Playwright 深度解析：从浏览器自动化到动态爬虫、自动化测试与 AI Agent

CSDN 原文镜像

本文为作者 CSDN 博客的全文镜像，原文发布于 2026-07-30。为适配本站结构，仅补充了站内元数据与来源说明，正文主体保持原文内容。

原文链接：https://blog.csdn.net/m0_63309778/article/details/163335924
站内分区：工程实践 / Playwright 自动化

在这里插入图片描述

前言

在传统爬虫中，我们通常使用 requests、httpx 或 curl 向服务器发送 HTTP 请求，然后解析返回的 HTML。

这种方式简单、快速、资源消耗低，但它有一个明显的问题：

它只能获取服务器直接返回的内容，无法像真实用户一样操作浏览器。

现代网站大量使用 React、Vue、Angular 等前端框架。很多页面打开时，服务器只返回一个简单的 HTML 外壳，真正的数据需要浏览器执行 JavaScript、调用后端接口后才能渲染出来。

除此之外，很多业务系统还包含：

登录状态与 Cookie；
动态 Token；
iframe 嵌套页面；
弹窗和新标签页；
文件上传与下载；
懒加载列表；
WebSocket 通信；
验证码和风控检测；
必须点击、滚动或输入后才加载的数据。

在这些场景下，仅使用 HTTP 请求库往往很难完成任务。

Playwright 正是为浏览器自动化而设计的工具。它能够通过统一的 API 控制 Chromium、Firefox 和 WebKit，并提供 TypeScript、JavaScript、Python、Java 和 .NET 等语言接口。目前 Playwright 不仅用于自动化测试，也广泛应用于浏览器脚本、动态网页采集和 AI Agent 浏览器操作。

一、Playwright 是什么

Playwright 是由 Microsoft 主导开发的开源浏览器自动化框架。

简单来说，Playwright 可以让程序像人一样操作浏览器，例如：

打开网站；
点击按钮；
输入账号密码；
登录系统；
切换页面；
操作 iframe；
等待数据加载；
监听接口请求；
上传和下载文件；
截图和录制视频；
提取页面中的数据。

它支持以下三类核心浏览器引擎：

浏览器引擎	常见浏览器
Chromium	Chrome、Edge、Chromium
Firefox	Mozilla Firefox
WebKit	Safari 使用的浏览器引擎

Playwright 可以使用同一套 API 控制 Chromium、Firefox 和 WebKit，也可以运行 Chrome、Edge 等品牌浏览器，并支持模拟桌面端、平板和移动设备。

可以把 Playwright 理解为：

一套通过代码控制真实浏览器的工具。

它不是一个简单的 HTML 解析器，也不是普通的 HTTP 请求库，而是真正启动并控制浏览器进程。

二、Playwright 能做什么

2.1 Web 自动化测试

Playwright 最典型的用途是端到端测试，也就是 E2E 测试。

例如测试一个登录流程：

打开登录页面；
输入用户名；
输入密码；
点击登录；
验证是否进入首页；
验证用户名称是否正确显示。

python

<span class="token keyword">from</span> playwright<span class="token punctuation">.</span>sync_api <span class="token keyword">import</span> sync_playwright<span class="token punctuation">,</span> expect

<span class="token keyword">def</span> <span class="token function">test_login</span><span class="token punctuation">(</span><span class="token punctuation">)</span><span class="token punctuation">:</span>
<span class="token keyword">with</span> sync_playwright<span class="token punctuation">(</span><span class="token punctuation">)</span> <span class="token keyword">as</span> p<span class="token punctuation">:</span>
browser <span class="token operator">=</span> p<span class="token punctuation">.</span>chromium<span class="token punctuation">.</span>launch<span class="token punctuation">(</span>headless<span class="token operator">=</span><span class="token boolean">True</span><span class="token punctuation">)</span>
page <span class="token operator">=</span> browser<span class="token punctuation">.</span>new_page<span class="token punctuation">(</span><span class="token punctuation">)</span>

page<span class="token punctuation">.</span>goto<span class="token punctuation">(</span><span class="token string">"https://example.com/login"</span><span class="token punctuation">)</span>

page<span class="token punctuation">.</span>get_by_label<span class="token punctuation">(</span><span class="token string">"用户名"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>fill<span class="token punctuation">(</span><span class="token string">"admin"</span><span class="token punctuation">)</span>
page<span class="token punctuation">.</span>get_by_label<span class="token punctuation">(</span><span class="token string">"密码"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>fill<span class="token punctuation">(</span><span class="token string">"123456"</span><span class="token punctuation">)</span>
page<span class="token punctuation">.</span>get_by_role<span class="token punctuation">(</span><span class="token string">"button"</span><span class="token punctuation">,</span> name<span class="token operator">=</span><span class="token string">"登录"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token punctuation">)</span>

expect<span class="token punctuation">(</span>page<span class="token punctuation">)</span><span class="token punctuation">.</span>to_have_url<span class="token punctuation">(</span><span class="token string">"https://example.com/home"</span><span class="token punctuation">)</span>
expect<span class="token punctuation">(</span>page<span class="token punctuation">.</span>get_by_text<span class="token punctuation">(</span><span class="token string">"欢迎回来"</span><span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">.</span>to_be_visible<span class="token punctuation">(</span><span class="token punctuation">)</span>

browser<span class="token punctuation">.</span>close<span class="token punctuation">(</span><span class="token punctuation">)</span>

这段代码不是直接调用登录接口，而是完整模拟用户操作浏览器。

因此，它能够验证的不仅是接口，还包括：

页面是否正常渲染；
前后端是否正确联调；
按钮是否可点击；
表单校验是否生效；
页面跳转是否正确；
登录状态是否保存；
权限控制是否正确；
不同浏览器中是否表现一致。

Playwright Test 本身还提供测试执行器、断言、自动等待、失败重试、并行执行和 Trace 调试等能力。

2.2 动态网页数据采集

Playwright 也经常被用于动态爬虫。

假设一个网页的初始 HTML 只有：

html

<span class="token tag"><span class="token tag"><span class="token punctuation"><</span>div</span> <span class="token attr-name">id</span><span class="token attr-value"><span class="token punctuation attr-equals">=</span><span class="token punctuation">"</span>app<span class="token punctuation">"</span></span><span class="token punctuation">></span></span><span class="token tag"><span class="token tag"><span class="token punctuation"></</span>div</span><span class="token punctuation">></span></span>
<span class="token tag"><span class="token tag"><span class="token punctuation"><</span>script</span> <span class="token attr-name">src</span><span class="token attr-value"><span class="token punctuation attr-equals">=</span><span class="token punctuation">"</span>/assets/index.js<span class="token punctuation">"</span></span><span class="token punctuation">></span></span><span class="token script"></span><span class="token tag"><span class="token tag"><span class="token punctuation"></</span>script</span><span class="token punctuation">></span></span>

真正的数据由 JavaScript 请求接口后渲染。

使用 requests 获取到的可能只是空壳页面，而 Playwright 会启动浏览器、加载 JavaScript、执行接口请求，最终拿到完整页面。

python

<span class="token keyword">from</span> playwright<span class="token punctuation">.</span>async_api <span class="token keyword">import</span> async_playwright

<span class="token keyword">async</span> <span class="token keyword">def</span> <span class="token function">fetch_page</span><span class="token punctuation">(</span><span class="token punctuation">)</span><span class="token punctuation">:</span>
<span class="token keyword">async</span> <span class="token keyword">with</span> async_playwright<span class="token punctuation">(</span><span class="token punctuation">)</span> <span class="token keyword">as</span> p<span class="token punctuation">:</span>
browser <span class="token operator">=</span> <span class="token keyword">await</span> p<span class="token punctuation">.</span>chromium<span class="token punctuation">.</span>launch<span class="token punctuation">(</span>headless<span class="token operator">=</span><span class="token boolean">True</span><span class="token punctuation">)</span>

context <span class="token operator">=</span> <span class="token keyword">await</span> browser<span class="token punctuation">.</span>new_context<span class="token punctuation">(</span>
locale<span class="token operator">=</span><span class="token string">"zh-CN"</span><span class="token punctuation">,</span>
viewport<span class="token operator">=</span><span class="token punctuation">{<!-- --></span><span class="token string">"width"</span><span class="token punctuation">:</span> <span class="token number">1440</span><span class="token punctuation">,</span> <span class="token string">"height"</span><span class="token punctuation">:</span> <span class="token number">900</span><span class="token punctuation">}</span>
<span class="token punctuation">)</span>

page <span class="token operator">=</span> <span class="token keyword">await</span> context<span class="token punctuation">.</span>new_page<span class="token punctuation">(</span><span class="token punctuation">)</span>

<span class="token keyword">await</span> page<span class="token punctuation">.</span>goto<span class="token punctuation">(</span>
<span class="token string">"https://example.com"</span><span class="token punctuation">,</span>
wait_until<span class="token operator">=</span><span class="token string">"domcontentloaded"</span>
<span class="token punctuation">)</span>

<span class="token keyword">await</span> page<span class="token punctuation">.</span>locator<span class="token punctuation">(</span><span class="token string">".result-list"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>wait_for<span class="token punctuation">(</span><span class="token punctuation">)</span>

items <span class="token operator">=</span> <span class="token keyword">await</span> page<span class="token punctuation">.</span>locator<span class="token punctuation">(</span><span class="token string">".result-item"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>all_inner_texts<span class="token punctuation">(</span><span class="token punctuation">)</span>

<span class="token keyword">print</span><span class="token punctuation">(</span>items<span class="token punctuation">)</span>

<span class="token keyword">await</span> context<span class="token punctuation">.</span>close<span class="token punctuation">(</span><span class="token punctuation">)</span>
<span class="token keyword">await</span> browser<span class="token punctuation">.</span>close<span class="token punctuation">(</span><span class="token punctuation">)</span>

这类方式适合处理：

JavaScript 动态渲染页面；
无限滚动列表；
点击“查看更多”后加载的数据；
登录后才能访问的页面；
需要切换筛选条件的页面；
数据位于 iframe 中的页面；
通过接口异步加载内容的页面。

不过，Playwright 的资源消耗通常比 requests 更高。因此，更合理的采集架构往往是：

text

普通静态页面
↓
requests / httpx 直接请求

动态页面或登录页面
↓
Playwright 获取登录状态、接口地址或页面内容

发现稳定数据接口
↓
尽可能切换回 HTTP 客户端批量请求

也就是说：

Playwright 不一定要承担全部采集工作，它也可以负责打开页面、完成登录和发现接口，再把后续批量请求交给更轻量的 HTTP 客户端。

2.3 自动登录和登录状态保存

很多系统需要登录后才能访问。

Playwright 可以操作登录页面，也可以保存 Cookie、LocalStorage 等浏览器状态。

python

context <span class="token operator">=</span> <span class="token keyword">await</span> browser<span class="token punctuation">.</span>new_context<span class="token punctuation">(</span><span class="token punctuation">)</span>

page <span class="token operator">=</span> <span class="token keyword">await</span> context<span class="token punctuation">.</span>new_page<span class="token punctuation">(</span><span class="token punctuation">)</span>

<span class="token keyword">await</span> page<span class="token punctuation">.</span>goto<span class="token punctuation">(</span><span class="token string">"https://example.com/login"</span><span class="token punctuation">)</span>

<span class="token keyword">await</span> page<span class="token punctuation">.</span>get_by_label<span class="token punctuation">(</span><span class="token string">"账号"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>fill<span class="token punctuation">(</span><span class="token string">"admin"</span><span class="token punctuation">)</span>
<span class="token keyword">await</span> page<span class="token punctuation">.</span>get_by_label<span class="token punctuation">(</span><span class="token string">"密码"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>fill<span class="token punctuation">(</span><span class="token string">"123456"</span><span class="token punctuation">)</span>
<span class="token keyword">await</span> page<span class="token punctuation">.</span>get_by_role<span class="token punctuation">(</span><span class="token string">"button"</span><span class="token punctuation">,</span> name<span class="token operator">=</span><span class="token string">"登录"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token punctuation">)</span>

<span class="token keyword">await</span> page<span class="token punctuation">.</span>wait_for_url<span class="token punctuation">(</span><span class="token string">"**/home"</span><span class="token punctuation">)</span>

<span class="token keyword">await</span> context<span class="token punctuation">.</span>storage_state<span class="token punctuation">(</span>path<span class="token operator">=</span><span class="token string">"auth.json"</span><span class="token punctuation">)</span>

下次执行时，可以直接加载登录状态：

python

context <span class="token operator">=</span> <span class="token keyword">await</span> browser<span class="token punctuation">.</span>new_context<span class="token punctuation">(</span>
storage_state<span class="token operator">=</span><span class="token string">"auth.json"</span>
<span class="token punctuation">)</span>

这样就不需要每次重新登录。

Playwright 的官方认证方案允许保存并复用已认证状态，从而减少重复登录并提高执行速度。需要注意的是，认证状态文件可能包含敏感 Cookie 和 Header，不应该提交到公开代码仓库。

2.4 操作 iframe

很多政务网站、企业系统和老旧平台都会使用 iframe。

页面结构可能是：

html

<span class="token tag"><span class="token tag"><span class="token punctuation"><</span>html</span><span class="token punctuation">></span></span>
<span class="token tag"><span class="token tag"><span class="token punctuation"><</span>body</span><span class="token punctuation">></span></span>
<span class="token tag"><span class="token tag"><span class="token punctuation"><</span>iframe</span> <span class="token attr-name">id</span><span class="token attr-value"><span class="token punctuation attr-equals">=</span><span class="token punctuation">"</span>content-frame<span class="token punctuation">"</span></span> <span class="token attr-name">src</span><span class="token attr-value"><span class="token punctuation attr-equals">=</span><span class="token punctuation">"</span>/detail/content<span class="token punctuation">"</span></span><span class="token punctuation">></span></span><span class="token tag"><span class="token tag"><span class="token punctuation"></</span>iframe</span><span class="token punctuation">></span></span>
<span class="token tag"><span class="token tag"><span class="token punctuation"></</span>body</span><span class="token punctuation">></span></span>
<span class="token tag"><span class="token tag"><span class="token punctuation"></</span>html</span><span class="token punctuation">></span></span>

正文并不在主页面中，而在 iframe 里。

Playwright 可以通过 frame_locator 进入 iframe：

python

frame <span class="token operator">=</span> page<span class="token punctuation">.</span>frame_locator<span class="token punctuation">(</span><span class="token string">"#content-frame"</span><span class="token punctuation">)</span>

title <span class="token operator">=</span> <span class="token keyword">await</span> frame<span class="token punctuation">.</span>locator<span class="token punctuation">(</span><span class="token string">"h1"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>inner_text<span class="token punctuation">(</span><span class="token punctuation">)</span>
content <span class="token operator">=</span> <span class="token keyword">await</span> frame<span class="token punctuation">.</span>locator<span class="token punctuation">(</span><span class="token string">".article-content"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>inner_text<span class="token punctuation">(</span><span class="token punctuation">)</span>

Playwright 的 FrameLocator 专门用于定位 iframe，并继续查找 iframe 内部的元素。

这对动态采集非常重要。

例如有些网站的详情页实际上分为两层：

text

外层壳页面
└── iframe
└── 真正的正文页面

如果直接在外层页面执行：

python

<span class="token keyword">await</span> page<span class="token punctuation">.</span>locator<span class="token punctuation">(</span><span class="token string">"body"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>inner_text<span class="token punctuation">(</span><span class="token punctuation">)</span>

拿到的可能是导航栏、菜单、版权信息和 iframe 周边文本，而不是正文。

更合理的方式是：

判断页面中是否存在 iframe；
找到当前激活的正文 iframe；
只在该 iframe 内提取正文；
iframe 不存在时，再回退到主页面提取。

2.5 监听和拦截网络请求

Playwright 不只能操作 DOM，还可以监听浏览器发送的网络请求。

python

page<span class="token punctuation">.</span>on<span class="token punctuation">(</span>
<span class="token string">"request"</span><span class="token punctuation">,</span>
<span class="token keyword">lambda</span> request<span class="token punctuation">:</span> <span class="token keyword">print</span><span class="token punctuation">(</span>
<span class="token string">"请求："</span><span class="token punctuation">,</span>
request<span class="token punctuation">.</span>method<span class="token punctuation">,</span>
request<span class="token punctuation">.</span>url
<span class="token punctuation">)</span>
<span class="token punctuation">)</span>

page<span class="token punctuation">.</span>on<span class="token punctuation">(</span>
<span class="token string">"response"</span><span class="token punctuation">,</span>
<span class="token keyword">lambda</span> response<span class="token punctuation">:</span> <span class="token keyword">print</span><span class="token punctuation">(</span>
<span class="token string">"响应："</span><span class="token punctuation">,</span>
response<span class="token punctuation">.</span>status<span class="token punctuation">,</span>
response<span class="token punctuation">.</span>url
<span class="token punctuation">)</span>
<span class="token punctuation">)</span>

这项能力在动态页面分析中非常实用。

例如页面上显示了一张列表，但 DOM 结构非常复杂。此时可以监听页面加载过程，找到真正返回 JSON 数据的接口：

python

<span class="token keyword">async</span> <span class="token keyword">with</span> page<span class="token punctuation">.</span>expect_response<span class="token punctuation">(</span>
<span class="token keyword">lambda</span> response<span class="token punctuation">:</span> <span class="token string">"/api/project/list"</span> <span class="token keyword">in</span> response<span class="token punctuation">.</span>url
<span class="token punctuation">)</span> <span class="token keyword">as</span> response_info<span class="token punctuation">:</span>
<span class="token keyword">await</span> page<span class="token punctuation">.</span>get_by_role<span class="token punctuation">(</span><span class="token string">"button"</span><span class="token punctuation">,</span> name<span class="token operator">=</span><span class="token string">"查询"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token punctuation">)</span>

response <span class="token operator">=</span> <span class="token keyword">await</span> response_info<span class="token punctuation">.</span>value
data <span class="token operator">=</span> <span class="token keyword">await</span> response<span class="token punctuation">.</span>json<span class="token punctuation">(</span><span class="token punctuation">)</span>

相比直接从页面文本中提取，读取接口 JSON 通常更加稳定。

Playwright 还支持：

修改请求 Header；
阻止图片、字体和视频；
Mock API 返回值；
修改接口响应；
监听 WebSocket；
Mock WebSocket 通信；
模拟接口异常和超时。

Playwright 官方网络能力覆盖 HTTP、HTTPS 和 WebSocket 的监听、修改与模拟。

例如，采集时可以阻止图片加载：

python

<span class="token keyword">async</span> <span class="token keyword">def</span> <span class="token function">handle_route</span><span class="token punctuation">(</span>route<span class="token punctuation">)</span><span class="token punctuation">:</span>
resource_type <span class="token operator">=</span> route<span class="token punctuation">.</span>request<span class="token punctuation">.</span>resource_type

<span class="token keyword">if</span> resource_type <span class="token keyword">in</span> <span class="token punctuation">{<!-- --></span><span class="token string">"image"</span><span class="token punctuation">,</span> <span class="token string">"font"</span><span class="token punctuation">,</span> <span class="token string">"media"</span><span class="token punctuation">}</span><span class="token punctuation">:</span>
<span class="token keyword">await</span> route<span class="token punctuation">.</span>abort<span class="token punctuation">(</span><span class="token punctuation">)</span>
<span class="token keyword">else</span><span class="token punctuation">:</span>
<span class="token keyword">await</span> route<span class="token punctuation">.</span>continue_<span class="token punctuation">(</span><span class="token punctuation">)</span>

<span class="token keyword">await</span> page<span class="token punctuation">.</span>route<span class="token punctuation">(</span><span class="token string">"**/*"</span><span class="token punctuation">,</span> handle_route<span class="token punctuation">)</span>

这样可以减少网络流量，但需要谨慎使用。有些页面会通过图片、字体或其他资源的加载结果判断页面状态，过度拦截可能导致页面逻辑异常。

2.6 文件上传与下载

Playwright 可以自动操作文件上传组件：

python

<span class="token keyword">await</span> page<span class="token punctuation">.</span>get_by_label<span class="token punctuation">(</span><span class="token string">"上传文件"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>set_input_files<span class="token punctuation">(</span>
<span class="token string">"documents/report.pdf"</span>
<span class="token punctuation">)</span>

也可以监听文件下载：

python

<span class="token keyword">async</span> <span class="token keyword">with</span> page<span class="token punctuation">.</span>expect_download<span class="token punctuation">(</span><span class="token punctuation">)</span> <span class="token keyword">as</span> download_info<span class="token punctuation">:</span>
<span class="token keyword">await</span> page<span class="token punctuation">.</span>get_by_role<span class="token punctuation">(</span><span class="token string">"button"</span><span class="token punctuation">,</span> name<span class="token operator">=</span><span class="token string">"导出"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token punctuation">)</span>

download <span class="token operator">=</span> <span class="token keyword">await</span> download_info<span class="token punctuation">.</span>value

<span class="token keyword">await</span> download<span class="token punctuation">.</span>save_as<span class="token punctuation">(</span>
<span class="token string-interpolation"><span class="token string">f"downloads/</span><span class="token interpolation"><span class="token punctuation">{<!-- --></span>download<span class="token punctuation">.</span>suggested_filename<span class="token punctuation">}</span></span><span class="token string">"</span></span>
<span class="token punctuation">)</span>

Playwright 会为下载事件创建对应的 Download 对象，可以获取文件名称、下载地址并保存文件。下载文件默认与产生它的 BrowserContext 生命周期相关。

因此，它非常适合：

自动下载报表；
自动导出 Excel；
上传标书或附件；
批量上传文档；
验证上传下载功能；
自动化处理后台管理系统。

2.7 截图、视频和 Trace

Playwright 可以对页面进行截图：

python

<span class="token keyword">await</span> page<span class="token punctuation">.</span>screenshot<span class="token punctuation">(</span>
path<span class="token operator">=</span><span class="token string">"page.png"</span><span class="token punctuation">,</span>
full_page<span class="token operator">=</span><span class="token boolean">True</span>
<span class="token punctuation">)</span>

也可以只截取某个元素：

python

<span class="token keyword">await</span> page<span class="token punctuation">.</span>locator<span class="token punctuation">(</span><span class="token string">".article-content"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>screenshot<span class="token punctuation">(</span>
path<span class="token operator">=</span><span class="token string">"article.png"</span>
<span class="token punctuation">)</span>

官方截图 API 支持全页面截图、指定区域截图和将图片读取为内存 Buffer。

除了截图，Playwright 还支持录制浏览器视频和 Trace。

Trace 可以记录：

每一步执行的操作；
操作前后的页面状态；
DOM 快照；
网络请求；
Console 日志；
页面截图；
操作耗时；
使用的 Locator；
错误发生的位置。

bash

npx playwright <span class="token builtin class-name">test</span> <span class="token parameter variable">--trace</span> on

运行完成后可以打开 Trace：

bash

npx playwright show-trace trace.zip

Trace Viewer 是 Playwright 提供的可视化分析工具，特别适合排查 CI 环境中偶发失败的问题。

2.8 模拟不同设备和网络环境

Playwright 可以模拟：

手机和平板；
不同屏幕尺寸；
不同 User-Agent；
不同语言；
不同时区；
地理位置；
摄像头和麦克风权限；
深色模式；
离线状态；
自定义 HTTP Header。

python

context <span class="token operator">=</span> <span class="token keyword">await</span> browser<span class="token punctuation">.</span>new_context<span class="token punctuation">(</span>
viewport<span class="token operator">=</span><span class="token punctuation">{<!-- --></span>
<span class="token string">"width"</span><span class="token punctuation">:</span> <span class="token number">390</span><span class="token punctuation">,</span>
<span class="token string">"height"</span><span class="token punctuation">:</span> <span class="token number">844</span>
<span class="token punctuation">}</span><span class="token punctuation">,</span>
user_agent<span class="token operator">=</span><span class="token string">"Mozilla/5.0 ..."</span><span class="token punctuation">,</span>
locale<span class="token operator">=</span><span class="token string">"zh-CN"</span><span class="token punctuation">,</span>
timezone_id<span class="token operator">=</span><span class="token string">"Asia/Shanghai"</span><span class="token punctuation">,</span>
color_scheme<span class="token operator">=</span><span class="token string">"dark"</span>
<span class="token punctuation">)</span>

这使 Playwright 不仅能测试桌面网页，还能验证移动端适配、国际化、权限和网络异常场景。浏览器和 BrowserContext 级别均可以配置设备模拟、网络和录制能力。

2.9 作为 AI Agent 的浏览器执行工具

随着 AI Agent 的发展，Playwright 也逐渐成为浏览器 Agent 的基础执行工具。

一个浏览器 Agent 的工作流程可能是：

text

用户提出任务
↓
大模型理解目标
↓
读取当前页面结构
↓
判断下一步操作
↓
调用 Playwright 点击、输入或跳转
↓
获取新的页面状态
↓
继续推理和执行

例如用户说：

登录后台，找到昨天失败的任务，把错误日志整理出来。

Agent 可以通过 Playwright：

打开系统；
填写账号密码；
进入任务中心；
选择时间范围；
筛选失败任务；
打开任务详情；
读取错误信息；
返回总结。

Microsoft 目前提供了 Playwright MCP Server，使大模型能够通过 Model Context Protocol 调用浏览器自动化能力。官方实现主要使用结构化的可访问性快照帮助模型理解页面，而不是完全依赖截图和视觉识别。

三、Playwright 的核心对象模型

学习 Playwright 时，最重要的是理解以下几个核心对象：

text

Playwright
├── Chromium
├── Firefox
└── WebKit
↓
Browser
↓
BrowserContext
↓
Page
↓
Frame / Locator

3.1 Playwright

Playwright 是整个框架的入口。

python

<span class="token keyword">async</span> <span class="token keyword">with</span> async_playwright<span class="token punctuation">(</span><span class="token punctuation">)</span> <span class="token keyword">as</span> p<span class="token punctuation">:</span>
browser <span class="token operator">=</span> <span class="token keyword">await</span> p<span class="token punctuation">.</span>chromium<span class="token punctuation">.</span>launch<span class="token punctuation">(</span><span class="token punctuation">)</span>

这里的 p 就是 Playwright 实例。

它提供三个主要的浏览器类型：

python

p<span class="token punctuation">.</span>chromium
p<span class="token punctuation">.</span>firefox
p<span class="token punctuation">.</span>webkit

3.2 Browser

Browser 表示一个浏览器进程。

python

browser <span class="token operator">=</span> <span class="token keyword">await</span> p<span class="token punctuation">.</span>chromium<span class="token punctuation">.</span>launch<span class="token punctuation">(</span>
headless<span class="token operator">=</span><span class="token boolean">True</span>
<span class="token punctuation">)</span>

常见配置包括：

python

browser <span class="token operator">=</span> <span class="token keyword">await</span> p<span class="token punctuation">.</span>chromium<span class="token punctuation">.</span>launch<span class="token punctuation">(</span>
headless<span class="token operator">=</span><span class="token boolean">False</span><span class="token punctuation">,</span>
slow_mo<span class="token operator">=</span><span class="token number">500</span>
<span class="token punctuation">)</span>

其中：

headless=True：无界面运行；
headless=False：显示浏览器界面；
slow_mo=500：每一步操作放慢 500 毫秒。

在开发和排查问题时，可以使用有界面模式；在服务器和生产环境中，通常使用无头模式。

3.3 BrowserContext

BrowserContext 是 Playwright 非常重要的设计。

可以把它理解为一个独立的无痕浏览器环境。

每个 BrowserContext 都有自己独立的：

Cookie；
LocalStorage；
SessionStorage；
页面；
权限；
User-Agent；
网络规则；
登录状态。

python

context_1 <span class="token operator">=</span> <span class="token keyword">await</span> browser<span class="token punctuation">.</span>new_context<span class="token punctuation">(</span><span class="token punctuation">)</span>
context_2 <span class="token operator">=</span> <span class="token keyword">await</span> browser<span class="token punctuation">.</span>new_context<span class="token punctuation">(</span><span class="token punctuation">)</span>

即使它们共用同一个 Browser 进程，也不会共享登录状态。

Playwright 使用 BrowserContext 实现测试隔离，每个测试可以拥有独立的浏览器上下文，从而降低状态污染和级联失败。

这意味着执行批量任务时，不一定要为每个任务都启动一个新的浏览器进程。

更常见的方式是：

text

一个 Browser
├── Context A：账号 A
├── Context B：账号 B
└── Context C：游客状态

这样既实现状态隔离，也能减少频繁启动浏览器的成本。

3.4 Page

Page 表示浏览器中的一个标签页或弹窗页面。

python

page <span class="token operator">=</span> <span class="token keyword">await</span> context<span class="token punctuation">.</span>new_page<span class="token punctuation">(</span><span class="token punctuation">)</span>

一个 BrowserContext 可以包含多个 Page。

python

page_1 <span class="token operator">=</span> <span class="token keyword">await</span> context<span class="token punctuation">.</span>new_page<span class="token punctuation">(</span><span class="token punctuation">)</span>
page_2 <span class="token operator">=</span> <span class="token keyword">await</span> context<span class="token punctuation">.</span>new_page<span class="token punctuation">(</span><span class="token punctuation">)</span>

Playwright 中的 Page 可以：

打开 URL；
获取 HTML；
执行 JavaScript；
操作页面元素；
监听请求；
监听弹窗；
监听下载；
截图；
获取 Console 日志。

官方将 Page 定义为 BrowserContext 中的一个标签页或弹出窗口。

3.5 Locator

Locator 是 Playwright 元素定位机制的核心。

python

login_button <span class="token operator">=</span> page<span class="token punctuation">.</span>get_by_role<span class="token punctuation">(</span>
<span class="token string">"button"</span><span class="token punctuation">,</span>
name<span class="token operator">=</span><span class="token string">"登录"</span>
<span class="token punctuation">)</span>

这里的 login_button 不是某一个固定 DOM 节点，而是一种“如何找到这个元素”的描述。

执行操作时，Playwright 会重新查找当前最新的 DOM 元素：

python

<span class="token keyword">await</span> login_button<span class="token punctuation">.</span>hover<span class="token punctuation">(</span><span class="token punctuation">)</span>
<span class="token keyword">await</span> login_button<span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token punctuation">)</span>

即使两次操作之间页面发生了重新渲染，Playwright 也会在每次操作前重新解析 Locator，而不是一直引用旧的 DOM 节点。

这对 React、Vue 等动态更新 DOM 的页面非常重要。

四、Playwright 的底层工作原理

4.1 整体调用链路

Playwright 的工作过程可以抽象为：

开发者编写：

python

<span class="token keyword">await</span> page<span class="token punctuation">.</span>get_by_role<span class="token punctuation">(</span>
<span class="token string">"button"</span><span class="token punctuation">,</span>
name<span class="token operator">=</span><span class="token string">"登录"</span>
<span class="token punctuation">)</span><span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token punctuation">)</span>

背后会经历以下过程：

Python 客户端接收调用；
将操作转换为 Playwright 协议消息；
Playwright Driver 接收消息；
Driver 将操作发送给对应浏览器；
浏览器查找元素并执行点击；
浏览器产生页面、网络和事件变化；
Driver 将结果返回给 Python 客户端；
Python 代码继续执行。

以 Playwright Python 为例，官方代码仓库说明，Python 客户端通过管道向内置的 Node.js Driver 发送 JSON 消息，通信协议由 Playwright 上游协议定义。

因此，Python 版本并不是完全用 Python 重新实现了一套浏览器控制核心。

更准确的理解是：

text

Python API
↓
Playwright Driver
↓
浏览器

Java 和 .NET 等语言版本也采用类似的语言绑定与 Driver 通信方式。

4.2 Playwright 并不只是 CDP 的简单封装

CDP 是 Chrome DevTools Protocol，也就是 Chromium 系浏览器使用的调试协议。

Playwright 确实支持通过 connect_over_cdp 连接现有 Chromium 浏览器：

python

browser <span class="token operator">=</span> <span class="token keyword">await</span> p<span class="token punctuation">.</span>chromium<span class="token punctuation">.</span>connect_over_cdp<span class="token punctuation">(</span>
<span class="token string">"http://localhost:9222"</span>
<span class="token punctuation">)</span>

但 CDP 连接仅适用于 Chromium 浏览器。Playwright 官方还明确指出，通过 CDP 建立的连接，相比 Playwright 自己的协议连接功能完整度更低；复杂功能更适合使用 Playwright 原生连接方式。

因此，不能简单地把 Playwright 理解为：

在 CDP 外面封装了一层 API。

更准确的说法是：

Playwright 提供统一的上层 API 和自己的通信协议，再针对 Chromium、Firefox 和 WebKit 实现浏览器控制能力。

4.3 事件驱动模型

浏览器中的很多行为不是同步发生的。

例如点击按钮后，可能产生：

页面跳转；
弹出新窗口；
文件下载；
接口请求；
iframe 加载；
WebSocket 消息；
DOM 更新。

因此 Playwright 大量使用事件驱动机制。

例如等待新标签页：

python

<span class="token keyword">async</span> <span class="token keyword">with</span> context<span class="token punctuation">.</span>expect_page<span class="token punctuation">(</span><span class="token punctuation">)</span> <span class="token keyword">as</span> page_info<span class="token punctuation">:</span>
<span class="token keyword">await</span> page<span class="token punctuation">.</span>get_by_text<span class="token punctuation">(</span><span class="token string">"打开详情"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token punctuation">)</span>

new_page <span class="token operator">=</span> <span class="token keyword">await</span> page_info<span class="token punctuation">.</span>value

等待下载：

python

<span class="token keyword">async</span> <span class="token keyword">with</span> page<span class="token punctuation">.</span>expect_download<span class="token punctuation">(</span><span class="token punctuation">)</span> <span class="token keyword">as</span> download_info<span class="token punctuation">:</span>
<span class="token keyword">await</span> page<span class="token punctuation">.</span>get_by_text<span class="token punctuation">(</span><span class="token string">"下载"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token punctuation">)</span>

download <span class="token operator">=</span> <span class="token keyword">await</span> download_info<span class="token punctuation">.</span>value

等待接口响应：

python

<span class="token keyword">async</span> <span class="token keyword">with</span> page<span class="token punctuation">.</span>expect_response<span class="token punctuation">(</span>
<span class="token keyword">lambda</span> response<span class="token punctuation">:</span> <span class="token string">"/api/detail"</span> <span class="token keyword">in</span> response<span class="token punctuation">.</span>url
<span class="token punctuation">)</span> <span class="token keyword">as</span> response_info<span class="token punctuation">:</span>
<span class="token keyword">await</span> page<span class="token punctuation">.</span>get_by_text<span class="token punctuation">(</span><span class="token string">"查看详情"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token punctuation">)</span>

response <span class="token operator">=</span> <span class="token keyword">await</span> response_info<span class="token punctuation">.</span>value

这里有一个重要原则：

先注册等待事件，再触发页面操作。

错误写法：

python

<span class="token keyword">await</span> page<span class="token punctuation">.</span>get_by_text<span class="token punctuation">(</span><span class="token string">"下载"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token punctuation">)</span>
download <span class="token operator">=</span> <span class="token keyword">await</span> page<span class="token punctuation">.</span>wait_for_event<span class="token punctuation">(</span><span class="token string">"download"</span><span class="token punctuation">)</span>

点击完成时，下载事件可能已经发生，后续等待就可能超时。

正确写法：

python

<span class="token keyword">async</span> <span class="token keyword">with</span> page<span class="token punctuation">.</span>expect_download<span class="token punctuation">(</span><span class="token punctuation">)</span> <span class="token keyword">as</span> download_info<span class="token punctuation">:</span>
<span class="token keyword">await</span> page<span class="token punctuation">.</span>get_by_text<span class="token punctuation">(</span><span class="token string">"下载"</span><span class="token punctuation">)</span><span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token punctuation">)</span>

五、Playwright 为什么比传统自动化脚本更稳定

5.1 自动等待

传统浏览器脚本中经常出现大量固定等待：

python

<span class="token keyword">await</span> page<span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token string">"#submit"</span><span class="token punctuation">)</span>
<span class="token keyword">await</span> asyncio<span class="token punctuation">.</span>sleep<span class="token punctuation">(</span><span class="token number">3</span><span class="token punctuation">)</span>

问题在于：

网络快时，浪费三秒；
网络慢时，三秒仍然不够；
CI 环境负载变化时容易偶发失败。

Playwright 在执行点击等操作之前，会自动检查元素状态。

以点击为例，Playwright 通常会检查：

元素是否已经找到；
元素是否可见；
元素位置是否稳定；
元素是否启用；
元素是否能够接收点击事件；
元素是否被其他弹窗或遮罩层挡住。

只有满足操作条件后，Playwright 才会执行点击。

python

<span class="token keyword">await</span> page<span class="token punctuation">.</span>get_by_role<span class="token punctuation">(</span>
<span class="token string">"button"</span><span class="token punctuation">,</span>
name<span class="token operator">=</span><span class="token string">"提交"</span>
<span class="token punctuation">)</span><span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token punctuation">)</span>

这行代码背后实际上包含了查找、等待、状态检查、滚动和点击等过程。

5.2 自动重试断言

Playwright 的 Web First Assertions 也会自动重试。

python

expect<span class="token punctuation">(</span>page<span class="token punctuation">.</span>locator<span class="token punctuation">(</span><span class="token string">".status"</span><span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">.</span>to_have_text<span class="token punctuation">(</span>
<span class="token string">"处理完成"</span>
<span class="token punctuation">)</span>

如果页面暂时显示的是：

text

处理中

Playwright 不会立即判定失败，而是会在超时时间内持续检查，直到：

text

处理完成

或者等待超时。

Playwright 官方断言支持对元素可见性、文本、属性、状态、URL 和响应等条件进行自动重试。

5.3 Locator 会重新解析元素

传统代码可能先找到 DOM 元素，然后长期持有这个引用。

但 React 或 Vue 重新渲染后，原来的元素可能已经被删除并替换。

Playwright 推荐使用 Locator：

python

button <span class="token operator">=</span> page<span class="token punctuation">.</span>get_by_role<span class="token punctuation">(</span>
<span class="token string">"button"</span><span class="token punctuation">,</span>
name<span class="token operator">=</span><span class="token string">"提交"</span>
<span class="token punctuation">)</span>

<span class="token keyword">await</span> button<span class="token punctuation">.</span>hover<span class="token punctuation">(</span><span class="token punctuation">)</span>
<span class="token keyword">await</span> button<span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token punctuation">)</span>

每次真正执行操作时，Locator 都会根据当前页面重新寻找元素，从而减少旧元素引用导致的问题。

六、Playwright 常见定位方式

Playwright 支持多种元素定位方式。

6.1 根据角色定位

python

page<span class="token punctuation">.</span>get_by_role<span class="token punctuation">(</span>
<span class="token string">"button"</span><span class="token punctuation">,</span>
name<span class="token operator">=</span><span class="token string">"登录"</span>
<span class="token punctuation">)</span>

适合按钮、链接、输入框、复选框等标准元素。

6.2 根据 Label 定位

python

page<span class="token punctuation">.</span>get_by_label<span class="token punctuation">(</span><span class="token string">"用户名"</span><span class="token punctuation">)</span>
page<span class="token punctuation">.</span>get_by_label<span class="token punctuation">(</span><span class="token string">"密码"</span><span class="token punctuation">)</span>

适合表单元素。

6.3 根据文本定位

python

page<span class="token punctuation">.</span>get_by_text<span class="token punctuation">(</span><span class="token string">"查看详情"</span><span class="token punctuation">)</span>

6.4 根据 Placeholder 定位

python

page<span class="token punctuation">.</span>get_by_placeholder<span class="token punctuation">(</span><span class="token string">"请输入关键词"</span><span class="token punctuation">)</span>

6.5 根据 Test ID 定位

python

page<span class="token punctuation">.</span>get_by_test_id<span class="token punctuation">(</span><span class="token string">"submit-button"</span><span class="token punctuation">)</span>

前端页面：

html

<span class="token tag"><span class="token tag"><span class="token punctuation"><</span>button</span> <span class="token attr-name">data-testid</span><span class="token attr-value"><span class="token punctuation attr-equals">=</span><span class="token punctuation">"</span>submit-button<span class="token punctuation">"</span></span><span class="token punctuation">></span></span>
提交
<span class="token tag"><span class="token tag"><span class="token punctuation"></</span>button</span><span class="token punctuation">></span></span>

6.6 CSS 选择器

python

page<span class="token punctuation">.</span>locator<span class="token punctuation">(</span><span class="token string">".article-list .article-item"</span><span class="token punctuation">)</span>

6.7 XPath

python

page<span class="token punctuation">.</span>locator<span class="token punctuation">(</span>
<span class="token string">"xpath=//button[contains(text(),'提交')]"</span>
<span class="token punctuation">)</span>

虽然 XPath 能力很强，但通常不应该优先使用过长的绝对 XPath：

python

<span class="token operator">/</span>html<span class="token operator">/</span>body<span class="token operator">/</span>div<span class="token punctuation">[</span><span class="token number">2</span><span class="token punctuation">]</span><span class="token operator">/</span>div<span class="token punctuation">[</span><span class="token number">3</span><span class="token punctuation">]</span><span class="token operator">/</span>div<span class="token operator">/</span>button

页面结构稍微变化，定位就可能失效。

Playwright 官方最佳实践建议优先使用用户可感知的属性和明确契约，例如 Role、Label、Text 和 Test ID，而不是依赖脆弱的 DOM 层级。

七、一个相对完整的 Python 示例

下面使用 Playwright 完成：

启动浏览器；
创建独立上下文；
打开页面；
监听接口；
点击查询；
提取列表；
保存截图；
捕获异常；
释放资源。

python

<span class="token keyword">import</span> asyncio
<span class="token keyword">from</span> pathlib <span class="token keyword">import</span> Path

<span class="token keyword">from</span> playwright<span class="token punctuation">.</span>async_api <span class="token keyword">import</span> <span class="token punctuation">(</span>
async_playwright<span class="token punctuation">,</span>
TimeoutError <span class="token keyword">as</span> PlaywrightTimeoutError<span class="token punctuation">,</span>
<span class="token punctuation">)</span>

<span class="token keyword">async</span> <span class="token keyword">def</span> <span class="token function">collect_data</span><span class="token punctuation">(</span><span class="token punctuation">)</span> <span class="token operator">-</span><span class="token operator">></span> <span class="token builtin">list</span><span class="token punctuation">[</span><span class="token builtin">dict</span><span class="token punctuation">[</span><span class="token builtin">str</span><span class="token punctuation">,</span> <span class="token builtin">str</span><span class="token punctuation">]</span><span class="token punctuation">]</span><span class="token punctuation">:</span>
output_dir <span class="token operator">=</span> Path<span class="token punctuation">(</span><span class="token string">"artifacts"</span><span class="token punctuation">)</span>
output_dir<span class="token punctuation">.</span>mkdir<span class="token punctuation">(</span>parents<span class="token operator">=</span><span class="token boolean">True</span><span class="token punctuation">,</span> exist_ok<span class="token operator">=</span><span class="token boolean">True</span><span class="token punctuation">)</span>

<span class="token keyword">async</span> <span class="token keyword">with</span> async_playwright<span class="token punctuation">(</span><span class="token punctuation">)</span> <span class="token keyword">as</span> playwright<span class="token punctuation">:</span>
browser <span class="token operator">=</span> <span class="token keyword">await</span> playwright<span class="token punctuation">.</span>chromium<span class="token punctuation">.</span>launch<span class="token punctuation">(</span>
headless<span class="token operator">=</span><span class="token boolean">True</span>
<span class="token punctuation">)</span>

context <span class="token operator">=</span> <span class="token keyword">await</span> browser<span class="token punctuation">.</span>new_context<span class="token punctuation">(</span>
locale<span class="token operator">=</span><span class="token string">"zh-CN"</span><span class="token punctuation">,</span>
viewport<span class="token operator">=</span><span class="token punctuation">{<!-- --></span>
<span class="token string">"width"</span><span class="token punctuation">:</span> <span class="token number">1440</span><span class="token punctuation">,</span>
<span class="token string">"height"</span><span class="token punctuation">:</span> <span class="token number">900</span><span class="token punctuation">,</span>
<span class="token punctuation">}</span><span class="token punctuation">,</span>
<span class="token punctuation">)</span>

page <span class="token operator">=</span> <span class="token keyword">await</span> context<span class="token punctuation">.</span>new_page<span class="token punctuation">(</span><span class="token punctuation">)</span>

<span class="token keyword">try</span><span class="token punctuation">:</span>
<span class="token keyword">await</span> page<span class="token punctuation">.</span>goto<span class="token punctuation">(</span>
<span class="token string">"https://example.com/projects"</span><span class="token punctuation">,</span>
wait_until<span class="token operator">=</span><span class="token string">"domcontentloaded"</span><span class="token punctuation">,</span>
timeout<span class="token operator">=</span><span class="token number">30_000</span><span class="token punctuation">,</span>
<span class="token punctuation">)</span>

keyword_input <span class="token operator">=</span> page<span class="token punctuation">.</span>get_by_placeholder<span class="token punctuation">(</span>
<span class="token string">"请输入项目名称"</span>
<span class="token punctuation">)</span>

<span class="token keyword">await</span> keyword_input<span class="token punctuation">.</span>fill<span class="token punctuation">(</span><span class="token string">"人工智能"</span><span class="token punctuation">)</span>

<span class="token keyword">async</span> <span class="token keyword">with</span> page<span class="token punctuation">.</span>expect_response<span class="token punctuation">(</span>
<span class="token keyword">lambda</span> response<span class="token punctuation">:</span> <span class="token punctuation">(</span>
<span class="token string">"/api/project/list"</span> <span class="token keyword">in</span> response<span class="token punctuation">.</span>url
<span class="token keyword">and</span> response<span class="token punctuation">.</span>status <span class="token operator">==</span> <span class="token number">200</span>
<span class="token punctuation">)</span><span class="token punctuation">,</span>
timeout<span class="token operator">=</span><span class="token number">20_000</span><span class="token punctuation">,</span>
<span class="token punctuation">)</span> <span class="token keyword">as</span> response_info<span class="token punctuation">:</span>
<span class="token keyword">await</span> page<span class="token punctuation">.</span>get_by_role<span class="token punctuation">(</span>
<span class="token string">"button"</span><span class="token punctuation">,</span>
name<span class="token operator">=</span><span class="token string">"查询"</span><span class="token punctuation">,</span>
<span class="token punctuation">)</span><span class="token punctuation">.</span>click<span class="token punctuation">(</span><span class="token punctuation">)</span>

response <span class="token operator">=</span> <span class="token keyword">await</span> response_info<span class="token punctuation">.</span>value
response_data <span class="token operator">=</span> <span class="token keyword">await</span> response<span class="token punctuation">.</span>json<span class="token punctuation">(</span><span class="token punctuation">)</span>

<span class="token keyword">await</span> page<span class="token punctuation">.</span>locator<span class="token punctuation">(</span>
<span class="token string">".project-list"</span>
<span class="token punctuation">)</span><span class="token punctuation">.</span>wait_for<span class="token punctuation">(</span>
state<span class="token operator">=</span><span class="token string">"visible"</span><span class="token punctuation">,</span>
timeout<span class="token operator">=</span><span class="token number">20_000</span><span class="token punctuation">,</span>
<span class="token punctuation">)</span>

rows <span class="token operator">=</span> page<span class="token punctuation">.</span>locator<span class="token punctuation">(</span><span class="token string">".project-item"</span><span class="token punctuation">)</span>
count <span class="token operator">=</span> <span class="token keyword">await</span> rows<span class="token punctuation">.</span>count<span class="token punctuation">(</span><span class="token punctuation">)</span>

results<span class="token punctuation">:</span> <span class="token builtin">list</span><span class="token punctuation">[</span><span class="token builtin">dict</span><span class="token punctuation">[</span><span class="token builtin">str</span><span class="token punctuation">,</span> <span class="token builtin">str</span><span class="token punctuation">]</span><span class="token punctuation">]</span> <span class="token operator">=</span> <span class="token punctuation">[</span><span class="token punctuation">]</span>

<span class="token keyword">for</span> index <span class="token keyword">in</span> <span class="token builtin">range</span><span class="token punctuation">(</span>count<span class="token punctuation">)</span><span class="token punctuation">:</span>
row <span class="token operator">=</span> rows<span class="token punctuation">.</span>nth<span class="token punctuation">(</span>index<span class="token punctuation">)</span>

title <span class="token operator">=</span> <span class="token keyword">await</span> row<span class="token punctuation">.</span>locator<span class="token punctuation">(</span>
<span class="token string">".project-title"</span>
<span class="token punctuation">)</span><span class="token punctuation">.</span>inner_text<span class="token punctuation">(</span><span class="token punctuation">)</span>

date <span class="token operator">=</span> <span class="token keyword">await</span> row<span class="token punctuation">.</span>locator<span class="token punctuation">(</span>
<span class="token string">".project-date"</span>
<span class="token punctuation">)</span><span class="token punctuation">.</span>inner_text<span class="token punctuation">(</span><span class="token punctuation">)</span>

results<span class="token punctuation">.</span>append<span class="token punctuation">(</span>
<span class="token punctuation">{<!-- --></span>
<span class="token string">"title"</span><span class="token punctuation">:</span> title<span class="token punctuation">.</span>strip<span class="token punctuation">(</span><span class="token punctuation">)</span><span class="token punctuation">,</span>
<span class="token string">"date"</span><span class="token punctuation">:</span> date<span class="token punctuation">.</span>strip<span class="token punctuation">(</span><span class="token punctuation">)</span><span class="token punctuation">,</span>
<span class="token punctuation">}</span>
<span class="token punctuation">)</span>

<span class="token keyword">await</span> page<span class="token punctuation">.</span>screenshot<span class="token punctuation">(</span>
path<span class="token operator">=</span>output_dir <span class="token operator">/</span> <span class="token string">"result.png"</span><span class="token punctuation">,</span>
full_page<span class="token operator">=</span><span class="token boolean">True</span><span class="token punctuation">,</span>
<span class="token punctuation">)</span>

<span class="token keyword">print</span><span class="token punctuation">(</span>
<span class="token string">"接口返回数据量："</span><span class="token punctuation">,</span>
<span class="token builtin">len</span><span class="token punctuation">(</span>response_data<span class="token punctuation">.</span>get<span class="token punctuation">(</span><span class="token string">"data"</span><span class="token punctuation">,</span> <span class="token punctuation">[</span><span class="token punctuation">]</span><span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">,</span>
<span class="token punctuation">)</span>

<span class="token keyword">return</span> results

<span class="token keyword">except</span> PlaywrightTimeoutError<span class="token punctuation">:</span>
<span class="token keyword">await</span> page<span class="token punctuation">.</span>screenshot<span class="token punctuation">(</span>
path<span class="token operator">=</span>output_dir <span class="token operator">/</span> <span class="token string">"timeout.png"</span><span class="token punctuation">,</span>
full_page<span class="token operator">=</span><span class="token boolean">True</span><span class="token punctuation">,</span>
<span class="token punctuation">)</span>
<span class="token keyword">raise</span> RuntimeError<span class="token punctuation">(</span><span class="token string">"页面操作超时"</span><span class="token punctuation">)</span>

<span class="token keyword">finally</span><span class="token punctuation">:</span>
<span class="token keyword">await</span> context<span class="token punctuation">.</span>close<span class="token punctuation">(</span><span class="token punctuation">)</span>
<span class="token keyword">await</span> browser<span class="token punctuation">.</span>close<span class="token punctuation">(</span><span class="token punctuation">)</span>

<span class="token keyword">if</span> __name__ <span class="token operator">==</span> <span class="token string">"__main__"</span><span class="token punctuation">:</span>
data <span class="token operator">=</span> asyncio<span class="token punctuation">.</span>run<span class="token punctuation">(</span>collect_data<span class="token punctuation">(</span><span class="token punctuation">)</span><span class="token punctuation">)</span>

<span class="token keyword">for</span> item <span class="token keyword">in</span> data<span class="token punctuation">:</span>
<span class="token keyword">print</span><span class="token punctuation">(</span>item<span class="token punctuation">)</span>

八、Playwright 在生产环境中的架构设计

直接写一个 Playwright 脚本并不难，难的是将它建设成稳定的生产系统。

推荐将系统拆分为以下几层：

8.1 Browser 不要频繁启动

错误方式：

text

每来一个任务
↓
启动浏览器
↓
执行任务
↓
关闭浏览器

浏览器启动属于相对昂贵的操作。

更合理的方式是：

text

Worker 启动
↓
创建 Browser
↓
任务到达
↓
创建 BrowserContext
↓
执行任务
↓
关闭 BrowserContext
↓
复用 Browser

但是 Browser 也不能永久运行，需要配置：

最大任务数；
最大存活时间；
内存阈值；
页面崩溃检测；
Browser 重启机制。

8.2 任务之间使用 BrowserContext 隔离

不同用户和任务不应该共用同一个 Context，否则可能产生：

Cookie 串用；
登录状态污染；
LocalStorage 污染；
页面残留；
权限相互影响。

8.3 设置分层超时

不要只有一个总超时。

建议分别设置：

text

任务总超时
├── 页面打开超时
├── 元素等待超时
├── 接口等待超时
├── 下载超时
└── 单步操作超时

例如：

python

context<span class="token punctuation">.</span>set_default_timeout<span class="token punctuation">(</span><span class="token number">10_000</span><span class="token punctuation">)</span>
context<span class="token punctuation">.</span>set_default_navigation_timeout<span class="token punctuation">(</span><span class="token number">30_000</span><span class="token punctuation">)</span>

任务调度层还应设置更高层的总超时，避免整个 Worker 被某个任务长期占用。

8.4 做好可观测性

生产环境中的 Playwright 任务失败时，只保存一条异常日志通常不够。

至少应该保存：

当前 URL；
任务 ID；
页面标题；
异常堆栈；
页面截图；
页面 HTML；
Console 日志；
失败请求；
Trace；
浏览器版本；
脚本版本；
账号或会话标识。

这样才能回答：

是页面改版了、接口失败了、登录失效了、元素被遮挡了，还是浏览器崩溃了？

8.5 设计任务状态机

一个 Playwright 任务通常不是简单的成功或失败。

可以设计为：

text

PENDING
↓
RUNNING
↓
LOGIN_REQUIRED
↓
COLLECTING
↓
PARSING
↓
SUCCESS

异常状态包括：

text

TIMEOUT
AUTH_EXPIRED
PAGE_CHANGED
NETWORK_ERROR
BROWSER_CRASHED
CAPTCHA_REQUIRED
FAILED

通过明确的状态机，可以决定哪些错误能够自动重试，哪些错误需要人工介入。

九、Playwright 的局限性

Playwright 很强，但并不是万能的。

9.1 资源消耗较大

每个浏览器进程和页面都需要消耗：

CPU；
内存；
文件描述符；
网络连接；
临时磁盘空间。

如果同时启动几百个浏览器实例，服务器很容易出现资源耗尽。

因此，需要限制：

Worker 数量；
Browser 数量；
Context 数量；
Page 数量；
单任务并发；
单浏览器存活时间。

9.2 无法天然解决验证码

Playwright 可以操作验证码页面，但它并不能天然解决：

图片验证码；
滑块验证码；
短信验证码；
扫码登录；
行为验证；
人机验证。

遇到这些场景时，通常需要：

人工介入；
官方接口；
合法授权的账号体系；
验证码识别服务；
会话状态复用。

9.3 不能天然绕过所有反爬系统

使用真实浏览器不等于完全无法被识别。

网站可能综合判断：

IP；
Cookie；
浏览器指纹；
请求频率；
鼠标轨迹；
页面停留时间；
账号行为；
TLS 指纹；
Header；
访问路径；
浏览器自动化特征。

因此，Playwright 的定位应该是：

浏览器自动化工具，而不是万能的风控绕过工具。

使用时也应遵守目标网站的服务条款、数据授权范围和相关法律法规。

9.4 页面改版仍然会导致脚本失效

如果页面按钮名称、DOM 结构或业务流程发生变化，自动化脚本仍然可能失败。

稳定性需要依赖：

更可靠的 Locator；
页面版本监控；
多种提取策略；
失败截图；
Trace；
结构变化检测；
回退逻辑；
自动化回归测试。

十、Playwright、HTTP 请求和人工浏览器如何选择

场景	推荐方案
静态 HTML 页面	requests、httpx
已知并且稳定的 JSON 接口	直接请求接口
JavaScript 动态渲染页面	Playwright
必须登录才能访问	Playwright 或复用登录 Cookie
iframe 页面	Playwright
需要点击、输入、滚动	Playwright
高并发批量接口采集	HTTP 客户端
自动化回归测试	Playwright Test
自动下载后台报表	Playwright
AI Agent 操作网页	Playwright、Playwright MCP
复杂验证码或人工审批	人机协同

一个成熟系统通常不会只使用一种方式。

更合理的组合是：

text

Playwright
负责登录、交互、页面分析和接口发现

HTTP 客户端
负责稳定接口的高并发数据请求

HTML 解析器
负责结构化提取

任务队列
负责并发、重试和调度

对象存储
负责保存截图、HTML、下载文件和 Trace

十一、最佳实践总结

在实际项目中使用 Playwright，可以重点遵循以下原则。

第一，优先使用 Locator，不要长期持有旧的 DOM ElementHandle。

第二，优先使用 Role、Label、Text 和 Test ID，尽量避免脆弱的绝对 XPath。

第三，减少固定 sleep，依赖 Locator 自动等待、接口事件和明确的页面状态。

第四，一个 Browser 可以复用，但不同任务尽量使用独立 BrowserContext。

第五，动态页面优先分析网络接口，不要所有数据都从页面文本中硬解析。

第六，出现 iframe 时，应明确进入目标 iframe，而不是直接扫描整个外层页面。

第七，等待下载、弹窗和接口时，要先注册事件，再触发操作。

第八，生产环境必须保存截图、HTML、Console、网络错误和 Trace。

第九，对浏览器、Context、Page 和任务设置资源上限与超时。

第十，不要把 Playwright 当作万能反爬工具，应在授权和合规范围内使用。

十二、结语

Playwright 的价值并不只是“自动点击网页”。

它真正解决的是：

如何让程序稳定地进入一个真实浏览器环境，并观察、操作和验证现代 Web 应用。

从能力上看，Playwright 横跨了多个领域：

text

自动化测试
动态网页采集
后台流程自动化
文件上传下载
接口调试与 Mock
页面截图与录制
浏览器 Agent
AI 自动化执行

它的核心优势来自几个关键设计：

统一控制 Chromium、Firefox 和 WebKit；
使用 BrowserContext 实现会话隔离；
使用 Locator 应对动态 DOM；
通过自动等待减少脚本偶发失败；
通过网络监听获取页面背后的真实数据；
通过 Trace 提升自动化任务的可调试性；
通过 MCP 等方式成为 AI Agent 的浏览器执行层。

如果只是采集一个简单的静态页面，使用 Playwright 可能显得过重。

但当你面对登录系统、动态渲染、iframe、复杂交互、文件下载、浏览器测试或 AI Agent 操作网页时，Playwright 往往是目前最值得掌握的浏览器自动化工具之一。