Skip to Content
UsageBiDiSelenium

Selenium (BiDi)

Use Lightpanda with Selenium  over WebDriver BiDi . Lightpanda only implements the BiDi handshake of the classic WebDriver protocol, not full classic automation, so a session is created the normal Selenium way but driven with BiDi commands (browsingContext, script) instead of driver.get() or driver.findElement().

⚠️

BiDi support is new and covers less than CDP: only browsingContext and a subset of script are implemented.

Start Lightpanda with the webdriver protocol enabled (find all options in the serve command reference):

lightpanda serve --host 127.0.0.1 --port 9222 --protocol webdriver

Request a BiDi session by setting the webSocketUrl capability to true (browserName can be any value; Lightpanda reports itself back regardless), then drive it with BiDi commands:

npm install selenium-webdriver
'use strict'
 
import { Builder, Capabilities } from 'selenium-webdriver'
import { BrowsingContext } from 'selenium-webdriver/bidi/generated/browsing_context.js'
import { Script } from 'selenium-webdriver/bidi/generated/script.js'
 
const caps = new Capabilities()
caps.set('browserName', 'lightpanda')
caps.set('webSocketUrl', true)
 
const driver = await new Builder()
  .usingServer('http://127.0.0.1:9222')
  .withCapabilities(caps)
  .build()
 
const bidi = await driver.getBidi()
await bidi.waitForConnection()
 
const browsingContext = new BrowsingContext(bidi)
const script = new Script(bidi)
 
const { context } = await browsingContext.create({ type: 'tab' })
await browsingContext.navigate({ context, url: 'https://wikipedia.com/', wait: 'complete' })
 
const title = await script.evaluate({
  expression: 'document.title',
  target: { context },
  awaitPromise: false,
})
console.log(title.result.value)
 
await browsingContext.close({ context })
await driver.quit()
PuppeteerAgent