@midscene/computer 1.8.0 → 1.8.1

package/dist/types/index.d.ts

@@ -3,11 +3,12 @@ import { Agent } from '@midscene/core/agent';
 import { AgentOpt } from '@midscene/core/agent';
 import { BaseMidsceneTools } from '@midscene/shared/mcp/base-tools';
 import { ChildProcessWithoutNullStreams } from 'node:child_process';
-import { DeviceAction } from '@midscene/core';
+import { ComputerInputPrimitives } from '@midscene/core/device';
+import type { DeviceAction } from '@midscene/core';
 import { InitArgSpec } from '@midscene/shared/mcp/base-tools';
-import { InterfaceType } from '@midscene/core';
+import type { InterfaceType } from '@midscene/core';
 import { overrideAIConfig } from '@midscene/shared/env';
-import { Size } from '@midscene/core';
+import type { Size } from '@midscene/core';
 import { SpawnOptionsWithoutStdio } from 'node:child_process';
 import type { ToolDefinition } from '@midscene/shared/mcp/types';
 
@@ -41,6 +42,15 @@ export declare function checkAccessibilityPermission(promptIfNeeded?: boolean):
  */
 export declare function checkComputerEnvironment(): Promise<EnvironmentCheck>;
 
+/**
+ * Check if macOS Screen Recording permission is granted (required for screencapture).
+ * Accessibility and Screen Recording are independent TCC permissions — granting
+ * one does not grant the other. On non-macOS platforms, always returns true.
+ *
+ * @param promptIfNeeded - If true, will trigger system prompt and open settings when permission is not granted (macOS only)
+ */
+export declare function checkScreenRecordingPermission(promptIfNeeded?: boolean): AccessibilityCheckResult;
+
 /**
  * Check if Xvfb is installed on the system
  */
@@ -67,6 +77,7 @@ export declare class ComputerDevice implements AbstractInterface {
      */
     private useAppleScript;
     uri?: string;
+    readonly inputPrimitives: ComputerInputPrimitives;
     constructor(options?: ComputerDeviceOpt);
     describe(): string;
     /**
@@ -97,6 +108,9 @@ export declare class ComputerDevice implements AbstractInterface {
      * which can swallow characters or convert them when a non-English IME is active.
      */
     private smartTypeString;
+    private selectAllAndDelete;
+    private pressKeyboardShortcut;
+    private performScroll;
     actionSpace(): DeviceAction<any>[];
     destroy(): Promise<void>;
     url(): Promise<string>;
@@ -124,10 +138,20 @@ export declare interface ComputerDeviceOpt {
     xvfbResolution?: string;
 }
 
-declare type ComputerInitArgs = Pick<ComputerDeviceOpt, 'displayId' | 'headless'>;
+/**
+ * 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.
+ */
+declare type ComputerInitArgs = ComputerLocalInitArgs | ComputerRDPInitArgs;
 
 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'>;
+
 /**
  * Computer-specific tools manager
  * Extends BaseMidsceneTools to provide desktop automation tools
@@ -143,6 +167,11 @@ export declare class ComputerMidsceneTools extends BaseMidsceneTools<ComputerAge
     protected preparePlatformTools(): ToolDefinition[];
 }
 
+/** Init args for the RDP remote-desktop agent. */
+declare type ComputerRDPInitArgs = {
+    mode: 'rdp';
+} & RDPConnectionConfig;
+
 export declare function createDefaultRDPBackendClient(): RDPBackendClient;
 
 export declare interface DisplayInfo {
@@ -252,6 +281,7 @@ export declare class RDPDevice implements AbstractInterface {
     private destroyed;
     private cursorPosition?;
     uri?: string;
+    readonly inputPrimitives: ComputerInputPrimitives;
     constructor(options: RDPDeviceOpt);
     describe(): string;
     connect(): Promise<void>;
@@ -261,7 +291,6 @@ export declare class RDPDevice implements AbstractInterface {
     actionSpace(): DeviceAction<any>[];
     private assertConnected;
     private throwIfDestroyed;
-    private requireLocate;
     private moveToElement;
     private clearInput;
     private edgeScrollDirection;

package/dist/types/mcp-server.d.ts

@@ -2,12 +2,13 @@ import { AbstractInterface } from '@midscene/core/device';
 import { Agent } from '@midscene/core/agent';
 import { BaseMCPServer } from '@midscene/shared/mcp';
 import { BaseMidsceneTools } from '@midscene/shared/mcp/base-tools';
-import { DeviceAction } from '@midscene/core';
+import { ComputerInputPrimitives } from '@midscene/core/device';
+import type { DeviceAction } from '@midscene/core';
 import { InitArgSpec } from '@midscene/shared/mcp/base-tools';
-import { InterfaceType } from '@midscene/core';
+import type { InterfaceType } from '@midscene/core';
 import { LaunchMCPServerOptions } from '@midscene/shared/mcp';
 import { LaunchMCPServerResult } from '@midscene/shared/mcp';
-import { Size } from '@midscene/core';
+import type { Size } from '@midscene/core';
 import { Tool } from '@midscene/shared/mcp';
 import type { ToolDefinition } from '@midscene/shared/mcp/types';
 
@@ -28,6 +29,7 @@ declare class ComputerDevice implements AbstractInterface {
      */
     private useAppleScript;
     uri?: string;
+    readonly inputPrimitives: ComputerInputPrimitives;
     constructor(options?: ComputerDeviceOpt);
     describe(): string;
     /**
@@ -58,6 +60,9 @@ declare class ComputerDevice implements AbstractInterface {
      * which can swallow characters or convert them when a non-English IME is active.
      */
     private smartTypeString;
+    private selectAllAndDelete;
+    private pressKeyboardShortcut;
+    private performScroll;
     actionSpace(): DeviceAction<any>[];
     destroy(): Promise<void>;
     url(): Promise<string>;
@@ -85,10 +90,20 @@ declare interface ComputerDeviceOpt {
     xvfbResolution?: string;
 }
 
-declare type ComputerInitArgs = Pick<ComputerDeviceOpt, 'displayId' | 'headless'>;
+/**
+ * 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.
+ */
+declare type ComputerInitArgs = ComputerLocalInitArgs | ComputerRDPInitArgs;
 
 declare type ComputerInterface = ComputerDevice | RDPDevice;
 
+/** Init args for the local desktop agent (macOS/Windows/Linux). */
+declare type ComputerLocalInitArgs = {
+    mode: 'local';
+} & Pick<ComputerDeviceOpt, 'displayId' | 'headless'>;
+
 /**
  * Computer MCP Server
  * Provides MCP tools for computer desktop automation
@@ -113,6 +128,11 @@ declare class ComputerMidsceneTools extends BaseMidsceneTools<ComputerAgent, Com
     protected preparePlatformTools(): ToolDefinition[];
 }
 
+/** Init args for the RDP remote-desktop agent. */
+declare type ComputerRDPInitArgs = {
+    mode: 'rdp';
+} & RDPConnectionConfig;
+
 declare interface DisplayInfo {
     id: string;
     name: string;
@@ -177,6 +197,7 @@ declare class RDPDevice implements AbstractInterface {
     private destroyed;
     private cursorPosition?;
     uri?: string;
+    readonly inputPrimitives: ComputerInputPrimitives;
     constructor(options: RDPDeviceOpt);
     describe(): string;
     connect(): Promise<void>;
@@ -186,7 +207,6 @@ declare class RDPDevice implements AbstractInterface {
     actionSpace(): DeviceAction<any>[];
     private assertConnected;
     private throwIfDestroyed;
-    private requireLocate;
     private moveToElement;
     private clearInput;
     private edgeScrollDirection;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@midscene/computer",
-  "version": "1.8.0",
+  "version": "1.8.1",
   "description": "Midscene.js Computer Desktop Automation",
   "license": "MIT",
   "repository": {
@@ -38,8 +38,8 @@
     "@computer-use/libnut": "^4.2.0",
     "clipboardy": "^4.0.0",
     "screenshot-desktop": "^1.15.3",
-    "@midscene/shared": "1.8.0",
-    "@midscene/core": "1.8.0"
+    "@midscene/core": "1.8.1",
+    "@midscene/shared": "1.8.1"
   },
   "optionalDependencies": {
     "node-mac-permissions": "2.5.0"