@midscene/core 0.9.1 → 0.9.2-beta-20250114083542.0

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

@@ -1,8 +1,9 @@
-import { g as AIUsageInfo } from './types-05c1f241.js';
+import { g as AIUsageInfo, y as PlanningAction } from './types-c4bec333.js';
 import { ChatCompletionMessageParam } from 'openai/resources';
 export { ChatCompletionMessageParam } from 'openai/resources';
-import { A as AIActionType } from './automation-81d96430.js';
-export { f as AiAssert, e as AiExtractElementInfo, b as AiInspectElement, c as callAiFn, d as describeUserPage, p as plan, t as transformElementPositionToId } from './automation-81d96430.js';
+import { A as AIActionType } from './automation-d7e10a4e.js';
+export { f as AiAssert, e as AiExtractElementInfo, b as AiInspectElement, c as callAiFn, d as describeUserPage, p as plan, t as transformElementPositionToId } from './automation-d7e10a4e.js';
+import '@midscene/shared/constants';
 
 declare function callToGetJSONObject<T>(messages: ChatCompletionMessageParam[], AIActionTypeValue: AIActionType): Promise<{
     content: T;
@@ -11,4 +12,59 @@ declare function callToGetJSONObject<T>(messages: ChatCompletionMessageParam[],
 
 declare function systemPromptToFindElement(): string;
 
-export { callToGetJSONObject, systemPromptToFindElement };
+type ActionType = 'click' | 'type' | 'hotkey' | 'finished' | 'scroll' | 'wait';
+declare function vlmPlanning(options: {
+    userInstruction: string;
+    conversationHistory: ChatCompletionMessageParam[];
+    size: {
+        width: number;
+        height: number;
+    };
+}): Promise<{
+    actions: PlanningAction<any>[];
+    realActions: Array<Action>;
+    action_summary: string;
+}>;
+interface BaseAction {
+    action_type: ActionType;
+    action_inputs: Record<string, any>;
+    reflection: string | null;
+    thought: string | null;
+}
+interface ClickAction extends BaseAction {
+    action_type: 'click';
+    action_inputs: {
+        start_box: string;
+    };
+}
+interface WaitAction extends BaseAction {
+    action_type: 'wait';
+    action_inputs: {
+        time: string;
+    };
+}
+interface TypeAction extends BaseAction {
+    action_type: 'type';
+    action_inputs: {
+        content: string;
+    };
+}
+interface HotkeyAction extends BaseAction {
+    action_type: 'hotkey';
+    action_inputs: {
+        key: string;
+    };
+}
+interface ScrollAction extends BaseAction {
+    action_type: 'scroll';
+    action_inputs: {
+        direction: 'up' | 'down';
+    };
+}
+interface FinishedAction extends BaseAction {
+    action_type: 'finished';
+    action_inputs: Record<string, never>;
+}
+type Action = ClickAction | TypeAction | HotkeyAction | ScrollAction | FinishedAction | WaitAction;
+
+export { callToGetJSONObject, systemPromptToFindElement, vlmPlanning };

package/dist/lib/types/{automation-81d96430.d.ts → automation-d7e10a4e.d.ts}

@@ -1,4 +1,4 @@
-import { g as AIUsageInfo, B as BaseElement, U as UIContext, l as AIElementResponse, A as AISingleElementResponse, u as ElementById, m as AISectionParseResponse, n as AIAssertionResponse, z as PlanningAIResponse } from './types-05c1f241.js';
+import { g as AIUsageInfo, B as BaseElement, U as UIContext, l as AIElementResponse, A as AISingleElementResponse, j as AISingleElementResponseByPosition, u as ElementById, m as AISectionParseResponse, n as AIAssertionResponse, z as PlanningAIResponse } from './types-c4bec333.js';
 import { ChatCompletionSystemMessageParam, ChatCompletionUserMessageParam } from 'openai/resources';
 
 type AIArgs = [
@@ -29,6 +29,10 @@ declare function describeUserPage<ElementType extends BaseElement = BaseElement>
         width: number;
         height: number;
     }): BaseElement | undefined;
+    insertElementByPosition(position: {
+        x: number;
+        y: number;
+    }): ElementType;
     size: {
         width: number;
         height: number;
@@ -61,7 +65,7 @@ declare function AiInspectElement<ElementType extends BaseElement = BaseElement>
     multi: boolean;
     targetElementDescription: string;
     callAI?: typeof callAiFn<AIElementResponse | [number, number]>;
-    quickAnswer?: AISingleElementResponse;
+    quickAnswer?: Partial<AISingleElementResponse | AISingleElementResponseByPosition>;
 }): Promise<{
     parseResult: AIElementResponse;
     rawResponse: any;

package/dist/lib/types/env.d.ts

@@ -10,6 +10,7 @@ declare const OPENAI_BASE_URL = "OPENAI_BASE_URL";
 declare const OPENAI_MAX_TOKENS = "OPENAI_MAX_TOKENS";
 declare const MIDSCENE_MODEL_TEXT_ONLY = "MIDSCENE_MODEL_TEXT_ONLY";
 declare const MIDSCENE_CACHE = "MIDSCENE_CACHE";
+declare const MIDSCENE_USE_VLM_UI_TARS = "MIDSCENE_USE_VLM_UI_TARS";
 declare const MATCH_BY_POSITION = "MATCH_BY_POSITION";
 declare const MIDSCENE_API_TYPE = "MIDSCENE-API-TYPE";
 declare const MIDSCENE_REPORT_TAG_NAME = "MIDSCENE_REPORT_TAG_NAME";
@@ -43,6 +44,7 @@ declare const allConfigFromEnv: () => {
     MIDSCENE_AZURE_OPENAI_SCOPE: string | undefined;
     MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON: string | undefined;
     MIDSCENE_USE_ANTHROPIC_SDK: string | undefined;
+    MIDSCENE_USE_VLM_UI_TARS: string | undefined;
     ANTHROPIC_API_KEY: string | undefined;
     AZURE_OPENAI_ENDPOINT: string | undefined;
     AZURE_OPENAI_KEY: string | undefined;
@@ -72,6 +74,7 @@ declare const allAIConfig: () => {
     MIDSCENE_AZURE_OPENAI_SCOPE: string | undefined;
     MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON: string | undefined;
     MIDSCENE_USE_ANTHROPIC_SDK: string | undefined;
+    MIDSCENE_USE_VLM_UI_TARS: string | undefined;
     ANTHROPIC_API_KEY: string | undefined;
     AZURE_OPENAI_ENDPOINT: string | undefined;
     AZURE_OPENAI_KEY: string | undefined;
@@ -80,4 +83,4 @@ declare const allAIConfig: () => {
 };
 declare const overrideAIConfig: (newConfig: ReturnType<typeof allConfigFromEnv>, extendMode?: boolean) => void;
 
-export { ANTHROPIC_API_KEY, AZURE_OPENAI_API_VERSION, AZURE_OPENAI_DEPLOYMENT, AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_KEY, MATCH_BY_POSITION, MIDSCENE_API_TYPE, MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON, MIDSCENE_AZURE_OPENAI_SCOPE, MIDSCENE_CACHE, MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG, MIDSCENE_DEBUG_AI_PROFILE, MIDSCENE_DEBUG_MODE, MIDSCENE_LANGSMITH_DEBUG, MIDSCENE_MODEL_NAME, MIDSCENE_MODEL_TEXT_ONLY, MIDSCENE_OPENAI_INIT_CONFIG_JSON, MIDSCENE_OPENAI_SOCKS_PROXY, MIDSCENE_REPORT_TAG_NAME, MIDSCENE_USE_ANTHROPIC_SDK, MIDSCENE_USE_AZURE_OPENAI, OPENAI_API_KEY, OPENAI_BASE_URL, OPENAI_MAX_TOKENS, OPENAI_USE_AZURE, allAIConfig, getAIConfig, getAIConfigInJson, overrideAIConfig };
+export { ANTHROPIC_API_KEY, AZURE_OPENAI_API_VERSION, AZURE_OPENAI_DEPLOYMENT, AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_KEY, MATCH_BY_POSITION, MIDSCENE_API_TYPE, MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON, MIDSCENE_AZURE_OPENAI_SCOPE, MIDSCENE_CACHE, MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG, MIDSCENE_DEBUG_AI_PROFILE, MIDSCENE_DEBUG_MODE, MIDSCENE_LANGSMITH_DEBUG, MIDSCENE_MODEL_NAME, MIDSCENE_MODEL_TEXT_ONLY, MIDSCENE_OPENAI_INIT_CONFIG_JSON, MIDSCENE_OPENAI_SOCKS_PROXY, MIDSCENE_REPORT_TAG_NAME, MIDSCENE_USE_ANTHROPIC_SDK, MIDSCENE_USE_AZURE_OPENAI, MIDSCENE_USE_VLM_UI_TARS, OPENAI_API_KEY, OPENAI_BASE_URL, OPENAI_MAX_TOKENS, OPENAI_USE_AZURE, allAIConfig, getAIConfig, getAIConfigInJson, overrideAIConfig };

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

@@ -1,9 +1,9 @@
-import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, B as BaseElement, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightOptions, e as InsightTaskInfo, A as AISingleElementResponse, f as InsightAssertionResponse } from './types-05c1f241.js';
-export { n as AIAssertionResponse, k as AIElementIdResponse, l as AIElementResponse, h as AIResponseFormat, m as AISectionParseResponse, i as AISingleElementResponseById, j as AISingleElementResponseByPosition, g as AIUsageInfo, w as AgentAssertOpt, v as AgentWaitForOpt, X as BaseAgentParserOpt, C as CallAIFn, W as Color, q as DumpMeta, u as ElementById, o as EnsureObject, _ as ExecutionRecorderItem, ag as ExecutionTaskAction, af as ExecutionTaskActionApply, ae as ExecutionTaskInsightAssertion, ad as ExecutionTaskInsightAssertionApply, ac as ExecutionTaskInsightAssertionParam, a5 as ExecutionTaskInsightDumpLog, a7 as ExecutionTaskInsightLocate, a6 as ExecutionTaskInsightLocateApply, a4 as ExecutionTaskInsightLocateOutput, a3 as ExecutionTaskInsightLocateParam, ab as ExecutionTaskInsightQuery, aa as ExecutionTaskInsightQueryApply, a9 as ExecutionTaskInsightQueryOutput, a8 as ExecutionTaskInsightQueryParam, ai as ExecutionTaskPlanning, ah as ExecutionTaskPlanningApply, a2 as ExecutionTaskReturn, $ as ExecutionTaskType, a0 as ExecutorContext, aj as GroupedActionDump, s as InsightDump, p as InsightExtractParam, L as LiteUISection, O as OnTaskStartTip, t as PartialInsightDumpFromSDK, z as PlanningAIResponse, y as PlanningAction, N as PlanningActionParamAssert, T as PlanningActionParamError, J as PlanningActionParamHover, K as PlanningActionParamInputOrKeyPress, G as PlanningActionParamPlan, M as PlanningActionParamScroll, Q as PlanningActionParamSleep, H as PlanningActionParamTap, V as PlanningActionParamWaitFor, F as PlanningFurtherPlan, x as PlanningLocateParam, Z as PlaywrightParserOpt, P as Point, Y as PuppeteerParserOpt, R as Rect, r as ReportDumpWithAttributes, S as Size, a1 as TaskCacheInfo } from './types-05c1f241.js';
-export { allAIConfig, getAIConfig, overrideAIConfig } from './env.js';
-import { c as callAiFn, r as retrieveElement, a as retrieveSection } from './automation-81d96430.js';
-export { p as plan, t as transformElementPositionToId } from './automation-81d96430.js';
+import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, B as BaseElement, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightOptions, e as InsightTaskInfo, f as InsightAssertionResponse, A as AISingleElementResponse } from './types-c4bec333.js';
+export { n as AIAssertionResponse, k as AIElementIdResponse, l as AIElementResponse, h as AIResponseFormat, m as AISectionParseResponse, i as AISingleElementResponseById, j as AISingleElementResponseByPosition, g as AIUsageInfo, w as AgentAssertOpt, v as AgentWaitForOpt, X as BaseAgentParserOpt, C as CallAIFn, W as Color, q as DumpMeta, u as ElementById, o as EnsureObject, _ as ExecutionRecorderItem, ag as ExecutionTaskAction, af as ExecutionTaskActionApply, ae as ExecutionTaskInsightAssertion, ad as ExecutionTaskInsightAssertionApply, ac as ExecutionTaskInsightAssertionParam, a5 as ExecutionTaskInsightDumpLog, a7 as ExecutionTaskInsightLocate, a6 as ExecutionTaskInsightLocateApply, a4 as ExecutionTaskInsightLocateOutput, a3 as ExecutionTaskInsightLocateParam, ab as ExecutionTaskInsightQuery, aa as ExecutionTaskInsightQueryApply, a9 as ExecutionTaskInsightQueryOutput, a8 as ExecutionTaskInsightQueryParam, ai as ExecutionTaskPlanning, ah as ExecutionTaskPlanningApply, a2 as ExecutionTaskReturn, $ as ExecutionTaskType, a0 as ExecutorContext, aj as GroupedActionDump, s as InsightDump, p as InsightExtractParam, L as LiteUISection, O as OnTaskStartTip, t as PartialInsightDumpFromSDK, z as PlanningAIResponse, y as PlanningAction, N as PlanningActionParamAssert, T as PlanningActionParamError, J as PlanningActionParamHover, K as PlanningActionParamInputOrKeyPress, G as PlanningActionParamPlan, M as PlanningActionParamScroll, Q as PlanningActionParamSleep, H as PlanningActionParamTap, V as PlanningActionParamWaitFor, F as PlanningFurtherPlan, x as PlanningLocateParam, Z as PlaywrightParserOpt, P as Point, Y as PuppeteerParserOpt, R as Rect, r as ReportDumpWithAttributes, S as Size, a1 as TaskCacheInfo } from './types-c4bec333.js';
+import { c as callAiFn, r as retrieveElement, a as retrieveSection } from './automation-d7e10a4e.js';
+export { p as plan, t as transformElementPositionToId } from './automation-d7e10a4e.js';
 export { getLogDirByType, getVersion, setLogDir } from './utils.js';
+import '@midscene/shared/constants';
 import 'openai/resources';
 
 interface MidsceneYamlScript {
@@ -97,6 +97,11 @@ declare class Executor {
     dump(): ExecutionDump;
 }
 
+interface LocateOpts {
+    multi?: boolean;
+    callAI?: typeof callAiFn<AIElementResponse>;
+    quickAnswer?: Partial<AISingleElementResponse>;
+}
 declare class Insight<ElementType extends BaseElement = BaseElement, ContextType extends UIContext<ElementType> = UIContext<ElementType>> {
     contextRetrieverFn: (action: InsightAction) => Promise<ContextType> | ContextType;
     aiVendorFn: (...args: Array<any>) => Promise<any>;
@@ -104,10 +109,7 @@ declare class Insight<ElementType extends BaseElement = BaseElement, ContextType
     generateElement: InsightOptions['generateElement'];
     taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;
     constructor(context: ContextType | ((action: InsightAction) => Promise<ContextType> | ContextType), opt?: InsightOptions);
-    locate(queryPrompt: string, opt?: {
-        callAI?: typeof callAiFn<AIElementResponse>;
-        quickAnswer?: AISingleElementResponse | null;
-    }): Promise<ElementType | null>;
+    locate(queryPrompt: string, opt?: LocateOpts): Promise<ElementType | null>;
     locate(queryPrompt: string, opt: {
         multi: true;
     }): Promise<ElementType[]>;

package/dist/lib/types/{types-05c1f241.d.ts → types-c4bec333.d.ts}

@@ -1,3 +1,4 @@
+import { NodeType } from '@midscene/shared/constants';
 import { ChatCompletionMessageParam } from 'openai/resources';
 
 interface Point {
@@ -12,13 +13,6 @@ interface Size {
 type Rect = Point & Size & {
     zoom?: number;
 };
-declare enum NodeType {
-    CONTAINER = "CONTAINER Node",
-    FORM_ITEM = "FORM_ITEM Node",
-    BUTTON = "BUTTON Node",
-    IMG = "IMG Node",
-    TEXT = "TEXT Node"
-}
 declare abstract class BaseElement {
     abstract id: string;
     abstract attributes: {
@@ -122,7 +116,7 @@ interface InsightDump extends DumpMeta {
         sections?: Record<string, string>;
         assertion?: string;
     };
-    quickAnswer?: AISingleElementResponse | null;
+    quickAnswer?: Partial<AISingleElementResponse> | null;
     matchedSection: [];
     matchedElement: BaseElement[];
     data: any;
@@ -166,14 +160,14 @@ interface PlanningLocateParam {
 }
 interface PlanningAction<ParamType = any> {
     thought?: string;
-    type: 'Locate' | 'Tap' | 'Hover' | 'Input' | 'KeyboardPress' | 'Scroll' | 'Error' | 'FalsyConditionStatement' | 'Assert' | 'AssertWithoutThrow' | 'Sleep';
+    type: 'Locate' | 'Tap' | 'Hover' | 'Input' | 'KeyboardPress' | 'Scroll' | 'Error' | 'FalsyConditionStatement' | 'Assert' | 'AssertWithoutThrow' | 'Sleep' | 'Finished';
     param: ParamType;
     locate: PlanningLocateParam | null;
 }
 interface PlanningAIResponse {
     actions: PlanningAction[];
     taskWillBeAccomplished: boolean;
-    furtherPlan: PlanningFurtherPlan | null;
+    furtherPlan?: PlanningFurtherPlan | null;
     error?: string;
 }
 interface PlanningFurtherPlan {
@@ -258,6 +252,7 @@ type ExecutionTask<E extends ExecutionTaskApply<any, any, any> = ExecutionTaskAp
         start: number;
         end?: number;
         cost?: number;
+        aiCost?: number;
     };
 };
 interface ExecutionDump extends DumpMeta {

package/dist/lib/types/utils.d.ts

@@ -1,4 +1,5 @@
-import { r as ReportDumpWithAttributes, R as Rect } from './types-05c1f241.js';
+import { r as ReportDumpWithAttributes, R as Rect } from './types-c4bec333.js';
+import '@midscene/shared/constants';
 import 'openai/resources';
 
 declare const insightDumpFileExt = "insight-dump.json";

package/dist/lib/utils.js

@@ -1,370 +1,40 @@
-"use strict";
-var __create = Object.create;
-var __defProp = Object.defineProperty;
-var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
-var __getOwnPropNames = Object.getOwnPropertyNames;
-var __getProtoOf = Object.getPrototypeOf;
-var __hasOwnProp = Object.prototype.hasOwnProperty;
-var __export = (target, all) => {
-  for (var name in all)
-    __defProp(target, name, { get: all[name], enumerable: true });
-};
-var __copyProps = (to, from, except, desc) => {
-  if (from && typeof from === "object" || typeof from === "function") {
-    for (let key of __getOwnPropNames(from))
-      if (!__hasOwnProp.call(to, key) && key !== except)
-        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
-  }
-  return to;
-};
-var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(
-  // If the importer is in node compatibility mode or this is not an ESM
-  // file that has been converted to a CommonJS file using a Babel-
-  // compatible transform (i.e. "__esModule" has not been set), then set
-  // "default" to the CommonJS "module.exports" for node compatibility.
-  isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", { value: mod, enumerable: true }) : target,
-  mod
-));
-var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
-
-// src/utils.ts
-var utils_exports = {};
-__export(utils_exports, {
-  getLogDir: () => getLogDir,
-  getLogDirByType: () => getLogDirByType,
-  getTmpDir: () => getTmpDir,
-  getTmpFile: () => getTmpFile,
-  getVersion: () => getVersion,
-  groupedActionDumpFileExt: () => groupedActionDumpFileExt,
-  insightDumpFileExt: () => insightDumpFileExt,
-  overlapped: () => overlapped,
-  replaceStringWithFirstAppearance: () => replaceStringWithFirstAppearance,
-  replacerForPageObject: () => replacerForPageObject,
-  reportHTMLContent: () => reportHTMLContent,
-  setLogDir: () => setLogDir,
-  sleep: () => sleep,
-  stringifyDumpData: () => stringifyDumpData,
-  uploadTestInfoToServer: () => uploadTestInfoToServer,
-  writeDumpReport: () => writeDumpReport,
-  writeLogFile: () => writeLogFile
-});
-module.exports = __toCommonJS(utils_exports);
-var import_node_assert = __toESM(require("assert"));
-var import_node_child_process = require("child_process");
-var import_node_fs = require("fs");
-var import_node_os = require("os");
-var import_node_path = require("path");
-var import_fs = require("@midscene/shared/fs");
-var import_utils = require("@midscene/shared/utils");
-
-// src/env.ts
-var MIDSCENE_OPENAI_INIT_CONFIG_JSON = "MIDSCENE_OPENAI_INIT_CONFIG_JSON";
-var MIDSCENE_MODEL_NAME = "MIDSCENE_MODEL_NAME";
-var MIDSCENE_LANGSMITH_DEBUG = "MIDSCENE_LANGSMITH_DEBUG";
-var MIDSCENE_DEBUG_AI_PROFILE = "MIDSCENE_DEBUG_AI_PROFILE";
-var MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG = "MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG";
-var MIDSCENE_DEBUG_MODE = "MIDSCENE_DEBUG_MODE";
-var MIDSCENE_OPENAI_SOCKS_PROXY = "MIDSCENE_OPENAI_SOCKS_PROXY";
-var OPENAI_API_KEY = "OPENAI_API_KEY";
-var OPENAI_BASE_URL = "OPENAI_BASE_URL";
-var OPENAI_MAX_TOKENS = "OPENAI_MAX_TOKENS";
-var MIDSCENE_MODEL_TEXT_ONLY = "MIDSCENE_MODEL_TEXT_ONLY";
-var MIDSCENE_CACHE = "MIDSCENE_CACHE";
-var MATCH_BY_POSITION = "MATCH_BY_POSITION";
-var MIDSCENE_REPORT_TAG_NAME = "MIDSCENE_REPORT_TAG_NAME";
-var MIDSCENE_USE_AZURE_OPENAI = "MIDSCENE_USE_AZURE_OPENAI";
-var MIDSCENE_AZURE_OPENAI_SCOPE = "MIDSCENE_AZURE_OPENAI_SCOPE";
-var MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON = "MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON";
-var AZURE_OPENAI_ENDPOINT = "AZURE_OPENAI_ENDPOINT";
-var AZURE_OPENAI_KEY = "AZURE_OPENAI_KEY";
-var AZURE_OPENAI_API_VERSION = "AZURE_OPENAI_API_VERSION";
-var AZURE_OPENAI_DEPLOYMENT = "AZURE_OPENAI_DEPLOYMENT";
-var MIDSCENE_USE_ANTHROPIC_SDK = "MIDSCENE_USE_ANTHROPIC_SDK";
-var ANTHROPIC_API_KEY = "ANTHROPIC_API_KEY";
-var OPENAI_USE_AZURE = "OPENAI_USE_AZURE";
-var allConfigFromEnv = () => {
-  return {
-    [MIDSCENE_OPENAI_INIT_CONFIG_JSON]: process.env[MIDSCENE_OPENAI_INIT_CONFIG_JSON] || void 0,
-    [MIDSCENE_MODEL_NAME]: process.env[MIDSCENE_MODEL_NAME] || void 0,
-    [MIDSCENE_DEBUG_MODE]: process.env[MIDSCENE_DEBUG_MODE] || void 0,
-    [MIDSCENE_LANGSMITH_DEBUG]: process.env[MIDSCENE_LANGSMITH_DEBUG] || void 0,
-    [MIDSCENE_DEBUG_AI_PROFILE]: process.env[MIDSCENE_DEBUG_AI_PROFILE] || void 0,
-    [MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG]: process.env[MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG] || void 0,
-    [OPENAI_API_KEY]: process.env[OPENAI_API_KEY] || void 0,
-    [OPENAI_BASE_URL]: process.env[OPENAI_BASE_URL] || void 0,
-    [MIDSCENE_MODEL_TEXT_ONLY]: process.env[MIDSCENE_MODEL_TEXT_ONLY] || void 0,
-    [OPENAI_MAX_TOKENS]: process.env[OPENAI_MAX_TOKENS] || void 0,
-    [OPENAI_USE_AZURE]: process.env[OPENAI_USE_AZURE] || void 0,
-    [MIDSCENE_CACHE]: process.env[MIDSCENE_CACHE] || void 0,
-    [MATCH_BY_POSITION]: process.env[MATCH_BY_POSITION] || void 0,
-    [MIDSCENE_REPORT_TAG_NAME]: process.env[MIDSCENE_REPORT_TAG_NAME] || void 0,
-    [MIDSCENE_OPENAI_SOCKS_PROXY]: process.env[MIDSCENE_OPENAI_SOCKS_PROXY] || void 0,
-    [MIDSCENE_USE_AZURE_OPENAI]: process.env[MIDSCENE_USE_AZURE_OPENAI] || void 0,
-    [MIDSCENE_AZURE_OPENAI_SCOPE]: process.env[MIDSCENE_AZURE_OPENAI_SCOPE] || void 0,
-    [MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON]: process.env[MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON] || void 0,
-    [MIDSCENE_USE_ANTHROPIC_SDK]: process.env[MIDSCENE_USE_ANTHROPIC_SDK] || void 0,
-    [ANTHROPIC_API_KEY]: process.env[ANTHROPIC_API_KEY] || void 0,
-    [AZURE_OPENAI_ENDPOINT]: process.env[AZURE_OPENAI_ENDPOINT] || void 0,
-    [AZURE_OPENAI_KEY]: process.env[AZURE_OPENAI_KEY] || void 0,
-    [AZURE_OPENAI_API_VERSION]: process.env[AZURE_OPENAI_API_VERSION] || void 0,
-    [AZURE_OPENAI_DEPLOYMENT]: process.env[AZURE_OPENAI_DEPLOYMENT] || void 0
-  };
-};
-var userConfig = {};
-var getAIConfig = (configKey) => {
-  if (typeof userConfig[configKey] !== "undefined") {
-    return userConfig[configKey];
-  }
-  return allConfigFromEnv()[configKey];
-};
-var getAIConfigInJson = (configKey) => {
-  const config = getAIConfig(configKey);
-  try {
-    return config ? JSON.parse(config) : void 0;
-  } catch (error) {
-    throw new Error(
-      `Failed to parse json config: ${configKey}. ${error.message}`,
-      {
-        cause: error
-      }
-    );
-  }
-};
-
-// src/utils.ts
-var logDir = (0, import_node_path.join)(process.cwd(), "./midscene_run/");
-var logEnvReady = false;
-var insightDumpFileExt = "insight-dump.json";
-var groupedActionDumpFileExt = "web-dump.json";
-function getLogDir() {
-  return logDir;
-}
-function setLogDir(dir) {
-  logDir = dir;
-}
-function getLogDirByType(type) {
-  const dir = (0, import_node_path.join)(getLogDir(), type);
-  if (!(0, import_node_fs.existsSync)(dir)) {
-    (0, import_node_fs.mkdirSync)(dir, { recursive: true });
-  }
-  return dir;
-}
-var reportTpl = null;
-function getReportTpl() {
-  if (import_utils.ifInBrowser) {
-    if (!reportTpl && window.get_midscene_report_tpl) {
-      reportTpl = window.get_midscene_report_tpl();
-    }
-    (0, import_node_assert.default)(
-      reportTpl,
-      "reportTpl should be set before writing report in browser"
-    );
-    return reportTpl;
-  }
-  if (!reportTpl) {
-    let reportPath = (0, import_node_path.join)(__dirname, "../../report/index.html");
-    if (!(0, import_node_fs.existsSync)(reportPath)) {
-      reportPath = (0, import_node_path.join)(__dirname, "../report/index.html");
-    }
-    reportTpl = (0, import_node_fs.readFileSync)(reportPath, "utf-8");
-  }
-  return reportTpl;
-}
-function replaceStringWithFirstAppearance(str, target, replacement) {
-  const index = str.indexOf(target);
-  return str.slice(0, index) + replacement + str.slice(index + target.length);
-}
-function reportHTMLContent(dumpData) {
-  const tpl = getReportTpl();
-  let reportContent;
-  if (Array.isArray(dumpData) && dumpData.length === 0 || typeof dumpData === "undefined") {
-    reportContent = replaceStringWithFirstAppearance(
-      tpl,
-      "{{dump}}",
-      `<script type="midscene_web_dump" type="application/json"></script>`
-    );
-  } else if (typeof dumpData === "string") {
-    reportContent = replaceStringWithFirstAppearance(
-      tpl,
-      "{{dump}}",
-      `<script type="midscene_web_dump" type="application/json">${dumpData}</script>`
-    );
-  } else {
-    const dumps = dumpData.map(({ dumpString, attributes }) => {
-      const attributesArr = Object.keys(attributes || {}).map((key) => {
-        return `${key}="${encodeURIComponent(attributes[key])}"`;
-      });
-      return `<script type="midscene_web_dump" type="application/json" ${attributesArr.join(
-        " "
-      )}
->${dumpString}
-</script>`;
-    });
-    reportContent = replaceStringWithFirstAppearance(
-      tpl,
-      "{{dump}}",
-      dumps.join("\n")
-    );
-  }
-  return reportContent;
-}
-function writeDumpReport(fileName, dumpData) {
-  if (import_utils.ifInBrowser) {
-    console.log("will not write report in browser");
-    return null;
-  }
-  const midscenePkgInfo = (0, import_fs.getRunningPkgInfo)(__dirname);
-  if (!midscenePkgInfo) {
-    console.warn("midscenePkgInfo not found, will not write report");
-    return null;
-  }
-  const reportPath = (0, import_node_path.join)(getLogDirByType("report"), `${fileName}.html`);
-  const reportContent = reportHTMLContent(dumpData);
-  (0, import_node_fs.writeFileSync)(reportPath, reportContent);
-  return reportPath;
-}
-function writeLogFile(opts) {
-  if (import_utils.ifInBrowser) {
-    return "/mock/report.html";
-  }
-  const { fileName, fileExt, fileContent, type = "dump" } = opts;
-  const targetDir = getLogDirByType(type);
-  if (!logEnvReady) {
-    (0, import_node_assert.default)(targetDir, "logDir should be set before writing dump file");
-    const gitIgnorePath = (0, import_node_path.join)(targetDir, "../../.gitignore");
-    let gitIgnoreContent = "";
-    if ((0, import_node_fs.existsSync)(gitIgnorePath)) {
-      gitIgnoreContent = (0, import_node_fs.readFileSync)(gitIgnorePath, "utf-8");
-    }
-    const logDirName = (0, import_node_path.basename)(logDir);
-    if (!gitIgnoreContent.includes(`${logDirName}/`)) {
-      (0, import_node_fs.writeFileSync)(
-        gitIgnorePath,
-        `${gitIgnoreContent}
-# Midscene.js dump files
-${logDirName}/report
-${logDirName}/dump
-${logDirName}/tmp
-`,
-        "utf-8"
-      );
-    }
-    logEnvReady = true;
-  }
-  const filePath = (0, import_node_path.join)(targetDir, `${fileName}.${fileExt}`);
-  const outputResourceDir = (0, import_node_path.dirname)(filePath);
-  if (!(0, import_node_fs.existsSync)(outputResourceDir)) {
-    (0, import_node_fs.mkdirSync)(outputResourceDir, { recursive: true });
-  }
-  (0, import_node_fs.writeFileSync)(filePath, fileContent);
-  if (opts == null ? void 0 : opts.generateReport) {
-    return writeDumpReport(fileName, fileContent);
-  }
-  return filePath;
-}
-function getTmpDir() {
-  if (import_utils.ifInBrowser) {
-    return null;
-  }
-  const runningPkgInfo = (0, import_fs.getRunningPkgInfo)();
-  if (!runningPkgInfo) {
-    return null;
-  }
-  const { name } = runningPkgInfo;
-  const path = (0, import_node_path.join)((0, import_node_os.tmpdir)(), name);
-  (0, import_node_fs.mkdirSync)(path, { recursive: true });
-  return path;
-}
-function getTmpFile(fileExtWithoutDot) {
-  if (import_utils.ifInBrowser) {
-    return null;
-  }
-  const tmpDir = getTmpDir();
-  const filename = `${(0, import_utils.uuid)()}.${fileExtWithoutDot}`;
-  return (0, import_node_path.join)(tmpDir, filename);
-}
-function overlapped(container, target) {
-  return container.left < target.left + target.width && container.left + container.width > target.left && container.top < target.top + target.height && container.top + container.height > target.top;
-}
-async function sleep(ms) {
-  return new Promise((resolve) => setTimeout(resolve, ms));
-}
-function replacerForPageObject(key, value) {
-  var _a, _b;
-  if (value && ((_a = value.constructor) == null ? void 0 : _a.name) === "Page") {
-    return "[Page object]";
-  }
-  if (value && ((_b = value.constructor) == null ? void 0 : _b.name) === "Browser") {
-    return "[Browser object]";
-  }
-  return value;
-}
-function stringifyDumpData(data, indents) {
-  return JSON.stringify(data, replacerForPageObject, indents);
-}
-function getVersion() {
-  return "0.9.1";
-}
-function debugLog(...message) {
-  const debugMode = getAIConfig(MIDSCENE_DEBUG_MODE);
-  if (debugMode) {
-    console.log("[Midscene]", ...message);
-  }
-}
-var lastReportedRepoUrl = "";
-function uploadTestInfoToServer({ testUrl }) {
-  let repoUrl = "";
-  let userEmail = "";
-  const extraConfig = getAIConfigInJson(MIDSCENE_OPENAI_INIT_CONFIG_JSON);
-  const serverUrl = extraConfig == null ? void 0 : extraConfig.REPORT_SERVER_URL;
-  try {
-    repoUrl = (0, import_node_child_process.execSync)("git config --get remote.origin.url").toString().trim();
-    userEmail = (0, import_node_child_process.execSync)("git config --get user.email").toString().trim();
-  } catch (error) {
-    debugLog("Failed to get git info:", error);
-  }
-  if (serverUrl && (repoUrl && repoUrl !== lastReportedRepoUrl || !repoUrl && testUrl)) {
-    debugLog("Uploading test info to server", {
-      serverUrl,
-      repoUrl,
-      testUrl,
-      userEmail
-    });
-    fetch(serverUrl, {
-      method: "POST",
-      headers: {
-        "Content-Type": "application/json"
-      },
-      body: JSON.stringify({
-        repo_url: repoUrl,
-        test_url: testUrl,
-        user_email: userEmail
-      })
-    }).then((response) => response.json()).then((data) => {
-      debugLog("Successfully uploaded test info to server:", data);
-    }).catch(
-      (error) => debugLog("Failed to upload test info to server:", error)
-    );
-    lastReportedRepoUrl = repoUrl;
-  }
-}
-// Annotate the CommonJS export names for ESM import in node:
-0 && (module.exports = {
-  getLogDir,
-  getLogDirByType,
-  getTmpDir,
-  getTmpFile,
-  getVersion,
-  groupedActionDumpFileExt,
-  insightDumpFileExt,
-  overlapped,
-  replaceStringWithFirstAppearance,
-  replacerForPageObject,
-  reportHTMLContent,
-  setLogDir,
-  sleep,
-  stringifyDumpData,
-  uploadTestInfoToServer,
-  writeDumpReport,
-  writeLogFile
-});
+"use strict";Object.defineProperty(exports, "__esModule", {value: true});
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+var _chunkZRCWDGK2js = require('./chunk-ZRCWDGK2.js');
+require('./chunk-SCNIHQKF.js');
+require('./chunk-G7A32JAG.js');
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+exports.getLogDir = _chunkZRCWDGK2js.getLogDir; exports.getLogDirByType = _chunkZRCWDGK2js.getLogDirByType; exports.getTmpDir = _chunkZRCWDGK2js.getTmpDir; exports.getTmpFile = _chunkZRCWDGK2js.getTmpFile; exports.getVersion = _chunkZRCWDGK2js.getVersion; exports.groupedActionDumpFileExt = _chunkZRCWDGK2js.groupedActionDumpFileExt; exports.insightDumpFileExt = _chunkZRCWDGK2js.insightDumpFileExt; exports.overlapped = _chunkZRCWDGK2js.overlapped; exports.replaceStringWithFirstAppearance = _chunkZRCWDGK2js.replaceStringWithFirstAppearance; exports.replacerForPageObject = _chunkZRCWDGK2js.replacerForPageObject; exports.reportHTMLContent = _chunkZRCWDGK2js.reportHTMLContent; exports.setLogDir = _chunkZRCWDGK2js.setLogDir; exports.sleep = _chunkZRCWDGK2js.sleep; exports.stringifyDumpData = _chunkZRCWDGK2js.stringifyDumpData; exports.uploadTestInfoToServer = _chunkZRCWDGK2js.uploadTestInfoToServer; exports.writeDumpReport = _chunkZRCWDGK2js.writeDumpReport; exports.writeLogFile = _chunkZRCWDGK2js.writeLogFile;