如何在CrewAI代理中解决Cloudflare Turnstile问题

# CapSolver 核心引擎
pip install git+https://github.com/capsolver-ai/capsolver-core.git
# CapSolver 代理工具
pip install git+https://github.com/capsolver-ai/capsolver-agent.git
# CrewAI
pip install crewai crewai-tools

export CAPSOLVER_API_KEY="your-capsolver-api-key"
export OPENAI_API_KEY="your-openai-api-key"

import asyncio
from crewai.tools import tool
from capsolver_agent.schema import create_executor

executor = create_executor(api_key="YOUR_CAPSOLVER_API_KEY")

@tool("解决 Cloudflare Turnstile")
def solve_turnstile(website_url: str, website_key: str) -> str:
    """解决 Cloudflare Turnstile 挑战并返回验证令牌。
    
    在网站受 Cloudflare Turnstile 保护时使用此工具。
    website_key 以 '0x' 开头，可以在页面的 cf-turnstile div 元素的 data-sitekey 属性中找到。
    
    参数:
        website_url: 受 Turnstile 保护的页面完整 URL
        website_key: Turnstile 站点密钥（以 0x 开头）
    
    返回:
        要与请求一起提交的 cf-turnstile-response 令牌
    """
    result = asyncio.run(executor.execute("solve_captcha", {
        "captcha_type": "cloudflare",
        "website_url": website_url,
        "website_key": website_key
    }))
    
    if result["success"]:
        return f"Turnstile 在 ~3 秒内解决。令牌: {result['solution']['token']}"
    return f"解决失败: {result['error']}"

from crewai import Agent, Task, Crew, Process

# 具备解决 Turnstile 的代理
researcher = Agent(
    role="网页研究专家",
    goal="访问受 Cloudflare 保护的网站并收集数据",
    backstory="""你研究受 Cloudflare Turnstile 保护的网站。 
    当你遇到 Turnstile 挑战（通过 cf-turnstile 元素或 Cloudflare 挑战页面识别时），使用页面 URL 和站点密钥（以 0x 开头）调用 solve_turnstile 工具。将返回的令牌作为 cf-turnstile-response 提交。""",
    tools=[solve_turnstile],
    verbose=True
)

# 需要解决 Turnstile 的任务
research_task = Task(
    description="""访问位于 {url} 的受 Cloudflare 保护的页面。
    Turnstile 站点密钥是 {site_key}。
    解决 Turnstile 挑战并报告令牌。""",
    expected_output="访问受保护资源的 Turnstile 令牌",
    agent=researcher
)

crew = Crew(agents=[researcher], tasks=[research_task], process=Process.sequential)
result = crew.kickoff(inputs={"url": "https://example.com", "site_key": "0x4AAAAAAA..."})

<!-- Turnstile 小部件 -->
<div class="cf-turnstile" data-sitekey="0x4AAAAAAABkMYinukE8nzYS"></div>

<!-- 或在 JavaScript 中 -->
<script>
  turnstile.render('#container', { sitekey: '0x4AAAAAAABkMYinukE8nzYS' });
</script>

from capsolver_core import create_capsolver
from playwright.async_api import async_playwright

async def solve_turnstile_on_page(url: str):
    cap = create_capsolver(api_key="YOUR_CAPSOLVER_API_KEY")
    
    async with async_playwright() as p:
        browser = await p.chromium.launch()
        page = await browser.new_page()
        await page.goto(url)
        
        # 自动检测并解决 Turnstile
        results = await cap.solve_on_page(page)
        for r in results:
            if r.solution:
                print(f"Turnstile 已解决并填写: {r.filled}")
        
        await browser.close()
    await cap.aclose()