Custom serialization that properly uses AgentHistory's model_dump

Get a description of the action for the prompt

Remove an agent from an automation flow

Simulates a right-click on a Mac UI element.

Asynchronously create an Agent, run it for a set number of steps, and return the string representation of the agent's final conversation his

Run the agent with the specified configuration

Save history to JSON file with proper serialization

Scrolls an element in the specified direction. direction can be: 'left', 'right', 'up', 'down'

Set up logging to capture terminal output

Set up logging to capture terminal output

Type text into element by index

Extend actions with custom actions

Wait for app to be ready with detailed status checking

Initialize SystemPrompt with action description, current date and max actions allowed per step. Args: action_des

Initialize AgentMessagePrompt with state and optional parameters. Args: state (str): Current system state

Representation of the AgentHistoryList object

Generate a unique path to this element including more identifiers

Decorator for registering custom actions @param description: Describe the LLM what the function does (better description == better function calling

Decorator for registering actions

Get all action names from history

Get the list of available actions for this element

Check if the element is enabled

Get all extracted content from history

Final result from history

Find an element using its accessibility path

Find all elements that support a specific action

Recursively build a detailed string representation of the UI tree

Get the index of the action

Get a description of all actions for the prompt

Check if the agent has any errors

Last action in history

Load history from JSON file

Get all model actions from history as JSON

Get all model outputs from history

Get all thoughts from history

Handle accessibility notifications

Execute a Mac accessibility action

Get the element's position

Uses the LLM to refine the user's prompt based on examples.

Reset the state between major steps

Synchronous wrapper for running the agent. Given a task description and optional parameters, run a limited number of steps, and return th

Wrapper for run_agent that adds share_terminal from preferences

Run an automation flow by executing its agents in sequence

Save the history to a file

Sends the prompt text and optional terminal output to a Google Form, which appends it to a linked Google Sheet.

Overwrite the index of the action

Get the element's size

Stop the running agent

Stream terminal output in real-time

Test adding browser state message

Test adding state with result that should be included in memory

Test adding state with result that should not be included in memory

Test that message manager initializes with system and task messages

Test handling of token overflow in a realistic message flow