The Dendrite class manages a browser instance using Playwright, allowing interactions with web pages using natural language. It handles initialization with API keys, manages browser contexts, and provides methods for navigation, authentication, and other browser-related tasks.

Initializing it starts a new browser instance.

​
Initialization

from dendrite import DendriteBrowser

browser = Dendrite()
# Perform actions in the browser
...

Arguments

  • auth (str | list[str], optional): Domains to authenticate on. Read more
  • playwright_options (Any, optional): Options for configuring Playwright. Defaults to running in non-headless mode with stealth arguments.

​
Properties

​
pages

Retrieves the list of active pages managed by the PageManager.

Usage

pages = browser.pages

Returns

List[Page]: The list of active pages in the browser instance.

​
Methods

​
goto

Navigates the page to the specified URL, optionally in a new tab.

Usage

page = browser.goto("https://example.com", expected_page="Feed of example articles")

Arguments

  • url (str): The URL to navigate to.
  • new_page (bool, optional): Whether to open the URL in a new page. Defaults to False.
  • timeout (float, optional): The maximum time (in milliseconds) to wait for the page to load. Defaults to 15000.
  • expected_page (str, optional): A description of the expected page type for verification at runtime. Defaults to an empty string.

Returns

  • Page: The page object after navigation.

This method exists on the Page and Dendrite class. If invoked from a Dendrite instance, it is performed on the active page

​
new_tab

Opens a new tab and navigates to the specified URL.

Usage

page = browser.new_tab("https://example.com", expected_page="Feed of example articles")

Arguments

  • url (str): The URL to navigate to.
  • timeout (float, optional): The maximum time (in milliseconds) to wait for the page to load. Defaults to 15000.
  • expected_page (str, optional): A description of the expected page type for verification at runtime. Defaults to an empty string. Returns

InteractionResponse: The response from the interaction.

Raises

DendriteException: If no suitable element is found or if the fill operation fails.

This method exists on the Page and Dendrite class. If invoked from a Dendrite instance, it is performed on the active page

​
click

Clicks an element on the active page based on the provided prompt.

Usage

browser.click("Search button in navbar")

Arguments

  • prompt (str): The prompt describing the element to be clicked.
  • expected_outcome (str, optional): The expected outcome of the click action.
  • use_cache (bool, optional): Whether to use cached results for element retrieval. Defaults to True.
  • timeout (int, optional): The timeout (in milliseconds) for the click operation. Defaults to 15000.
  • force (bool, optional): Whether to force the click operation. Defaults to False.

Returns

InteractionResponse

This method exists on the Page and Dendrite class. If invoked from a Dendrite instance, it is performed on the active page

​
fill

Fills an element on the active page with the provided value based on the given prompt.

This method combines the functionality of get_element and element.fill, allowing for a more concise way to interact with elements on the page.

Usage

query = "AI agent builders"
browser.fill("Search input", value=query)

Arguments

  • prompt (str): The prompt describing the element to be filled.
  • value (str): The value to fill the element with.
  • expected_outcome (str, optional): The expected outcome of the fill action.
  • use_cache (bool, optional): Whether to use cached results for element retrieval. Defaults to True.
  • max_retries (int, optional): The maximum number of retry attempts for element retrieval. Defaults to 3.
  • timeout (int, optional): The timeout (in milliseconds) for the fill operation. Defaults to 15000.

Returns

InteractionResponse: The response from the interaction.

Raises

DendriteException: If no suitable element is found or if the fill operation fails.

This method exists on the Page and Dendrite class. If invoked from a Dendrite instance, it is performed on the active page

​
fill_fields

Fills multiple fields on the active page with the provided values.

This method iterates through the given dictionary of fields and their corresponding values, making a separate fill request for each key-value pair.

Usage

browser.fill_fields({
  "First name": "John",
  "Last name:": "Doe",
  "Age": 36,
  "Occupation": "Firefighter"
})

Paramaters

fields (Dict[str, Any]): A dictionary where each key is a field identifier (e.g., a prompt or selector) and each value is the content to fill in that field.

Returns

None

This method exists on the Page and Dendrite class. If invoked from a Dendrite instance, it is performed on the active page

​
extract

Extract data from the active page based on a prompt and optional type specification.

Usage


class Article(BaseModel):
  title: str
  content: str
  author: str
  published_at: Optional[datetime] = None
  views: Optional[int] = 0

article = browser.extract("The article information", type_spec=Article)
print(article.title)

Arguments

  • prompt (str, optional): The prompt to guide the extraction.
  • type_spec (TypeSpec, optional): The type specification for the extracted data.
  • use_cache (bool, optional): Whether to use cached results. Defaults to True.

Returns

TypeSpec | Any: The response data as the specified type. Defaults to Any.

This method exists on the Page and Dendrite class. If invoked from a Dendrite instance, it is performed on the active page

​
ask

Asks a question about the active page and processes the response based on the specified type.

Usage


class Conversation(BaseModel):
  username: str
  unread_messages: int

signed_in = browser.ask("Is this a signed in dashboard?", type_spec=bool)

if signed_in:
  conversation = browser.ask(
      "Which conversation has most unread emails", type_spec=Conversation
  )
  print(conversation.username)

Arguments

  • prompt (str): The question or prompt to be asked.
  • type_spec (TypeSpec, optional): The expected return type, which can be a type or a schema. Defaults to None.

Returns

TypeSpec | Any: The response data as the specified type. Defaults to Any.

Raises

DendriteException: If the request fails, the exception includes the failure message and a screenshot.

This method exists on the Page and Dendrite class. If invoked from a Dendrite instance, it is performed on the active page

​
wait_for

Waits for the condition specified in the prompt to become true by periodically checking the page content.

This method retrieves the page information and evaluates whether the specified condition (provided in the prompt) is met. It continues to retry until the total elapsed time exceeds the timeout.

Usage

try:
  browser.wait_for("Loading to complete", timeout=10000)
  # continue with page interactions
except PageConditionNotMet as e:
  print(f"Condition was not met before timeout: {e.message}")

Arguments

  • prompt (str): The prompt to determine the condition to wait for on the page.
  • timeout (float, optional): The maximum time (in milliseconds) to wait for the condition. Defaults to 30000.

Returns

True if condition is met before timeout.

Raises

PageConditionNotMet if condition is not met before timeout

This method exists on the Page and Dendrite class. If invoked from a Dendrite instance, it is performed on the active page

​
press

Presses a keyboard key on the active page, optionally with modifier keys.

Usage

browser.fill("Search field", value="AI agents")
browser.press("Enter")

Arguments

  • key (str): The main key to be pressed.
  • hold_shift (bool, optional): Whether to hold the Shift key. Defaults to False.
  • hold_ctrl (bool, optional): Whether to hold the Control key. Defaults to False.
  • hold_alt (bool, optional): Whether to hold the Alt key. Defaults to False.
  • hold_cmd (bool, optional): Whether to hold the Command key (Meta on some systems). Defaults to False.

Returns

None

Raises

DendriteException if key press fails

This method exists on the Page and Dendrite class. If invoked from a Dendrite instance, it is performed on the active page

​
upload_files

Uploads files to the page using the file chooser.

Usage

browser.click("Upload file button")
browser.upload_files("path/to/file.png")

Arguments

  • files (Union[str, pathlib.Path, FilePayload, Sequence[Union[str, pathlib.Path]], Sequence[FilePayload]]): The file(s) to be uploaded. This can be a file path, a FilePayload object, or a sequence of file paths or FilePayload objects.
  • timeout (float, optional): The maximum amount of time (in milliseconds) to wait for the file chooser to be ready. Defaults to 30000.

Returns

  • None

This method exists on the Page and Dendrite class. If invoked from a Dendrite instance, it is performed on the active page

​
get_download

Retrieves the most recent reference to a downloaded file. To actually save the file to the local machine, you need to invoke the download.save_as method on the object returned from get_download.

Usage

# Download the file and get a reference to it
browser.click("Download File")
download = browser.get_download(timeout=5000)

# Save the downloaded file to the suggested location
download_path = "path/to/file/" + download.suggested_filename
download.save_as(download_path)

Arguments

  • timeout (float, optional): The maximum amount of time (in milliseconds) to wait for the download to complete. Defaults to 30000.

Returns

  • Download

This method exists on the Page and Dendrite class. If invoked from a Dendrite instance, it is performed on the active page

​
get_active_page

Retrieves the currently active page managed by the PageManager.

Usage

active_page = browser.get_active_page()

Returns

  • Page: The active page object.

​
new_page

Opens a new page in the browser.

Usage

new_page = browser.new_page()

Returns

  • Page: The newly opened page.

​
add_cookies

Adds cookies to the current browser context.

Usage

cookies = [
    {"name": "session_id", "value": "123456", "domain": "example.com"}
] browser.add_cookies(cookies)

Arguments

  • cookies (List[Dict[str, Any]]): A list of cookies to be added to the browser context.

​
close

Closes the browser and uploads authentication session data if available.

Usage

browser.close()

This ensures that the browser is properly closed when you’re done using it.

LangChain AgentPage