Create Browserbase integration

[adeprez]

We'd like to have an easy integration with browserbase. A code example is provided here: https://docs.lavague.ai/en/latest/docs/examples/medical_appointment_booking/

[paulpalmieri]

A few questions:

• Do we want this to work with the playwright driver ?
• We could make it a argument to the driver like so (provided the user has set an api key. What do you think ?

  selenium_driver = SeleniumDriver(browserbase=True)

[adeprez]

It can be done with playwright too: https://docs.browserbase.com/quickstart/playwright

We can accept a new remote_connection argument of type selenium.webdriver.remote.remote_connection.RemoteConnection, then provide and document an implementation for BrowserbaseRemoteConnection.

BrowserbaseRemoteConnection would accept api_key as parameter or default to env BROWSERBASE_API_KEY

[paulpalmieri]

@adeprez What's the purpose of the following code in the driver init:

self.driver.execute_cdp_cmd(
    "Page.addScriptToEvaluateOnNewDocument",
    {"source": JS_SETUP_GET_EVENTS},
)

Seems like the webdriver.Remote doesn't support this and browserbase can't start if it's run on the driver.

[paulpalmieri]

@adeprez added a basic implementation for SeleniumDriver on 538-browserbase-integration

As I understand we should reuse the following Browserbase code for Playwright

def create_session():
    url = 'https://www.browserbase.com/v1/sessions'
    headers = {'Content-Type': 'application/json', 'x-bb-api-key': os.environ["BROWSERBASE_API_KEY"]}
    response = requests.post(url, json={ "projectId": os.environ["BROWSERBASE_PROJECT_ID"] }, headers=headers)
    # print(response.json())
    return response.json()['id']

class CustomRemoteConnection(RemoteConnection):
    _session_id = None

    def __init__(self, remote_server_addr: str, session_id: str):
        super().__init__(remote_server_addr)
        self._session_id = session_id

    def get_remote_connection_headers(self, parsed_url, keep_alive=False):
        headers = super().get_remote_connection_headers(parsed_url, keep_alive)
        headers.update({'x-bb-api-key': os.environ["BROWSERBASE_API_KEY"]})
        headers.update({'session-id': self._session_id})
        return headers

Where would you put it so that it's usable by both drivers ?

Also not sure what we want to accept from the user if not a flag. Do we want to create a wrapper BrowserbaseRemoteConnection that accepts keys ? And if we pass it to a driver, it uses the common logic above to instantiate a driver ?

[adeprez]

@adeprez What's the purpose of the following code in the driver init:

self.driver.execute_cdp_cmd(
    "Page.addScriptToEvaluateOnNewDocument",
    {"source": JS_SETUP_GET_EVENTS},
)

Seems like the webdriver.Remote doesn't support this and browserbase can't start if it's run on the driver.

It is some JavaScript code injected before page loads, we use it to track events listeners added to elements. It's needed to determine which elements are interactive

@adeprez added a basic implementation for SeleniumDriver on 538-browserbase-integration

As I understand we should reuse the following Browserbase code for Playwright

def create_session():
    url = 'https://www.browserbase.com/v1/sessions'
    headers = {'Content-Type': 'application/json', 'x-bb-api-key': os.environ["BROWSERBASE_API_KEY"]}
    response = requests.post(url, json={ "projectId": os.environ["BROWSERBASE_PROJECT_ID"] }, headers=headers)
    # print(response.json())
    return response.json()['id']

class CustomRemoteConnection(RemoteConnection):
    _session_id = None

    def __init__(self, remote_server_addr: str, session_id: str):
        super().__init__(remote_server_addr)
        self._session_id = session_id

    def get_remote_connection_headers(self, parsed_url, keep_alive=False):
        headers = super().get_remote_connection_headers(parsed_url, keep_alive)
        headers.update({'x-bb-api-key': os.environ["BROWSERBASE_API_KEY"]})
        headers.update({'session-id': self._session_id})
        return headers

Where would you put it so that it's usable by both drivers ?

Also not sure what we want to accept from the user if not a flag. Do we want to create a wrapper BrowserbaseRemoteConnection that accepts keys ? And if we pass it to a driver, it uses the common logic above to instantiate a driver ?

I think we could accept any RemoteConnection that matches the requirements from selenium. And our BrowserbaseRemoteConnection implementation could fit for both Playwright and Selenium use cases.

The class itself can be in core, either in existing base_driver.py or in a new file.

[paulpalmieri]

• in Selenium base.py: rename CustomRemoteConnection to BrowserBaseCustomConnection
• add api key to BrowserBaseCustomConnection
• remove bool in SeleniumDriver, accept remote_connection: BrowserBaseCustomConnection
• skip exec_cdp for now if not handled by Browserbase.

connection = BrowserBaseCustomConnection(addr, id, KEY: opt)
driver = SeleniumDriver(remote_connection=connection)

For playwright: Browserbase gives a browser with an already setup context, make sure to rearange the logic in base.py

driver = PlaywrightDriver(remote_url)