Sets up and returns a Playwright Browser instance with anti-detection measures.

Sets up and returns a Playwright Browser instance with anti-detection measures.

Sets up and returns a Playwright Browser instance with anti-detection measures. Firefox has no longer CDP support.

Sets up and returns a Playwright Browser instance with anti-detection measures.

Sets up and returns a Playwright Browser instance with anti-detection measures.

Update action indices based on current page state. Returns updated action or None if element cannot be found.

Update action models with page-specific actions

Update and return state.

Verify that the LLM API keys are working properly by sending a simple test prompt and checking that the response contains the expected answer.

Write messages to conversation file

Write model response to conversation file

Get all results from history

Comprehensively adds a new logging level to the `logging` module and the currently configured logging class. `levelName` becomes an attribute of

Analyze and print the structure of a webpage with enhanced debugging

Send *prompt* to OpenAI and return the assistant text. Parameters ---------- prompt : str User prompt / question. system_prom

* Creates a node data object for a given node and its descendants.

Check if all required environment variables are set

Return the correct template file for a given action_type.

Cleanup all httpx clients

Clean up webdriver cache directories.

* Recursively clone a node, but only include nodes that are visible in the viewport. * For element nodes, if the node itself is not visible (per isIn

Count the number of tokens in a string using a specified model.

Create a GIF from the agent's history with overlaid task and goal text.

Creates an MCP server file with the specified parameters. Args: ultimate_goal: The goal description for the MCP function

Create a new tab and optionally navigate to a URL

Create a procedural memory if needed based on the current step. Args: current_step: The current step number of the agent

Get current message list, potentially trimmed to max tokens

Ask the user (once) for host permission to the current origin and, * when granted, inject the content-script.

Execute a registered action

Execute a tool with retry mechanism. Args: tool_name: Name of the tool to execute. arguments: Tool arguments.

Split the incoming JSON dict into: action {type, eventTarget} change_events list from `allEvents` page_content

Try to extract ultimate_goal from either exact_replay or smart_replay file. Args: task_id: The task ID to look for in the replay

Walk sub-directories recursively, load every *.json file, convert each to replay actions via `record_metadata_to_actions`, and return the

Format error message based on error type and optionally include trace

Format tool information for LLM. Returns: A formatted string describing the tool.

Load all .md files under *dir_path*, query the LLM, and return results. Parameters ---------- dir_path : str | Path Directory con

* Gets the HTML string for the current page but only for visible nodes. * * @param {number} viewportExpansion - How much extra area to consider as v

* Returns an XPath tree string for an element.

Generator that creates fresh model instances each time

Locates an element on the page using the provided text. If `nth` is provided, it returns the nth matching element (0-based). If `element_type`

Locates an element on the page using the provided XPath.

Get a debug view of the page structure including iframes

Get a browser context

Get scroll position information for the current page.

Get information about all tabs

Get information about all open tabs. Returns: list: A list of TabInfo objects containing details about each tab.

Returns recommended x, y offsets for window positioning

Navigate back in history

Initialize the server connection.

* Checks if an element is visible.

* Checks if an element is within the expanded viewport.

* Checks if an element is interactive.

* Checks if a text node is visible.

* Checks if an element is the topmost element at its position.

Check if element or its children are file uploaders

Load server configuration from JSON file. Args: file_path: Path to the JSON configuration file. Returns: Dic

Load environment variables from .env file.

Load history from JSON file

Utility function to log the model's response.

Initialize and run the chat session.

Get all model actions from history as JSON

Custom serialization that properly uses AgentHistory's model_dump

Process the LLM response and execute tools if needed. Args: llm_response: The response from the LLM. Returns:

Process a single task asynchronously.