GetHistory returns the agent's execution history.

GetSystemPrompt returns the system prompt.

GetURL returns the current page URL.

GoBack navigates back in history.

GoForward navigates forward in history.

Hover moves the mouse to hover over an element.

IsEmpty returns true if the bounding box has no area.

NewAgentHistory creates a new agent history manager.

NewBrowserAgent creates a new browser agent using ADK.

NewBrowserToolkit creates a new browser toolkit.

NewElementMap creates a new empty element map.

NewMessageManager creates a new message manager.

NewSensitiveDataFilter creates a new sensitive data filter.

Reload reloads the current page.

ScreenshotAfterActionWithAnnotations captures an annotated screenshot after an action.

Scroll scrolls the page or an element.

ScrollToElement scrolls an element into view.

SendKeys sends keyboard keys to the page.

SetTask sets the task description for context.

SwitchTab switches to a different tab by ID.

SystemPrompt returns the system prompt for the browser agent. Ported and adapted from browser-use Python project.

ToDescription generates a text description of the history for the LLM. It implements truncation: keeps first item + most recent items within maxItems.

ToTokenString serializes the element map for LLM consumption. Uses a compact format to minimize token usage.

TypeText types text into an element by index.

UpdateLastItem updates the result and success status of the last history item.

applyDefaults fills in default values for the config.

captureScreenshotAfterAction captures a screenshot after an action has completed. Uses enhanced waiting for page stability after the action.

drawBoundingBoxFromInfo draws a rectangle border around the bounding box using the interface.

drawDigit draws a single digit using a simple 5x7 pixel pattern.

drawIndexLabelFromInfo draws the element index at the top center of the bounding box using the interface. Uses a simple built-in digit renderer for mi

formatElement formats a single element as a compact string.

getElementColorFromInfo returns the appropriate color for an element type using the interface.

isImplicitRole returns true if the role is implied by the tag.

jsonMarshal wraps json.Marshal.

truncate truncates a string to maxLen characters.

validate checks that required configuration is provided.

waitForPageReady waits for the page to be ready for screenshot.

AddPattern adds a custom sensitive data pattern.

AnnotateBrowserUseStyle annotates following browser-use pattern: Only show index labels for elements without text information.

AnnotateForLLM annotates a screenshot optimized for LLM vision consumption. Uses default annotation config with appropriate sizing.

Available returns tokens available after subtracting used tokens.

BuildContinuationPrompt creates a prompt for continuing after an action.

BuildErrorRecoveryMessage builds a message for recovering from an error.

CaptureAfterAction captures a screenshot after an action has been performed. It waits for the page to stabilize after the action before capturing.

CaptureAfterActionWithAnnotations captures an annotated screenshot after an action.

CaptureElement takes a screenshot of a specific element.

CaptureFullPage takes a screenshot of the entire page.

CaptureViewport takes a screenshot of the current viewport only.

Clear removes all elements.

Clear resets the message manager state.

ClickAt clicks at specific coordinates.

Close cleans up the agent resources.

CloseTab closes a tab by ID.

DefaultConfig returns a default browser configuration.

DefaultOptions returns sensible defaults for LLM consumption.

EstimateFromElements estimates tokens for element descriptions. Elements have a predictable format so we can be more precise.

ExtractElementMap is a convenience function for extracting elements.

FindBySelector returns the first element matching the selector.

FindByText returns elements containing the given text.

ForLLM captures a screenshot optimized for LLM consumption. Uses JPEG format with reasonable compression for token efficiency. Includes page readiness

ForLLMSafe captures a screenshot for LLM consumption with full validation. Returns nil data (not error) if page is blank or screenshot is empty. This

ForLLMSafeWithAnnotations captures an annotated screenshot, returning nil for blank pages.

ForLLMWithAnnotations captures an annotated screenshot optimized for LLM vision. Includes bounding boxes around interactive elements with index labels

GetBoundingBox implements ElementInfo interface for screenshot annotations. Returns a BoundingBoxInfoInterface that wraps the BoundingBox.

GetCurrentMemory returns the accumulated memory from history.

GetElements returns elements as screenshot.ElementInfo slice.

GetElements returns all elements as ElementInfoInterface for screenshot annotations. This implements ElementMapInterface from the screenshot package.

GetHeight implements BoundingBoxInfo interface.

GetHistory returns the agent history.

GetIndex implements ElementInfo interface for screenshot annotations.

GetIsEmpty implements BoundingBoxInfo interface.

GetIsVisible implements ElementInfo interface for screenshot annotations.

GetItems returns all history items.

GetLastItem returns the most recent history item, or nil if empty.