@ui-tars/operator-browser/dist/browser-operator.d.ts

Native-browser operator for UI-TARS

import { Operator } from '@ui-tars/sdk/core';
import { BrowserType } from '@agent-infra/browser';
import type { ScreenshotOutput, ExecuteParams, ExecuteOutput } from '@ui-tars/sdk/core';
import { BrowserOperatorOptions, SearchEngine } from './types';
/**
 * BrowserOperator class that extends the base Operator
 * Provides functionality to control a browser instance for UI automation
 */
export declare class BrowserOperator extends Operator {
    private options;
    private browser;
    private currentPage;
    private logger;
    private uiHelper;
    private highlightClickableElements;
    private showActionInfo;
    private showWaterFlowEffect;
    private deviceScaleFactor?;
    /**
     * Creates a new BrowserOperator instance
     * @param options Configuration options for the browser operator
     */
    constructor(options: BrowserOperatorOptions);
    /**
     * Gets the currently active browser page
     * @returns Promise resolving to the active Page object
     * @throws Error if no active page is found
     */
    private getActivePage;
    setHighlightClickableElements(enable: boolean): void;
    /**
     * Sets whether to show the water flow effect during screenshots
     * @param enable Whether to enable the water flow effect
     */
    setShowWaterFlow(enable: boolean): void;
    /**
     * Takes a screenshot of the current browser viewport
     * @returns Promise resolving to screenshot data
     */
    screenshot(): Promise<ScreenshotOutput>;
    /**
     * Executes a specified action based on the parsed prediction
     * @param params Parameters containing action details
     * @returns Promise resolving to execution output
     */
    execute(params: ExecuteParams): Promise<ExecuteOutput>;
    private handleClick;
    private handleDoubleClick;
    private handleRightClick;
    private handleType;
    private handleHotkey;
    private handlePress;
    private handleRelease;
    private handleScroll;
    private delay;
    private handleNavigate;
    private handleDrag;
    private handleNavigateBack;
    /**
     * A helper function to wait for possible navigation to complete.
     * @param page
     */
    private waitForPossibleNavigation;
    private getDeviceScaleFactor;
    cleanup(): Promise<void>;
}
export declare class DefaultBrowserOperator extends BrowserOperator {
    private static instance;
    private static browser;
    private static browserPath;
    private static browserType;
    private static logger;
    private constructor();
    /**
     * Check whether the local environment has a browser available
     * @returns {boolean}
     */
    static hasBrowser(browser?: BrowserType): boolean;
    static getInstance(highlight?: boolean, showActionInfo?: boolean, showWaterFlow?: boolean, isCallUser?: boolean, searchEngine?: SearchEngine): Promise<DefaultBrowserOperator>;
    static destroyInstance(): Promise<void>;
}
export declare class RemoteBrowserOperator extends BrowserOperator {
    private static instance;
    private static browser;
    private static browserType;
    private static logger;
    private constructor();
    static getInstance(wsEndpoint: string, highlight?: boolean, showActionInfo?: boolean, showWaterFlow?: boolean, isCallUser?: boolean): Promise<DefaultBrowserOperator>;
    static destroyInstance(): Promise<void>;
}
//# sourceMappingURL=browser-operator.d.ts.map