
            const all = document.querySelectorAll('a, button, input, select, textarea');
            
            all.forEach((el, index) => {
                const rect = el.getBoundingClientRect();
                if (rect.width > 0 && rect.height > 0) {
                    results.push({
                        index: index,
                        rect: {x: rect.x, y: rect.y, width: rect.width, height: rect.height}
                    });
                }
            });
            
            return results;
        }""")
        
        for el in elements:
            if el["index"] == action["index"]:
                # 点击元素中心
                x = el["rect"]["x"] + el["rect"]["width"] / 2
                y = el["rect"]["y"] + el["rect"]["height"] / 2
                await page.mouse.click(x, y)
                return
    
    elif action["action"] == "input":
        # 先点击输入框，再输入文本
        await execute_action(page, {"action": "click", "index": action["index"]})
        await page.keyboard.type(action["text"])
    
    elif action["action"] == "goto":
        # 导航到指定URL
        await page.goto(action["url"])

if __name__ == "__main__":
    asyncio.run(main())