wuying-agentbay-sdk 0.1.0 → 0.4.0

package/dist/src/ui/index.js

@@ -1,17 +0,0 @@
-"use strict";
-var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
-    if (k2 === undefined) k2 = k;
-    var desc = Object.getOwnPropertyDescriptor(m, k);
-    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
-      desc = { enumerable: true, get: function() { return m[k]; } };
-    }
-    Object.defineProperty(o, k2, desc);
-}) : (function(o, m, k, k2) {
-    if (k2 === undefined) k2 = k;
-    o[k2] = m[k];
-}));
-var __exportStar = (this && this.__exportStar) || function(m, exports) {
-    for (var p in m) if (p !== "default" && !Object.prototype.hasOwnProperty.call(exports, p)) __createBinding(exports, m, p);
-};
-Object.defineProperty(exports, "__esModule", { value: true });
-__exportStar(require("./ui"), exports);

package/dist/src/ui/ui.d.ts

@@ -1,95 +0,0 @@
-import { Session } from '../session';
-/**
- * KeyCode constants for mobile device input
- */
-export declare const KeyCode: {
-    HOME: number;
-    BACK: number;
-    VOLUME_UP: number;
-    VOLUME_DOWN: number;
-    POWER: number;
-    MENU: number;
-};
-/**
- * UI handles UI operations in the AgentBay cloud environment.
- */
-export declare class UI {
-    private session;
-    /**
-     * Initialize a UI object.
-     *
-     * @param session - The Session instance that this UI belongs to.
-     */
-    constructor(session: Session);
-    /**
-     * Helper method to call MCP tools and handle common response processing
-     *
-     * @param toolName - Name of the MCP tool to call
-     * @param args - Arguments to pass to the tool
-     * @param defaultErrorMsg - Default error message if specific error details are not available
-     * @returns A CallMcpToolResult with the response data
-     * @throws APIError if the call fails
-     */
-    private callMcpTool;
-    /**
-     * Retrieves all clickable UI elements within the specified timeout.
-     *
-     * @param timeoutMs - The timeout in milliseconds. Default is 2000ms.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    getClickableUIElements(timeoutMs?: number): Promise<any>;
-    /**
-     * Retrieves all UI elements within the specified timeout.
-     *
-     * @param timeoutMs - The timeout in milliseconds. Default is 2000ms.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    getAllUIElements(timeoutMs?: number): Promise<any>;
-    /**
-     * Sends a key press event.
-     *
-     * @param key - The key code to send.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    sendKey(key: number): Promise<any>;
-    /**
-     * Inputs text into the active field.
-     *
-     * @param text - The text to input.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    inputText(text: string): Promise<any>;
-    /**
-     * Performs a swipe gesture on the screen.
-     *
-     * @param startX - The starting X coordinate.
-     * @param startY - The starting Y coordinate.
-     * @param endX - The ending X coordinate.
-     * @param endY - The ending Y coordinate.
-     * @param durationMs - The duration of the swipe in milliseconds. Default is 300ms.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    swipe(startX: number, startY: number, endX: number, endY: number, durationMs?: number): Promise<any>;
-    /**
-     * Clicks on the screen at the specified coordinates.
-     *
-     * @param x - The X coordinate.
-     * @param y - The Y coordinate.
-     * @param button - The mouse button to use. Default is 'left'.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    click(x: number, y: number, button?: string): Promise<any>;
-    /**
-     * Takes a screenshot of the current screen.
-     *
-     * @returns The content field from the API response containing the base64 encoded screenshot
-     * @throws Error if the operation fails.
-     */
-    screenshot(): Promise<any>;
-}

package/dist/src/ui/ui.js

@@ -1,206 +0,0 @@
-"use strict";
-Object.defineProperty(exports, "__esModule", { value: true });
-exports.UI = exports.KeyCode = void 0;
-const exceptions_1 = require("../exceptions");
-const model_1 = require("../api/models/model");
-const logger_1 = require("../utils/logger");
-/**
- * KeyCode constants for mobile device input
- */
-exports.KeyCode = {
-    HOME: 3,
-    BACK: 4,
-    VOLUME_UP: 24,
-    VOLUME_DOWN: 25,
-    POWER: 26,
-    MENU: 82
-};
-/**
- * UI handles UI operations in the AgentBay cloud environment.
- */
-class UI {
-    /**
-     * Initialize a UI object.
-     *
-     * @param session - The Session instance that this UI belongs to.
-     */
-    constructor(session) {
-        this.session = session;
-    }
-    /**
-     * Helper method to call MCP tools and handle common response processing
-     *
-     * @param toolName - Name of the MCP tool to call
-     * @param args - Arguments to pass to the tool
-     * @param defaultErrorMsg - Default error message if specific error details are not available
-     * @returns A CallMcpToolResult with the response data
-     * @throws APIError if the call fails
-     */
-    async callMcpTool(toolName, args, defaultErrorMsg) {
-        try {
-            const argsJSON = JSON.stringify(args);
-            const callToolRequest = new model_1.CallMcpToolRequest({
-                authorization: `Bearer ${this.session.getAPIKey()}`,
-                sessionId: this.session.getSessionId(),
-                name: toolName,
-                args: argsJSON
-            });
-            // Log API request
-            (0, logger_1.log)(`API Call: CallMcpTool - ${toolName}`);
-            (0, logger_1.log)(`Request: SessionId=${this.session.getSessionId()}, Args=${argsJSON}`);
-            const response = await this.session.getClient().callMcpTool(callToolRequest);
-            // Log API response
-            (0, logger_1.log)(`Response from CallMcpTool - ${toolName}:`, response.body);
-            if (!response.body?.data) {
-                throw new Error('Invalid response data format');
-            }
-            // Extract data from response
-            const data = response.body.data;
-            // Create result object
-            const result = {
-                data,
-                statusCode: response.statusCode || 0,
-                isError: false
-            };
-            // Check if there's an error in the response
-            if (data.isError === true) {
-                result.isError = true;
-                // Try to extract the error message from the content field
-                const contentArray = data.content;
-                if (contentArray && contentArray.length > 0) {
-                    result.content = contentArray;
-                    // Extract error message from the first content item
-                    if (contentArray[0]?.text) {
-                        result.errorMsg = contentArray[0].text;
-                        throw new Error(contentArray[0].text);
-                    }
-                }
-                throw new Error(defaultErrorMsg);
-            }
-            // Extract content array if it exists
-            if (Array.isArray(data.content)) {
-                result.content = data.content;
-            }
-            return result;
-        }
-        catch (error) {
-            (0, logger_1.logError)(`Error calling CallMcpTool - ${toolName}:`, error);
-            throw new exceptions_1.APIError(`Failed to call ${toolName}: ${error}`);
-        }
-    }
-    /**
-     * Retrieves all clickable UI elements within the specified timeout.
-     *
-     * @param timeoutMs - The timeout in milliseconds. Default is 2000ms.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    async getClickableUIElements(timeoutMs = 2000) {
-        const args = {
-            timeout_ms: timeoutMs
-        };
-        const result = await this.callMcpTool('get_clickable_ui_elements', args, 'Failed to get clickable UI elements');
-        // Return the raw content field for the caller to parse
-        return result.data.content;
-    }
-    /**
-     * Retrieves all UI elements within the specified timeout.
-     *
-     * @param timeoutMs - The timeout in milliseconds. Default is 2000ms.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    async getAllUIElements(timeoutMs = 2000) {
-        const args = {
-            timeout_ms: timeoutMs
-        };
-        const result = await this.callMcpTool('get_all_ui_elements', args, 'Failed to get all UI elements');
-        // Return the raw content field for the caller to parse
-        return result.data.content;
-    }
-    /**
-     * Sends a key press event.
-     *
-     * @param key - The key code to send.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    async sendKey(key) {
-        const args = {
-            key
-        };
-        const result = await this.callMcpTool('send_key', args, 'Failed to send key');
-        // Return the raw content field for the caller to parse
-        return result.data.content;
-    }
-    /**
-     * Inputs text into the active field.
-     *
-     * @param text - The text to input.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    async inputText(text) {
-        const args = {
-            text
-        };
-        const result = await this.callMcpTool('input_text', args, 'Failed to input text');
-        // Return the raw content field for the caller to parse
-        return result.data.content;
-    }
-    /**
-     * Performs a swipe gesture on the screen.
-     *
-     * @param startX - The starting X coordinate.
-     * @param startY - The starting Y coordinate.
-     * @param endX - The ending X coordinate.
-     * @param endY - The ending Y coordinate.
-     * @param durationMs - The duration of the swipe in milliseconds. Default is 300ms.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    async swipe(startX, startY, endX, endY, durationMs = 300) {
-        const args = {
-            start_x: startX,
-            start_y: startY,
-            end_x: endX,
-            end_y: endY,
-            duration_ms: durationMs
-        };
-        const result = await this.callMcpTool('swipe', args, 'Failed to perform swipe');
-        // Return the raw content field for the caller to parse
-        return result.data.content;
-    }
-    /**
-     * Clicks on the screen at the specified coordinates.
-     *
-     * @param x - The X coordinate.
-     * @param y - The Y coordinate.
-     * @param button - The mouse button to use. Default is 'left'.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    async click(x, y, button = 'left') {
-        const args = {
-            x,
-            y,
-            button
-        };
-        const result = await this.callMcpTool('click', args, 'Failed to click');
-        // Return the raw content field for the caller to parse
-        return result.data.content;
-    }
-    /**
-     * Takes a screenshot of the current screen.
-     *
-     * @returns The content field from the API response containing the base64 encoded screenshot
-     * @throws Error if the operation fails.
-     */
-    async screenshot() {
-        const args = {};
-        const result = await this.callMcpTool('screenshot', args, 'Failed to take screenshot');
-        // Return the raw content field for the caller to parse
-        return result.data.content;
-    }
-}
-exports.UI = UI;

package/dist/src/utils/logger.d.ts

@@ -1,15 +0,0 @@
-/**
- * Utility functions for logging in a clean format
- */
-/**
- * Log a message without the log prefix and file location
- * @param message The message to log
- * @param args Optional arguments to log
- */
-export declare function log(message: string, ...args: any[]): void;
-/**
- * Log an error message
- * @param message The error message to log
- * @param error Optional error object
- */
-export declare function logError(message: string, error?: any): void;

package/dist/src/utils/logger.js

@@ -1,48 +0,0 @@
-"use strict";
-/**
- * Utility functions for logging in a clean format
- */
-Object.defineProperty(exports, "__esModule", { value: true });
-exports.log = log;
-exports.logError = logError;
-/**
- * Log a message without the log prefix and file location
- * @param message The message to log
- * @param args Optional arguments to log
- */
-function log(message, ...args) {
-    // Use process.stdout.write instead of log to avoid the prefix and newline
-    process.stdout.write(message + '\n');
-    // If there are additional arguments, print them on new lines with proper formatting
-    if (args.length > 0) {
-        for (const arg of args) {
-            if (typeof arg === 'object') {
-                // For objects, format them as JSON with indentation
-                process.stdout.write(JSON.stringify(arg, null, 2) + '\n');
-            }
-            else {
-                // For other types, just convert to string
-                process.stdout.write(String(arg) + '\n');
-            }
-        }
-    }
-}
-/**
- * Log an error message
- * @param message The error message to log
- * @param error Optional error object
- */
-function logError(message, error) {
-    process.stderr.write(`ERROR: ${message}\n`);
-    if (error) {
-        if (error instanceof Error) {
-            process.stderr.write(`${error.message}\n`);
-        }
-        else if (typeof error === 'object') {
-            process.stderr.write(JSON.stringify(error, null, 2) + '\n');
-        }
-        else {
-            process.stderr.write(String(error) + '\n');
-        }
-    }
-}

package/dist/src/window/index.d.ts

@@ -1 +0,0 @@
-export * from './window';

package/dist/src/window/index.js

@@ -1,17 +0,0 @@
-"use strict";
-var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
-    if (k2 === undefined) k2 = k;
-    var desc = Object.getOwnPropertyDescriptor(m, k);
-    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
-      desc = { enumerable: true, get: function() { return m[k]; } };
-    }
-    Object.defineProperty(o, k2, desc);
-}) : (function(o, m, k, k2) {
-    if (k2 === undefined) k2 = k;
-    o[k2] = m[k];
-}));
-var __exportStar = (this && this.__exportStar) || function(m, exports) {
-    for (var p in m) if (p !== "default" && !Object.prototype.hasOwnProperty.call(exports, p)) __createBinding(exports, m, p);
-};
-Object.defineProperty(exports, "__esModule", { value: true });
-__exportStar(require("./window"), exports);

package/dist/src/window/window.d.ts

@@ -1,109 +0,0 @@
-/**
- * Represents a window in the system.
- */
-export interface Window {
-    window_id: number;
-    title: string;
-    absolute_upper_left_x?: number;
-    absolute_upper_left_y?: number;
-    width?: number;
-    height?: number;
-    pid?: number;
-    pname?: string;
-    child_windows?: Window[];
-}
-/**
- * Handles window management operations in the AgentBay cloud environment.
- */
-export declare class WindowManager {
-    private session;
-    /**
-     * Creates a new WindowManager instance.
-     * @param session The session object that provides access to the AgentBay API.
-     */
-    constructor(session: {
-        getAPIKey(): string;
-        getClient(): any;
-        getSessionId(): string;
-    });
-    /**
-     * Helper method to call MCP tools and handle common response processing
-     *
-     * @param toolName - Name of the MCP tool to call
-     * @param args - Arguments to pass to the tool
-     * @param defaultErrorMsg - Default error message if specific error details are not available
-     * @returns A CallMcpToolResult with the response data
-     * @throws APIError if the call fails
-     */
-    private callMcpTool;
-    /**
-     * Lists all root windows in the system.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    listRootWindows(): Promise<any>;
-    /**
-     * Gets the currently active window.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    getActiveWindow(): Promise<any>;
-    /**
-     * Activates a window by ID.
-     * @param windowId The ID of the window to activate.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    activateWindow(windowId: number): Promise<any>;
-    /**
-     * Maximizes a window by ID.
-     * @param windowId The ID of the window to maximize.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    maximizeWindow(windowId: number): Promise<any>;
-    /**
-     * Minimizes a window by ID.
-     * @param windowId The ID of the window to minimize.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    minimizeWindow(windowId: number): Promise<any>;
-    /**
-     * Restores a window by ID.
-     * @param windowId The ID of the window to restore.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    restoreWindow(windowId: number): Promise<any>;
-    /**
-     * Closes a window by ID.
-     * @param windowId The ID of the window to close.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    closeWindow(windowId: number): Promise<any>;
-    /**
-     * Sets a window to fullscreen mode.
-     * @param windowId The ID of the window to set to fullscreen.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    fullscreenWindow(windowId: number): Promise<any>;
-    /**
-     * Resizes a window by ID.
-     * @param windowId The ID of the window to resize.
-     * @param width The new width of the window.
-     * @param height The new height of the window.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    resizeWindow(windowId: number, width: number, height: number): Promise<any>;
-    /**
-     * Enables or disables focus mode.
-     * @param on Whether to enable (true) or disable (false) focus mode.
-     * @returns The content field from the API response
-     * @throws Error if the operation fails.
-     */
-    focusMode(on: boolean): Promise<any>;
-}