@midscene/computer 1.9.8 → 1.10.1-beta-20260624112700.0

package/dist/lib/index.js

@@ -424,6 +424,42 @@ function sendKeyViaAppleScript(key, modifiers = []) {
         script
     ]);
 }
+const POWERSHELL_TIMEOUT_MS = 15000;
+const POWERSHELL_MAX_BUFFER = 67108864;
+function escapePowershellSingleQuoted(value) {
+    return value.replace(/'/g, "''");
+}
+function runPowershell(script) {
+    const encoded = Buffer.from(script, 'utf16le').toString('base64');
+    return (0, external_node_child_process_namespaceObject.execFileSync)('powershell.exe', [
+        '-NoProfile',
+        '-NonInteractive',
+        '-ExecutionPolicy',
+        'Bypass',
+        '-EncodedCommand',
+        encoded
+    ], {
+        encoding: 'utf8',
+        timeout: POWERSHELL_TIMEOUT_MS,
+        maxBuffer: POWERSHELL_MAX_BUFFER,
+        windowsHide: true
+    });
+}
+function listWindowsDisplays() {
+    const script = `
+Add-Type -AssemblyName System.Windows.Forms
+$s = [System.Windows.Forms.Screen]::AllScreens | ForEach-Object {
+  [PSCustomObject]@{ id = $_.DeviceName; name = $_.DeviceName; primary = $_.Primary }
+}
+ConvertTo-Json @($s) -Compress
+`.trim();
+    const parsed = JSON.parse(runPowershell(script).trim());
+    return parsed.map((d)=>({
+            id: String(d.id),
+            name: d.name || String(d.id),
+            primary: d.primary || false
+        }));
+}
 let device_libnut = null;
 let libnutLoadError = null;
 async function getLibnut() {
@@ -690,6 +726,7 @@ class ComputerDevice {
     }
     static async listDisplays() {
         try {
+            if ('win32' === process.platform) return listWindowsDisplays();
             const displays = await external_screenshot_desktop_default().listDisplays();
             return displays.map((d)=>({
                     id: String(d.id),
@@ -747,7 +784,7 @@ Available Displays: ${displays.length > 0 ? displays.map((d)=>d.name).join(', ')
     }
     async healthCheck() {
         console.log('[HealthCheck] Starting health check...');
-        console.log("[HealthCheck] @midscene/computer v1.9.8");
+        console.log("[HealthCheck] @midscene/computer v1.10.1-beta-20260624112700.0");
         console.log('[HealthCheck] Taking screenshot...');
         const screenshotTimeout = 15000;
         let timeoutId;
@@ -816,6 +853,7 @@ Available Displays: ${displays.length > 0 ? displays.map((d)=>d.name).join(', ')
         debugDevice('Taking screenshot', {
             displayId: this.displayId
         });
+        if ('win32' === process.platform) return this.screenshotViaPowershell();
         const options = {
             format: 'png'
         };
@@ -849,6 +887,40 @@ Please follow these steps:
 Original error: ${lastRawMessage}`);
         throw new Error(`Failed to take screenshot: ${lastRawMessage}`);
     }
+    screenshotViaPowershell() {
+        const deviceName = this.displayId ? String(this.displayId) : '';
+        const target = deviceName ? `[System.Windows.Forms.Screen]::AllScreens | Where-Object { $_.DeviceName -eq '${escapePowershellSingleQuoted(deviceName)}' } | Select-Object -First 1` : '$null';
+        const script = `
+$ErrorActionPreference = 'Stop'
+Add-Type -AssemblyName System.Windows.Forms, System.Drawing
+Add-Type @"
+using System;
+using System.Runtime.InteropServices;
+public class NutDpi { [DllImport("user32.dll")] public static extern bool SetProcessDPIAware(); }
+"@
+[NutDpi]::SetProcessDPIAware() | Out-Null
+$screen = ${target}
+if (-not $screen) { $screen = [System.Windows.Forms.Screen]::PrimaryScreen }
+$b = $screen.Bounds
+$bmp = New-Object System.Drawing.Bitmap($b.Width, $b.Height)
+$g = [System.Drawing.Graphics]::FromImage($bmp)
+$g.CopyFromScreen($b.X, $b.Y, 0, 0, $bmp.Size)
+$ms = New-Object System.IO.MemoryStream
+$bmp.Save($ms, [System.Drawing.Imaging.ImageFormat]::Png)
+[Console]::Out.Write([Convert]::ToBase64String($ms.ToArray()))
+$g.Dispose(); $bmp.Dispose(); $ms.Dispose()
+`.trim();
+        let stdout;
+        try {
+            stdout = runPowershell(script);
+        } catch (error) {
+            const message = error instanceof Error ? error.message : String(error);
+            throw new Error(`Failed to take screenshot on Windows: ${message}`);
+        }
+        const body = stdout.trim();
+        if (!body) throw new Error('Failed to take screenshot on Windows: PowerShell returned no image data');
+        return (0, img_namespaceObject.createImgBase64ByFormat)('png', body);
+    }
     async size() {
         if (this.displayGeometry) return {
             width: Math.round(this.displayGeometry.bounds.width),
@@ -1972,9 +2044,9 @@ async function agentForRDPComputer(opts) {
     return new ComputerAgent(device, opts);
 }
 const core_namespaceObject = require("@midscene/core");
-const agent_behavior_init_args_namespaceObject = require("@midscene/shared/mcp/agent-behavior-init-args");
-const base_tools_namespaceObject = require("@midscene/shared/mcp/base-tools");
-function mcp_tools_define_property(obj, key, value) {
+const agent_behavior_init_args_namespaceObject = require("@midscene/shared/agent-tools/agent-behavior-init-args");
+const base_tools_namespaceObject = require("@midscene/shared/agent-tools/base-tools");
+function agent_tools_define_property(obj, key, value) {
     if (key in obj) Object.defineProperty(obj, key, {
         value: value,
         enumerable: true,
@@ -1984,7 +2056,7 @@ function mcp_tools_define_property(obj, key, value) {
     else obj[key] = value;
     return obj;
 }
-const mcp_tools_debug = (0, logger_namespaceObject.getDebug)('mcp:computer-tools');
+const agent_tools_debug = (0, logger_namespaceObject.getDebug)('agent-tools:computer');
 const RDP_SECURITY_PROTOCOLS = [
     'auto',
     'tls',
@@ -2052,14 +2124,14 @@ class ComputerMidsceneTools extends base_tools_namespaceObject.BaseMidsceneTools
             try {
                 await this.agent.destroy?.();
             } catch (error) {
-                mcp_tools_debug('Failed to destroy agent during cleanup:', error);
+                agent_tools_debug('Failed to destroy agent during cleanup:', error);
             }
             this.agent = void 0;
         }
         if (this.agent) return this.agent;
         const reportOptions = this.readCliReportAgentOptions();
         if (opts?.mode === 'rdp') {
-            mcp_tools_debug('Creating RDP Computer agent for host:', opts.host);
+            agent_tools_debug('Creating RDP Computer agent for host:', opts.host);
             const { mode: _mode, ...rdpFields } = opts;
             const agent = await agentForRDPComputer({
                 ...rdpFields,
@@ -2071,7 +2143,7 @@ class ComputerMidsceneTools extends base_tools_namespaceObject.BaseMidsceneTools
         }
         const displayId = opts?.mode === 'local' ? opts.displayId : void 0;
         const headless = opts?.mode === 'local' ? opts.headless : void 0;
-        mcp_tools_debug('Creating Computer agent with displayId:', displayId || 'primary');
+        agent_tools_debug('Creating Computer agent with displayId:', displayId || 'primary');
         const agentOpts = {
             ...displayId ? {
                 displayId
@@ -2102,7 +2174,7 @@ class ComputerMidsceneTools extends base_tools_namespaceObject.BaseMidsceneTools
                         try {
                             await this.agent.destroy?.();
                         } catch (error) {
-                            mcp_tools_debug('Failed to destroy agent during connect:', error);
+                            agent_tools_debug('Failed to destroy agent during connect:', error);
                         }
                         this.agent = void 0;
                         this.lastInitArgsSignature = void 0;
@@ -2145,7 +2217,7 @@ class ComputerMidsceneTools extends base_tools_namespaceObject.BaseMidsceneTools
         ];
     }
     constructor(...args){
-        super(...args), mcp_tools_define_property(this, "lastInitArgsSignature", void 0), mcp_tools_define_property(this, "initArgSpec", {
+        super(...args), agent_tools_define_property(this, "lastInitArgsSignature", void 0), agent_tools_define_property(this, "initArgSpec", {
             namespace: 'computer',
             shape: computerInitArgShape,
             cli: {
@@ -2157,7 +2229,7 @@ class ComputerMidsceneTools extends base_tools_namespaceObject.BaseMidsceneTools
 }
 const env_namespaceObject = require("@midscene/shared/env");
 function version() {
-    const currentVersion = "1.9.8";
+    const currentVersion = "1.10.1-beta-20260624112700.0";
     console.log(`@midscene/computer v${currentVersion}`);
     return currentVersion;
 }

package/dist/types/index.d.ts

@@ -1,17 +1,17 @@
 import { AbstractInterface } from '@midscene/core/device';
 import { Agent } from '@midscene/core/agent';
-import { AgentBehaviorInitArgs } from '@midscene/shared/mcp/agent-behavior-init-args';
+import { AgentBehaviorInitArgs } from '@midscene/shared/agent-tools/agent-behavior-init-args';
 import { AgentOpt } from '@midscene/core/agent';
-import { BaseMidsceneTools } from '@midscene/shared/mcp/base-tools';
+import { BaseMidsceneTools } from '@midscene/shared/agent-tools/base-tools';
 import { ChildProcessWithoutNullStreams } from 'node:child_process';
 import { ComputerInputPrimitives } from '@midscene/core/device';
 import type { DeviceAction } from '@midscene/core';
-import { InitArgSpec } from '@midscene/shared/mcp/base-tools';
+import { InitArgSpec } from '@midscene/shared/agent-tools/base-tools';
 import type { InterfaceType } from '@midscene/core';
 import { overrideAIConfig } from '@midscene/shared/env';
 import type { Size } from '@midscene/core';
 import { SpawnOptionsWithoutStdio } from 'node:child_process';
-import type { ToolDefinition } from '@midscene/shared/mcp/types';
+import type { ToolDefinition } from '@midscene/shared/agent-tools/types';
 
 declare interface AccessibilityCheckResult {
     hasPermission: boolean;
@@ -101,6 +101,20 @@ export declare class ComputerDevice implements AbstractInterface {
      */
     private isRunningAsAdmin;
     screenshotBase64(): Promise<string>;
+    /**
+     * Windows screenshot path that bypasses screenshot-desktop's polyglot .bat
+     * (see screenshotBase64 for the rationale). Captures via PowerShell +
+     * System.Drawing, which honors `displayId` by matching the monitor's
+     * DeviceName and captures in virtual-desktop coordinates, so secondary
+     * displays — including those at negative offsets — are supported.
+     *
+     * Note: the script makes the process system-DPI-aware (SetProcessDPIAware)
+     * so captures return physical pixels. On mixed-DPI multi-monitor setups a
+     * non-primary monitor may still be off by its per-monitor scale factor;
+     * system DPI awareness covers the common single-scale and uniform-scale
+     * cases that #2150 reports.
+     */
+    private screenshotViaPowershell;
     size(): Promise<Size>;
     private toGlobalPoint;
     /**
@@ -153,7 +167,7 @@ export declare interface ComputerDeviceOpt {
 /**
  * Discriminated union describing the two ways `computer_*` tools can spawn an
  * agent. `mode` is filled in by `initArgSpec.adapt` based on whether `host` is
- * set, so callers (CLI/MCP/YAML) never have to provide it explicitly.
+ * set, so callers (CLI/YAML) never have to provide it explicitly.
  */
 declare type ComputerInitArgs = ComputerLocalInitArgs | ComputerRDPInitArgs;
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@midscene/computer",
-  "version": "1.9.8",
+  "version": "1.10.1-beta-20260624112700.0",
   "description": "Midscene.js Computer Desktop Automation",
   "license": "MIT",
   "repository": {
@@ -27,19 +27,14 @@
       "import": "./dist/es/index.mjs",
       "require": "./dist/lib/index.js"
     },
-    "./mcp-server": {
-      "types": "./dist/types/mcp-server.d.ts",
-      "import": "./dist/es/mcp-server.mjs",
-      "require": "./dist/lib/mcp-server.js"
-    },
     "./package.json": "./package.json"
   },
   "dependencies": {
     "@computer-use/libnut": "^4.2.0",
     "clipboardy": "^4.0.0",
     "screenshot-desktop": "^1.15.3",
-    "@midscene/core": "1.9.8",
-    "@midscene/shared": "1.9.8"
+    "@midscene/shared": "1.10.1-beta-20260624112700.0",
+    "@midscene/core": "1.10.1-beta-20260624112700.0"
   },
   "optionalDependencies": {
     "node-mac-permissions": "2.5.0"