@midscene/computer 1.12.4-beta-20260904041024.0 → 1.12.4

package/dist/lib/index.js

@@ -68,210 +68,6 @@ const img_namespaceObject = require("@midscene/shared/img");
 const logger_namespaceObject = require("@midscene/shared/logger");
 const external_screenshot_desktop_namespaceObject = require("screenshot-desktop");
 var external_screenshot_desktop_default = /*#__PURE__*/ __webpack_require__.n(external_screenshot_desktop_namespaceObject);
-const US_SHIFTED_CHARACTER_KEYS = new Map([
-    [
-        '~',
-        '`'
-    ],
-    [
-        '!',
-        '1'
-    ],
-    [
-        '@',
-        '2'
-    ],
-    [
-        '#',
-        '3'
-    ],
-    [
-        '$',
-        '4'
-    ],
-    [
-        '%',
-        '5'
-    ],
-    [
-        '^',
-        '6'
-    ],
-    [
-        '&',
-        '7'
-    ],
-    [
-        '*',
-        '8'
-    ],
-    [
-        '(',
-        '9'
-    ],
-    [
-        ')',
-        '0'
-    ],
-    [
-        '_',
-        '-'
-    ],
-    [
-        '+',
-        '='
-    ],
-    [
-        '{',
-        '['
-    ],
-    [
-        '}',
-        ']'
-    ],
-    [
-        '|',
-        '\\'
-    ],
-    [
-        ':',
-        ';'
-    ],
-    [
-        '"',
-        "'"
-    ],
-    [
-        '<',
-        ','
-    ],
-    [
-        '>',
-        '.'
-    ],
-    [
-        '?',
-        '/'
-    ]
-]);
-const debugKeyboard = (0, logger_namespaceObject.getDebug)('computer:keyboard');
-const APPLE_SCRIPT_KEY_CODES = {
-    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 APPLE_SCRIPT_MODIFIER_KEYS = {
-    command: 'command',
-    cmd: 'command',
-    control: 'control',
-    ctrl: 'control',
-    shift: 'shift',
-    alt: 'option',
-    option: 'option',
-    meta: 'command'
-};
-function buildKeyCommand(key) {
-    const keyCode = APPLE_SCRIPT_KEY_CODES[key.toLowerCase()];
-    if (void 0 !== keyCode) return `key code ${keyCode}`;
-    const escapedKey = key.replace(/\\/g, '\\\\').replace(/"/g, '\\"');
-    return `keystroke "${escapedKey}"`;
-}
-function resolveModifierKeys(modifiers) {
-    return modifiers.map((modifier)=>APPLE_SCRIPT_MODIFIER_KEYS[modifier.toLowerCase()]).filter((modifier)=>void 0 !== modifier);
-}
-function buildLogicalKeyPress(key, modifiers) {
-    const modifierKeys = resolveModifierKeys(modifiers);
-    const modifierClause = modifierKeys.length ? ` using {${modifierKeys.map((modifier)=>`${modifier} down`).join(', ')}}` : '';
-    return `tell application "System Events" to ${buildKeyCommand(key)}${modifierClause}`;
-}
-function resolvePhysicalKey(key, modifiers) {
-    const resolvedModifiers = [
-        ...modifiers
-    ];
-    const shiftedBaseKey = US_SHIFTED_CHARACTER_KEYS.get(key);
-    if (/^[A-Z]$/.test(key)) {
-        resolvedModifiers.push('shift');
-        return {
-            key: key.toLowerCase(),
-            modifiers: resolvedModifiers
-        };
-    }
-    if (void 0 !== shiftedBaseKey) {
-        resolvedModifiers.push('shift');
-        return {
-            key: shiftedBaseKey,
-            modifiers: resolvedModifiers
-        };
-    }
-    return {
-        key,
-        modifiers: resolvedModifiers
-    };
-}
-function buildPhysicalKeyPress(key, modifiers) {
-    const resolved = resolvePhysicalKey(key, modifiers);
-    const modifierKeys = [
-        ...new Set(resolveModifierKeys(resolved.modifiers))
-    ];
-    const keyCommand = buildKeyCommand(resolved.key);
-    if (0 === modifierKeys.length) return `tell application "System Events" to ${keyCommand}`;
-    const releaseCommands = [
-        ...modifierKeys
-    ].reverse().map((modifier)=>`key up ${modifier}`);
-    return [
-        'tell application "System Events"',
-        'try',
-        ...modifierKeys.map((modifier)=>`key down ${modifier}`),
-        keyCommand,
-        'on error errorMessage number errorNumber',
-        ...releaseCommands,
-        'error errorMessage number errorNumber',
-        'end try',
-        ...releaseCommands,
-        'end tell'
-    ].join('\n');
-}
-function buildAppleScriptKeyPress(key, modifiers = [], eventMode = 'logical') {
-    return 'physical' === eventMode ? buildPhysicalKeyPress(key, modifiers) : buildLogicalKeyPress(key, modifiers);
-}
-function sendKeyViaAppleScript(key, modifiers = [], eventMode = 'logical') {
-    const script = buildAppleScriptKeyPress(key, modifiers, eventMode);
-    debugKeyboard('sendKeyViaAppleScript', {
-        key,
-        modifiers,
-        eventMode,
-        script
-    });
-    (0, external_node_child_process_namespaceObject.execFileSync)("osascript", [
-        '-e',
-        script
-    ]);
-}
 const external_node_assert_namespaceObject = require("node:assert");
 var external_node_assert_default = /*#__PURE__*/ __webpack_require__.n(external_node_assert_namespaceObject);
 function _define_property(obj, key, value) {
@@ -797,6 +593,92 @@ const LIBNUT_FALLBACK_PIXELS_PER_DETENT = 100;
 const LIBNUT_FALLBACK_TICK_DELAY_MS = 30;
 const LIBNUT_FALLBACK_MAX_DETENTS = 200;
 const LIBNUT_FALLBACK_DETENT_AMOUNT = 'win32' === process.platform ? 120 : 1;
+const LINUX_SHIFTED_CHARACTER_KEYS = new Map([
+    [
+        '~',
+        '`'
+    ],
+    [
+        '!',
+        '1'
+    ],
+    [
+        '@',
+        '2'
+    ],
+    [
+        '#',
+        '3'
+    ],
+    [
+        '$',
+        '4'
+    ],
+    [
+        '%',
+        '5'
+    ],
+    [
+        '^',
+        '6'
+    ],
+    [
+        '&',
+        '7'
+    ],
+    [
+        '*',
+        '8'
+    ],
+    [
+        '(',
+        '9'
+    ],
+    [
+        ')',
+        '0'
+    ],
+    [
+        '_',
+        '-'
+    ],
+    [
+        '+',
+        '='
+    ],
+    [
+        '{',
+        '['
+    ],
+    [
+        '}',
+        ']'
+    ],
+    [
+        '|',
+        '\\'
+    ],
+    [
+        ':',
+        ';'
+    ],
+    [
+        '"',
+        "'"
+    ],
+    [
+        '<',
+        ','
+    ],
+    [
+        '>',
+        '.'
+    ],
+    [
+        '?',
+        '/'
+    ]
+]);
 const LIBNUT_FALLBACK_EDGE_DETENTS = Math.min(LIBNUT_FALLBACK_MAX_DETENTS, Math.max(1, Math.ceil(EDGE_SCROLL_TOTAL_PX / LIBNUT_FALLBACK_PIXELS_PER_DETENT)));
 const DEFAULT_SCROLL_VIEWPORT_RATIO = 0.7;
 const EDGE_SCROLL_SPEC = {
@@ -833,6 +715,67 @@ const EDGE_SCROLL_SPEC = {
         ]
     }
 };
+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 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 {
+        const escapedKey = key.replace(/\\/g, '\\\\').replace(/"/g, '\\"');
+        script = `tell application "System Events" to keystroke "${escapedKey}"${modifierStr}`;
+    }
+    debugDevice('sendKeyViaAppleScript', {
+        key,
+        modifiers,
+        script
+    });
+    (0, external_node_child_process_namespaceObject.execFileSync)("osascript", [
+        '-e',
+        script
+    ]);
+}
 function escapePowershellSingleQuoted(value) {
     return value.replace(/'/g, "''");
 }
@@ -1260,7 +1203,7 @@ Available Displays: ${displays.length > 0 ? displays.map((d)=>d.name).join(', ')
     }
     async healthCheck(displays) {
         console.log('[HealthCheck] Starting health check...');
-        console.log("[HealthCheck] @midscene/computer v1.12.4-beta-20260904041024.0");
+        console.log("[HealthCheck] @midscene/computer v1.12.4");
         console.log('[HealthCheck] Taking screenshot...');
         const screenshotTimeout = 15000;
         let timeoutId;
@@ -1455,7 +1398,7 @@ $g.Dispose(); $bmp.Dispose(); $ms.Dispose()
     async typeStringWithDelay(text, keyboardTypeDelay) {
         await (0, device_namespaceObject.sendTextSequentially)(text.replace(/\r\n?/g, '\n'), {
             sendCharacter: (character)=>{
-                const linuxShiftedKey = 'linux' === process.platform ? US_SHIFTED_CHARACTER_KEYS.get(character) : void 0;
+                const linuxShiftedKey = 'linux' === process.platform ? LINUX_SHIFTED_CHARACTER_KEYS.get(character) : void 0;
                 if ('\n' === character) this.inputDriver.sendKey('return');
                 else if ('\t' === character) this.inputDriver.sendKey('tab');
                 else if (' ' === character) this.inputDriver.sendKey('space');
@@ -1649,7 +1592,7 @@ $g.Dispose(); $bmp.Dispose(); $ms.Dispose()
         device_define_property(this, "inputDriver", new ComputerInputDriver({
             getLibnut: ()=>device_libnut,
             useAppleScript: ()=>this.useAppleScript,
-            sendKeyViaAppleScript: (key, modifiers)=>sendKeyViaAppleScript(key, modifiers, this.options?.keyboardEventMode ?? 'logical'),
+            sendKeyViaAppleScript,
             runPhasedScroll,
             debug: (message)=>debugDevice(message)
         }));
@@ -2519,7 +2462,6 @@ function createLocalComputerDevice(opts) {
         keyboardTypeDelay: opts?.keyboardTypeDelay,
         inputStrategy: opts?.inputStrategy,
         keyboardDriver: opts?.keyboardDriver,
-        keyboardEventMode: opts?.keyboardEventMode,
         headless: opts?.headless,
         xvfbResolution: opts?.xvfbResolution,
         keepXvfbAliveUntilProcessExit: opts?.keepXvfbAliveUntilProcessExit
@@ -2580,10 +2522,6 @@ const computerInitArgShape = {
     headless: core_namespaceObject.z.boolean().optional().describe('Start virtual display via Xvfb (Linux local mode only). Ignored when host is set.'),
     keyboardTypeDelay: core_namespaceObject.z.number().finite().nonnegative().optional().describe('Finite non-negative delay in milliseconds between keystrokes. In "legacy" mode, positive values enable key-by-key input; zero or omitted uses clipboard input.'),
     inputStrategy: core_namespaceObject.z["enum"](device_namespaceObject.inputStrategies).optional().describe('Text input strategy. "legacy" (default) preserves current Computer behavior, "sequential" sends one Unicode code point at a time, and "bulk" uses one backend input operation. "bulk" requires keyboardTypeDelay to be omitted or set to 0.'),
-    keyboardEventMode: core_namespaceObject.z["enum"]([
-        'logical',
-        'physical'
-    ]).optional().describe('macOS AppleScript keyboard event mode for local control. "logical" (default) targets native apps; "physical" is only for VNC clients, assumes an en-US layout for shifted punctuation, and requires sequential text input or a positive keyboardTypeDelay. Ignored in RDP mode and outside the macOS AppleScript driver.'),
     host: core_namespaceObject.z.string().optional().describe('RDP host (FQDN or IP). Set this to switch into RDP mode.'),
     port: core_namespaceObject.z.number().optional().describe('RDP port (default 3389). Requires host.'),
     username: core_namespaceObject.z.string().optional().describe('RDP username. Requires host.'),
@@ -2600,7 +2538,7 @@ const computerInitArgShape = {
 function adaptComputerInitArgs(extracted) {
     if (!extracted || 0 === Object.keys(extracted).length) return;
     if (extracted.host) {
-        const { displayId: _d, headless: _h, keyboardEventMode: _k, ...rdpFields } = extracted;
+        const { displayId: _d, headless: _h, ...rdpFields } = extracted;
         const host = normalizeRdpHost(extracted.host);
         return {
             mode: 'rdp',
@@ -2614,7 +2552,6 @@ function adaptComputerInitArgs(extracted) {
         headless: extracted.headless,
         keyboardTypeDelay: extracted.keyboardTypeDelay,
         inputStrategy: extracted.inputStrategy,
-        keyboardEventMode: extracted.keyboardEventMode,
         ...(0, agent_behavior_init_args_namespaceObject.extractAgentBehaviorInitArgs)(extracted) ?? {}
     };
 }
@@ -2666,7 +2603,6 @@ class ComputerMidsceneTools extends base_tools_namespaceObject.BaseMidsceneTools
         const headless = opts?.mode === 'local' ? opts.headless : void 0;
         const keyboardTypeDelay = opts?.keyboardTypeDelay;
         const inputStrategy = opts?.inputStrategy;
-        const keyboardEventMode = opts?.mode === 'local' ? opts.keyboardEventMode : void 0;
         agent_tools_debug('Creating Computer agent with displayId:', displayId || 'primary');
         const agentOpts = {
             ...displayId ? {
@@ -2681,9 +2617,6 @@ class ComputerMidsceneTools extends base_tools_namespaceObject.BaseMidsceneTools
             ...void 0 !== inputStrategy ? {
                 inputStrategy
             } : {},
-            ...void 0 !== keyboardEventMode ? {
-                keyboardEventMode
-            } : {},
             ...this.options.keepXvfbAliveUntilProcessExit ? {
                 keepXvfbAliveUntilProcessExit: true
             } : {},
@@ -2765,7 +2698,7 @@ class ComputerMidsceneTools extends base_tools_namespaceObject.BaseMidsceneTools
 }
 const env_namespaceObject = require("@midscene/shared/env");
 function version() {
-    const currentVersion = "1.12.4-beta-20260904041024.0";
+    const currentVersion = "1.12.4";
     console.log(`@midscene/computer v${currentVersion}`);
     return currentVersion;
 }

package/dist/types/index.d.ts

@@ -177,19 +177,6 @@ export declare interface ComputerDeviceOpt extends ComputerDeviceInputOpt {
      * - 'libnut': Use libnut's keyTap (faster but may not work with some TUI apps)
      */
     keyboardDriver?: 'applescript' | 'libnut';
-    /**
-     * How the macOS AppleScript keyboard driver represents modifier keys.
-     * `logical` keeps the default compact `keystroke ... using` behavior.
-     * `physical` emits explicit modifier key-down/key-up transitions for apps
-     * such as VNC clients that forward physical keyboard events. Physical mode
-     * assumes an en-US layout for shifted punctuation and may type base keys in
-     * native macOS applications. Text input must use sequential input or a
-     * positive `keyboardTypeDelay` to emit individual keys.
-     *
-     * Ignored outside macOS and when `keyboardDriver` is `libnut`.
-     * @default 'logical'
-     */
-    keyboardEventMode?: KeyboardEventMode;
     /**
      * Headless mode via Xvfb (Linux only).
      * - true: start Xvfb virtual display
@@ -223,7 +210,7 @@ export declare type ComputerInterface = ComputerDevice | RDPDevice;
 /** Init args for the local desktop agent (macOS/Windows/Linux). */
 declare type ComputerLocalInitArgs = {
     mode: 'local';
-} & Pick<ComputerDeviceOpt, 'displayId' | 'headless'> & Pick<ComputerDeviceOpt, 'inputStrategy' | 'keyboardTypeDelay' | 'keyboardEventMode'> & AgentBehaviorInitArgs;
+} & Pick<ComputerDeviceOpt, 'displayId' | 'headless'> & Pick<ComputerDeviceOpt, 'inputStrategy' | 'keyboardTypeDelay'> & AgentBehaviorInitArgs;
 
 /**
  * Computer-specific tools manager
@@ -310,15 +297,6 @@ export declare class HelperProcessRDPBackendClient implements RDPBackendClient {
     private shutdownHelper;
 }
 
-/**
- * Modifier delivery mode for the macOS AppleScript keyboard backend.
- *
- * `logical` is the default for native macOS applications. `physical` emits
- * explicit modifier transitions for VNC clients and assumes an en-US mapping
- * for shifted punctuation; it should not be enabled for native applications.
- */
-export declare type KeyboardEventMode = 'logical' | 'physical';
-
 export declare type LocalComputerAgentOpt = BaseComputerAgentOpt & Omit<ComputerDeviceOpt, keyof ComputerAgentSharedDeviceOpt>;
 
 /**

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@midscene/computer",
-  "version": "1.12.4-beta-20260904041024.0",
+  "version": "1.12.4",
   "description": "Midscene.js Computer Desktop Automation",
   "license": "MIT",
   "repository": {
@@ -33,8 +33,8 @@
     "@computer-use/libnut": "^4.2.0",
     "clipboardy": "^4.0.0",
     "screenshot-desktop": "^1.15.3",
-    "@midscene/core": "1.12.4-beta-20260904041024.0",
-    "@midscene/shared": "1.12.4-beta-20260904041024.0"
+    "@midscene/core": "1.12.4",
+    "@midscene/shared": "1.12.4"
   },
   "optionalDependencies": {
     "node-mac-permissions": "2.5.0"