@midscene/core 0.26.5-beta-20250814125155.0 → 0.26.6-beta-20250815074212.0

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

@@ -1,4 +1,4 @@
-import type { AIUsageInfo, BaseElement, ElementTreeNode, MidsceneYamlFlowItem, PlanningAction, Rect, Size } from '../types';
+import type { AIUsageInfo, BaseElement, DeviceAction, ElementTreeNode, MidsceneYamlFlowItem, PlanningAction, Rect, Size } from '../types';
 import type { ChatCompletionSystemMessageParam, ChatCompletionUserMessageParam } from 'openai/resources/index';
 import type { PlanningLocateParam } from '../types';
 export type AIArgs = [
@@ -12,6 +12,7 @@ export declare enum AIActionType {
     PLAN = 3,
     DESCRIBE_ELEMENT = 4
 }
+export declare const actionSpaceTypePrefix = "action_space_";
 export declare function callAiFn<T>(msgs: AIArgs, AIActionTypeValue: AIActionType): Promise<{
     content: T;
     usage?: AIUsageInfo;
@@ -31,4 +32,4 @@ export declare function mergeRects(rects: Rect[]): {
 };
 export declare function expandSearchArea(rect: Rect, screenSize: Size): Rect;
 export declare function markupImageForLLM(screenshotBase64: string, tree: ElementTreeNode<BaseElement>, size: Size): Promise<string>;
-export declare function buildYamlFlowFromPlans(plans: PlanningAction[], sleep?: number): MidsceneYamlFlowItem[];
+export declare function buildYamlFlowFromPlans(plans: PlanningAction[], actionSpace: DeviceAction[], sleep?: number): MidsceneYamlFlowItem[];

package/dist/types/types.d.ts

@@ -390,6 +390,7 @@ export type TUserPrompt = string | ({
 } & Partial<TMultimodalPrompt>);
 export interface DeviceAction<ParamType = any> {
     name: string;
+    interfaceAlias?: string;
     description?: string;
     paramSchema?: string;
     paramDescription?: string;

package/dist/types/yaml.d.ts

@@ -110,16 +110,15 @@ export interface MidsceneYamlFlowItemAIHover extends LocateOption {
     aiHover: TUserPrompt;
 }
 export interface MidsceneYamlFlowItemAIInput extends LocateOption {
-    aiInput: string;
-    locate: TUserPrompt;
+    aiInput: TUserPrompt | undefined;
+    value: string;
 }
 export interface MidsceneYamlFlowItemAIKeyboardPress extends LocateOption {
-    aiKeyboardPress: string;
-    locate?: TUserPrompt;
+    aiKeyboardPress: TUserPrompt | undefined;
+    key: string;
 }
 export interface MidsceneYamlFlowItemAIScroll extends LocateOption, ScrollParam {
-    aiScroll: null;
-    locate?: TUserPrompt;
+    aiScroll: TUserPrompt | undefined;
 }
 export interface MidsceneYamlFlowItemEvaluateJavaScript {
     javascript: string;

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": "0.26.5-beta-20250814125155.0",
+  "version": "0.26.6-beta-20250815074212.0",
   "repository": "https://github.com/web-infra-dev/midscene",
   "homepage": "https://midscenejs.com/",
   "main": "./dist/lib/index.js",
@@ -60,8 +60,8 @@
     "langsmith": "0.3.7",
     "openai": "4.81.0",
     "socks-proxy-agent": "8.0.4",
-    "@midscene/recorder": "0.26.5-beta-20250814125155.0",
-    "@midscene/shared": "0.26.5-beta-20250814125155.0"
+    "@midscene/recorder": "0.26.6-beta-20250815074212.0",
+    "@midscene/shared": "0.26.6-beta-20250815074212.0"
   },
   "devDependencies": {
     "@microsoft/api-extractor": "^7.52.10",