@midscene/core 0.17.4-beta-20250526154612.0 → 0.17.4-beta-20250530041415.0

package/dist/lib/index.d.ts

@@ -1,7 +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 InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightAssertionResponse, R as Rect, A as AIDescribeElementResponse } from './types-81bb2c02.js';
-export { u as AIAssertionResponse, s as AIDataExtractionResponse, q as AIElementCoordinatesResponse, p as AIElementLocatorResponse, r as AIElementResponse, l as AIResponseFormat, t as AISectionLocatorResponse, o as AISingleElementResponse, m as AISingleElementResponseById, n as AISingleElementResponseByPosition, k as AIUsageInfo, V as AgentAssertOpt, x as AgentDescribeElementAtPointResult, T as AgentWaitForOpt, a6 as BaseAgentParserOpt, C as CallAIFn, a5 as Color, G as DumpMeta, O as ElementById, j as ElementTreeNode, y as EnsureObject, a9 as ExecutionRecorderItem, ar as ExecutionTaskAction, aq as ExecutionTaskActionApply, ap as ExecutionTaskInsightAssertion, ao as ExecutionTaskInsightAssertionApply, an as ExecutionTaskInsightAssertionParam, ag as ExecutionTaskInsightDumpLog, ai as ExecutionTaskInsightLocate, ah as ExecutionTaskInsightLocateApply, af as ExecutionTaskInsightLocateOutput, ae as ExecutionTaskInsightLocateParam, am as ExecutionTaskInsightQuery, al as ExecutionTaskInsightQueryApply, ak as ExecutionTaskInsightQueryOutput, aj as ExecutionTaskInsightQueryParam, at as ExecutionTaskPlanning, as as ExecutionTaskPlanningApply, ad as ExecutionTaskReturn, aa as ExecutionTaskType, ab as ExecutorContext, aS as FreeFn, au as GroupedActionDump, J as InsightDump, z as InsightExtractParam, N as LiteUISection, aw as LocateOption, F as LocateResultElement, w as LocateValidatorResult, v as LocatorValidatorOption, i as MidsceneYamlFlowItem, aD as MidsceneYamlFlowItemAIAction, aE as MidsceneYamlFlowItemAIAssert, aI as MidsceneYamlFlowItemAIBoolean, aM as MidsceneYamlFlowItemAIHover, aN as MidsceneYamlFlowItemAIInput, aO as MidsceneYamlFlowItemAIKeyboardPress, aJ as MidsceneYamlFlowItemAILocate, aH as MidsceneYamlFlowItemAINString, aG as MidsceneYamlFlowItemAINumber, aF as MidsceneYamlFlowItemAIQuery, aP as MidsceneYamlFlowItemAIScroll, aL as MidsceneYamlFlowItemAITap, aK as MidsceneYamlFlowItemAIWaitFor, aQ as MidsceneYamlFlowItemEvaluateJavaScript, aR as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aB as MidsceneYamlScriptAndroidEnv, aC as MidsceneYamlScriptEnv, az as MidsceneYamlScriptEnvBase, aA as MidsceneYamlScriptWebEnv, h as MidsceneYamlTask, Q as OnTaskStartTip, av as PageType, K as PartialInsightDumpFromSDK, Y as PlanningAIResponse, X as PlanningAction, a1 as PlanningActionParamAssert, a3 as PlanningActionParamError, _ as PlanningActionParamHover, $ as PlanningActionParamInputOrKeyPress, a0 as PlanningActionParamScroll, a2 as PlanningActionParamSleep, Z as PlanningActionParamTap, a4 as PlanningActionParamWaitFor, W as PlanningLocateParam, a8 as PlaywrightParserOpt, P as Point, a7 as PuppeteerParserOpt, ax as ReferenceImage, H as ReportDumpWithAttributes, aU as ScriptPlayerStatusValue, aT as ScriptPlayerTaskStatus, S as Size, ac as TaskCacheInfo, ay as scrollParam } from './types-81bb2c02.js';
-import { c as callAiFn } from './llm-planning-9c97f5c2.js';
-export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-9c97f5c2.js';
+import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightExtractOption, h as InsightAssertionResponse, A as AIDescribeElementResponse } from './types-8c197f92.js';
+export { v as AIAssertionResponse, t as AIDataExtractionResponse, r as AIElementCoordinatesResponse, q as AIElementLocatorResponse, s as AIElementResponse, m as AIResponseFormat, u as AISectionLocatorResponse, p as AISingleElementResponse, n as AISingleElementResponseById, o as AISingleElementResponseByPosition, l as AIUsageInfo, Q as AgentAssertOpt, y as AgentDescribeElementAtPointResult, N as AgentWaitForOpt, a4 as BaseAgentParserOpt, C as CallAIFn, a3 as Color, G as DumpMeta, K as ElementById, z as EnsureObject, a7 as ExecutionRecorderItem, ap as ExecutionTaskAction, ao as ExecutionTaskActionApply, an as ExecutionTaskInsightAssertion, am as ExecutionTaskInsightAssertionApply, al as ExecutionTaskInsightAssertionParam, ae as ExecutionTaskInsightDumpLog, ag as ExecutionTaskInsightLocate, af as ExecutionTaskInsightLocateApply, ad as ExecutionTaskInsightLocateOutput, ac as ExecutionTaskInsightLocateParam, ak as ExecutionTaskInsightQuery, aj as ExecutionTaskInsightQueryApply, ai as ExecutionTaskInsightQueryOutput, ah as ExecutionTaskInsightQueryParam, ar as ExecutionTaskPlanning, aq as ExecutionTaskPlanningApply, ab as ExecutionTaskReturn, a8 as ExecutionTaskType, a9 as ExecutorContext, aQ as FreeFn, as as GroupedActionDump, H as InsightDump, B as InsightExtractParam, J as LiteUISection, au as LocateOption, F as LocateResultElement, x as LocateValidatorResult, w as LocatorValidatorOption, j as MidsceneYamlFlowItem, aB as MidsceneYamlFlowItemAIAction, aC as MidsceneYamlFlowItemAIAssert, aG as MidsceneYamlFlowItemAIBoolean, aK as MidsceneYamlFlowItemAIHover, aL as MidsceneYamlFlowItemAIInput, aM as MidsceneYamlFlowItemAIKeyboardPress, aH as MidsceneYamlFlowItemAILocate, aF as MidsceneYamlFlowItemAINString, aE as MidsceneYamlFlowItemAINumber, aD as MidsceneYamlFlowItemAIQuery, k as MidsceneYamlFlowItemAIRightClick, aN as MidsceneYamlFlowItemAIScroll, aJ as MidsceneYamlFlowItemAITap, aI as MidsceneYamlFlowItemAIWaitFor, aO as MidsceneYamlFlowItemEvaluateJavaScript, aP as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, az as MidsceneYamlScriptAndroidEnv, aA as MidsceneYamlScriptEnv, ax as MidsceneYamlScriptEnvBase, ay as MidsceneYamlScriptWebEnv, i as MidsceneYamlTask, O as OnTaskStartTip, at as PageType, P as PartialInsightDumpFromSDK, V as PlanningAIResponse, T as PlanningAction, $ as PlanningActionParamAssert, a1 as PlanningActionParamError, X as PlanningActionParamHover, Z as PlanningActionParamInputOrKeyPress, Y as PlanningActionParamRightClick, _ as PlanningActionParamScroll, a0 as PlanningActionParamSleep, W as PlanningActionParamTap, a2 as PlanningActionParamWaitFor, S as PlanningLocateParam, a6 as PlaywrightParserOpt, a5 as PuppeteerParserOpt, av as ReferenceImage, R as ReportDumpWithAttributes, aS as ScriptPlayerStatusValue, aR as ScriptPlayerTaskStatus, aa as TaskCacheInfo, aw as scrollParam } from './types-8c197f92.js';
+import { c as callAiFn } from './llm-planning-573b9b34.js';
+export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-573b9b34.js';
+import { BaseElement, Rect } from '@midscene/shared/types';
+export { BaseElement, ElementTreeNode, Point, Rect, Size } from '@midscene/shared/types';
 export { getVersion } from './utils.js';
 export { MIDSCENE_MODEL_NAME, getAIConfig } from '@midscene/shared/env';
 import '@midscene/shared/constants';
@@ -34,13 +36,13 @@ declare class Insight<ElementType extends BaseElement = BaseElement, ContextType
     taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;
     constructor(context: ContextType | ((action: InsightAction) => Promise<ContextType> | ContextType), opt?: InsightOptions);
     locate(query: DetailedLocateParam, opt?: LocateOpts): Promise<LocateResult>;
-    extract<T = any>(input: string): Promise<T>;
-    extract<T extends Record<string, string>>(input: T): Promise<Record<keyof T, any>>;
-    extract<T extends object>(input: Record<keyof T, string>): Promise<T>;
+    extract<T = any>(input: string, opt?: InsightExtractOption): Promise<T>;
+    extract<T extends Record<string, string>>(input: T, opt?: InsightExtractOption): Promise<Record<keyof T, any>>;
+    extract<T extends object>(input: Record<keyof T, string>, opt?: InsightExtractOption): Promise<T>;
     assert(assertion: string): Promise<InsightAssertionResponse>;
     describe(target: Rect | [number, number], opt?: {
         deepThink?: boolean;
     }): Promise<Pick<AIDescribeElementResponse, 'description'>>;
 }
 
-export { AIDescribeElementResponse, BaseElement, DetailedLocateParam, DumpSubscriber, ExecutionDump, ExecutionTask, ExecutionTaskApply, ExecutionTaskProgressOptions, Executor, Insight, InsightAction, InsightAssertionResponse, InsightOptions, InsightTaskInfo, LocateResult, Rect, UIContext, Insight as default };
+export { AIDescribeElementResponse, DetailedLocateParam, DumpSubscriber, ExecutionDump, ExecutionTask, ExecutionTaskApply, ExecutionTaskProgressOptions, Executor, Insight, InsightAction, InsightAssertionResponse, InsightExtractOption, InsightOptions, InsightTaskInfo, LocateResult, UIContext, Insight as default };

package/dist/lib/index.js

@@ -1,6 +1,6 @@
 "use strict";Object.defineProperty(exports, "__esModule", {value: true}); function _optionalChain(ops) { let lastAccessLHS = undefined; let value = ops[0]; let i = 1; while (i < ops.length) { const op = ops[i]; const fn = ops[i + 1]; i += 2; if ((op === 'optionalAccess' || op === 'optionalCall') && value == null) { return undefined; } if (op === 'access' || op === 'optionalAccess') { lastAccessLHS = value; value = fn(value); } else if (op === 'call' || op === 'optionalCall') { value = fn((...args) => value.call(lastAccessLHS, ...args)); lastAccessLHS = undefined; } } return value; }
 
-var _chunk3YRZWYIBjs = require('./chunk-3YRZWYIB.js');
+var _chunkDR6LJT3Kjs = require('./chunk-DR6LJT3K.js');
 
 
 
@@ -11,7 +11,7 @@ var _chunk3YRZWYIBjs = require('./chunk-3YRZWYIB.js');
 
 
 
-var _chunk2AS6R2ARjs = require('./chunk-2AS6R2AR.js');
+var _chunk7MNTWX2Ajs = require('./chunk-7MNTWX2A.js');
 
 // src/ai-model/action-executor.ts
 
@@ -172,7 +172,7 @@ ${_optionalChain([this, 'access', _7 => _7.latestErrorTask, 'call', _8 => _8(),
       }
     }
     const dumpData = {
-      sdkVersion: _chunk3YRZWYIBjs.getVersion.call(void 0, ),
+      sdkVersion: _chunkDR6LJT3Kjs.getVersion.call(void 0, ),
       model_name: _env.getAIConfig.call(void 0, _env.MIDSCENE_MODEL_NAME) || "",
       model_description: modelDescription,
       logTime: Date.now(),
@@ -227,7 +227,7 @@ var _logger = require('@midscene/shared/logger');
 
 function emitInsightDump(data, dumpSubscriber) {
   const baseData = {
-    sdkVersion: _chunk3YRZWYIBjs.getVersion.call(void 0, ),
+    sdkVersion: _chunkDR6LJT3Kjs.getVersion.call(void 0, ),
     logTime: Date.now(),
     model_name: _env.getAIConfig.call(void 0, _env.MIDSCENE_MODEL_NAME) || ""
   };
@@ -243,7 +243,7 @@ function emitInsightDump(data, dumpSubscriber) {
 var debug = _logger.getDebug.call(void 0, "ai:insight");
 var Insight = class {
   constructor(context, opt) {
-    this.aiVendorFn = _chunk2AS6R2ARjs.callAiFn;
+    this.aiVendorFn = _chunk7MNTWX2Ajs.callAiFn;
     _utils.assert.call(void 0, context, "context is required for Insight");
     if (typeof context === "function") {
       this.contextRetrieverFn = context;
@@ -286,7 +286,7 @@ var Insight = class {
     let searchAreaUsage = void 0;
     let searchAreaResponse = void 0;
     if (searchAreaPrompt) {
-      searchAreaResponse = await _chunk2AS6R2ARjs.AiLocateSection.call(void 0, {
+      searchAreaResponse = await _chunk7MNTWX2Ajs.AiLocateSection.call(void 0, {
         context,
         sectionDescription: searchAreaPrompt
       });
@@ -299,7 +299,7 @@ var Insight = class {
       searchArea = searchAreaResponse.rect;
     }
     const startTime = Date.now();
-    const { parseResult, rect, elementById, rawResponse, usage } = await _chunk2AS6R2ARjs.AiLocateElement.call(void 0, {
+    const { parseResult, rect, elementById, rawResponse, usage } = await _chunk7MNTWX2Ajs.AiLocateElement.call(void 0, {
       callAI: callAI || this.aiVendorFn,
       context,
       targetElementDescription: queryPrompt,
@@ -378,7 +378,7 @@ ${parseResult.errors.join("\n")}`;
       rect
     };
   }
-  async extract(dataDemand) {
+  async extract(dataDemand, opt) {
     _utils.assert.call(void 0, 
       typeof dataDemand === "object" || typeof dataDemand === "string",
       `dataDemand should be object or string, but get ${typeof dataDemand}`
@@ -387,9 +387,10 @@ ${parseResult.errors.join("\n")}`;
     this.onceDumpUpdatedFn = void 0;
     const context = await this.contextRetrieverFn("extract");
     const startTime = Date.now();
-    const { parseResult, usage } = await _chunk2AS6R2ARjs.AiExtractElementInfo.call(void 0, {
+    const { parseResult, usage } = await _chunk7MNTWX2Ajs.AiExtractElementInfo.call(void 0, {
       context,
-      dataQuery: dataDemand
+      dataQuery: dataDemand,
+      extractOption: opt
     });
     const timeCost = Date.now() - startTime;
     const taskInfo = {
@@ -438,7 +439,7 @@ ${parseResult.errors.join("\n")}`;
     this.onceDumpUpdatedFn = void 0;
     const context = await this.contextRetrieverFn("assert");
     const startTime = Date.now();
-    const assertResult = await _chunk2AS6R2ARjs.AiAssert.call(void 0, {
+    const assertResult = await _chunk7MNTWX2Ajs.AiAssert.call(void 0, {
       assertion,
       context
     });
@@ -491,7 +492,7 @@ ${parseResult.errors.join("\n")}`;
       borderThickness: 3
     });
     if (_optionalChain([opt, 'optionalAccess', _23 => _23.deepThink])) {
-      const searchArea = _chunk2AS6R2ARjs.expandSearchArea.call(void 0, targetRect, context.size);
+      const searchArea = _chunk7MNTWX2Ajs.expandSearchArea.call(void 0, targetRect, context.size);
       debug("describe: set searchArea", searchArea);
       imagePayload = await _img.cropByRect.call(void 0, 
         imagePayload,
@@ -514,7 +515,7 @@ ${parseResult.errors.join("\n")}`;
         ]
       }
     ];
-    const callAIFn = this.aiVendorFn || _chunk2AS6R2ARjs.callToGetJSONObject;
+    const callAIFn = this.aiVendorFn || _chunk7MNTWX2Ajs.callToGetJSONObject;
     const res = await callAIFn(msgs, 4 /* DESCRIBE_ELEMENT */);
     const { content } = res;
     _utils.assert.call(void 0, !content.error, `describe failed: ${content.error}`);
@@ -537,6 +538,6 @@ var src_default = Insight;
 
 
 
-exports.AiAssert = _chunk2AS6R2ARjs.AiAssert; exports.AiLocateElement = _chunk2AS6R2ARjs.AiLocateElement; exports.Executor = Executor; exports.Insight = Insight; exports.MIDSCENE_MODEL_NAME = _env.MIDSCENE_MODEL_NAME; exports.default = src_default; exports.describeUserPage = _chunk2AS6R2ARjs.describeUserPage; exports.getAIConfig = _env.getAIConfig; exports.getVersion = _chunk3YRZWYIBjs.getVersion; exports.plan = _chunk2AS6R2ARjs.plan;
+exports.AiAssert = _chunk7MNTWX2Ajs.AiAssert; exports.AiLocateElement = _chunk7MNTWX2Ajs.AiLocateElement; exports.Executor = Executor; exports.Insight = Insight; exports.MIDSCENE_MODEL_NAME = _env.MIDSCENE_MODEL_NAME; exports.default = src_default; exports.describeUserPage = _chunk7MNTWX2Ajs.describeUserPage; exports.getAIConfig = _env.getAIConfig; exports.getVersion = _chunkDR6LJT3Kjs.getVersion; exports.plan = _chunk7MNTWX2Ajs.plan;
 
 //# sourceMappingURL=index.js.map

package/dist/lib/index.js.map

@@ -1 +1 @@
-{"version":3,"mappings":";;;;;;;;;;;;;;;;AAUA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,cAAc;AAEhB,IAAM,WAAN,MAAe;AAAA,EAUpB,YACE,MACA,SAGA;AACA,SAAK,SACH,SAAS,SAAS,QAAQ,MAAM,SAAS,IAAI,YAAY;AAC3D,SAAK,OAAO;AACZ,SAAK,SAAS,SAAS,SAAS,CAAC,GAAG;AAAA,MAAI,CAAC,SACvC,KAAK,kBAAkB,IAAI;AAAA,IAC7B;AACA,SAAK,cAAc,SAAS;AAAA,EAC9B;AAAA,EAEQ,kBAAkB,MAAyC;AACjE,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,MAAgE;AAC3E;AAAA,MACE,KAAK,WAAW;AAAA,MAChB;AAAA,QAAyD,KAAK,gBAAgB,GAAG,KAAK;AAAA,EAAK,KAAK,gBAAgB,GAAG,UAAU;AAAA,IAC/H;AACA,QAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,WAAK,MAAM,KAAK,GAAG,KAAK,IAAI,CAAC,SAAS,KAAK,kBAAkB,IAAI,CAAC,CAAC;AAAA,IACrE,OAAO;AACL,WAAK,MAAM,KAAK,KAAK,kBAAkB,IAAI,CAAC;AAAA,IAC9C;AACA,QAAI,KAAK,WAAW,WAAW;AAC7B,WAAK,SAAS;AAAA,IAChB;AAAA,EACF;AAAA,EAEA,MAAM,QAAsB;AAC1B,QAAI,KAAK,WAAW,UAAU,KAAK,MAAM,SAAS,GAAG;AACnD,cAAQ;AAAA,QACN;AAAA,MACF;AAAA,IACF;AAEA,WAAO,KAAK,WAAW,WAAW,6BAA6B;AAC/D,WAAO,KAAK,WAAW,aAAa,+BAA+B;AACnE,WAAO,KAAK,WAAW,SAAS,4BAA4B;AAE5D,UAAM,mBAAmB,KAAK,MAAM;AAAA,MAClC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,mBAAmB,GAAG;AAExB;AAAA,IACF;AAEA,SAAK,SAAS;AACd,QAAI,YAAY;AAChB,QAAI,wBAAwB;AAE5B,QAAI;AAEJ,WAAO,YAAY,KAAK,MAAM,QAAQ;AACpC,YAAM,OAAO,KAAK,MAAM,SAAS;AACjC;AAAA,QACE,KAAK,WAAW;AAAA,QAChB,2CAA2C,KAAK,MAAM;AAAA,MACxD;AACA,WAAK,SAAS;AAAA,QACZ,OAAO,KAAK,IAAI;AAAA,MAClB;AACA,UAAI;AACF,aAAK,SAAS;AACd,YAAI;AACF,cAAI,KAAK,aAAa;AACpB,kBAAM,KAAK,YAAY,IAAI;AAAA,UAC7B;AAAA,QACF,SAAS,GAAG;AACV,kBAAQ,MAAM,wBAAwB,CAAC;AAAA,QACzC;AACA;AAAA,UACE,CAAC,WAAW,UAAU,UAAU,EAAE,QAAQ,KAAK,IAAI,KAAK;AAAA,UACxD,0BAA0B,KAAK,IAAI;AAAA,QACrC;AAEA,cAAM,EAAE,UAAU,MAAM,IAAI;AAC5B,eAAO,UAAU,uCAAuC,KAAK,IAAI,EAAE;AAEnE,YAAI;AACJ,cAAM,kBAAmC;AAAA,UACvC;AAAA,UACA,SAAS,oBAAoB;AAAA,QAC/B;AAEA,YAAI,KAAK,SAAS,WAAW;AAC3B;AAAA,YACE,KAAK,YAAY,YACf,KAAK,YAAY,WACjB,KAAK,YAAY,YACjB,KAAK,YAAY,aACjB,KAAK,YAAY,YACjB,KAAK,YAAY;AAAA,YACnB,gCAAgC,KAAK,OAAO;AAAA,UAC9C;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AACxD,cAAI,KAAK,YAAY,UAAU;AAC7B,iCACE,aACC;AAAA,UACL;AAAA,QACF,WAAW,KAAK,SAAS,YAAY,KAAK,SAAS,YAAY;AAC7D,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D,OAAO;AACL,kBAAQ;AAAA,YACN,0BAA0B,KAAK,IAAI;AAAA,UACrC;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D;AAEA,eAAO,OAAO,MAAM,WAAW;AAC/B,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD,aAAK,OAAO,SAAU,aAAqB,UAAU;AACrD;AAAA,MACF,SAAS,GAAQ;AACf,gCAAwB;AACxB,aAAK,QACH,GAAG,YAAY,OAAO,MAAM,WAAW,IAAI;AAC7C,aAAK,aAAa,EAAE;AAEpB,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF;AAAA,IACF;AAGA,aAAS,IAAI,YAAY,GAAG,IAAI,KAAK,MAAM,QAAQ,KAAK;AACtD,WAAK,MAAM,CAAC,EAAE,SAAS;AAAA,IACzB;AAEA,QAAI,uBAAuB;AACzB,WAAK,SAAS;AAAA,IAChB,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AAEA,QAAI,KAAK,MAAM,QAAQ;AAErB,YAAM,cAAc,KAAK,IAAI,WAAW,KAAK,MAAM,SAAS,CAAC;AAC7D,aAAO,KAAK,MAAM,WAAW,EAAE;AAAA,IACjC;AAAA,EACF;AAAA,EAEA,iBAA0B;AACxB,WAAO,KAAK,WAAW;AAAA,EACzB;AAAA,EAEA,kBAAwC;AACtC,QAAI,KAAK,WAAW,SAAS;AAC3B,aAAO;AAAA,IACT;AACA,UAAM,iBAAiB,KAAK,MAAM;AAAA,MAChC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,kBAAkB,GAAG;AACvB,aAAO,KAAK,MAAM,cAAc;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AAAA,EAEA,OAAsB;AACpB,QAAI,mBAAmB;AAEvB,QAAI,aAAa,GAAG;AAClB,YAAM,iBAAiB,mBAAmB;AAC1C,UAAI,gBAAgB;AAClB,2BAAmB,WAAW,cAAc;AAAA,MAC9C,OAAO;AACL,2BAAmB,GAAG,aAAa,CAAC;AAAA,MACtC;AAAA,IACF;AACA,UAAM,WAA0B;AAAA,MAC9B,YAAY,WAAW;AAAA,MACvB,YAAY,YAAY,mBAAmB,KAAK;AAAA,MAChD,mBAAmB;AAAA,MACnB,SAAS,KAAK,IAAI;AAAA,MAClB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACF;;;AC9NA,SAAS,4BAA4B;AAE9B,IAAM,8BAA8B,MAAM;AAC/C,SAAO,mNAAmN,qBAAqB,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAoBlP;;;ACQA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA,gBAAAA;AAAA,OACK;AACP,SAAS,yBAAyB,kBAAkB;AACpD,SAAS,gBAAgB;AACzB,SAAS,UAAAC,eAAc;;;AChCvB;AAAA,EACE,uBAAAC;AAAA,EACA,eAAAC;AAAA,OAGK;AACP,SAAS,YAAY;AAEd,SAAS,gBACd,MACA,gBACA;AACA,QAAM,WAAqB;AAAA,IACzB,YAAY,WAAW;AAAA,IACvB,SAAS,KAAK,IAAI;AAAA,IAClB,YAAYA,aAAYD,oBAAmB,KAAK;AAAA,EAClD;AACA,QAAM,YAAyB;AAAA,IAC7B,OAAO,KAAK;AAAA,IACZ,GAAG;AAAA,IACH,GAAG;AAAA,EACL;AAEA,mBAAiB,SAAS;AAC5B;;;ADoBA,IAAM,QAAQ,SAAS,YAAY;AACnC,IAAqB,UAArB,MAGE;AAAA,EAWA,YACE,SAGA,KACA;AAXF,sBAAoD;AAYlD,IAAAD,QAAO,SAAS,iCAAiC;AACjD,QAAI,OAAO,YAAY,YAAY;AACjC,WAAK,qBAAqB;AAAA,IAC5B,OAAO;AACL,WAAK,qBAAqB,MAAM,QAAQ,QAAQ,OAAO;AAAA,IACzD;AAEA,QAAI,OAAO,KAAK,eAAe,aAAa;AAC1C,WAAK,aAAa,IAAI;AAAA,IACxB;AACA,QAAI,OAAO,KAAK,aAAa,aAAa;AACxC,WAAK,WAAW,IAAI;AAAA,IACtB;AAAA,EACF;AAAA,EAEA,MAAM,OACJ,OACA,KACuB;AACvB,UAAM,EAAE,OAAO,IAAI,OAAO,CAAC;AAC3B,UAAM,cAAc,OAAO,UAAU,WAAW,QAAQ,MAAM;AAC9D,IAAAA,QAAO,aAAa,8BAA8B;AAClD,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,IAAAA,QAAO,OAAO,UAAU,UAAU,sCAAsC;AAExE,UAAM,wBAAwB;AAAA,MAC5B;AAAA,IACF;AACA,QAAI,uBAAuB;AACzB,YAAM,yBAAyB,qBAAqB;AAAA,IACtD;AACA,QAAI;AACJ,QAAI,MAAM,aAAa,uBAAuB;AAC5C,yBAAmB,MAAM;AAAA,IAC3B;AAEA,QAAI,oBAAoB,CAACD,cAAa,GAAG;AACvC,cAAQ;AAAA,QACN;AAAA,MACF;AACA,yBAAmB;AAAA,IACrB;AAEA,UAAM,UAAU,KAAK,WAAY,MAAM,KAAK,mBAAmB,QAAQ;AAEvE,QAAI,aAA+B;AACnC,QAAI,wBAA4C;AAChD,QAAI,kBAA2C;AAC/C,QAAI,qBAEY;AAChB,QAAI,kBAAkB;AACpB,2BAAqB,MAAM,gBAAgB;AAAA,QACzC;AAAA,QACA,oBAAoB;AAAA,MACtB,CAAC;AACD,MAAAC;AAAA,QACE,mBAAmB;AAAA,QACnB,gCAAgC,gBAAgB,IAC9C,mBAAmB,QAAQ,KAAK,mBAAmB,KAAK,KAAK,EAC/D;AAAA,MACF;AACA,8BAAwB,mBAAmB;AAC3C,wBAAkB,mBAAmB;AACrC,mBAAa,mBAAmB;AAAA,IAClC;AAEA,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,aAAa,aAAa,MAAM,IACzD,MAAM,gBAAgB;AAAA,MACpB,QAAQ,UAAU,KAAK;AAAA,MACvB;AAAA,MACA,0BAA0B;AAAA,MAC1B,cAAc;AAAA,IAChB,CAAC;AAEH,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,MACvC,gBAAgB,KAAK,UAAU,WAAW;AAAA,MAC1C;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAgC,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAC1E;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT,SAAS;AAAA,MACX;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,aAAa;AAAA,MACb,MAAM;AAAA,MACN;AAAA,MACA,WAAW,CAAC,CAAC;AAAA,MACb,OAAO;AAAA,IACT;AAEA,UAAM,WAA0B,CAAC;AACjC,KAAC,YAAY,YAAY,CAAC,GAAG,QAAQ,CAAC,SAAS;AAC7C,UAAI,QAAQ,MAAM;AAChB,cAAM,UAAU,YAAY,MAAM,EAAE;AAEpC,YAAI,CAAC,SAAS;AACZ,kBAAQ;AAAA,YACN,kCAAkC,KAAK,EAAE;AAAA,UAC3C;AACA;AAAA,QACF;AACA,iBAAS,KAAK,OAAO;AAAA,MACvB;AAAA,IACF,CAAC;AAED;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,IACF;AAEA,QAAI,UAAU;AACZ,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,IAAAA;AAAA,MACE,SAAS,UAAU;AAAA,MACnB,6CAA6C,SAAS,MAAM;AAAA,IAC9D;AAEA,QAAI,SAAS,WAAW,GAAG;AACzB,aAAO;AAAA,QACL,SAAS;AAAA,UACP,IAAI,SAAS,CAAC,EAAG;AAAA,UACjB,SAAS,SAAS,CAAC,EAAG;AAAA,UACtB,QAAQ,SAAS,CAAC,EAAG;AAAA,UACrB,MAAM,SAAS,CAAC,EAAG;AAAA,UACnB,QAAQ,SAAS,CAAC,EAAG,UAAU,CAAC;AAAA,UAChC,YAAY,SAAS,CAAC,EAAG;AAAA,QAC3B;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,WAAO;AAAA,MACL,SAAS;AAAA,MACT;AAAA,IACF;AAAA,EACF;AAAA,EAQA,MAAM,QAAW,YAA+C;AAC9D,IAAAA;AAAA,MACE,OAAO,eAAe,YAAY,OAAO,eAAe;AAAA,MACxD,kDAAkD,OAAO,UAAU;AAAA,IACrE;AACA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,SAAS;AAEvD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,IAAI,MAAM,qBAAwB;AAAA,MAC3D;AAAA,MACA,WAAW;AAAA,IACb,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,IACzC;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAwB,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAClE;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,OAAO;AAAA,IACT;AAEA,UAAM,EAAE,KAAK,IAAI,eAAe,CAAC;AAGjC;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH;AAAA,MACF;AAAA,MACA;AAAA,IACF;AAEA,QAAI,YAAY,CAAC,MAAM;AACrB,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,WAAO;AAAA,MACL;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,WAAsD;AACjE,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,QAAQ;AACtD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,eAAe,MAAM,SAAS;AAAA,MAClC;AAAA,MACA;AAAA,IACF,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,aAAa,OAAO;AAAA,IAClD;AAEA,UAAM,EAAE,SAAS,KAAK,IAAI,aAAa;AACvC,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,eAAe;AAAA,MACf,kBAAkB;AAAA,MAClB,OAAO,OAAO,SAAY;AAAA,IAC5B;AACA,oBAAgB,UAAU,cAAc;AAExC,WAAO;AAAA,MACL;AAAA,MACA;AAAA,MACA,OAAO,aAAa;AAAA,IACtB;AAAA,EACF;AAAA,EACA,MAAM,SACJ,QACA,KAGyD;AACzD,IAAAA,QAAO,QAAQ,yCAAyC;AACxD,UAAM,UAAU,MAAM,KAAK,mBAAmB,UAAU;AACxD,UAAM,EAAE,iBAAiB,IAAI;AAC7B,IAAAA,QAAO,kBAAkB,6CAA6C;AAEtE,UAAM,eAAe,4BAA4B;AAGjD,UAAM,kBAAkB;AACxB,UAAM,aAAmB,MAAM,QAAQ,MAAM,IACzC;AAAA,MACE,MAAM,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAChD,KAAK,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAC/C,OAAO;AAAA,MACP,QAAQ;AAAA,IACV,IACA;AAEJ,QAAI,eAAe,MAAM,wBAAwB;AAAA,MAC/C,gBAAgB;AAAA,MAChB,sBAAsB;AAAA,QACpB;AAAA,UACE,MAAM;AAAA,QACR;AAAA,MACF;AAAA,MACA,iBAAiB;AAAA,IACnB,CAAC;AAED,QAAI,KAAK,WAAW;AAClB,YAAM,aAAa,iBAAiB,YAAY,QAAQ,IAAI;AAC5D,YAAM,4BAA4B,UAAU;AAC5C,qBAAe,MAAM;AAAA,QACnB;AAAA,QACA;AAAA,QACA,qBAAqB,oBAAoB;AAAA,MAC3C;AAAA,IACF;AAEA,UAAM,OAAe;AAAA,MACnB,EAAE,MAAM,UAAU,SAAS,aAAa;AAAA,MACxC;AAAA,QACE,MAAM;AAAA,QACN,SAAS;AAAA,UACP;AAAA,YACE,MAAM;AAAA,YACN,WAAW;AAAA,cACT,KAAK;AAAA,cACL,QAAQ;AAAA,YACV;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,WACJ,KAAK,cAAc;AAErB,UAAM,MAAM,MAAM,SAAS,8BAAmC;AAE9D,UAAM,EAAE,QAAQ,IAAI;AACpB,IAAAA,QAAO,CAAC,QAAQ,OAAO,oBAAoB,QAAQ,KAAK,EAAE;AAC1D,IAAAA,QAAO,QAAQ,aAAa,gCAAgC;AAC5D,WAAO;AAAA,EACT;AACF;;;AEjZA,SAAS,eAAAE,cAAa,uBAAAD,4BAA2B;AAGjD,IAAO,cAAQ","names":["vlLocateMode","assert","MIDSCENE_MODEL_NAME","getAIConfig"],"ignoreList":[],"sources":["../../src/ai-model/action-executor.ts","../../src/ai-model/prompt/describe.ts","../../src/insight/index.ts","../../src/insight/utils.ts","../../src/index.ts"],"sourcesContent":["import type {\n  ExecutionDump,\n  ExecutionTask,\n  ExecutionTaskApply,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskProgressOptions,\n  ExecutionTaskReturn,\n  ExecutorContext,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n  MIDSCENE_MODEL_NAME,\n  getAIConfig,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\n\nexport class Executor {\n  name: string;\n\n  tasks: ExecutionTask[];\n\n  // status of executor\n  status: 'init' | 'pending' | 'running' | 'completed' | 'error';\n\n  onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n\n  constructor(\n    name: string,\n    options?: ExecutionTaskProgressOptions & {\n      tasks?: ExecutionTaskApply[];\n    },\n  ) {\n    this.status =\n      options?.tasks && options.tasks.length > 0 ? 'pending' : 'init';\n    this.name = name;\n    this.tasks = (options?.tasks || []).map((item) =>\n      this.markTaskAsPending(item),\n    );\n    this.onTaskStart = options?.onTaskStart;\n  }\n\n  private markTaskAsPending(task: ExecutionTaskApply): ExecutionTask {\n    return {\n      status: 'pending',\n      ...task,\n    };\n  }\n\n  async append(task: ExecutionTaskApply[] | ExecutionTaskApply): Promise<void> {\n    assert(\n      this.status !== 'error',\n      `executor is in error state, cannot append task\\nerror=${this.latestErrorTask()?.error}\\n${this.latestErrorTask()?.errorStack}`,\n    );\n    if (Array.isArray(task)) {\n      this.tasks.push(...task.map((item) => this.markTaskAsPending(item)));\n    } else {\n      this.tasks.push(this.markTaskAsPending(task));\n    }\n    if (this.status !== 'running') {\n      this.status = 'pending';\n    }\n  }\n\n  async flush(): Promise<any> {\n    if (this.status === 'init' && this.tasks.length > 0) {\n      console.warn(\n        'illegal state for executor, status is init but tasks are not empty',\n      );\n    }\n\n    assert(this.status !== 'running', 'executor is already running');\n    assert(this.status !== 'completed', 'executor is already completed');\n    assert(this.status !== 'error', 'executor is in error state');\n\n    const nextPendingIndex = this.tasks.findIndex(\n      (task) => task.status === 'pending',\n    );\n    if (nextPendingIndex < 0) {\n      // all tasks are completed\n      return;\n    }\n\n    this.status = 'running';\n    let taskIndex = nextPendingIndex;\n    let successfullyCompleted = true;\n\n    let previousFindOutput: ExecutionTaskInsightLocateOutput | undefined;\n\n    while (taskIndex < this.tasks.length) {\n      const task = this.tasks[taskIndex];\n      assert(\n        task.status === 'pending',\n        `task status should be pending, but got: ${task.status}`,\n      );\n      task.timing = {\n        start: Date.now(),\n      };\n      try {\n        task.status = 'running';\n        try {\n          if (this.onTaskStart) {\n            await this.onTaskStart(task);\n          }\n        } catch (e) {\n          console.error('error in onTaskStart', e);\n        }\n        assert(\n          ['Insight', 'Action', 'Planning'].indexOf(task.type) >= 0,\n          `unsupported task type: ${task.type}`,\n        );\n\n        const { executor, param } = task;\n        assert(executor, `executor is required for task type: ${task.type}`);\n\n        let returnValue;\n        const executorContext: ExecutorContext = {\n          task,\n          element: previousFindOutput?.element,\n        };\n\n        if (task.type === 'Insight') {\n          assert(\n            task.subType === 'Locate' ||\n              task.subType === 'Query' ||\n              task.subType === 'Assert' ||\n              task.subType === 'Boolean' ||\n              task.subType === 'Number' ||\n              task.subType === 'String',\n            `unsupported insight subType: ${task.subType}`,\n          );\n          returnValue = await task.executor(param, executorContext);\n          if (task.subType === 'Locate') {\n            previousFindOutput = (\n              returnValue as ExecutionTaskReturn<ExecutionTaskInsightLocateOutput>\n            )?.output;\n          }\n        } else if (task.type === 'Action' || task.type === 'Planning') {\n          returnValue = await task.executor(param, executorContext);\n        } else {\n          console.warn(\n            `unsupported task type: ${task.type}, will try to execute it directly`,\n          );\n          returnValue = await task.executor(param, executorContext);\n        }\n\n        Object.assign(task, returnValue);\n        task.status = 'finished';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        task.timing.aiCost = (returnValue as any)?.aiCost || 0;\n        taskIndex++;\n      } catch (e: any) {\n        successfullyCompleted = false;\n        task.error =\n          e?.message || (typeof e === 'string' ? e : 'error-without-message');\n        task.errorStack = e.stack;\n\n        task.status = 'failed';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        break;\n      }\n    }\n\n    // set all remaining tasks as cancelled\n    for (let i = taskIndex + 1; i < this.tasks.length; i++) {\n      this.tasks[i].status = 'cancelled';\n    }\n\n    if (successfullyCompleted) {\n      this.status = 'completed';\n    } else {\n      this.status = 'error';\n    }\n\n    if (this.tasks.length) {\n      // return the last output\n      const outputIndex = Math.min(taskIndex, this.tasks.length - 1);\n      return this.tasks[outputIndex].output;\n    }\n  }\n\n  isInErrorState(): boolean {\n    return this.status === 'error';\n  }\n\n  latestErrorTask(): ExecutionTask | null {\n    if (this.status !== 'error') {\n      return null;\n    }\n    const errorTaskIndex = this.tasks.findIndex(\n      (task) => task.status === 'failed',\n    );\n    if (errorTaskIndex >= 0) {\n      return this.tasks[errorTaskIndex];\n    }\n    return null;\n  }\n\n  dump(): ExecutionDump {\n    let modelDescription = '';\n\n    if (vlLocateMode()) {\n      const uiTarsModelVer = uiTarsModelVersion();\n      if (uiTarsModelVer) {\n        modelDescription = `UI-TARS=${uiTarsModelVer}`;\n      } else {\n        modelDescription = `${vlLocateMode()} mode`;\n      }\n    }\n    const dumpData: ExecutionDump = {\n      sdkVersion: getVersion(),\n      model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n      model_description: modelDescription,\n      logTime: Date.now(),\n      name: this.name,\n      tasks: this.tasks,\n    };\n    return dumpData;\n  }\n}\n","import { getPreferredLanguage } from '@midscene/shared/env';\n\nexport const elementDescriberInstruction = () => {\n  return `Tell what is the content of the element wrapped by the read rectangle in the screenshot. Your description is expected to be used to precisely locate the element from other similar elements on screenshot. Use ${getPreferredLanguage()} in the description.\n\nPlease follow the following rules:\n1. The description should be start with a brief description, like \"a button for confirming the action\".\n\n2. Include these information in the description to distinguish the element from its siblings and other similar elements, as much as possible:\n- The text of the element, like \"with text 'Confirm'\"\n- What the element looks like if it's an image, like \"with image '...'\"\n- The relative position of the element, like \"on the left of ..., around ...\"\n- How to distinguish the element from its siblings elements, like \"it is the icon instead of the text\"\n\n3. Do NOT mention the red rectangle in the description.\n\n4. Use the error field to describe the unexpected situations, if any. If not, put null.\n\nReturn in JSON:\n{\n  \"description\": \"[{brief description}]: {text of the element} {image of the element} {relative position of the element} ... \",\n  \"error\"?: \"...\"\n}`;\n};\n","import {\n  AIActionType,\n  type AIArgs,\n  callAiFn,\n  expandSearchArea,\n} from '@/ai-model/common';\nimport {\n  AiExtractElementInfo,\n  AiLocateElement,\n  callToGetJSONObject,\n} from '@/ai-model/index';\nimport { AiAssert, AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n  AIDescribeElementResponse,\n  AIElementResponse,\n  AISingleElementResponse,\n  AIUsageInfo,\n  BaseElement,\n  DetailedLocateParam,\n  DumpSubscriber,\n  InsightAction,\n  InsightAssertionResponse,\n  InsightExtractParam,\n  InsightOptions,\n  InsightTaskInfo,\n  LocateResult,\n  PartialInsightDumpFromSDK,\n  Rect,\n  UIContext,\n} from '@/types';\nimport {\n  MIDSCENE_FORCE_DEEP_THINK,\n  MIDSCENE_USE_QWEN_VL,\n  getAIConfigInBoolean,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext<BaseElement>;\n  callAI?: typeof callAiFn<AIElementResponse>;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\nconst debug = getDebug('ai:insight');\nexport default class Insight<\n  ElementType extends BaseElement = BaseElement,\n  ContextType extends UIContext<ElementType> = UIContext<ElementType>,\n> {\n  contextRetrieverFn: (\n    action: InsightAction,\n  ) => Promise<ContextType> | ContextType;\n\n  aiVendorFn: (...args: Array<any>) => Promise<any> = callAiFn;\n\n  onceDumpUpdatedFn?: DumpSubscriber;\n\n  taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n  constructor(\n    context:\n      | ContextType\n      | ((action: InsightAction) => Promise<ContextType> | ContextType),\n    opt?: InsightOptions,\n  ) {\n    assert(context, 'context is required for Insight');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\n\n    if (typeof opt?.aiVendorFn !== 'undefined') {\n      this.aiVendorFn = opt.aiVendorFn;\n    }\n    if (typeof opt?.taskInfo !== 'undefined') {\n      this.taskInfo = opt.taskInfo;\n    }\n  }\n\n  async locate(\n    query: DetailedLocateParam,\n    opt?: LocateOpts,\n  ): Promise<LocateResult> {\n    const { callAI } = opt || {};\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    const globalDeepThinkSwitch = getAIConfigInBoolean(\n      MIDSCENE_FORCE_DEEP_THINK,\n    );\n    if (globalDeepThinkSwitch) {\n      debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n    }\n    let searchAreaPrompt;\n    if (query.deepThink || globalDeepThinkSwitch) {\n      searchAreaPrompt = query.prompt;\n    }\n\n    if (searchAreaPrompt && !vlLocateMode()) {\n      console.warn(\n        'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n      );\n      searchAreaPrompt = undefined;\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn('locate'));\n\n    let searchArea: Rect | undefined = undefined;\n    let searchAreaRawResponse: string | undefined = undefined;\n    let searchAreaUsage: AIUsageInfo | undefined = undefined;\n    let searchAreaResponse:\n      | Awaited<ReturnType<typeof AiLocateSection>>\n      | undefined = undefined;\n    if (searchAreaPrompt) {\n      searchAreaResponse = await AiLocateSection({\n        context,\n        sectionDescription: searchAreaPrompt,\n      });\n      assert(\n        searchAreaResponse.rect,\n        `cannot find search area for \"${searchAreaPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const { parseResult, rect, elementById, rawResponse, usage } =\n      await AiLocateElement({\n        callAI: callAI || this.aiVendorFn,\n        context,\n        targetElementDescription: queryPrompt,\n        searchConfig: searchAreaResponse,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      matchedElement: [],\n      matchedRect: rect,\n      data: null,\n      taskInfo,\n      deepThink: !!searchArea,\n      error: errorLog,\n    };\n\n    const elements: BaseElement[] = [];\n    (parseResult.elements || []).forEach((item) => {\n      if ('id' in item) {\n        const element = elementById(item?.id);\n\n        if (!element) {\n          console.warn(\n            `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\n          );\n          return;\n        }\n        elements.push(element);\n      }\n    });\n\n    emitInsightDump(\n      {\n        ...dumpData,\n        matchedElement: elements,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog) {\n      throw new Error(errorLog);\n    }\n\n    assert(\n      elements.length <= 1,\n      `locate: multiple elements found, length = ${elements.length}`,\n    );\n\n    if (elements.length === 1) {\n      return {\n        element: {\n          id: elements[0]!.id,\n          indexId: elements[0]!.indexId,\n          center: elements[0]!.center,\n          rect: elements[0]!.rect,\n          xpaths: elements[0]!.xpaths || [],\n          attributes: elements[0]!.attributes,\n        },\n        rect,\n      };\n    }\n    return {\n      element: null,\n      rect,\n    };\n  }\n\n  async extract<T = any>(input: string): Promise<T>;\n  async extract<T extends Record<string, string>>(\n    input: T,\n  ): Promise<Record<keyof T, any>>;\n  async extract<T extends object>(input: Record<keyof T, string>): Promise<T>;\n\n  async extract<T>(dataDemand: InsightExtractParam): Promise<any> {\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const context = await this.contextRetrieverFn('extract');\n\n    const startTime = Date.now();\n    const { parseResult, usage } = await AiExtractElementInfo<T>({\n      context,\n      dataQuery: dataDemand,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(parseResult),\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'extract',\n      userQuery: {\n        dataDemand,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      error: errorLog,\n    };\n\n    const { data } = parseResult || {};\n\n    // 4\n    emitInsightDump(\n      {\n        ...dumpData,\n        data,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog && !data) {\n      throw new Error(errorLog);\n    }\n\n    return {\n      data,\n      usage,\n    };\n  }\n\n  async assert(assertion: string): Promise<InsightAssertionResponse> {\n    if (typeof assertion !== 'string') {\n      throw new Error(\n        'This is the assert method for Midscene, the first argument should be a string. If you want to use the assert method from Node.js, please import it from the Node.js assert module.',\n      );\n    }\n\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const context = await this.contextRetrieverFn('assert');\n    const startTime = Date.now();\n    const assertResult = await AiAssert({\n      assertion,\n      context,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(assertResult.content),\n    };\n\n    const { thought, pass } = assertResult.content;\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'assert',\n      userQuery: {\n        assertion,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      assertionPass: pass,\n      assertionThought: thought,\n      error: pass ? undefined : thought,\n    };\n    emitInsightDump(dumpData, dumpSubscriber);\n\n    return {\n      pass,\n      thought,\n      usage: assertResult.usage,\n    };\n  }\n  async describe(\n    target: Rect | [number, number],\n    opt?: {\n      deepThink?: boolean;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for insight.describe');\n    const context = await this.contextRetrieverFn('describe');\n    const { screenshotBase64 } = context;\n    assert(screenshotBase64, 'screenshot is required for insight.describe');\n\n    const systemPrompt = elementDescriberInstruction();\n\n    // Convert [x,y] center point to Rect if needed\n    const defaultRectSize = 30;\n    const targetRect: Rect = Array.isArray(target)\n      ? {\n          left: Math.floor(target[0] - defaultRectSize / 2),\n          top: Math.floor(target[1] - defaultRectSize / 2),\n          width: defaultRectSize,\n          height: defaultRectSize,\n        }\n      : target;\n\n    let imagePayload = await compositeElementInfoImg({\n      inputImgBase64: screenshotBase64,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepThink) {\n      const searchArea = expandSearchArea(targetRect, context.size);\n      debug('describe: set searchArea', searchArea);\n      imagePayload = await cropByRect(\n        imagePayload,\n        searchArea,\n        getAIConfigInBoolean(MIDSCENE_USE_QWEN_VL),\n      );\n    }\n\n    const msgs: AIArgs = [\n      { role: 'system', content: systemPrompt },\n      {\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: imagePayload,\n              detail: 'high',\n            },\n          },\n        ],\n      },\n    ];\n\n    const callAIFn =\n      this.aiVendorFn || callToGetJSONObject<AIDescribeElementResponse>;\n\n    const res = await callAIFn(msgs, AIActionType.DESCRIBE_ELEMENT);\n\n    const { content } = res;\n    assert(!content.error, `describe failed: ${content.error}`);\n    assert(content.description, 'failed to describe the element');\n    return content;\n  }\n}\n","import type {\n  DumpMeta,\n  DumpSubscriber,\n  InsightDump,\n  PartialInsightDumpFromSDK,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n  MIDSCENE_MODEL_NAME,\n  getAIConfig,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { uuid } from '@midscene/shared/utils';\n\nexport function emitInsightDump(\n  data: PartialInsightDumpFromSDK,\n  dumpSubscriber?: DumpSubscriber,\n) {\n  const baseData: DumpMeta = {\n    sdkVersion: getVersion(),\n    logTime: Date.now(),\n    model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n  };\n  const finalData: InsightDump = {\n    logId: uuid(),\n    ...baseData,\n    ...data,\n  };\n\n  dumpSubscriber?.(finalData);\n}\n","import { Executor } from './ai-model/action-executor';\nimport Insight from './insight/index';\nimport { getVersion } from './utils';\n\nexport {\n  plan,\n  describeUserPage,\n  AiLocateElement,\n  AiAssert,\n} from './ai-model/index';\n\nexport { getAIConfig, MIDSCENE_MODEL_NAME } from '@midscene/shared/env';\n\nexport type * from './types';\nexport default Insight;\nexport { Executor, Insight, getVersion };\n\nexport type {\n  MidsceneYamlScript,\n  MidsceneYamlTask,\n  MidsceneYamlFlowItem,\n} from './yaml';\n"]}
+{"version":3,"mappings":";;;;;;;;;;;;;;;;AAUA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,cAAc;AAEhB,IAAM,WAAN,MAAe;AAAA,EAUpB,YACE,MACA,SAGA;AACA,SAAK,SACH,SAAS,SAAS,QAAQ,MAAM,SAAS,IAAI,YAAY;AAC3D,SAAK,OAAO;AACZ,SAAK,SAAS,SAAS,SAAS,CAAC,GAAG;AAAA,MAAI,CAAC,SACvC,KAAK,kBAAkB,IAAI;AAAA,IAC7B;AACA,SAAK,cAAc,SAAS;AAAA,EAC9B;AAAA,EAEQ,kBAAkB,MAAyC;AACjE,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,MAAgE;AAC3E;AAAA,MACE,KAAK,WAAW;AAAA,MAChB;AAAA,QAAyD,KAAK,gBAAgB,GAAG,KAAK;AAAA,EAAK,KAAK,gBAAgB,GAAG,UAAU;AAAA,IAC/H;AACA,QAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,WAAK,MAAM,KAAK,GAAG,KAAK,IAAI,CAAC,SAAS,KAAK,kBAAkB,IAAI,CAAC,CAAC;AAAA,IACrE,OAAO;AACL,WAAK,MAAM,KAAK,KAAK,kBAAkB,IAAI,CAAC;AAAA,IAC9C;AACA,QAAI,KAAK,WAAW,WAAW;AAC7B,WAAK,SAAS;AAAA,IAChB;AAAA,EACF;AAAA,EAEA,MAAM,QAAsB;AAC1B,QAAI,KAAK,WAAW,UAAU,KAAK,MAAM,SAAS,GAAG;AACnD,cAAQ;AAAA,QACN;AAAA,MACF;AAAA,IACF;AAEA,WAAO,KAAK,WAAW,WAAW,6BAA6B;AAC/D,WAAO,KAAK,WAAW,aAAa,+BAA+B;AACnE,WAAO,KAAK,WAAW,SAAS,4BAA4B;AAE5D,UAAM,mBAAmB,KAAK,MAAM;AAAA,MAClC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,mBAAmB,GAAG;AAExB;AAAA,IACF;AAEA,SAAK,SAAS;AACd,QAAI,YAAY;AAChB,QAAI,wBAAwB;AAE5B,QAAI;AAEJ,WAAO,YAAY,KAAK,MAAM,QAAQ;AACpC,YAAM,OAAO,KAAK,MAAM,SAAS;AACjC;AAAA,QACE,KAAK,WAAW;AAAA,QAChB,2CAA2C,KAAK,MAAM;AAAA,MACxD;AACA,WAAK,SAAS;AAAA,QACZ,OAAO,KAAK,IAAI;AAAA,MAClB;AACA,UAAI;AACF,aAAK,SAAS;AACd,YAAI;AACF,cAAI,KAAK,aAAa;AACpB,kBAAM,KAAK,YAAY,IAAI;AAAA,UAC7B;AAAA,QACF,SAAS,GAAG;AACV,kBAAQ,MAAM,wBAAwB,CAAC;AAAA,QACzC;AACA;AAAA,UACE,CAAC,WAAW,UAAU,UAAU,EAAE,QAAQ,KAAK,IAAI,KAAK;AAAA,UACxD,0BAA0B,KAAK,IAAI;AAAA,QACrC;AAEA,cAAM,EAAE,UAAU,MAAM,IAAI;AAC5B,eAAO,UAAU,uCAAuC,KAAK,IAAI,EAAE;AAEnE,YAAI;AACJ,cAAM,kBAAmC;AAAA,UACvC;AAAA,UACA,SAAS,oBAAoB;AAAA,QAC/B;AAEA,YAAI,KAAK,SAAS,WAAW;AAC3B;AAAA,YACE,KAAK,YAAY,YACf,KAAK,YAAY,WACjB,KAAK,YAAY,YACjB,KAAK,YAAY,aACjB,KAAK,YAAY,YACjB,KAAK,YAAY;AAAA,YACnB,gCAAgC,KAAK,OAAO;AAAA,UAC9C;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AACxD,cAAI,KAAK,YAAY,UAAU;AAC7B,iCACE,aACC;AAAA,UACL;AAAA,QACF,WAAW,KAAK,SAAS,YAAY,KAAK,SAAS,YAAY;AAC7D,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D,OAAO;AACL,kBAAQ;AAAA,YACN,0BAA0B,KAAK,IAAI;AAAA,UACrC;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D;AAEA,eAAO,OAAO,MAAM,WAAW;AAC/B,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD,aAAK,OAAO,SAAU,aAAqB,UAAU;AACrD;AAAA,MACF,SAAS,GAAQ;AACf,gCAAwB;AACxB,aAAK,QACH,GAAG,YAAY,OAAO,MAAM,WAAW,IAAI;AAC7C,aAAK,aAAa,EAAE;AAEpB,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF;AAAA,IACF;AAGA,aAAS,IAAI,YAAY,GAAG,IAAI,KAAK,MAAM,QAAQ,KAAK;AACtD,WAAK,MAAM,CAAC,EAAE,SAAS;AAAA,IACzB;AAEA,QAAI,uBAAuB;AACzB,WAAK,SAAS;AAAA,IAChB,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AAEA,QAAI,KAAK,MAAM,QAAQ;AAErB,YAAM,cAAc,KAAK,IAAI,WAAW,KAAK,MAAM,SAAS,CAAC;AAC7D,aAAO,KAAK,MAAM,WAAW,EAAE;AAAA,IACjC;AAAA,EACF;AAAA,EAEA,iBAA0B;AACxB,WAAO,KAAK,WAAW;AAAA,EACzB;AAAA,EAEA,kBAAwC;AACtC,QAAI,KAAK,WAAW,SAAS;AAC3B,aAAO;AAAA,IACT;AACA,UAAM,iBAAiB,KAAK,MAAM;AAAA,MAChC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,kBAAkB,GAAG;AACvB,aAAO,KAAK,MAAM,cAAc;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AAAA,EAEA,OAAsB;AACpB,QAAI,mBAAmB;AAEvB,QAAI,aAAa,GAAG;AAClB,YAAM,iBAAiB,mBAAmB;AAC1C,UAAI,gBAAgB;AAClB,2BAAmB,WAAW,cAAc;AAAA,MAC9C,OAAO;AACL,2BAAmB,GAAG,aAAa,CAAC;AAAA,MACtC;AAAA,IACF;AACA,UAAM,WAA0B;AAAA,MAC9B,YAAY,WAAW;AAAA,MACvB,YAAY,YAAY,mBAAmB,KAAK;AAAA,MAChD,mBAAmB;AAAA,MACnB,SAAS,KAAK,IAAI;AAAA,MAClB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACF;;;AC9NA,SAAS,4BAA4B;AAE9B,IAAM,8BAA8B,MAAM;AAC/C,SAAO,mNAAmN,qBAAqB,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAoBlP;;;ACQA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA,gBAAAA;AAAA,OACK;AACP,SAAS,yBAAyB,kBAAkB;AACpD,SAAS,gBAAgB;AACzB,SAAS,UAAAC,eAAc;;;AChCvB;AAAA,EACE,uBAAAC;AAAA,EACA,eAAAC;AAAA,OAGK;AACP,SAAS,YAAY;AAEd,SAAS,gBACd,MACA,gBACA;AACA,QAAM,WAAqB;AAAA,IACzB,YAAY,WAAW;AAAA,IACvB,SAAS,KAAK,IAAI;AAAA,IAClB,YAAYA,aAAYD,oBAAmB,KAAK;AAAA,EAClD;AACA,QAAM,YAAyB;AAAA,IAC7B,OAAO,KAAK;AAAA,IACZ,GAAG;AAAA,IACH,GAAG;AAAA,EACL;AAEA,mBAAiB,SAAS;AAC5B;;;ADoBA,IAAM,QAAQ,SAAS,YAAY;AACnC,IAAqB,UAArB,MAGE;AAAA,EAWA,YACE,SAGA,KACA;AAXF,sBAAoD;AAYlD,IAAAD,QAAO,SAAS,iCAAiC;AACjD,QAAI,OAAO,YAAY,YAAY;AACjC,WAAK,qBAAqB;AAAA,IAC5B,OAAO;AACL,WAAK,qBAAqB,MAAM,QAAQ,QAAQ,OAAO;AAAA,IACzD;AAEA,QAAI,OAAO,KAAK,eAAe,aAAa;AAC1C,WAAK,aAAa,IAAI;AAAA,IACxB;AACA,QAAI,OAAO,KAAK,aAAa,aAAa;AACxC,WAAK,WAAW,IAAI;AAAA,IACtB;AAAA,EACF;AAAA,EAEA,MAAM,OACJ,OACA,KACuB;AACvB,UAAM,EAAE,OAAO,IAAI,OAAO,CAAC;AAC3B,UAAM,cAAc,OAAO,UAAU,WAAW,QAAQ,MAAM;AAC9D,IAAAA,QAAO,aAAa,8BAA8B;AAClD,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,IAAAA,QAAO,OAAO,UAAU,UAAU,sCAAsC;AAExE,UAAM,wBAAwB;AAAA,MAC5B;AAAA,IACF;AACA,QAAI,uBAAuB;AACzB,YAAM,yBAAyB,qBAAqB;AAAA,IACtD;AACA,QAAI;AACJ,QAAI,MAAM,aAAa,uBAAuB;AAC5C,yBAAmB,MAAM;AAAA,IAC3B;AAEA,QAAI,oBAAoB,CAACD,cAAa,GAAG;AACvC,cAAQ;AAAA,QACN;AAAA,MACF;AACA,yBAAmB;AAAA,IACrB;AAEA,UAAM,UAAU,KAAK,WAAY,MAAM,KAAK,mBAAmB,QAAQ;AAEvE,QAAI,aAA+B;AACnC,QAAI,wBAA4C;AAChD,QAAI,kBAA2C;AAC/C,QAAI,qBAEY;AAChB,QAAI,kBAAkB;AACpB,2BAAqB,MAAM,gBAAgB;AAAA,QACzC;AAAA,QACA,oBAAoB;AAAA,MACtB,CAAC;AACD,MAAAC;AAAA,QACE,mBAAmB;AAAA,QACnB,gCAAgC,gBAAgB,IAC9C,mBAAmB,QAAQ,KAAK,mBAAmB,KAAK,KAAK,EAC/D;AAAA,MACF;AACA,8BAAwB,mBAAmB;AAC3C,wBAAkB,mBAAmB;AACrC,mBAAa,mBAAmB;AAAA,IAClC;AAEA,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,aAAa,aAAa,MAAM,IACzD,MAAM,gBAAgB;AAAA,MACpB,QAAQ,UAAU,KAAK;AAAA,MACvB;AAAA,MACA,0BAA0B;AAAA,MAC1B,cAAc;AAAA,IAChB,CAAC;AAEH,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,MACvC,gBAAgB,KAAK,UAAU,WAAW;AAAA,MAC1C;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAgC,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAC1E;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT,SAAS;AAAA,MACX;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,aAAa;AAAA,MACb,MAAM;AAAA,MACN;AAAA,MACA,WAAW,CAAC,CAAC;AAAA,MACb,OAAO;AAAA,IACT;AAEA,UAAM,WAA0B,CAAC;AACjC,KAAC,YAAY,YAAY,CAAC,GAAG,QAAQ,CAAC,SAAS;AAC7C,UAAI,QAAQ,MAAM;AAChB,cAAM,UAAU,YAAY,MAAM,EAAE;AAEpC,YAAI,CAAC,SAAS;AACZ,kBAAQ;AAAA,YACN,kCAAkC,KAAK,EAAE;AAAA,UAC3C;AACA;AAAA,QACF;AACA,iBAAS,KAAK,OAAO;AAAA,MACvB;AAAA,IACF,CAAC;AAED;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,IACF;AAEA,QAAI,UAAU;AACZ,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,IAAAA;AAAA,MACE,SAAS,UAAU;AAAA,MACnB,6CAA6C,SAAS,MAAM;AAAA,IAC9D;AAEA,QAAI,SAAS,WAAW,GAAG;AACzB,aAAO;AAAA,QACL,SAAS;AAAA,UACP,IAAI,SAAS,CAAC,EAAG;AAAA,UACjB,SAAS,SAAS,CAAC,EAAG;AAAA,UACtB,QAAQ,SAAS,CAAC,EAAG;AAAA,UACrB,MAAM,SAAS,CAAC,EAAG;AAAA,UACnB,QAAQ,SAAS,CAAC,EAAG,UAAU,CAAC;AAAA,UAChC,YAAY,SAAS,CAAC,EAAG;AAAA,QAC3B;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,WAAO;AAAA,MACL,SAAS;AAAA,MACT;AAAA,IACF;AAAA,EACF;AAAA,EAYA,MAAM,QACJ,YACA,KACc;AACd,IAAAA;AAAA,MACE,OAAO,eAAe,YAAY,OAAO,eAAe;AAAA,MACxD,kDAAkD,OAAO,UAAU;AAAA,IACrE;AACA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,SAAS;AAEvD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,IAAI,MAAM,qBAAwB;AAAA,MAC3D;AAAA,MACA,WAAW;AAAA,MACX,eAAe;AAAA,IACjB,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,IACzC;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAwB,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAClE;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,OAAO;AAAA,IACT;AAEA,UAAM,EAAE,KAAK,IAAI,eAAe,CAAC;AAGjC;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH;AAAA,MACF;AAAA,MACA;AAAA,IACF;AAEA,QAAI,YAAY,CAAC,MAAM;AACrB,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,WAAO;AAAA,MACL;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,WAAsD;AACjE,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,QAAQ;AACtD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,eAAe,MAAM,SAAS;AAAA,MAClC;AAAA,MACA;AAAA,IACF,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,aAAa,OAAO;AAAA,IAClD;AAEA,UAAM,EAAE,SAAS,KAAK,IAAI,aAAa;AACvC,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,eAAe;AAAA,MACf,kBAAkB;AAAA,MAClB,OAAO,OAAO,SAAY;AAAA,IAC5B;AACA,oBAAgB,UAAU,cAAc;AAExC,WAAO;AAAA,MACL;AAAA,MACA;AAAA,MACA,OAAO,aAAa;AAAA,IACtB;AAAA,EACF;AAAA,EACA,MAAM,SACJ,QACA,KAGyD;AACzD,IAAAA,QAAO,QAAQ,yCAAyC;AACxD,UAAM,UAAU,MAAM,KAAK,mBAAmB,UAAU;AACxD,UAAM,EAAE,iBAAiB,IAAI;AAC7B,IAAAA,QAAO,kBAAkB,6CAA6C;AAEtE,UAAM,eAAe,4BAA4B;AAGjD,UAAM,kBAAkB;AACxB,UAAM,aAAmB,MAAM,QAAQ,MAAM,IACzC;AAAA,MACE,MAAM,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAChD,KAAK,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAC/C,OAAO;AAAA,MACP,QAAQ;AAAA,IACV,IACA;AAEJ,QAAI,eAAe,MAAM,wBAAwB;AAAA,MAC/C,gBAAgB;AAAA,MAChB,sBAAsB;AAAA,QACpB;AAAA,UACE,MAAM;AAAA,QACR;AAAA,MACF;AAAA,MACA,iBAAiB;AAAA,IACnB,CAAC;AAED,QAAI,KAAK,WAAW;AAClB,YAAM,aAAa,iBAAiB,YAAY,QAAQ,IAAI;AAC5D,YAAM,4BAA4B,UAAU;AAC5C,qBAAe,MAAM;AAAA,QACnB;AAAA,QACA;AAAA,QACA,qBAAqB,oBAAoB;AAAA,MAC3C;AAAA,IACF;AAEA,UAAM,OAAe;AAAA,MACnB,EAAE,MAAM,UAAU,SAAS,aAAa;AAAA,MACxC;AAAA,QACE,MAAM;AAAA,QACN,SAAS;AAAA,UACP;AAAA,YACE,MAAM;AAAA,YACN,WAAW;AAAA,cACT,KAAK;AAAA,cACL,QAAQ;AAAA,YACV;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,WACJ,KAAK,cAAc;AAErB,UAAM,MAAM,MAAM,SAAS,8BAAmC;AAE9D,UAAM,EAAE,QAAQ,IAAI;AACpB,IAAAA,QAAO,CAAC,QAAQ,OAAO,oBAAoB,QAAQ,KAAK,EAAE;AAC1D,IAAAA,QAAO,QAAQ,aAAa,gCAAgC;AAC5D,WAAO;AAAA,EACT;AACF;;;AEzZA,SAAS,eAAAE,cAAa,uBAAAD,4BAA2B;AAGjD,IAAO,cAAQ","names":["vlLocateMode","assert","MIDSCENE_MODEL_NAME","getAIConfig"],"ignoreList":[],"sources":["../../src/ai-model/action-executor.ts","../../src/ai-model/prompt/describe.ts","../../src/insight/index.ts","../../src/insight/utils.ts","../../src/index.ts"],"sourcesContent":["import type {\n  ExecutionDump,\n  ExecutionTask,\n  ExecutionTaskApply,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskProgressOptions,\n  ExecutionTaskReturn,\n  ExecutorContext,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n  MIDSCENE_MODEL_NAME,\n  getAIConfig,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\n\nexport class Executor {\n  name: string;\n\n  tasks: ExecutionTask[];\n\n  // status of executor\n  status: 'init' | 'pending' | 'running' | 'completed' | 'error';\n\n  onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n\n  constructor(\n    name: string,\n    options?: ExecutionTaskProgressOptions & {\n      tasks?: ExecutionTaskApply[];\n    },\n  ) {\n    this.status =\n      options?.tasks && options.tasks.length > 0 ? 'pending' : 'init';\n    this.name = name;\n    this.tasks = (options?.tasks || []).map((item) =>\n      this.markTaskAsPending(item),\n    );\n    this.onTaskStart = options?.onTaskStart;\n  }\n\n  private markTaskAsPending(task: ExecutionTaskApply): ExecutionTask {\n    return {\n      status: 'pending',\n      ...task,\n    };\n  }\n\n  async append(task: ExecutionTaskApply[] | ExecutionTaskApply): Promise<void> {\n    assert(\n      this.status !== 'error',\n      `executor is in error state, cannot append task\\nerror=${this.latestErrorTask()?.error}\\n${this.latestErrorTask()?.errorStack}`,\n    );\n    if (Array.isArray(task)) {\n      this.tasks.push(...task.map((item) => this.markTaskAsPending(item)));\n    } else {\n      this.tasks.push(this.markTaskAsPending(task));\n    }\n    if (this.status !== 'running') {\n      this.status = 'pending';\n    }\n  }\n\n  async flush(): Promise<any> {\n    if (this.status === 'init' && this.tasks.length > 0) {\n      console.warn(\n        'illegal state for executor, status is init but tasks are not empty',\n      );\n    }\n\n    assert(this.status !== 'running', 'executor is already running');\n    assert(this.status !== 'completed', 'executor is already completed');\n    assert(this.status !== 'error', 'executor is in error state');\n\n    const nextPendingIndex = this.tasks.findIndex(\n      (task) => task.status === 'pending',\n    );\n    if (nextPendingIndex < 0) {\n      // all tasks are completed\n      return;\n    }\n\n    this.status = 'running';\n    let taskIndex = nextPendingIndex;\n    let successfullyCompleted = true;\n\n    let previousFindOutput: ExecutionTaskInsightLocateOutput | undefined;\n\n    while (taskIndex < this.tasks.length) {\n      const task = this.tasks[taskIndex];\n      assert(\n        task.status === 'pending',\n        `task status should be pending, but got: ${task.status}`,\n      );\n      task.timing = {\n        start: Date.now(),\n      };\n      try {\n        task.status = 'running';\n        try {\n          if (this.onTaskStart) {\n            await this.onTaskStart(task);\n          }\n        } catch (e) {\n          console.error('error in onTaskStart', e);\n        }\n        assert(\n          ['Insight', 'Action', 'Planning'].indexOf(task.type) >= 0,\n          `unsupported task type: ${task.type}`,\n        );\n\n        const { executor, param } = task;\n        assert(executor, `executor is required for task type: ${task.type}`);\n\n        let returnValue;\n        const executorContext: ExecutorContext = {\n          task,\n          element: previousFindOutput?.element,\n        };\n\n        if (task.type === 'Insight') {\n          assert(\n            task.subType === 'Locate' ||\n              task.subType === 'Query' ||\n              task.subType === 'Assert' ||\n              task.subType === 'Boolean' ||\n              task.subType === 'Number' ||\n              task.subType === 'String',\n            `unsupported insight subType: ${task.subType}`,\n          );\n          returnValue = await task.executor(param, executorContext);\n          if (task.subType === 'Locate') {\n            previousFindOutput = (\n              returnValue as ExecutionTaskReturn<ExecutionTaskInsightLocateOutput>\n            )?.output;\n          }\n        } else if (task.type === 'Action' || task.type === 'Planning') {\n          returnValue = await task.executor(param, executorContext);\n        } else {\n          console.warn(\n            `unsupported task type: ${task.type}, will try to execute it directly`,\n          );\n          returnValue = await task.executor(param, executorContext);\n        }\n\n        Object.assign(task, returnValue);\n        task.status = 'finished';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        task.timing.aiCost = (returnValue as any)?.aiCost || 0;\n        taskIndex++;\n      } catch (e: any) {\n        successfullyCompleted = false;\n        task.error =\n          e?.message || (typeof e === 'string' ? e : 'error-without-message');\n        task.errorStack = e.stack;\n\n        task.status = 'failed';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        break;\n      }\n    }\n\n    // set all remaining tasks as cancelled\n    for (let i = taskIndex + 1; i < this.tasks.length; i++) {\n      this.tasks[i].status = 'cancelled';\n    }\n\n    if (successfullyCompleted) {\n      this.status = 'completed';\n    } else {\n      this.status = 'error';\n    }\n\n    if (this.tasks.length) {\n      // return the last output\n      const outputIndex = Math.min(taskIndex, this.tasks.length - 1);\n      return this.tasks[outputIndex].output;\n    }\n  }\n\n  isInErrorState(): boolean {\n    return this.status === 'error';\n  }\n\n  latestErrorTask(): ExecutionTask | null {\n    if (this.status !== 'error') {\n      return null;\n    }\n    const errorTaskIndex = this.tasks.findIndex(\n      (task) => task.status === 'failed',\n    );\n    if (errorTaskIndex >= 0) {\n      return this.tasks[errorTaskIndex];\n    }\n    return null;\n  }\n\n  dump(): ExecutionDump {\n    let modelDescription = '';\n\n    if (vlLocateMode()) {\n      const uiTarsModelVer = uiTarsModelVersion();\n      if (uiTarsModelVer) {\n        modelDescription = `UI-TARS=${uiTarsModelVer}`;\n      } else {\n        modelDescription = `${vlLocateMode()} mode`;\n      }\n    }\n    const dumpData: ExecutionDump = {\n      sdkVersion: getVersion(),\n      model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n      model_description: modelDescription,\n      logTime: Date.now(),\n      name: this.name,\n      tasks: this.tasks,\n    };\n    return dumpData;\n  }\n}\n","import { getPreferredLanguage } from '@midscene/shared/env';\n\nexport const elementDescriberInstruction = () => {\n  return `Tell what is the content of the element wrapped by the read rectangle in the screenshot. Your description is expected to be used to precisely locate the element from other similar elements on screenshot. Use ${getPreferredLanguage()} in the description.\n\nPlease follow the following rules:\n1. The description should be start with a brief description, like \"a button for confirming the action\".\n\n2. Include these information in the description to distinguish the element from its siblings and other similar elements, as much as possible:\n- The text of the element, like \"with text 'Confirm'\"\n- What the element looks like if it's an image, like \"with image '...'\"\n- The relative position of the element, like \"on the left of ..., around ...\"\n- How to distinguish the element from its siblings elements, like \"it is the icon instead of the text\"\n\n3. Do NOT mention the red rectangle in the description.\n\n4. Use the error field to describe the unexpected situations, if any. If not, put null.\n\nReturn in JSON:\n{\n  \"description\": \"[{brief description}]: {text of the element} {image of the element} {relative position of the element} ... \",\n  \"error\"?: \"...\"\n}`;\n};\n","import {\n  AIActionType,\n  type AIArgs,\n  callAiFn,\n  expandSearchArea,\n} from '@/ai-model/common';\nimport {\n  AiExtractElementInfo,\n  AiLocateElement,\n  callToGetJSONObject,\n} from '@/ai-model/index';\nimport { AiAssert, AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n  AIDescribeElementResponse,\n  AIElementResponse,\n  AIUsageInfo,\n  BaseElement,\n  DetailedLocateParam,\n  DumpSubscriber,\n  InsightAction,\n  InsightAssertionResponse,\n  InsightExtractOption,\n  InsightExtractParam,\n  InsightOptions,\n  InsightTaskInfo,\n  LocateResult,\n  PartialInsightDumpFromSDK,\n  Rect,\n  UIContext,\n} from '@/types';\nimport {\n  MIDSCENE_FORCE_DEEP_THINK,\n  MIDSCENE_USE_QWEN_VL,\n  getAIConfigInBoolean,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext<BaseElement>;\n  callAI?: typeof callAiFn<AIElementResponse>;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\nconst debug = getDebug('ai:insight');\nexport default class Insight<\n  ElementType extends BaseElement = BaseElement,\n  ContextType extends UIContext<ElementType> = UIContext<ElementType>,\n> {\n  contextRetrieverFn: (\n    action: InsightAction,\n  ) => Promise<ContextType> | ContextType;\n\n  aiVendorFn: (...args: Array<any>) => Promise<any> = callAiFn;\n\n  onceDumpUpdatedFn?: DumpSubscriber;\n\n  taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n  constructor(\n    context:\n      | ContextType\n      | ((action: InsightAction) => Promise<ContextType> | ContextType),\n    opt?: InsightOptions,\n  ) {\n    assert(context, 'context is required for Insight');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\n\n    if (typeof opt?.aiVendorFn !== 'undefined') {\n      this.aiVendorFn = opt.aiVendorFn;\n    }\n    if (typeof opt?.taskInfo !== 'undefined') {\n      this.taskInfo = opt.taskInfo;\n    }\n  }\n\n  async locate(\n    query: DetailedLocateParam,\n    opt?: LocateOpts,\n  ): Promise<LocateResult> {\n    const { callAI } = opt || {};\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    const globalDeepThinkSwitch = getAIConfigInBoolean(\n      MIDSCENE_FORCE_DEEP_THINK,\n    );\n    if (globalDeepThinkSwitch) {\n      debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n    }\n    let searchAreaPrompt;\n    if (query.deepThink || globalDeepThinkSwitch) {\n      searchAreaPrompt = query.prompt;\n    }\n\n    if (searchAreaPrompt && !vlLocateMode()) {\n      console.warn(\n        'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n      );\n      searchAreaPrompt = undefined;\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn('locate'));\n\n    let searchArea: Rect | undefined = undefined;\n    let searchAreaRawResponse: string | undefined = undefined;\n    let searchAreaUsage: AIUsageInfo | undefined = undefined;\n    let searchAreaResponse:\n      | Awaited<ReturnType<typeof AiLocateSection>>\n      | undefined = undefined;\n    if (searchAreaPrompt) {\n      searchAreaResponse = await AiLocateSection({\n        context,\n        sectionDescription: searchAreaPrompt,\n      });\n      assert(\n        searchAreaResponse.rect,\n        `cannot find search area for \"${searchAreaPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const { parseResult, rect, elementById, rawResponse, usage } =\n      await AiLocateElement({\n        callAI: callAI || this.aiVendorFn,\n        context,\n        targetElementDescription: queryPrompt,\n        searchConfig: searchAreaResponse,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      matchedElement: [],\n      matchedRect: rect,\n      data: null,\n      taskInfo,\n      deepThink: !!searchArea,\n      error: errorLog,\n    };\n\n    const elements: BaseElement[] = [];\n    (parseResult.elements || []).forEach((item) => {\n      if ('id' in item) {\n        const element = elementById(item?.id);\n\n        if (!element) {\n          console.warn(\n            `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\n          );\n          return;\n        }\n        elements.push(element);\n      }\n    });\n\n    emitInsightDump(\n      {\n        ...dumpData,\n        matchedElement: elements,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog) {\n      throw new Error(errorLog);\n    }\n\n    assert(\n      elements.length <= 1,\n      `locate: multiple elements found, length = ${elements.length}`,\n    );\n\n    if (elements.length === 1) {\n      return {\n        element: {\n          id: elements[0]!.id,\n          indexId: elements[0]!.indexId,\n          center: elements[0]!.center,\n          rect: elements[0]!.rect,\n          xpaths: elements[0]!.xpaths || [],\n          attributes: elements[0]!.attributes,\n        },\n        rect,\n      };\n    }\n    return {\n      element: null,\n      rect,\n    };\n  }\n\n  async extract<T = any>(input: string, opt?: InsightExtractOption): Promise<T>;\n  async extract<T extends Record<string, string>>(\n    input: T,\n    opt?: InsightExtractOption,\n  ): Promise<Record<keyof T, any>>;\n  async extract<T extends object>(\n    input: Record<keyof T, string>,\n    opt?: InsightExtractOption,\n  ): Promise<T>;\n\n  async extract<T>(\n    dataDemand: InsightExtractParam,\n    opt?: InsightExtractOption,\n  ): Promise<any> {\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const context = await this.contextRetrieverFn('extract');\n\n    const startTime = Date.now();\n    const { parseResult, usage } = await AiExtractElementInfo<T>({\n      context,\n      dataQuery: dataDemand,\n      extractOption: opt,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(parseResult),\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'extract',\n      userQuery: {\n        dataDemand,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      error: errorLog,\n    };\n\n    const { data } = parseResult || {};\n\n    // 4\n    emitInsightDump(\n      {\n        ...dumpData,\n        data,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog && !data) {\n      throw new Error(errorLog);\n    }\n\n    return {\n      data,\n      usage,\n    };\n  }\n\n  async assert(assertion: string): Promise<InsightAssertionResponse> {\n    if (typeof assertion !== 'string') {\n      throw new Error(\n        'This is the assert method for Midscene, the first argument should be a string. If you want to use the assert method from Node.js, please import it from the Node.js assert module.',\n      );\n    }\n\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const context = await this.contextRetrieverFn('assert');\n    const startTime = Date.now();\n    const assertResult = await AiAssert({\n      assertion,\n      context,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(assertResult.content),\n    };\n\n    const { thought, pass } = assertResult.content;\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'assert',\n      userQuery: {\n        assertion,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      assertionPass: pass,\n      assertionThought: thought,\n      error: pass ? undefined : thought,\n    };\n    emitInsightDump(dumpData, dumpSubscriber);\n\n    return {\n      pass,\n      thought,\n      usage: assertResult.usage,\n    };\n  }\n  async describe(\n    target: Rect | [number, number],\n    opt?: {\n      deepThink?: boolean;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for insight.describe');\n    const context = await this.contextRetrieverFn('describe');\n    const { screenshotBase64 } = context;\n    assert(screenshotBase64, 'screenshot is required for insight.describe');\n\n    const systemPrompt = elementDescriberInstruction();\n\n    // Convert [x,y] center point to Rect if needed\n    const defaultRectSize = 30;\n    const targetRect: Rect = Array.isArray(target)\n      ? {\n          left: Math.floor(target[0] - defaultRectSize / 2),\n          top: Math.floor(target[1] - defaultRectSize / 2),\n          width: defaultRectSize,\n          height: defaultRectSize,\n        }\n      : target;\n\n    let imagePayload = await compositeElementInfoImg({\n      inputImgBase64: screenshotBase64,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepThink) {\n      const searchArea = expandSearchArea(targetRect, context.size);\n      debug('describe: set searchArea', searchArea);\n      imagePayload = await cropByRect(\n        imagePayload,\n        searchArea,\n        getAIConfigInBoolean(MIDSCENE_USE_QWEN_VL),\n      );\n    }\n\n    const msgs: AIArgs = [\n      { role: 'system', content: systemPrompt },\n      {\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: imagePayload,\n              detail: 'high',\n            },\n          },\n        ],\n      },\n    ];\n\n    const callAIFn =\n      this.aiVendorFn || callToGetJSONObject<AIDescribeElementResponse>;\n\n    const res = await callAIFn(msgs, AIActionType.DESCRIBE_ELEMENT);\n\n    const { content } = res;\n    assert(!content.error, `describe failed: ${content.error}`);\n    assert(content.description, 'failed to describe the element');\n    return content;\n  }\n}\n","import type {\n  DumpMeta,\n  DumpSubscriber,\n  InsightDump,\n  PartialInsightDumpFromSDK,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n  MIDSCENE_MODEL_NAME,\n  getAIConfig,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { uuid } from '@midscene/shared/utils';\n\nexport function emitInsightDump(\n  data: PartialInsightDumpFromSDK,\n  dumpSubscriber?: DumpSubscriber,\n) {\n  const baseData: DumpMeta = {\n    sdkVersion: getVersion(),\n    logTime: Date.now(),\n    model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n  };\n  const finalData: InsightDump = {\n    logId: uuid(),\n    ...baseData,\n    ...data,\n  };\n\n  dumpSubscriber?.(finalData);\n}\n","import { Executor } from './ai-model/action-executor';\nimport Insight from './insight/index';\nimport { getVersion } from './utils';\n\nexport {\n  plan,\n  describeUserPage,\n  AiLocateElement,\n  AiAssert,\n} from './ai-model/index';\n\nexport { getAIConfig, MIDSCENE_MODEL_NAME } from '@midscene/shared/env';\n\nexport type * from './types';\nexport default Insight;\nexport { Executor, Insight, getVersion };\n\nexport type {\n  MidsceneYamlScript,\n  MidsceneYamlTask,\n  MidsceneYamlFlowItem,\n  MidsceneYamlFlowItemAIRightClick,\n} from './yaml';\n"]}

package/dist/lib/{llm-planning-9c97f5c2.d.ts → llm-planning-573b9b34.d.ts}

@@ -1,4 +1,5 @@
-import { k as AIUsageInfo, R as Rect, j as ElementTreeNode, B as BaseElement, U as UIContext, ax as ReferenceImage, p as AIElementLocatorResponse, O as ElementById, s as AIDataExtractionResponse, u as AIAssertionResponse, av as PageType, Y as PlanningAIResponse } from './types-81bb2c02.js';
+import { l as AIUsageInfo, U as UIContext, av as ReferenceImage, q as AIElementLocatorResponse, K as ElementById, g as InsightExtractOption, t as AIDataExtractionResponse, v as AIAssertionResponse, at as PageType, V as PlanningAIResponse } from './types-8c197f92.js';
+import { Rect, ElementTreeNode, BaseElement } from '@midscene/shared/types';
 import { ChatCompletionSystemMessageParam, ChatCompletionUserMessageParam } from 'openai/resources';
 
 type AIArgs = [
@@ -28,6 +29,8 @@ declare function elementByPositionWithElementInfo(treeRoot: ElementTreeNode<Base
 declare function describeUserPage<ElementType extends BaseElement = BaseElement>(context: Omit<UIContext<ElementType>, 'describer'>, opt?: {
     truncateTextLength?: number;
     filterNonTextContent?: boolean;
+    domIncluded?: boolean;
+    visibleOnly?: boolean;
 }): Promise<{
     description: string;
     elementById(idOrIndexId: string): ElementType;
@@ -75,6 +78,7 @@ declare function AiLocateSection(options: {
 declare function AiExtractElementInfo<T, ElementType extends BaseElement = BaseElement>(options: {
     dataQuery: string | Record<string, string>;
     context: UIContext<ElementType>;
+    extractOption?: InsightExtractOption;
 }): Promise<{
     parseResult: AIDataExtractionResponse<T>;
     elementById: (idOrIndexId: string) => ElementType;

package/dist/lib/tree.d.ts

@@ -1,12 +1 @@
-import * as _midscene_shared_constants from '@midscene/shared/constants';
-import { B as BaseElement, j as ElementTreeNode } from './types-81bb2c02.js';
-import 'openai/resources';
-
-declare function truncateText(text: string | number | object | undefined, maxLength?: number): string;
-declare function trimAttributes(attributes: Record<string, any>, truncateTextLength?: number): {
-    [key: string]: string;
-    nodeType: _midscene_shared_constants.NodeType;
-};
-declare function descriptionOfTree<ElementType extends BaseElement = BaseElement>(tree: ElementTreeNode<ElementType>, truncateTextLength?: number, filterNonTextContent?: boolean): string;
-
-export { descriptionOfTree, trimAttributes, truncateText };
+export { descriptionOfTree, trimAttributes, truncateText } from '@midscene/shared/extractor';

package/dist/lib/tree.js

@@ -1,102 +1,13 @@
-"use strict";Object.defineProperty(exports, "__esModule", {value: true}); function _optionalChain(ops) { let lastAccessLHS = undefined; let value = ops[0]; let i = 1; while (i < ops.length) { const op = ops[i]; const fn = ops[i + 1]; i += 2; if ((op === 'optionalAccess' || op === 'optionalCall') && value == null) { return undefined; } if (op === 'access' || op === 'optionalAccess') { lastAccessLHS = value; value = fn(value); } else if (op === 'call' || op === 'optionalCall') { value = fn((...args) => value.call(lastAccessLHS, ...args)); lastAccessLHS = undefined; } } return value; }// src/tree.ts
-function truncateText(text, maxLength = 150) {
-  if (typeof text === "undefined") {
-    return "";
-  }
-  if (typeof text === "object") {
-    text = JSON.stringify(text);
-  }
-  if (typeof text === "number") {
-    return text.toString();
-  }
-  if (typeof text === "string" && text.length > maxLength) {
-    return `${text.slice(0, maxLength)}...`;
-  }
-  if (typeof text === "string") {
-    return text.trim();
-  }
-  return "";
-}
-function trimAttributes(attributes, truncateTextLength) {
-  const tailorAttributes = Object.keys(attributes).reduce(
-    (res, currentKey) => {
-      const attributeVal = attributes[currentKey];
-      if (currentKey === "style" || currentKey === "src" || currentKey === "htmlTagName" || currentKey === "nodeType") {
-        return res;
-      }
-      res[currentKey] = truncateText(attributeVal, truncateTextLength);
-      return res;
-    },
-    {}
-  );
-  return tailorAttributes;
-}
-var nodeSizeThreshold = 4;
-function descriptionOfTree(tree, truncateTextLength, filterNonTextContent = false) {
-  const attributesString = (kv) => {
-    return Object.entries(kv).map(
-      ([key, value]) => `${key}="${truncateText(value, truncateTextLength)}"`
-    ).join(" ");
-  };
-  function buildContentTree(node, indent = 0) {
-    let before = "";
-    let contentWithIndent = "";
-    let after = "";
-    let emptyNode = true;
-    const indentStr = "  ".repeat(indent);
-    let children = "";
-    for (let i = 0; i < (node.children || []).length; i++) {
-      const childContent = buildContentTree(node.children[i], indent + 1);
-      if (childContent) {
-        children += `
-${childContent}`;
-      }
-    }
-    if (node.node && node.node.rect.width > nodeSizeThreshold && node.node.rect.height > nodeSizeThreshold && (!filterNonTextContent || filterNonTextContent && node.node.content)) {
-      emptyNode = false;
-      let nodeTypeString;
-      if (_optionalChain([node, 'access', _ => _.node, 'access', _2 => _2.attributes, 'optionalAccess', _3 => _3.htmlTagName])) {
-        nodeTypeString = node.node.attributes.htmlTagName.replace(/[<>]/g, "");
-      } else {
-        nodeTypeString = node.node.attributes.nodeType.replace(/\sNode$/, "").toLowerCase();
-      }
-      const markerId = node.node.indexId;
-      const markerIdString = typeof markerId !== "undefined" ? `markerId="${markerId}"` : "";
-      const rectAttribute = node.node.rect ? {
-        left: node.node.rect.left,
-        top: node.node.rect.top,
-        width: node.node.rect.width,
-        height: node.node.rect.height
-      } : {};
-      before = `<${nodeTypeString} id="${node.node.id}" ${markerIdString} ${attributesString(trimAttributes(node.node.attributes || {}, truncateTextLength))} ${attributesString(rectAttribute)}>`;
-      const content = truncateText(node.node.content, truncateTextLength);
-      contentWithIndent = content ? `
-${indentStr}  ${content}` : "";
-      after = `</${nodeTypeString}>`;
-    } else if (!filterNonTextContent) {
-      if (!children.trim().startsWith("<>")) {
-        before = "<>";
-        contentWithIndent = "";
-        after = "</>";
-      }
-    }
-    if (emptyNode && !children.trim()) {
-      return "";
-    }
-    const result2 = `${indentStr}${before}${contentWithIndent}${children}
-${indentStr}${after}`;
-    if (result2.trim()) {
-      return result2;
-    }
-    return "";
-  }
-  const result = buildContentTree(tree);
-  return result.replace(/^\s*\n/gm, "");
-}
+"use strict";Object.defineProperty(exports, "__esModule", {value: true});// src/tree.ts
 
 
 
 
-exports.descriptionOfTree = descriptionOfTree; exports.trimAttributes = trimAttributes; exports.truncateText = truncateText;
+var _extractor = require('@midscene/shared/extractor');
+
+
+
+
+exports.descriptionOfTree = _extractor.descriptionOfTree; exports.trimAttributes = _extractor.trimAttributes; exports.truncateText = _extractor.truncateText;
 
 //# sourceMappingURL=tree.js.map

package/dist/lib/tree.js.map

@@ -1 +1 @@
-{"version":3,"mappings":";AAEO,SAAS,aACd,MACA,YAAY,KACZ;AACA,MAAI,OAAO,SAAS,aAAa;AAC/B,WAAO;AAAA,EACT;AAEA,MAAI,OAAO,SAAS,UAAU;AAC5B,WAAO,KAAK,UAAU,IAAI;AAAA,EAC5B;AAEA,MAAI,OAAO,SAAS,UAAU;AAC5B,WAAO,KAAK,SAAS;AAAA,EACvB;AAEA,MAAI,OAAO,SAAS,YAAY,KAAK,SAAS,WAAW;AACvD,WAAO,GAAG,KAAK,MAAM,GAAG,SAAS,CAAC;AAAA,EACpC;AAEA,MAAI,OAAO,SAAS,UAAU;AAC5B,WAAO,KAAK,KAAK;AAAA,EACnB;AAEA,SAAO;AACT;AAEO,SAAS,eACd,YACA,oBACA;AACA,QAAM,mBAAmB,OAAO,KAAK,UAAU,EAAE;AAAA,IAC/C,CAAC,KAAK,eAAuB;AAC3B,YAAM,eAAgB,WAAmB,UAAU;AACnD,UACE,eAAe,WACf,eAAe,SACf,eAAe,iBACf,eAAe,YACf;AACA,eAAO;AAAA,MACT;AAEA,UAAI,UAAU,IAAI,aAAa,cAAc,kBAAkB;AAC/D,aAAO;AAAA,IACT;AAAA,IACA,CAAC;AAAA,EACH;AACA,SAAO;AACT;AAEA,IAAM,oBAAoB;AACnB,SAAS,kBAGd,MACA,oBACA,uBAAuB,OACvB;AACA,QAAM,mBAAmB,CAAC,OAA4B;AACpD,WAAO,OAAO,QAAQ,EAAE,EACrB;AAAA,MACC,CAAC,CAAC,KAAK,KAAK,MAAM,GAAG,GAAG,KAAK,aAAa,OAAO,kBAAkB,CAAC;AAAA,IACtE,EACC,KAAK,GAAG;AAAA,EACb;AAEA,WAAS,iBACP,MACA,SAAS,GACD;AACR,QAAI,SAAS;AACb,QAAI,oBAAoB;AACxB,QAAI,QAAQ;AACZ,QAAI,YAAY;AAChB,UAAM,YAAY,KAAK,OAAO,MAAM;AAEpC,QAAI,WAAW;AACf,aAAS,IAAI,GAAG,KAAK,KAAK,YAAY,CAAC,GAAG,QAAQ,KAAK;AACrD,YAAM,eAAe,iBAAiB,KAAK,SAAS,CAAC,GAAG,SAAS,CAAC;AAClE,UAAI,cAAc;AAChB,oBAAY;AAAA,EAAK,YAAY;AAAA,MAC/B;AAAA,IACF;AAEA,QACE,KAAK,QACL,KAAK,KAAK,KAAK,QAAQ,qBACvB,KAAK,KAAK,KAAK,SAAS,sBACvB,CAAC,wBAAyB,wBAAwB,KAAK,KAAK,UAC7D;AACA,kBAAY;AACZ,UAAI;AACJ,UAAI,KAAK,KAAK,YAAY,aAAa;AACrC,yBAAiB,KAAK,KAAK,WAAW,YAAY,QAAQ,SAAS,EAAE;AAAA,MACvE,OAAO;AACL,yBAAiB,KAAK,KAAK,WAAW,SACnC,QAAQ,WAAW,EAAE,EACrB,YAAY;AAAA,MACjB;AACA,YAAM,WAAW,KAAK,KAAK;AAC3B,YAAM,iBACJ,OAAO,aAAa,cAAc,aAAa,QAAQ,MAAM;AAC/D,YAAM,gBAAgB,KAAK,KAAK,OAC5B;AAAA,QACE,MAAM,KAAK,KAAK,KAAK;AAAA,QACrB,KAAK,KAAK,KAAK,KAAK;AAAA,QACpB,OAAO,KAAK,KAAK,KAAK;AAAA,QACtB,QAAQ,KAAK,KAAK,KAAK;AAAA,MACzB,IACA,CAAC;AACL,eAAS,IAAI,cAAc,QAAQ,KAAK,KAAK,EAAE,KAAK,cAAc,IAAI,iBAAiB,eAAe,KAAK,KAAK,cAAc,CAAC,GAAG,kBAAkB,CAAC,CAAC,IAAI,iBAAiB,aAAa,CAAC;AACzL,YAAM,UAAU,aAAa,KAAK,KAAK,SAAS,kBAAkB;AAClE,0BAAoB,UAAU;AAAA,EAAK,SAAS,KAAK,OAAO,KAAK;AAC7D,cAAQ,KAAK,cAAc;AAAA,IAC7B,WAAW,CAAC,sBAAsB;AAChC,UAAI,CAAC,SAAS,KAAK,EAAE,WAAW,IAAI,GAAG;AACrC,iBAAS;AACT,4BAAoB;AACpB,gBAAQ;AAAA,MACV;AAAA,IACF;AAEA,QAAI,aAAa,CAAC,SAAS,KAAK,GAAG;AACjC,aAAO;AAAA,IACT;AAEA,UAAMA,UAAS,GAAG,SAAS,GAAG,MAAM,GAAG,iBAAiB,GAAG,QAAQ;AAAA,EAAK,SAAS,GAAG,KAAK;AACzF,QAAIA,QAAO,KAAK,GAAG;AACjB,aAAOA;AAAA,IACT;AACA,WAAO;AAAA,EACT;AAEA,QAAM,SAAS,iBAAiB,IAAI;AACpC,SAAO,OAAO,QAAQ,YAAY,EAAE;AACtC","names":["result"],"ignoreList":[],"sources":["../../src/tree.ts"],"sourcesContent":["import type { BaseElement, ElementTreeNode, Size, UIContext } from '@/types';\n\nexport function truncateText(\n  text: string | number | object | undefined,\n  maxLength = 150,\n) {\n  if (typeof text === 'undefined') {\n    return '';\n  }\n\n  if (typeof text === 'object') {\n    text = JSON.stringify(text);\n  }\n\n  if (typeof text === 'number') {\n    return text.toString();\n  }\n\n  if (typeof text === 'string' && text.length > maxLength) {\n    return `${text.slice(0, maxLength)}...`;\n  }\n\n  if (typeof text === 'string') {\n    return text.trim();\n  }\n\n  return '';\n}\n\nexport function trimAttributes(\n  attributes: Record<string, any>,\n  truncateTextLength?: number,\n) {\n  const tailorAttributes = Object.keys(attributes).reduce(\n    (res, currentKey: string) => {\n      const attributeVal = (attributes as any)[currentKey];\n      if (\n        currentKey === 'style' ||\n        currentKey === 'src' ||\n        currentKey === 'htmlTagName' ||\n        currentKey === 'nodeType'\n      ) {\n        return res;\n      }\n\n      res[currentKey] = truncateText(attributeVal, truncateTextLength);\n      return res;\n    },\n    {} as BaseElement['attributes'],\n  );\n  return tailorAttributes;\n}\n\nconst nodeSizeThreshold = 4;\nexport function descriptionOfTree<\n  ElementType extends BaseElement = BaseElement,\n>(\n  tree: ElementTreeNode<ElementType>,\n  truncateTextLength?: number,\n  filterNonTextContent = false,\n) {\n  const attributesString = (kv: Record<string, any>) => {\n    return Object.entries(kv)\n      .map(\n        ([key, value]) => `${key}=\"${truncateText(value, truncateTextLength)}\"`,\n      )\n      .join(' ');\n  };\n\n  function buildContentTree(\n    node: ElementTreeNode<ElementType>,\n    indent = 0,\n  ): string {\n    let before = '';\n    let contentWithIndent = '';\n    let after = '';\n    let emptyNode = true;\n    const indentStr = '  '.repeat(indent);\n\n    let children = '';\n    for (let i = 0; i < (node.children || []).length; i++) {\n      const childContent = buildContentTree(node.children[i], indent + 1);\n      if (childContent) {\n        children += `\\n${childContent}`;\n      }\n    }\n\n    if (\n      node.node &&\n      node.node.rect.width > nodeSizeThreshold &&\n      node.node.rect.height > nodeSizeThreshold &&\n      (!filterNonTextContent || (filterNonTextContent && node.node.content))\n    ) {\n      emptyNode = false;\n      let nodeTypeString: string;\n      if (node.node.attributes?.htmlTagName) {\n        nodeTypeString = node.node.attributes.htmlTagName.replace(/[<>]/g, '');\n      } else {\n        nodeTypeString = node.node.attributes.nodeType\n          .replace(/\\sNode$/, '')\n          .toLowerCase();\n      }\n      const markerId = node.node.indexId;\n      const markerIdString =\n        typeof markerId !== 'undefined' ? `markerId=\"${markerId}\"` : '';\n      const rectAttribute = node.node.rect\n        ? {\n            left: node.node.rect.left,\n            top: node.node.rect.top,\n            width: node.node.rect.width,\n            height: node.node.rect.height,\n          }\n        : {};\n      before = `<${nodeTypeString} id=\"${node.node.id}\" ${markerIdString} ${attributesString(trimAttributes(node.node.attributes || {}, truncateTextLength))} ${attributesString(rectAttribute)}>`;\n      const content = truncateText(node.node.content, truncateTextLength);\n      contentWithIndent = content ? `\\n${indentStr}  ${content}` : '';\n      after = `</${nodeTypeString}>`;\n    } else if (!filterNonTextContent) {\n      if (!children.trim().startsWith('<>')) {\n        before = '<>';\n        contentWithIndent = '';\n        after = '</>';\n      }\n    }\n\n    if (emptyNode && !children.trim()) {\n      return '';\n    }\n\n    const result = `${indentStr}${before}${contentWithIndent}${children}\\n${indentStr}${after}`;\n    if (result.trim()) {\n      return result;\n    }\n    return '';\n  }\n\n  const result = buildContentTree(tree);\n  return result.replace(/^\\s*\\n/gm, '');\n}\n"]}
+{"version":3,"mappings":";AAAA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK","names":[],"ignoreList":[],"sources":["../../src/tree.ts"],"sourcesContent":["export {\n  truncateText,\n  trimAttributes,\n  descriptionOfTree,\n} from '@midscene/shared/extractor';\n"]}

package/dist/{types/types-81bb2c02.d.ts → lib/types-8c197f92.d.ts}

@@ -1,4 +1,5 @@
 import { NodeType } from '@midscene/shared/constants';
+import { Rect, BaseElement, ElementTreeNode, Size } from '@midscene/shared/types';
 import { ChatCompletionMessageParam } from 'openai/resources';
 
 interface LocateOption {
@@ -6,6 +7,10 @@ interface LocateOption {
     deepThink?: boolean;
     cacheable?: boolean;
 }
+interface InsightExtractOption {
+    domIncluded?: boolean;
+    screenshotIncluded?: boolean;
+}
 interface ReferenceImage {
     base64: string;
     rect?: Rect;
@@ -64,20 +69,21 @@ interface MidsceneYamlFlowItemAIAction {
 }
 interface MidsceneYamlFlowItemAIAssert {
     aiAssert: string;
+    errorMessage?: string;
 }
-interface MidsceneYamlFlowItemAIQuery {
+interface MidsceneYamlFlowItemAIQuery extends InsightExtractOption {
     aiQuery: string;
     name?: string;
 }
-interface MidsceneYamlFlowItemAINumber {
+interface MidsceneYamlFlowItemAINumber extends InsightExtractOption {
     aiNumber: string;
     name?: string;
 }
-interface MidsceneYamlFlowItemAINString {
+interface MidsceneYamlFlowItemAINString extends InsightExtractOption {
     aiString: string;
     name?: string;
 }
-interface MidsceneYamlFlowItemAIBoolean {
+interface MidsceneYamlFlowItemAIBoolean extends InsightExtractOption {
     aiBoolean: string;
     name?: string;
 }
@@ -92,6 +98,9 @@ interface MidsceneYamlFlowItemAIWaitFor {
 interface MidsceneYamlFlowItemAITap extends LocateOption {
     aiTap: string;
 }
+interface MidsceneYamlFlowItemAIRightClick extends LocateOption {
+    aiRightClick: string;
+}
 interface MidsceneYamlFlowItemAIHover extends LocateOption {
     aiHover: string;
 }
@@ -114,7 +123,7 @@ interface MidsceneYamlFlowItemEvaluateJavaScript {
 interface MidsceneYamlFlowItemSleep {
     sleep: number;
 }
-type MidsceneYamlFlowItem = MidsceneYamlFlowItemAIAction | MidsceneYamlFlowItemAIAssert | MidsceneYamlFlowItemAIQuery | MidsceneYamlFlowItemAIWaitFor | MidsceneYamlFlowItemAITap | MidsceneYamlFlowItemAIHover | MidsceneYamlFlowItemAIInput | MidsceneYamlFlowItemAIKeyboardPress | MidsceneYamlFlowItemAIScroll | MidsceneYamlFlowItemSleep;
+type MidsceneYamlFlowItem = MidsceneYamlFlowItemAIAction | MidsceneYamlFlowItemAIAssert | MidsceneYamlFlowItemAIQuery | MidsceneYamlFlowItemAIWaitFor | MidsceneYamlFlowItemAITap | MidsceneYamlFlowItemAIRightClick | MidsceneYamlFlowItemAIHover | MidsceneYamlFlowItemAIInput | MidsceneYamlFlowItemAIKeyboardPress | MidsceneYamlFlowItemAIScroll | MidsceneYamlFlowItemSleep;
 interface FreeFn {
     name: string;
     fn: () => void;
@@ -127,35 +136,6 @@ interface ScriptPlayerTaskStatus extends MidsceneYamlTask {
 }
 type ScriptPlayerStatusValue = 'init' | 'running' | 'done' | 'error';
 
-interface Point {
-    left: number;
-    top: number;
-}
-interface Size {
-    width: number;
-    height: number;
-    dpr?: number;
-}
-type Rect = Point & Size & {
-    zoom?: number;
-};
-declare abstract class BaseElement {
-    abstract id: string;
-    abstract indexId?: number;
-    abstract attributes: {
-        nodeType: NodeType;
-        [key: string]: string;
-    };
-    abstract content: string;
-    abstract rect: Rect;
-    abstract center: [number, number];
-    abstract locator?: string;
-    abstract xpaths?: string[];
-}
-interface ElementTreeNode<ElementType extends BaseElement = BaseElement> {
-    node: ElementType | null;
-    children: ElementTreeNode<ElementType>[];
-}
 type AIUsageInfo = Record<string, any> & {
     prompt_tokens: number;
     completion_tokens: number;
@@ -338,7 +318,7 @@ interface PlanningLocateParam extends DetailedLocateParam {
 }
 interface PlanningAction<ParamType = any> {
     thought?: string;
-    type: 'Locate' | 'Tap' | 'Drag' | 'Hover' | 'Input' | 'KeyboardPress' | 'Scroll' | 'Error' | 'ExpectedFalsyCondition' | 'Assert' | 'AssertWithoutThrow' | 'Sleep' | 'Finished' | 'AndroidBackButton' | 'AndroidHomeButton' | 'AndroidRecentAppsButton';
+    type: 'Locate' | 'Tap' | 'RightClick' | 'Hover' | 'Drag' | 'Input' | 'KeyboardPress' | 'Scroll' | 'Error' | 'ExpectedFalsyCondition' | 'Assert' | 'AssertWithoutThrow' | 'Sleep' | 'Finished' | 'AndroidBackButton' | 'AndroidHomeButton' | 'AndroidRecentAppsButton';
     param: ParamType;
     locate?: PlanningLocateParam | null;
 }
@@ -356,6 +336,7 @@ interface PlanningAIResponse {
 }
 type PlanningActionParamTap = null;
 type PlanningActionParamHover = null;
+type PlanningActionParamRightClick = null;
 interface PlanningActionParamInputOrKeyPress {
     value: string;
 }
@@ -472,4 +453,4 @@ interface GroupedActionDump {
 }
 type PageType = 'puppeteer' | 'playwright' | 'static' | 'chrome-extension-proxy' | 'android';
 
-export { type PlanningActionParamInputOrKeyPress as $, type AIDescribeElementResponse as A, BaseElement as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type LocateResultElement as F, type DumpMeta as G, type ReportDumpWithAttributes as H, type InsightAction as I, type InsightDump as J, type PartialInsightDumpFromSDK as K, type LocateResult as L, type MidsceneYamlScript as M, type LiteUISection as N, type ElementById as O, type Point as P, type OnTaskStartTip as Q, type Rect as R, type Size as S, type AgentWaitForOpt as T, UIContext as U, type AgentAssertOpt as V, type PlanningLocateParam as W, type PlanningAction as X, type PlanningAIResponse as Y, type PlanningActionParamTap as Z, type PlanningActionParamHover as _, type ExecutionTaskProgressOptions as a, type PlanningActionParamScroll as a0, type PlanningActionParamAssert as a1, type PlanningActionParamSleep as a2, type PlanningActionParamError as a3, type PlanningActionParamWaitFor as a4, type Color as a5, type BaseAgentParserOpt as a6, type PuppeteerParserOpt as a7, type PlaywrightParserOpt as a8, type ExecutionRecorderItem as a9, type MidsceneYamlScriptWebEnv as aA, type MidsceneYamlScriptAndroidEnv as aB, type MidsceneYamlScriptEnv as aC, type MidsceneYamlFlowItemAIAction as aD, type MidsceneYamlFlowItemAIAssert as aE, type MidsceneYamlFlowItemAIQuery as aF, type MidsceneYamlFlowItemAINumber as aG, type MidsceneYamlFlowItemAINString as aH, type MidsceneYamlFlowItemAIBoolean as aI, type MidsceneYamlFlowItemAILocate as aJ, type MidsceneYamlFlowItemAIWaitFor as aK, type MidsceneYamlFlowItemAITap as aL, type MidsceneYamlFlowItemAIHover as aM, type MidsceneYamlFlowItemAIInput as aN, type MidsceneYamlFlowItemAIKeyboardPress as aO, type MidsceneYamlFlowItemAIScroll as aP, type MidsceneYamlFlowItemEvaluateJavaScript as aQ, type MidsceneYamlFlowItemSleep as aR, type FreeFn as aS, type ScriptPlayerTaskStatus as aT, type ScriptPlayerStatusValue as aU, type ExecutionTaskType as aa, type ExecutorContext as ab, type TaskCacheInfo as ac, type ExecutionTaskReturn as ad, type ExecutionTaskInsightLocateParam as ae, type ExecutionTaskInsightLocateOutput as af, type ExecutionTaskInsightDumpLog as ag, type ExecutionTaskInsightLocateApply as ah, type ExecutionTaskInsightLocate as ai, type ExecutionTaskInsightQueryParam as aj, type ExecutionTaskInsightQueryOutput as ak, type ExecutionTaskInsightQueryApply as al, type ExecutionTaskInsightQuery as am, type ExecutionTaskInsightAssertionParam as an, type ExecutionTaskInsightAssertionApply as ao, type ExecutionTaskInsightAssertion as ap, type ExecutionTaskActionApply as aq, type ExecutionTaskAction as ar, type ExecutionTaskPlanningApply as as, type ExecutionTaskPlanning as at, type GroupedActionDump as au, type PageType as av, type LocateOption as aw, type ReferenceImage as ax, type scrollParam as ay, type MidsceneYamlScriptEnvBase as az, type ExecutionTaskApply as b, type ExecutionDump as c, type InsightTaskInfo as d, type InsightOptions as e, type DetailedLocateParam as f, type InsightAssertionResponse as g, type MidsceneYamlTask as h, type MidsceneYamlFlowItem as i, type ElementTreeNode as j, type AIUsageInfo as k, AIResponseFormat as l, type AISingleElementResponseById as m, type AISingleElementResponseByPosition as n, type AISingleElementResponse as o, type AIElementLocatorResponse as p, type AIElementCoordinatesResponse as q, type AIElementResponse as r, type AIDataExtractionResponse as s, type AISectionLocatorResponse as t, type AIAssertionResponse as u, type LocatorValidatorOption as v, type LocateValidatorResult as w, type AgentDescribeElementAtPointResult as x, type EnsureObject as y, type InsightExtractParam as z };
+export { type PlanningActionParamAssert as $, type AIDescribeElementResponse as A, type InsightExtractParam as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type LocateResultElement as F, type DumpMeta as G, type InsightDump as H, type InsightAction as I, type LiteUISection as J, type ElementById as K, type LocateResult as L, type MidsceneYamlScript as M, type AgentWaitForOpt as N, type OnTaskStartTip as O, type PartialInsightDumpFromSDK as P, type AgentAssertOpt as Q, type ReportDumpWithAttributes as R, type PlanningLocateParam as S, type PlanningAction as T, UIContext as U, type PlanningAIResponse as V, type PlanningActionParamTap as W, type PlanningActionParamHover as X, type PlanningActionParamRightClick as Y, type PlanningActionParamInputOrKeyPress as Z, type PlanningActionParamScroll as _, type ExecutionTaskProgressOptions as a, type PlanningActionParamSleep as a0, type PlanningActionParamError as a1, type PlanningActionParamWaitFor as a2, type Color as a3, type BaseAgentParserOpt as a4, type PuppeteerParserOpt as a5, type PlaywrightParserOpt as a6, type ExecutionRecorderItem as a7, type ExecutionTaskType as a8, type ExecutorContext as a9, type MidsceneYamlScriptEnv as aA, type MidsceneYamlFlowItemAIAction as aB, type MidsceneYamlFlowItemAIAssert as aC, type MidsceneYamlFlowItemAIQuery as aD, type MidsceneYamlFlowItemAINumber as aE, type MidsceneYamlFlowItemAINString as aF, type MidsceneYamlFlowItemAIBoolean as aG, type MidsceneYamlFlowItemAILocate as aH, type MidsceneYamlFlowItemAIWaitFor as aI, type MidsceneYamlFlowItemAITap as aJ, type MidsceneYamlFlowItemAIHover as aK, type MidsceneYamlFlowItemAIInput as aL, type MidsceneYamlFlowItemAIKeyboardPress as aM, type MidsceneYamlFlowItemAIScroll as aN, type MidsceneYamlFlowItemEvaluateJavaScript as aO, type MidsceneYamlFlowItemSleep as aP, type FreeFn as aQ, type ScriptPlayerTaskStatus as aR, type ScriptPlayerStatusValue as aS, type TaskCacheInfo as aa, type ExecutionTaskReturn as ab, type ExecutionTaskInsightLocateParam as ac, type ExecutionTaskInsightLocateOutput as ad, type ExecutionTaskInsightDumpLog as ae, type ExecutionTaskInsightLocateApply as af, type ExecutionTaskInsightLocate as ag, type ExecutionTaskInsightQueryParam as ah, type ExecutionTaskInsightQueryOutput as ai, type ExecutionTaskInsightQueryApply as aj, type ExecutionTaskInsightQuery as ak, type ExecutionTaskInsightAssertionParam as al, type ExecutionTaskInsightAssertionApply as am, type ExecutionTaskInsightAssertion as an, type ExecutionTaskActionApply as ao, type ExecutionTaskAction as ap, type ExecutionTaskPlanningApply as aq, type ExecutionTaskPlanning as ar, type GroupedActionDump as as, type PageType as at, type LocateOption as au, type ReferenceImage as av, type scrollParam as aw, type MidsceneYamlScriptEnvBase as ax, type MidsceneYamlScriptWebEnv as ay, type MidsceneYamlScriptAndroidEnv as az, type ExecutionTaskApply as b, type ExecutionDump as c, type InsightTaskInfo as d, type InsightOptions as e, type DetailedLocateParam as f, type InsightExtractOption as g, type InsightAssertionResponse as h, type MidsceneYamlTask as i, type MidsceneYamlFlowItem as j, type MidsceneYamlFlowItemAIRightClick as k, type AIUsageInfo as l, AIResponseFormat as m, type AISingleElementResponseById as n, type AISingleElementResponseByPosition as o, type AISingleElementResponse as p, type AIElementLocatorResponse as q, type AIElementCoordinatesResponse as r, type AIElementResponse as s, type AIDataExtractionResponse as t, type AISectionLocatorResponse as u, type AIAssertionResponse as v, type LocatorValidatorOption as w, type LocateValidatorResult as x, type AgentDescribeElementAtPointResult as y, type EnsureObject as z };