repository·main·Indexed 18 days ago
https://github.com/cloudflare/puppeteerA specialized fork of Puppeteer core optimized for Cloudflare Browser Run, providing a lightweight version for use within Cloudflare Workers. It utilizes the standard Chrome DevTools Protocol (CDP) for communication and includes tools such as @puppeteer/browsers for browser installation, @puppeteer/ng-schematics for Angular E2E testing, and @pptr/testserver for local HTTP/HTTPS testing servers.
The @puppeteer/browsers programmatic API allows you to install, launch, and manage browsers directly from your code.
install(options) to download browsers and uninstall(options) to remove them. Use canDownload(options) to check availability.launch(opts) to start a browser instance. Note that launching system browsers is only supported for Chrome/Chromium.computeExecutablePath(options) for managed browsers or computeSystemExecutablePath(options) for system-installed browsers.getInstalledBrowsers(options) to retrieve metadata about browsers currently in your cache directory.createProfile(browser, opts) to manage browser profiles.When calling these functions, you will interact with several option interfaces:
InstallOptionsLaunchOptionsUninstallOptionsGetInstalledBrowsersOptionsProfileOptionsSystemOptionsThis repository is a specialized fork of Puppeteer core designed for use with Cloudflare Browser Run (formerly Browser Rendering). It aims to minimize library size for Workers and provide a seamless experience in the Cloudflare ecosystem.
Starting with @cloudflare/puppeteer version 1.1.0, the library uses the standard Chrome DevTools Protocol (CDP) to communicate with Browser Run. Most existing Puppeteer code should work without modification.
Puppeteer raises a Request is already handled! exception if abort, continue, or respond are called more than once for the same request.
Because 3rd party packages or other listeners might resolve a request while your handler is awaiting an asynchronous operation, you must follow these rules:
request.isInterceptResolutionHandled() (or request.interceptResolutionState()) immediately before calling a resolution method.abort/continue/respond) within the same synchronous code block to avoid race conditions.async, you must re-verify the resolution status after every await before proceeding with a resolution.page.on('request', async interceptedRequest => {
// 1. Initial check
if (interceptedRequest.isInterceptResolutionHandled()) return;
await someLongAsyncOperation();
// 2. Re-check after async operation
if (interceptedRequest.isInterceptResolutionHandled()) return;
interceptedRequest.continue();
});Puppeteer is not a direct replacement for Selenium WebDriver, as they serve different primary purposes:
| Feature | Selenium WebDriver | Puppeteer |
|---|---|---|
| Primary Focus | Cross-browser automation | Chromium-based browsers |
| Language Support | Multiple languages | JavaScript only |
| Setup | Requires driver configuration | Zero setup; bundles compatible browser |
| Architecture | Request/Response | Event-driven (reduces flakiness) |
Use Puppeteer when you want deep integration with Chromium features, high-speed execution, and an event-driven architecture that avoids the need for manual sleep() calls.
Puppeteer treats itself and Chromium as an indivisible entity. Each version of Puppeteer bundles a specific version of Chromium that is guaranteed to work with it.
If you encounter compatibility issues, it is because Puppeteer is tied to a specific Chromium revision. You can find the specific Chrome version used by a Puppeteer release by checking the chrome entry in the revisions.ts file in the source repository.
A BrowserContext represents individual user contexts within a Browser.
Key behaviors:
localStorage.Browser has a single context by default when launched. You can create additional isolated contexts using Browser.createBrowserContext().Page opens another page (e.g., via window.open), the popup belongs to the parent page's BrowserContext.Puppeteer manages certain operations to prevent interference while a screenshot is being captured within a BrowserContext.
Automatic Waiting: The following methods will automatically wait for an ongoing screenshot to finish before executing:
BrowserContext.newPage()Browser.newPage()Page.close()No Automatic Waiting:
Page.bringToFront() does not wait for existing screenshot operations to complete.Running Puppeteer in a Chrome extension environment is experimental and differs significantly from a standard Node.js environment.
chrome.debugger API, which provides restricted access and allows attaching to only one page at a time.ExtensionTransport) instead of the standard Node.js transport.chrome.tabs API and then establish a new Puppeteer connection for that specific tab.Since version 18.2.0, the project is split into two packages:
puppeteer: The full package that includes both the API and the automatic downloading of a compatible browser (Chromium).puppeteer-core: A lightweight version that provides only the API and does not download any browsers. This is ideal for environments where you want to connect to an existing browser instance or manage downloads manually.21.6.0 introduced experimental support for the WebDriver BiDi protocol when using Firefox. This is an alternative to the CDP protocol and includes a BiDi implementation of Puppeteer.connect for Firefox.