@midscene/core 1.2.2 → 1.2.3-beta-20260120082504.0

package/dist/types/agent/task-builder.d.ts

@@ -9,6 +9,7 @@ interface TaskBuilderDeps {
     service: Service;
     taskCache?: TaskCache;
     actionSpace: DeviceAction[];
+    waitAfterAction?: number;
 }
 interface BuildOptions {
     cacheable?: boolean;
@@ -19,7 +20,8 @@ export declare class TaskBuilder {
     private readonly service;
     private readonly taskCache?;
     private readonly actionSpace;
-    constructor({ interfaceInstance, service, taskCache, actionSpace, }: TaskBuilderDeps);
+    private readonly waitAfterAction?;
+    constructor({ interfaceInstance, service, taskCache, actionSpace, waitAfterAction, }: TaskBuilderDeps);
     build(plans: PlanningAction[], modelConfigForPlanning: IModelConfig, modelConfigForDefaultIntent: IModelConfig, options?: BuildOptions): Promise<{
         tasks: ExecutionTaskApply[];
     }>;

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

@@ -26,11 +26,13 @@ export declare class TaskExecutor {
     onTaskStartCallback?: ExecutionTaskProgressOptions['onTaskStart'];
     private readonly hooks?;
     replanningCycleLimit?: number;
+    waitAfterAction?: number;
     get page(): AbstractInterface;
     constructor(interfaceInstance: AbstractInterface, service: Service, opts: {
         taskCache?: TaskCache;
         onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];
         replanningCycleLimit?: number;
+        waitAfterAction?: number;
         hooks?: TaskExecutorHooks;
         actionSpace: DeviceAction[];
     });

package/dist/types/ai-model/inspect.d.ts

@@ -1,4 +1,4 @@
-import type { AIDataExtractionResponse, AIElementResponse, AIUsageInfo, Rect, ServiceExtractOption, UIContext } from '../types';
+import type { AIElementResponse, AIUsageInfo, Rect, ServiceExtractOption, UIContext } from '../types';
 import type { IModelConfig } from '@midscene/shared/env';
 import type { LocateResultElement } from '@midscene/shared/types';
 import type { ChatCompletionSystemMessageParam, ChatCompletionUserMessageParam } from 'openai/resources/index';
@@ -43,7 +43,7 @@ export declare function AiExtractElementInfo<T>(options: {
     extractOption?: ServiceExtractOption;
     modelConfig: IModelConfig;
 }): Promise<{
-    parseResult: AIDataExtractionResponse<T>;
+    parseResult: import("../types").AIDataExtractionResponse<T>;
     usage: AIUsageInfo | undefined;
     reasoning_content: string | undefined;
 }>;

package/dist/types/ai-model/llm-planning.d.ts

@@ -1,6 +1,10 @@
-import type { DeepThinkOption, DeviceAction, InterfaceType, PlanningAIResponse, UIContext } from '../types';
-import type { IModelConfig } from '@midscene/shared/env';
+import type { DeepThinkOption, DeviceAction, InterfaceType, PlanningAIResponse, RawResponsePlanningAIResponse, UIContext } from '../types';
+import type { IModelConfig, TModelFamily } from '@midscene/shared/env';
 import type { ConversationHistory } from './conversation-history';
+/**
+ * Parse XML response from LLM and convert to RawResponsePlanningAIResponse
+ */
+export declare function parseXMLPlanningResponse(xmlString: string, modelFamily: TModelFamily | undefined): RawResponsePlanningAIResponse;
 export declare function plan(userInstruction: string, opts: {
     context: UIContext;
     interfaceType: InterfaceType;

package/dist/types/ai-model/prompt/extraction.d.ts

@@ -1,4 +1,7 @@
-import type { ResponseFormatJSONSchema } from 'openai/resources/index';
+import type { AIDataExtractionResponse } from '../../types';
+/**
+ * Parse XML response from LLM and convert to AIDataExtractionResponse
+ */
+export declare function parseXMLExtractionResponse<T>(xmlString: string): AIDataExtractionResponse<T>;
 export declare function systemPromptToExtract(): string;
 export declare const extractDataQueryPrompt: (pageDescription: string, dataQuery: string | Record<string, string>) => string;
-export declare const extractDataSchema: ResponseFormatJSONSchema;

package/dist/types/ai-model/prompt/util.d.ts

@@ -1,4 +1,11 @@
 import type { BaseElement, Size, UIContext } from '../../types';
+/**
+ * Extract content from an XML tag in a string
+ * @param xmlString - The XML string to parse
+ * @param tagName - The name of the tag to extract (case-insensitive)
+ * @returns The trimmed content of the tag, or undefined if not found
+ */
+export declare function extractXMLTag(xmlString: string, tagName: string): string | undefined;
 export declare function describeSize(size: Size): string;
 export declare function describeElement(elements: (Pick<BaseElement, 'rect' | 'content'> & {
     id: string;

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

@@ -2228,16 +2228,5 @@ export type ActionSleepParam = {
     millisecond?: number;
 };
 export declare const defineActionSleep: () => DeviceAction<ActionSleepParam>;
-export declare const actionFinalizeParamSchema: z.ZodObject<{
-    message: z.ZodOptional<z.ZodString>;
-}, "strip", z.ZodTypeAny, {
-    message?: string | undefined;
-}, {
-    message?: string | undefined;
-}>;
-export type ActionFinalizeParam = {
-    message?: string;
-};
-export declare const defineActionFinalize: () => DeviceAction<ActionFinalizeParam>;
 export type { DeviceAction } from '../types';
 export type { AndroidDeviceOpt, AndroidDeviceInputOpt, IOSDeviceOpt, IOSDeviceInputOpt, } from './device-options';

package/dist/types/types.d.ts

@@ -174,6 +174,8 @@ export interface RawResponsePlanningAIResponse {
     log: string;
     note?: string;
     error?: string;
+    finalizeMessage?: string;
+    finalizeSuccess?: boolean;
 }
 export interface PlanningAIResponse extends Omit<RawResponsePlanningAIResponse, 'action'> {
     actions?: PlanningAction[];
@@ -456,6 +458,12 @@ export interface AgentOpt {
      * If omitted, the agent will also read `MIDSCENE_REPLANNING_CYCLE_LIMIT` for backward compatibility.
      */
     replanningCycleLimit?: number;
+    /**
+     * Wait time in milliseconds after each action execution.
+     * This allows the UI to settle and stabilize before the next action.
+     * Defaults to 300ms when not provided.
+     */
+    waitAfterAction?: number;
     /**
      * Custom OpenAI client factory function
      *

package/dist/types/yaml.d.ts

@@ -1,6 +1,6 @@
 import type { TUserPrompt } from './common';
 import type { AndroidDeviceOpt, IOSDeviceOpt } from './device';
-import type { AgentOpt, LocateResultElement, Rect } from './types';
+import type { AgentOpt, LocateResultElement } from './types';
 import type { UIContext } from './types';
 export interface LocateOption {
     prompt?: TUserPrompt;
@@ -15,10 +15,6 @@ export interface ServiceExtractOption {
     screenshotIncluded?: boolean;
     [key: string]: unknown;
 }
-export interface ReferenceImage {
-    base64: string;
-    rect?: Rect;
-}
 export interface DetailedLocateParam extends LocateOption {
     prompt: TUserPrompt;
 }

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": "1.2.2",
+  "version": "1.2.3-beta-20260120082504.0",
   "repository": "https://github.com/web-infra-dev/midscene",
   "homepage": "https://midscenejs.com/",
   "main": "./dist/lib/index.js",
@@ -89,7 +89,7 @@
     "semver": "7.5.2",
     "undici": "^6.0.0",
     "zod": "3.24.3",
-    "@midscene/shared": "1.2.2"
+    "@midscene/shared": "1.2.3-beta-20260120082504.0"
   },
   "devDependencies": {
     "@rslib/core": "^0.18.3",