@midscene/computer 1.3.12-beta-20260212063915.0 → 1.4.0

package/bin/midscene-computer

@@ -0,0 +1,2 @@
+#!/usr/bin/env node
+require('../dist/lib/cli.js');

package/dist/es/cli.mjs

@@ -0,0 +1,626 @@
+import { CLIError, runToolsCLI } from "@midscene/shared/cli";
+import { getMidsceneLocationSchema, z } from "@midscene/core";
+import { getDebug } from "@midscene/shared/logger";
+import { BaseMidsceneTools } from "@midscene/shared/mcp";
+import { Agent } from "@midscene/core/agent";
+import node_assert from "node:assert";
+import { execSync } from "node:child_process";
+import { createRequire } from "node:module";
+import { actionHoverParamSchema, defineAction, defineActionClearInput, defineActionDoubleClick, defineActionDragAndDrop, defineActionKeyboardPress, defineActionRightClick, defineActionScroll, defineActionTap } from "@midscene/core/device";
+import { sleep } from "@midscene/core/utils";
+import { createImgBase64ByFormat } from "@midscene/shared/img";
+import screenshot_desktop from "screenshot-desktop";
+function _define_property(obj, key, value) {
+    if (key in obj) Object.defineProperty(obj, key, {
+        value: value,
+        enumerable: true,
+        configurable: true,
+        writable: true
+    });
+    else obj[key] = value;
+    return obj;
+}
+const SMOOTH_MOVE_STEPS_TAP = 8;
+const SMOOTH_MOVE_STEPS_MOUSE_MOVE = 10;
+const SMOOTH_MOVE_DELAY_TAP = 8;
+const SMOOTH_MOVE_DELAY_MOUSE_MOVE = 10;
+const MOUSE_MOVE_EFFECT_WAIT = 300;
+const CLICK_HOLD_DURATION = 50;
+const INPUT_FOCUS_DELAY = 300;
+const INPUT_CLEAR_DELAY = 150;
+const SCROLL_REPEAT_COUNT = 10;
+const SCROLL_STEP_DELAY = 100;
+const SCROLL_COMPLETE_DELAY = 500;
+const INPUT_STRATEGY_ALWAYS_CLIPBOARD = 'always-clipboard';
+const INPUT_STRATEGY_CLIPBOARD_FOR_NON_ASCII = 'clipboard-for-non-ascii';
+const APPLESCRIPT_KEY_CODE_MAP = {
+    return: 36,
+    enter: 36,
+    tab: 48,
+    space: 49,
+    backspace: 51,
+    delete: 51,
+    escape: 53,
+    forwarddelete: 117,
+    left: 123,
+    right: 124,
+    down: 125,
+    up: 126,
+    home: 115,
+    end: 119,
+    pageup: 116,
+    pagedown: 121,
+    f1: 122,
+    f2: 120,
+    f3: 99,
+    f4: 118,
+    f5: 96,
+    f6: 97,
+    f7: 98,
+    f8: 100,
+    f9: 101,
+    f10: 109,
+    f11: 103,
+    f12: 111
+};
+const APPLESCRIPT_MODIFIER_MAP = {
+    command: 'command down',
+    cmd: 'command down',
+    control: 'control down',
+    ctrl: 'control down',
+    shift: 'shift down',
+    alt: 'option down',
+    option: 'option down',
+    meta: 'command down'
+};
+function typeStringViaAppleScript(text) {
+    const escaped = text.replace(/\\/g, '\\\\').replace(/"/g, '\\"');
+    const script = `tell application "System Events" to keystroke "${escaped}"`;
+    debugDevice('typeStringViaAppleScript', {
+        text
+    });
+    execSync(`osascript -e '${script}'`);
+}
+function sendKeyViaAppleScript(key, modifiers = []) {
+    const lowerKey = key.toLowerCase();
+    const keyCode = APPLESCRIPT_KEY_CODE_MAP[lowerKey];
+    const modifierParts = modifiers.map((m)=>APPLESCRIPT_MODIFIER_MAP[m.toLowerCase()]).filter(Boolean);
+    const modifierStr = modifierParts.length > 0 ? ` using {${modifierParts.join(', ')}}` : '';
+    let script;
+    if (void 0 !== keyCode) script = `tell application "System Events" to key code ${keyCode}${modifierStr}`;
+    else lowerKey.length, script = `tell application "System Events" to keystroke "${key}"${modifierStr}`;
+    debugDevice('sendKeyViaAppleScript', {
+        key,
+        modifiers,
+        script
+    });
+    execSync(`osascript -e '${script}'`);
+}
+let libnut = null;
+let libnutLoadError = null;
+async function getLibnut() {
+    if (libnut) return libnut;
+    if (libnutLoadError) throw libnutLoadError;
+    try {
+        const require = createRequire(import.meta.url);
+        const libnutModule = require('@computer-use/libnut/dist/import_libnut');
+        libnut = libnutModule.libnut;
+        if (!libnut) throw new Error('libnut module loaded but libnut object is undefined');
+        return libnut;
+    } catch (error) {
+        libnutLoadError = error;
+        throw new Error(`Failed to load @computer-use/libnut. Make sure it is properly installed and compiled for your platform. Error: ${error}`);
+    }
+}
+const debugDevice = getDebug('computer:device');
+async function smoothMoveMouse(targetX, targetY, steps, stepDelay) {
+    node_assert(libnut, 'libnut not initialized');
+    const currentPos = libnut.getMousePos();
+    for(let i = 1; i <= steps; i++){
+        const stepX = Math.round(currentPos.x + (targetX - currentPos.x) * i / steps);
+        const stepY = Math.round(currentPos.y + (targetY - currentPos.y) * i / steps);
+        libnut.moveMouse(stepX, stepY);
+        await sleep(stepDelay);
+    }
+}
+const KEY_NAME_MAP = {
+    windows: 'win',
+    win: 'win',
+    ctrl: 'control',
+    esc: 'escape',
+    del: 'delete',
+    ins: 'insert',
+    pgup: 'pageup',
+    pgdn: 'pagedown',
+    arrowup: 'up',
+    arrowdown: 'down',
+    arrowleft: 'left',
+    arrowright: 'right',
+    volumedown: 'audio_vol_down',
+    volumeup: 'audio_vol_up',
+    mediavolumedown: 'audio_vol_down',
+    mediavolumeup: 'audio_vol_up',
+    mute: 'audio_mute',
+    mediamute: 'audio_mute',
+    mediaplay: 'audio_play',
+    mediapause: 'audio_pause',
+    mediaplaypause: 'audio_play',
+    mediastop: 'audio_stop',
+    medianexttrack: 'audio_next',
+    mediaprevioustrack: 'audio_prev',
+    medianext: 'audio_next',
+    mediaprev: 'audio_prev'
+};
+const PRIMARY_KEY_MAP = {
+    command: 'cmd',
+    cmd: 'cmd',
+    meta: 'meta',
+    control: 'control',
+    ctrl: 'control',
+    shift: 'shift',
+    alt: 'alt',
+    option: 'alt'
+};
+function normalizeKeyName(key) {
+    const lowerKey = key.toLowerCase();
+    return KEY_NAME_MAP[lowerKey] || lowerKey;
+}
+function normalizePrimaryKey(key) {
+    const lowerKey = key.toLowerCase();
+    if (PRIMARY_KEY_MAP[lowerKey]) return PRIMARY_KEY_MAP[lowerKey];
+    return KEY_NAME_MAP[lowerKey] || lowerKey;
+}
+class ComputerDevice {
+    describe() {
+        return this.description || 'Computer Device';
+    }
+    static async listDisplays() {
+        try {
+            const displays = await screenshot_desktop.listDisplays();
+            return displays.map((d)=>({
+                    id: String(d.id),
+                    name: d.name || `Display ${d.id}`,
+                    primary: d.primary || false
+                }));
+        } catch (error) {
+            debugDevice(`Failed to list displays: ${error}`);
+            return [];
+        }
+    }
+    async connect() {
+        debugDevice('Connecting to computer device');
+        try {
+            libnut = await getLibnut();
+            const size = await this.size();
+            const displays = await ComputerDevice.listDisplays();
+            this.description = `
+Type: Computer
+Platform: ${process.platform}
+Display: ${this.displayId || 'Primary'}
+Screen Size: ${size.width}x${size.height}
+Available Displays: ${displays.length > 0 ? displays.map((d)=>d.name).join(', ') : 'Unknown'}
+`;
+            debugDevice('Computer device connected', this.description);
+        } catch (error) {
+            debugDevice(`Failed to connect: ${error}`);
+            throw new Error(`Unable to connect to computer device: ${error}`);
+        }
+    }
+    async screenshotBase64() {
+        debugDevice('Taking screenshot', {
+            displayId: this.displayId
+        });
+        try {
+            const options = {
+                format: 'png'
+            };
+            if (void 0 !== this.displayId) if ('darwin' === process.platform) {
+                const screenIndex = Number(this.displayId);
+                if (!Number.isNaN(screenIndex)) options.screen = screenIndex;
+            } else options.screen = this.displayId;
+            debugDevice('Screenshot options', options);
+            const buffer = await screenshot_desktop(options);
+            return createImgBase64ByFormat('png', buffer.toString('base64'));
+        } catch (error) {
+            debugDevice(`Screenshot failed: ${error}`);
+            throw new Error(`Failed to take screenshot: ${error}`);
+        }
+    }
+    async size() {
+        node_assert(libnut, 'libnut not initialized');
+        try {
+            const screenSize = libnut.getScreenSize();
+            return {
+                width: screenSize.width,
+                height: screenSize.height,
+                dpr: 1
+            };
+        } catch (error) {
+            debugDevice(`Failed to get screen size: ${error}`);
+            throw new Error(`Failed to get screen size: ${error}`);
+        }
+    }
+    shouldUseClipboardForText(text) {
+        const hasNonAscii = /[\x80-\uFFFF]/.test(text);
+        return hasNonAscii;
+    }
+    async typeViaClipboard(text) {
+        node_assert(libnut, 'libnut not initialized');
+        debugDevice('Using clipboard to input text', {
+            textLength: text.length,
+            preview: text.substring(0, 20)
+        });
+        const clipboardy = await import("clipboardy");
+        const oldClipboard = await clipboardy.default.read().catch(()=>'');
+        try {
+            await clipboardy.default.write(text);
+            await sleep(50);
+            if (this.useAppleScript) sendKeyViaAppleScript('v', [
+                'command'
+            ]);
+            else {
+                const modifier = 'darwin' === process.platform ? 'command' : 'control';
+                libnut.keyTap('v', [
+                    modifier
+                ]);
+            }
+            await sleep(100);
+        } finally{
+            if (oldClipboard) await clipboardy.default.write(oldClipboard).catch(()=>{
+                debugDevice('Failed to restore clipboard content');
+            });
+        }
+    }
+    async smartTypeString(text) {
+        node_assert(libnut, 'libnut not initialized');
+        if ('darwin' === process.platform) {
+            if (this.useAppleScript) if (this.shouldUseClipboardForText(text)) await this.typeViaClipboard(text);
+            else typeStringViaAppleScript(text);
+            else libnut.typeString(text);
+            return;
+        }
+        const inputStrategy = this.options?.inputStrategy ?? INPUT_STRATEGY_CLIPBOARD_FOR_NON_ASCII;
+        if (inputStrategy === INPUT_STRATEGY_ALWAYS_CLIPBOARD) return void await this.typeViaClipboard(text);
+        const shouldUseClipboard = this.shouldUseClipboardForText(text);
+        if (shouldUseClipboard) await this.typeViaClipboard(text);
+        else libnut.typeString(text);
+    }
+    actionSpace() {
+        const defaultActions = [
+            defineActionTap(async (param)=>{
+                node_assert(libnut, 'libnut not initialized');
+                const element = param.locate;
+                node_assert(element, 'Element not found, cannot tap');
+                const [x, y] = element.center;
+                const targetX = Math.round(x);
+                const targetY = Math.round(y);
+                await smoothMoveMouse(targetX, targetY, SMOOTH_MOVE_STEPS_TAP, SMOOTH_MOVE_DELAY_TAP);
+                libnut.mouseToggle('down', 'left');
+                await sleep(CLICK_HOLD_DURATION);
+                libnut.mouseToggle('up', 'left');
+            }),
+            defineActionDoubleClick(async (param)=>{
+                node_assert(libnut, 'libnut not initialized');
+                const element = param.locate;
+                node_assert(element, 'Element not found, cannot double click');
+                const [x, y] = element.center;
+                libnut.moveMouse(Math.round(x), Math.round(y));
+                libnut.mouseClick('left', true);
+            }),
+            defineActionRightClick(async (param)=>{
+                node_assert(libnut, 'libnut not initialized');
+                const element = param.locate;
+                node_assert(element, 'Element not found, cannot right click');
+                const [x, y] = element.center;
+                libnut.moveMouse(Math.round(x), Math.round(y));
+                libnut.mouseClick('right');
+            }),
+            defineAction({
+                name: 'MouseMove',
+                description: 'Move the mouse to the element',
+                interfaceAlias: 'aiHover',
+                paramSchema: actionHoverParamSchema,
+                call: async (param)=>{
+                    node_assert(libnut, 'libnut not initialized');
+                    const element = param.locate;
+                    node_assert(element, 'Element not found, cannot move mouse');
+                    const [x, y] = element.center;
+                    const targetX = Math.round(x);
+                    const targetY = Math.round(y);
+                    await smoothMoveMouse(targetX, targetY, SMOOTH_MOVE_STEPS_MOUSE_MOVE, SMOOTH_MOVE_DELAY_MOUSE_MOVE);
+                    await sleep(MOUSE_MOVE_EFFECT_WAIT);
+                }
+            }),
+            defineAction({
+                name: 'Input',
+                description: 'Input text into the input field',
+                interfaceAlias: 'aiInput',
+                paramSchema: z.object({
+                    value: z.string().describe('The text to input'),
+                    mode: z["enum"]([
+                        'replace',
+                        'clear',
+                        'append'
+                    ]).default('replace').optional().describe('Input mode: replace, clear, or append'),
+                    locate: getMidsceneLocationSchema().describe('The input field to be filled').optional()
+                }),
+                call: async (param)=>{
+                    node_assert(libnut, 'libnut not initialized');
+                    const element = param.locate;
+                    if (element) {
+                        const [x, y] = element.center;
+                        libnut.moveMouse(Math.round(x), Math.round(y));
+                        libnut.mouseClick('left');
+                        await sleep(INPUT_FOCUS_DELAY);
+                        if ('append' !== param.mode) {
+                            if (this.useAppleScript) {
+                                sendKeyViaAppleScript('a', [
+                                    'command'
+                                ]);
+                                await sleep(50);
+                                sendKeyViaAppleScript('backspace', []);
+                            } else {
+                                const modifier = 'darwin' === process.platform ? 'command' : 'control';
+                                libnut.keyTap('a', [
+                                    modifier
+                                ]);
+                                await sleep(50);
+                                libnut.keyTap('backspace');
+                            }
+                            await sleep(INPUT_CLEAR_DELAY);
+                        }
+                    }
+                    if ('clear' === param.mode) return;
+                    if (!param.value) return;
+                    await this.smartTypeString(param.value);
+                }
+            }),
+            defineActionScroll(async (param)=>{
+                node_assert(libnut, 'libnut not initialized');
+                if (param.locate) {
+                    const element = param.locate;
+                    const [x, y] = element.center;
+                    libnut.moveMouse(Math.round(x), Math.round(y));
+                }
+                const scrollType = param?.scrollType;
+                const scrollToEdgeActions = {
+                    scrollToTop: [
+                        0,
+                        10
+                    ],
+                    scrollToBottom: [
+                        0,
+                        -10
+                    ],
+                    scrollToLeft: [
+                        -10,
+                        0
+                    ],
+                    scrollToRight: [
+                        10,
+                        0
+                    ]
+                };
+                const edgeAction = scrollToEdgeActions[scrollType || ''];
+                if (edgeAction) {
+                    const [dx, dy] = edgeAction;
+                    for(let i = 0; i < SCROLL_REPEAT_COUNT; i++){
+                        libnut.scrollMouse(dx, dy);
+                        await sleep(SCROLL_STEP_DELAY);
+                    }
+                    return;
+                }
+                if ('singleAction' === scrollType || !scrollType) {
+                    const distance = param?.distance || 500;
+                    const ticks = Math.ceil(distance / 100);
+                    const direction = param?.direction || 'down';
+                    const directionMap = {
+                        up: [
+                            0,
+                            ticks
+                        ],
+                        down: [
+                            0,
+                            -ticks
+                        ],
+                        left: [
+                            -ticks,
+                            0
+                        ],
+                        right: [
+                            ticks,
+                            0
+                        ]
+                    };
+                    const [dx, dy] = directionMap[direction] || [
+                        0,
+                        -ticks
+                    ];
+                    libnut.scrollMouse(dx, dy);
+                    await sleep(SCROLL_COMPLETE_DELAY);
+                    return;
+                }
+                throw new Error(`Unknown scroll type: ${scrollType}, param: ${JSON.stringify(param)}`);
+            }),
+            defineActionKeyboardPress(async (param)=>{
+                node_assert(libnut, 'libnut not initialized');
+                if (param.locate) {
+                    const [x, y] = param.locate.center;
+                    libnut.moveMouse(Math.round(x), Math.round(y));
+                    libnut.mouseClick('left');
+                    await sleep(50);
+                }
+                const keys = param.keyName.split('+');
+                const modifiers = keys.slice(0, -1).map(normalizeKeyName);
+                const key = normalizePrimaryKey(keys[keys.length - 1]);
+                debugDevice('KeyboardPress', {
+                    original: param.keyName,
+                    key,
+                    modifiers,
+                    driver: this.useAppleScript ? "applescript" : 'libnut'
+                });
+                if (this.useAppleScript) sendKeyViaAppleScript(key, modifiers);
+                else if (modifiers.length > 0) libnut.keyTap(key, modifiers);
+                else libnut.keyTap(key);
+            }),
+            defineActionDragAndDrop(async (param)=>{
+                node_assert(libnut, 'libnut not initialized');
+                const from = param.from;
+                const to = param.to;
+                node_assert(from, 'missing "from" param for drag and drop');
+                node_assert(to, 'missing "to" param for drag and drop');
+                const [fromX, fromY] = from.center;
+                const [toX, toY] = to.center;
+                libnut.moveMouse(Math.round(fromX), Math.round(fromY));
+                libnut.mouseToggle('down', 'left');
+                await sleep(100);
+                libnut.moveMouse(Math.round(toX), Math.round(toY));
+                await sleep(100);
+                libnut.mouseToggle('up', 'left');
+            }),
+            defineActionClearInput(async (param)=>{
+                node_assert(libnut, 'libnut not initialized');
+                const element = param.locate;
+                node_assert(element, 'Element not found, cannot clear input');
+                const [x, y] = element.center;
+                libnut.moveMouse(Math.round(x), Math.round(y));
+                libnut.mouseClick('left');
+                await sleep(100);
+                if (this.useAppleScript) {
+                    sendKeyViaAppleScript('a', [
+                        'command'
+                    ]);
+                    await sleep(50);
+                    sendKeyViaAppleScript('backspace', []);
+                } else {
+                    const modifier = 'darwin' === process.platform ? 'command' : 'control';
+                    libnut.keyTap('a', [
+                        modifier
+                    ]);
+                    libnut.keyTap('backspace');
+                }
+                await sleep(50);
+            })
+        ];
+        const platformActions = Object.values(createPlatformActions());
+        const customActions = this.options?.customActions || [];
+        return [
+            ...defaultActions,
+            ...platformActions,
+            ...customActions
+        ];
+    }
+    async destroy() {
+        if (this.destroyed) return;
+        this.destroyed = true;
+        debugDevice('Computer device destroyed');
+    }
+    async url() {
+        return '';
+    }
+    constructor(options){
+        _define_property(this, "interfaceType", 'computer');
+        _define_property(this, "options", void 0);
+        _define_property(this, "displayId", void 0);
+        _define_property(this, "description", void 0);
+        _define_property(this, "destroyed", false);
+        _define_property(this, "useAppleScript", void 0);
+        _define_property(this, "uri", void 0);
+        this.options = options;
+        this.displayId = options?.displayId;
+        this.useAppleScript = 'darwin' === process.platform && options?.keyboardDriver !== 'libnut';
+    }
+}
+function createPlatformActions() {
+    return {
+        ListDisplays: defineAction({
+            name: 'ListDisplays',
+            description: 'List all available displays/monitors',
+            call: async ()=>await ComputerDevice.listDisplays()
+        })
+    };
+}
+class ComputerAgent extends Agent {
+}
+async function agentFromComputer(opts) {
+    const device = new ComputerDevice(opts || {});
+    await device.connect();
+    return new ComputerAgent(device, opts);
+}
+const debug = getDebug('mcp:computer-tools');
+class ComputerMidsceneTools extends BaseMidsceneTools {
+    createTemporaryDevice() {
+        return new ComputerDevice({});
+    }
+    async ensureAgent(displayId) {
+        if (this.agent && displayId) {
+            try {
+                await this.agent.destroy?.();
+            } catch (error) {
+                debug('Failed to destroy agent during cleanup:', error);
+            }
+            this.agent = void 0;
+        }
+        if (this.agent) return this.agent;
+        debug('Creating Computer agent with displayId:', displayId || 'primary');
+        const opts = displayId ? {
+            displayId
+        } : void 0;
+        const agent = await agentFromComputer(opts);
+        this.agent = agent;
+        return agent;
+    }
+    preparePlatformTools() {
+        return [
+            {
+                name: 'computer_connect',
+                description: 'Connect to computer desktop. If displayId not provided, uses the primary display.',
+                schema: {
+                    displayId: z.string().optional().describe('Display ID (from list_displays)')
+                },
+                handler: async ({ displayId })=>{
+                    const agent = await this.ensureAgent(displayId);
+                    const screenshot = await agent.interface.screenshotBase64();
+                    return {
+                        content: [
+                            {
+                                type: 'text',
+                                text: `Connected to computer${displayId ? ` (Display: ${displayId})` : ' (Primary display)'}`
+                            },
+                            ...this.buildScreenshotContent(screenshot)
+                        ]
+                    };
+                }
+            },
+            {
+                name: 'computer_disconnect',
+                description: 'Disconnect from computer and release resources',
+                schema: {},
+                handler: this.createDisconnectHandler('computer')
+            },
+            {
+                name: 'computer_list_displays',
+                description: 'List all available displays/monitors',
+                schema: {},
+                handler: async ()=>{
+                    const displays = await ComputerDevice.listDisplays();
+                    return {
+                        content: [
+                            {
+                                type: 'text',
+                                text: `Available displays:\n${displays.map((d)=>`- ${d.name} (ID: ${d.id})${d.primary ? ' [PRIMARY]' : ''}`).join('\n')}`
+                            }
+                        ]
+                    };
+                }
+            }
+        ];
+    }
+}
+const tools = new ComputerMidsceneTools();
+runToolsCLI(tools, 'midscene-computer', {
+    stripPrefix: 'computer_'
+}).catch((e)=>{
+    if (!(e instanceof CLIError)) console.error(e);
+    process.exit(e instanceof CLIError ? e.exitCode : 1);
+});