@midscene/core 1.0.1-beta-20251024064637.0 → 1.0.1-beta-20251027033034.0

package/dist/types/agent/agent.d.ts

@@ -118,7 +118,7 @@ export declare class Agent<InterfaceType extends AbstractInterface = AbstractInt
         deepThink?: boolean;
     } & LocatorValidatorOption): Promise<AgentDescribeElementAtPointResult>;
     verifyLocator(prompt: string, locateOpt: LocateOption | undefined, expectCenter: [number, number], verifyLocateOption?: LocatorValidatorOption): Promise<LocateValidatorResult>;
-    aiLocate(prompt: TUserPrompt, opt?: LocateOption): Promise<Pick<LocateResultElement, "center" | "rect"> & {
+    aiLocate(prompt: TUserPrompt, opt?: LocateOption): Promise<Pick<LocateResultElement, "rect" | "center"> & {
         dpr?: number;
     }>;
     aiAssert(assertion: TUserPrompt, msg?: string, opt?: AgentAssertOpt & InsightExtractOption): Promise<{

package/dist/types/device/device-options.d.ts

@@ -0,0 +1,57 @@
+import type { DeviceAction } from '../types';
+/**
+ * Android device input options
+ */
+export type AndroidDeviceInputOpt = {
+    /** Automatically dismiss the keyboard after input is completed */
+    autoDismissKeyboard?: boolean;
+    /** Strategy for dismissing the keyboard: 'esc-first' tries ESC before BACK, 'back-first' tries BACK before ESC */
+    keyboardDismissStrategy?: 'esc-first' | 'back-first';
+};
+/**
+ * Android device options
+ */
+export type AndroidDeviceOpt = {
+    /** Path to the ADB executable */
+    androidAdbPath?: string;
+    /** Remote ADB host address */
+    remoteAdbHost?: string;
+    /** Remote ADB port */
+    remoteAdbPort?: number;
+    /** Input method editor strategy: 'always-yadb' always uses yadb, 'yadb-for-non-ascii' uses yadb only for non-ASCII characters */
+    imeStrategy?: 'always-yadb' | 'yadb-for-non-ascii';
+    /** Display ID to use for this device */
+    displayId?: number;
+    /** Use physical display ID for screenshot operations */
+    usePhysicalDisplayIdForScreenshot?: boolean;
+    /** Use physical display ID when looking up display information */
+    usePhysicalDisplayIdForDisplayLookup?: boolean;
+    /** Custom device actions to register */
+    customActions?: DeviceAction<any>[];
+    /** Screenshot resize scale factor */
+    screenshotResizeScale?: number;
+    /** Always fetch screen info on each call; if false, cache the first result */
+    alwaysRefreshScreenInfo?: boolean;
+} & AndroidDeviceInputOpt;
+/**
+ * iOS device input options
+ */
+export type IOSDeviceInputOpt = {
+    /** Automatically dismiss the keyboard after input is completed */
+    autoDismissKeyboard?: boolean;
+};
+/**
+ * iOS device options
+ */
+export type IOSDeviceOpt = {
+    /** Device ID (UDID) to connect to */
+    deviceId?: string;
+    /** Custom device actions to register */
+    customActions?: DeviceAction<any>[];
+    /** WebDriverAgent port (default: 8100) */
+    wdaPort?: number;
+    /** WebDriverAgent host (default: 'localhost') */
+    wdaHost?: string;
+    /** Whether to use WebDriverAgent */
+    useWDA?: boolean;
+} & IOSDeviceInputOpt;

package/dist/types/device/index.d.ts

@@ -1106,7 +1106,7 @@ export declare const actionScrollParamSchema: z.ZodObject<{
         xpath: z.ZodOptional<z.ZodUnion<[z.ZodString, z.ZodBoolean]>>;
     }, z.ZodTypeAny, "passthrough">>>;
 }, "strip", z.ZodTypeAny, {
-    direction: "down" | "up" | "right" | "left";
+    direction: "up" | "down" | "right" | "left";
     scrollType: "once" | "untilBottom" | "untilTop" | "untilRight" | "untilLeft";
     locate?: z.objectOutputType<{
         prompt: z.ZodUnion<[z.ZodString, z.ZodIntersection<z.ZodObject<{
@@ -1182,7 +1182,7 @@ export declare const actionScrollParamSchema: z.ZodObject<{
         cacheable: z.ZodOptional<z.ZodBoolean>;
         xpath: z.ZodOptional<z.ZodUnion<[z.ZodString, z.ZodBoolean]>>;
     }, z.ZodTypeAny, "passthrough"> | undefined;
-    direction?: "down" | "up" | "right" | "left" | undefined;
+    direction?: "up" | "down" | "right" | "left" | undefined;
     scrollType?: "once" | "untilBottom" | "untilTop" | "untilRight" | "untilLeft" | undefined;
     distance?: number | null | undefined;
 }>;
@@ -1848,7 +1848,7 @@ export declare const ActionSwipeParamSchema: z.ZodObject<{
 }, "strip", z.ZodTypeAny, {
     duration: number;
     repeat?: number | undefined;
-    direction?: "down" | "up" | "right" | "left" | undefined;
+    direction?: "up" | "down" | "right" | "left" | undefined;
     distance?: number | undefined;
     start?: z.objectOutputType<{
         prompt: z.ZodUnion<[z.ZodString, z.ZodIntersection<z.ZodObject<{
@@ -1924,7 +1924,7 @@ export declare const ActionSwipeParamSchema: z.ZodObject<{
     }, z.ZodTypeAny, "passthrough"> | undefined;
 }, {
     repeat?: number | undefined;
-    direction?: "down" | "up" | "right" | "left" | undefined;
+    direction?: "up" | "down" | "right" | "left" | undefined;
     distance?: number | undefined;
     duration?: number | undefined;
     start?: z.objectInputType<{
@@ -2156,3 +2156,4 @@ export type ActionClearInputParam = {
 };
 export declare const defineActionClearInput: (call: (param: ActionClearInputParam) => Promise<void>) => DeviceAction<ActionClearInputParam>;
 export type { DeviceAction } from '../types';
+export type { AndroidDeviceOpt, AndroidDeviceInputOpt, IOSDeviceOpt, IOSDeviceInputOpt, } from './device-options';

package/dist/types/types.d.ts

@@ -397,14 +397,9 @@ export interface AgentOpt {
      *
      * @example
      * ```typescript
-     * createOpenAIClient: (config) => {
-     *   const openai = new OpenAI({
-     *     apiKey: config.openaiApiKey,
-     *     baseURL: config.openaiBaseURL,
-     *   });
-     *
+     * createOpenAIClient: async (openai, opts) => {
      *   // Wrap with langsmith for planning tasks
-     *   if (config.intent === 'planning') {
+     *   if (opts.baseURL?.includes('planning')) {
      *     return wrapOpenAI(openai, { metadata: { task: 'planning' } });
      *   }
      *

package/dist/types/yaml.d.ts

@@ -1,4 +1,5 @@
 import type { TUserPrompt } from './ai-model/common';
+import type { AndroidDeviceOpt, IOSDeviceOpt } from './device';
 import type { AgentOpt, Rect } from './types';
 import type { UIContext } from './types';
 export interface LocateOption {
@@ -68,14 +69,11 @@ export interface MidsceneYamlScriptWebEnv extends MidsceneYamlScriptConfig, Mids
     bridgeMode?: false | 'newTabWithUrl' | 'currentTab';
     closeNewTabsAfterDisconnect?: boolean;
 }
-export interface MidsceneYamlScriptAndroidEnv extends MidsceneYamlScriptConfig {
+export interface MidsceneYamlScriptAndroidEnv extends MidsceneYamlScriptConfig, Omit<AndroidDeviceOpt, 'customActions'> {
     deviceId?: string;
     launch?: string;
 }
-export interface MidsceneYamlScriptIOSEnv extends MidsceneYamlScriptConfig {
-    wdaPort?: number;
-    wdaHost?: string;
-    autoDismissKeyboard?: boolean;
+export interface MidsceneYamlScriptIOSEnv extends MidsceneYamlScriptConfig, Omit<IOSDeviceOpt, 'customActions'> {
     launch?: string;
 }
 export type MidsceneYamlScriptEnv = MidsceneYamlScriptWebEnv | MidsceneYamlScriptAndroidEnv | MidsceneYamlScriptIOSEnv;

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@midscene/core",
   "description": "Automate browser actions, extract data, and perform assertions using AI. It offers JavaScript SDK, Chrome extension, and support for scripting in YAML. See https://midscenejs.com/ for details.",
-  "version": "1.0.1-beta-20251024064637.0",
+  "version": "1.0.1-beta-20251027033034.0",
   "repository": "https://github.com/web-infra-dev/midscene",
   "homepage": "https://midscenejs.com/",
   "main": "./dist/lib/index.js",
@@ -89,8 +89,8 @@
     "zod": "3.24.3",
     "semver": "7.5.2",
     "js-yaml": "4.1.0",
-    "@midscene/recorder": "1.0.1-beta-20251024064637.0",
-    "@midscene/shared": "1.0.1-beta-20251024064637.0"
+    "@midscene/recorder": "1.0.1-beta-20251027033034.0",
+    "@midscene/shared": "1.0.1-beta-20251027033034.0"
   },
   "devDependencies": {
     "@rslib/core": "^0.11.2",