@midscene/core 0.16.11-beta-20250514094713.0 → 0.16.11-beta-20250516035357.0

package/dist/lib/index.d.ts

@@ -1,7 +1,7 @@
-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, A as AISingleElementResponse } from './types-8c83481c.js';
-export { t as AIAssertionResponse, r as AIDataExtractionResponse, p as AIElementCoordinatesResponse, o as AIElementLocatorResponse, q as AIElementResponse, l as AIResponseFormat, s as AISectionLocatorResponse, m as AISingleElementResponseById, n as AISingleElementResponseByPosition, k as AIUsageInfo, K as AgentAssertOpt, J as AgentWaitForOpt, a2 as BaseAgentParserOpt, C as CallAIFn, a1 as Color, x as DumpMeta, H as ElementById, j as ElementTreeNode, u as EnsureObject, a5 as ExecutionRecorderItem, an as ExecutionTaskAction, am as ExecutionTaskActionApply, al as ExecutionTaskInsightAssertion, ak as ExecutionTaskInsightAssertionApply, aj as ExecutionTaskInsightAssertionParam, ac as ExecutionTaskInsightDumpLog, ae as ExecutionTaskInsightLocate, ad as ExecutionTaskInsightLocateApply, ab as ExecutionTaskInsightLocateOutput, aa as ExecutionTaskInsightLocateParam, ai as ExecutionTaskInsightQuery, ah as ExecutionTaskInsightQueryApply, ag as ExecutionTaskInsightQueryOutput, af as ExecutionTaskInsightQueryParam, ap as ExecutionTaskPlanning, ao as ExecutionTaskPlanningApply, a9 as ExecutionTaskReturn, a6 as ExecutionTaskType, a7 as ExecutorContext, aN as FreeFn, aq as GroupedActionDump, z as InsightDump, v as InsightExtractParam, G as LiteUISection, as as LocateOption, w as LocateResultElement, i as MidsceneYamlFlowItem, ay as MidsceneYamlFlowItemAIAction, az as MidsceneYamlFlowItemAIAssert, aD as MidsceneYamlFlowItemAIBoolean, aH as MidsceneYamlFlowItemAIHover, aI as MidsceneYamlFlowItemAIInput, aJ as MidsceneYamlFlowItemAIKeyboardPress, aE as MidsceneYamlFlowItemAILocate, aC as MidsceneYamlFlowItemAINString, aB as MidsceneYamlFlowItemAINumber, aA as MidsceneYamlFlowItemAIQuery, aK as MidsceneYamlFlowItemAIScroll, aG as MidsceneYamlFlowItemAITap, aF as MidsceneYamlFlowItemAIWaitFor, aL as MidsceneYamlFlowItemEvaluateJavaScript, aM as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aw as MidsceneYamlScriptAndroidEnv, ax as MidsceneYamlScriptEnv, au as MidsceneYamlScriptEnvBase, av as MidsceneYamlScriptWebEnv, h as MidsceneYamlTask, O as OnTaskStartTip, ar as PageType, F as PartialInsightDumpFromSDK, T as PlanningAIResponse, Q as PlanningAction, Z as PlanningActionParamAssert, $ as PlanningActionParamError, W as PlanningActionParamHover, X as PlanningActionParamInputOrKeyPress, Y as PlanningActionParamScroll, _ as PlanningActionParamSleep, V as PlanningActionParamTap, a0 as PlanningActionParamWaitFor, N as PlanningLocateParam, a4 as PlaywrightParserOpt, P as Point, a3 as PuppeteerParserOpt, R as Rect, y as ReportDumpWithAttributes, aP as ScriptPlayerStatusValue, aO as ScriptPlayerTaskStatus, S as Size, a8 as TaskCacheInfo, at as scrollParam } from './types-8c83481c.js';
-import { c as callAiFn } from './llm-planning-e16d1814.js';
-export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-e16d1814.js';
+import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, B as BaseElement, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightAssertionResponse } from './types-cbcbeb4e.js';
+export { t as AIAssertionResponse, r as AIDataExtractionResponse, p as AIElementCoordinatesResponse, o as AIElementLocatorResponse, q as AIElementResponse, k as AIResponseFormat, s as AISectionLocatorResponse, n as AISingleElementResponse, l as AISingleElementResponseById, m as AISingleElementResponseByPosition, A as AIUsageInfo, K as AgentAssertOpt, J as AgentWaitForOpt, a2 as BaseAgentParserOpt, C as CallAIFn, a1 as Color, x as DumpMeta, H as ElementById, j as ElementTreeNode, u as EnsureObject, a5 as ExecutionRecorderItem, an as ExecutionTaskAction, am as ExecutionTaskActionApply, al as ExecutionTaskInsightAssertion, ak as ExecutionTaskInsightAssertionApply, aj as ExecutionTaskInsightAssertionParam, ac as ExecutionTaskInsightDumpLog, ae as ExecutionTaskInsightLocate, ad as ExecutionTaskInsightLocateApply, ab as ExecutionTaskInsightLocateOutput, aa as ExecutionTaskInsightLocateParam, ai as ExecutionTaskInsightQuery, ah as ExecutionTaskInsightQueryApply, ag as ExecutionTaskInsightQueryOutput, af as ExecutionTaskInsightQueryParam, ap as ExecutionTaskPlanning, ao as ExecutionTaskPlanningApply, a9 as ExecutionTaskReturn, a6 as ExecutionTaskType, a7 as ExecutorContext, aN as FreeFn, aq as GroupedActionDump, z as InsightDump, v as InsightExtractParam, G as LiteUISection, as as LocateOption, w as LocateResultElement, i as MidsceneYamlFlowItem, ay as MidsceneYamlFlowItemAIAction, az as MidsceneYamlFlowItemAIAssert, aD as MidsceneYamlFlowItemAIBoolean, aH as MidsceneYamlFlowItemAIHover, aI as MidsceneYamlFlowItemAIInput, aJ as MidsceneYamlFlowItemAIKeyboardPress, aE as MidsceneYamlFlowItemAILocate, aC as MidsceneYamlFlowItemAINString, aB as MidsceneYamlFlowItemAINumber, aA as MidsceneYamlFlowItemAIQuery, aK as MidsceneYamlFlowItemAIScroll, aG as MidsceneYamlFlowItemAITap, aF as MidsceneYamlFlowItemAIWaitFor, aL as MidsceneYamlFlowItemEvaluateJavaScript, aM as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aw as MidsceneYamlScriptAndroidEnv, ax as MidsceneYamlScriptEnv, au as MidsceneYamlScriptEnvBase, av as MidsceneYamlScriptWebEnv, h as MidsceneYamlTask, O as OnTaskStartTip, ar as PageType, F as PartialInsightDumpFromSDK, T as PlanningAIResponse, Q as PlanningAction, Z as PlanningActionParamAssert, $ as PlanningActionParamError, W as PlanningActionParamHover, X as PlanningActionParamInputOrKeyPress, Y as PlanningActionParamScroll, _ as PlanningActionParamSleep, V as PlanningActionParamTap, a0 as PlanningActionParamWaitFor, N as PlanningLocateParam, a4 as PlaywrightParserOpt, P as Point, a3 as PuppeteerParserOpt, R as Rect, y as ReportDumpWithAttributes, aP as ScriptPlayerStatusValue, aO as ScriptPlayerTaskStatus, S as Size, a8 as TaskCacheInfo, at as scrollParam } from './types-cbcbeb4e.js';
+import { c as callAiFn } from './llm-planning-26db5c81.js';
+export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-26db5c81.js';
 export { getVersion } from './utils.js';
 export { MIDSCENE_MODEL_NAME, getAIConfig } from '@midscene/shared/env';
 import '@midscene/shared/constants';
@@ -24,8 +24,8 @@ declare class Executor {
 }
 
 interface LocateOpts {
+    context?: UIContext<BaseElement>;
     callAI?: typeof callAiFn<AIElementResponse>;
-    quickAnswer?: Partial<AISingleElementResponse>;
 }
 declare class Insight<ElementType extends BaseElement = BaseElement, ContextType extends UIContext<ElementType> = UIContext<ElementType>> {
     contextRetrieverFn: (action: InsightAction) => Promise<ContextType> | ContextType;
@@ -40,4 +40,4 @@ declare class Insight<ElementType extends BaseElement = BaseElement, ContextType
     assert(assertion: string): Promise<InsightAssertionResponse>;
 }
 
-export { AISingleElementResponse, BaseElement, DetailedLocateParam, DumpSubscriber, ExecutionDump, ExecutionTask, ExecutionTaskApply, ExecutionTaskProgressOptions, Executor, Insight, InsightAction, InsightAssertionResponse, InsightOptions, InsightTaskInfo, LocateResult, UIContext, Insight as default };
+export { BaseElement, DetailedLocateParam, DumpSubscriber, ExecutionDump, ExecutionTask, ExecutionTaskApply, ExecutionTaskProgressOptions, Executor, Insight, InsightAction, InsightAssertionResponse, 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 _chunkHLRMNAQKjs = require('./chunk-HLRMNAQK.js');
+var _chunkTOTKWQNQjs = require('./chunk-TOTKWQNQ.js');
 
 
 
@@ -9,7 +9,7 @@ var _chunkHLRMNAQKjs = require('./chunk-HLRMNAQK.js');
 
 
 
-var _chunkFYD4T7QDjs = require('./chunk-FYD4T7QD.js');
+var _chunkH3B2KXXRjs = require('./chunk-H3B2KXXR.js');
 
 // src/ai-model/action-executor.ts
 var _env = require('@midscene/shared/env');
@@ -156,7 +156,7 @@ ${_optionalChain([this, 'access', _7 => _7.latestErrorTask, 'call', _8 => _8(),
   }
   dump() {
     const dumpData = {
-      sdkVersion: _chunkHLRMNAQKjs.getVersion.call(void 0, ),
+      sdkVersion: _chunkTOTKWQNQjs.getVersion.call(void 0, ),
       model_name: _env.getAIConfig.call(void 0, _env.MIDSCENE_MODEL_NAME) || "",
       logTime: Date.now(),
       name: this.name,
@@ -194,7 +194,7 @@ function emitInsightDump(data, dumpSubscriber) {
     }
   }
   const baseData = {
-    sdkVersion: _chunkHLRMNAQKjs.getVersion.call(void 0, ),
+    sdkVersion: _chunkTOTKWQNQjs.getVersion.call(void 0, ),
     logTime: Date.now(),
     model_name: _env.getAIConfig.call(void 0, _env.MIDSCENE_MODEL_NAME) || "",
     model_description: modelDescription
@@ -211,7 +211,7 @@ function emitInsightDump(data, dumpSubscriber) {
 var debug = _logger.getDebug.call(void 0, "ai:insight");
 var Insight = class {
   constructor(context, opt) {
-    this.aiVendorFn = _chunkFYD4T7QDjs.callAiFn;
+    this.aiVendorFn = _chunkH3B2KXXRjs.callAiFn;
     _utils.assert.call(void 0, context, "context is required for Insight");
     if (typeof context === "function") {
       this.contextRetrieverFn = context;
@@ -228,10 +228,7 @@ var Insight = class {
   async locate(query, opt) {
     const { callAI } = opt || {};
     const queryPrompt = typeof query === "string" ? query : query.prompt;
-    _utils.assert.call(void 0, 
-      queryPrompt || _optionalChain([opt, 'optionalAccess', _17 => _17.quickAnswer]),
-      "query or quickAnswer is required for locate"
-    );
+    _utils.assert.call(void 0, queryPrompt, "query is required for locate");
     const dumpSubscriber = this.onceDumpUpdatedFn;
     this.onceDumpUpdatedFn = void 0;
     _utils.assert.call(void 0, typeof query === "object", "query should be an object for locate");
@@ -251,13 +248,13 @@ var Insight = class {
       );
       searchAreaPrompt = void 0;
     }
-    const context = await this.contextRetrieverFn("locate");
+    const context = _optionalChain([opt, 'optionalAccess', _17 => _17.context]) || await this.contextRetrieverFn("locate");
     let searchArea = void 0;
     let searchAreaRawResponse = void 0;
     let searchAreaUsage = void 0;
     let searchAreaResponse = void 0;
     if (searchAreaPrompt) {
-      searchAreaResponse = await _chunkFYD4T7QDjs.AiLocateSection.call(void 0, {
+      searchAreaResponse = await _chunkH3B2KXXRjs.AiLocateSection.call(void 0, {
         context,
         sectionDescription: searchAreaPrompt
       });
@@ -270,11 +267,10 @@ var Insight = class {
       searchArea = searchAreaResponse.rect;
     }
     const startTime = Date.now();
-    const { parseResult, rect, elementById, rawResponse, usage } = await _chunkFYD4T7QDjs.AiLocateElement.call(void 0, {
+    const { parseResult, rect, elementById, rawResponse, usage } = await _chunkH3B2KXXRjs.AiLocateElement.call(void 0, {
       callAI: callAI || this.aiVendorFn,
       context,
       targetElementDescription: queryPrompt,
-      quickAnswer: _optionalChain([opt, 'optionalAccess', _18 => _18.quickAnswer]),
       searchConfig: searchAreaResponse
     });
     const timeCost = Date.now() - startTime;
@@ -289,7 +285,7 @@ var Insight = class {
       searchAreaUsage
     };
     let errorLog;
-    if (_optionalChain([parseResult, 'access', _19 => _19.errors, 'optionalAccess', _20 => _20.length])) {
+    if (_optionalChain([parseResult, 'access', _18 => _18.errors, 'optionalAccess', _19 => _19.length])) {
       errorLog = `AI model failed to locate: 
 ${parseResult.errors.join("\n")}`;
     }
@@ -298,7 +294,6 @@ ${parseResult.errors.join("\n")}`;
       userQuery: {
         element: queryPrompt
       },
-      quickAnswer: _optionalChain([opt, 'optionalAccess', _21 => _21.quickAnswer]),
       matchedElement: [],
       matchedRect: rect,
       data: null,
@@ -309,7 +304,7 @@ ${parseResult.errors.join("\n")}`;
     const elements = [];
     (parseResult.elements || []).forEach((item) => {
       if ("id" in item) {
-        const element = elementById(item.id);
+        const element = elementById(_optionalChain([item, 'optionalAccess', _20 => _20.id]));
         if (!element) {
           console.warn(
             `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`
@@ -339,7 +334,9 @@ ${parseResult.errors.join("\n")}`;
           id: elements[0].id,
           indexId: elements[0].indexId,
           center: elements[0].center,
-          rect: elements[0].rect
+          rect: elements[0].rect,
+          xpaths: elements[0].xpaths || [],
+          attributes: elements[0].attributes
         },
         rect
       };
@@ -358,7 +355,7 @@ ${parseResult.errors.join("\n")}`;
     this.onceDumpUpdatedFn = void 0;
     const context = await this.contextRetrieverFn("extract");
     const startTime = Date.now();
-    const { parseResult, usage } = await _chunkFYD4T7QDjs.AiExtractElementInfo.call(void 0, {
+    const { parseResult, usage } = await _chunkH3B2KXXRjs.AiExtractElementInfo.call(void 0, {
       context,
       dataQuery: dataDemand
     });
@@ -369,7 +366,7 @@ ${parseResult.errors.join("\n")}`;
       rawResponse: JSON.stringify(parseResult)
     };
     let errorLog;
-    if (_optionalChain([parseResult, 'access', _22 => _22.errors, 'optionalAccess', _23 => _23.length])) {
+    if (_optionalChain([parseResult, 'access', _21 => _21.errors, 'optionalAccess', _22 => _22.length])) {
       errorLog = `AI response error: 
 ${parseResult.errors.join("\n")}`;
     }
@@ -409,7 +406,7 @@ ${parseResult.errors.join("\n")}`;
     this.onceDumpUpdatedFn = void 0;
     const context = await this.contextRetrieverFn("assert");
     const startTime = Date.now();
-    const assertResult = await _chunkFYD4T7QDjs.AiAssert.call(void 0, {
+    const assertResult = await _chunkH3B2KXXRjs.AiAssert.call(void 0, {
       assertion,
       context
     });
@@ -455,6 +452,6 @@ var src_default = Insight;
 
 
 
-exports.AiAssert = _chunkFYD4T7QDjs.AiAssert; exports.AiLocateElement = _chunkFYD4T7QDjs.AiLocateElement; exports.Executor = Executor; exports.Insight = Insight; exports.MIDSCENE_MODEL_NAME = _env.MIDSCENE_MODEL_NAME; exports.default = src_default; exports.describeUserPage = _chunkFYD4T7QDjs.describeUserPage; exports.getAIConfig = _env.getAIConfig; exports.getVersion = _chunkHLRMNAQKjs.getVersion; exports.plan = _chunkFYD4T7QDjs.plan;
+exports.AiAssert = _chunkH3B2KXXRjs.AiAssert; exports.AiLocateElement = _chunkH3B2KXXRjs.AiLocateElement; exports.Executor = Executor; exports.Insight = Insight; exports.MIDSCENE_MODEL_NAME = _env.MIDSCENE_MODEL_NAME; exports.default = src_default; exports.describeUserPage = _chunkH3B2KXXRjs.describeUserPage; exports.getAIConfig = _env.getAIConfig; exports.getVersion = _chunkTOTKWQNQjs.getVersion; exports.plan = _chunkH3B2KXXRjs.plan;
 
 //# sourceMappingURL=index.js.map

package/dist/lib/index.js.map

@@ -1 +1 @@
-{"version":3,"mappings":";;;;;;;;;;;;;;AAUA,SAAS,qBAAqB,mBAAmB;AACjD,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,UAAM,WAA0B;AAAA,MAC9B,YAAY,WAAW;AAAA,MACvB,YAAY,YAAY,mBAAmB,KAAK;AAAA,MAChD,SAAS,KAAK,IAAI;AAAA,MAClB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACF;;;AC1LA;AAAA,EACE;AAAA,EACA;AAAA,EACA,gBAAAA;AAAA,OACK;AACP,SAAS,gBAAgB;AACzB,SAAS,UAAAC,eAAc;;;ACnBvB;AAAA,EACE,uBAAAC;AAAA,EACA,eAAAC;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,YAAY;AAEd,SAAS,gBACd,MACA,gBACA;AACA,MAAI,mBAAmB;AAEvB,MAAI,aAAa,GAAG;AAClB,UAAM,iBAAiB,mBAAmB;AAC1C,QAAI,gBAAgB;AAClB,yBAAmB,WAAW,cAAc;AAAA,IAC9C,OAAO;AACL,yBAAmB,GAAG,aAAa,CAAC;AAAA,IACtC;AAAA,EACF;AAEA,QAAM,WAAqB;AAAA,IACzB,YAAY,WAAW;AAAA,IACvB,SAAS,KAAK,IAAI;AAAA,IAClB,YAAYA,aAAYD,oBAAmB,KAAK;AAAA,IAChD,mBAAmB;AAAA,EACrB;AACA,QAAM,YAAyB;AAAA,IAC7B,OAAO,KAAK;AAAA,IACZ,GAAG;AAAA,IACH,GAAG;AAAA,EACL;AAEA,mBAAiB,SAAS;AAC5B;;;ADLA,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;AAAA,MACE,eAAe,KAAK;AAAA,MACpB;AAAA,IACF;AACA,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,MAAM,KAAK,mBAAmB,QAAQ;AAEtD,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,aAAa,KAAK;AAAA,MAClB,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,aAAa,KAAK;AAAA,MAClB,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,KAAK,EAAE;AAEnC,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,QACrB;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;AACF;;;AEjUA,SAAS,eAAAE,cAAa,uBAAAD,4BAA2B;AAGjD,IAAO,cAAQ","names":["vlLocateMode","assert","MIDSCENE_MODEL_NAME","getAIConfig"],"ignoreList":[],"sources":["../../src/ai-model/action-executor.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 { MIDSCENE_MODEL_NAME, getAIConfig } 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    const dumpData: ExecutionDump = {\n      sdkVersion: getVersion(),\n      model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n      logTime: Date.now(),\n      name: this.name,\n      tasks: this.tasks,\n    };\n    return dumpData;\n  }\n}\n","import { callAiFn } from '@/ai-model/common';\nimport { AiExtractElementInfo, AiLocateElement } from '@/ai-model/index';\nimport { AiAssert, AiLocateSection } from '@/ai-model/inspect';\nimport type {\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  getAIConfigInBoolean,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n  callAI?: typeof callAiFn<AIElementResponse>;\n  quickAnswer?: Partial<AISingleElementResponse>;\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(\n      queryPrompt || opt?.quickAnswer,\n      'query or quickAnswer is required for locate',\n    );\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 = 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 ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const { parseResult, rect, elementById, rawResponse, usage } =\n      await AiLocateElement({\n        callAI: callAI || this.aiVendorFn,\n        context,\n        targetElementDescription: queryPrompt,\n        quickAnswer: opt?.quickAnswer,\n        searchConfig: searchAreaResponse,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      quickAnswer: opt?.quickAnswer,\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        },\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}\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  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\n  const baseData: DumpMeta = {\n    sdkVersion: getVersion(),\n    logTime: Date.now(),\n    model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n    model_description: modelDescription,\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,SAAS,qBAAqB,mBAAmB;AACjD,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,UAAM,WAA0B;AAAA,MAC9B,YAAY,WAAW;AAAA,MACvB,YAAY,YAAY,mBAAmB,KAAK;AAAA,MAChD,SAAS,KAAK,IAAI;AAAA,MAClB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACF;;;AC1LA;AAAA,EACE;AAAA,EACA;AAAA,EACA,gBAAAA;AAAA,OACK;AACP,SAAS,gBAAgB;AACzB,SAAS,UAAAC,eAAc;;;ACnBvB;AAAA,EACE,uBAAAC;AAAA,EACA,eAAAC;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,YAAY;AAEd,SAAS,gBACd,MACA,gBACA;AACA,MAAI,mBAAmB;AAEvB,MAAI,aAAa,GAAG;AAClB,UAAM,iBAAiB,mBAAmB;AAC1C,QAAI,gBAAgB;AAClB,yBAAmB,WAAW,cAAc;AAAA,IAC9C,OAAO;AACL,yBAAmB,GAAG,aAAa,CAAC;AAAA,IACtC;AAAA,EACF;AAEA,QAAM,WAAqB;AAAA,IACzB,YAAY,WAAW;AAAA,IACvB,SAAS,KAAK,IAAI;AAAA,IAClB,YAAYA,aAAYD,oBAAmB,KAAK;AAAA,IAChD,mBAAmB;AAAA,EACrB;AACA,QAAM,YAAyB;AAAA,IAC7B,OAAO,KAAK;AAAA,IACZ,GAAG;AAAA,IACH,GAAG;AAAA,EACL;AAEA,mBAAiB,SAAS;AAC5B;;;ADLA,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;AACF;;;AE9TA,SAAS,eAAAE,cAAa,uBAAAD,4BAA2B;AAGjD,IAAO,cAAQ","names":["vlLocateMode","assert","MIDSCENE_MODEL_NAME","getAIConfig"],"ignoreList":[],"sources":["../../src/ai-model/action-executor.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 { MIDSCENE_MODEL_NAME, getAIConfig } 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    const dumpData: ExecutionDump = {\n      sdkVersion: getVersion(),\n      model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n      logTime: Date.now(),\n      name: this.name,\n      tasks: this.tasks,\n    };\n    return dumpData;\n  }\n}\n","import { callAiFn } from '@/ai-model/common';\nimport { AiExtractElementInfo, AiLocateElement } from '@/ai-model/index';\nimport { AiAssert, AiLocateSection } from '@/ai-model/inspect';\nimport type {\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  getAIConfigInBoolean,\n  vlLocateMode,\n} from '@midscene/shared/env';\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}\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  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\n  const baseData: DumpMeta = {\n    sdkVersion: getVersion(),\n    logTime: Date.now(),\n    model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n    model_description: modelDescription,\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"]}

package/dist/lib/{llm-planning-e16d1814.d.ts → llm-planning-26db5c81.d.ts}

@@ -1,4 +1,4 @@
-import { k as AIUsageInfo, R as Rect, B as BaseElement, U as UIContext, A as AISingleElementResponse, n as AISingleElementResponseByPosition, o as AIElementLocatorResponse, H as ElementById, r as AIDataExtractionResponse, t as AIAssertionResponse, ar as PageType, T as PlanningAIResponse } from './types-8c83481c.js';
+import { A as AIUsageInfo, R as Rect, j as ElementTreeNode, B as BaseElement, U as UIContext, o as AIElementLocatorResponse, H as ElementById, r as AIDataExtractionResponse, t as AIAssertionResponse, ar as PageType, T as PlanningAIResponse } from './types-cbcbeb4e.js';
 import { ChatCompletionSystemMessageParam, ChatCompletionUserMessageParam } from 'openai/resources';
 
 type AIArgs = [
@@ -17,6 +17,13 @@ declare function callAiFn<T>(msgs: AIArgs, AIActionTypeValue: AIActionType): Pro
 }>;
 declare function adaptBboxToRect(bbox: number[], width: number, height: number, offsetX?: number, offsetY?: number): Rect;
 
+declare function elementByPositionWithElementInfo(treeRoot: ElementTreeNode<BaseElement>, position: {
+    x: number;
+    y: number;
+}, options?: {
+    requireStrictDistance?: boolean;
+    filterPositionElements?: boolean;
+}): BaseElement | undefined;
 declare function describeUserPage<ElementType extends BaseElement = BaseElement>(context: Omit<UIContext<ElementType>, 'describer'>, opt?: {
     truncateTextLength?: number;
     filterNonTextContent?: boolean;
@@ -44,7 +51,6 @@ declare function AiLocateElement<ElementType extends BaseElement = BaseElement>(
     context: UIContext<ElementType>;
     targetElementDescription: string;
     callAI?: typeof callAiFn<AIElementResponse | [number, number]>;
-    quickAnswer?: Partial<AISingleElementResponse | AISingleElementResponseByPosition>;
     searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;
 }): Promise<{
     parseResult: AIElementLocatorResponse;
@@ -88,4 +94,4 @@ declare function plan(userInstruction: string, opts: {
     actionContext?: string;
 }): Promise<PlanningAIResponse>;
 
-export { AiLocateElement as A, AiAssert as a, AIActionType as b, callAiFn as c, describeUserPage as d, AiExtractElementInfo as e, AiLocateSection as f, adaptBboxToRect as g, plan as p };
+export { AiLocateElement as A, AiAssert as a, AIActionType as b, callAiFn as c, describeUserPage as d, elementByPositionWithElementInfo as e, AiExtractElementInfo as f, AiLocateSection as g, adaptBboxToRect as h, plan as p };

package/dist/lib/tree.d.ts

@@ -1,5 +1,5 @@
 import * as _midscene_shared_constants from '@midscene/shared/constants';
-import { B as BaseElement, j as ElementTreeNode } from './types-8c83481c.js';
+import { B as BaseElement, j as ElementTreeNode } from './types-cbcbeb4e.js';
 import 'openai/resources';
 
 declare function truncateText(text: string | number | object | undefined, maxLength?: number): string;

package/dist/{types/types-8c83481c.d.ts → lib/types-cbcbeb4e.d.ts}

@@ -143,6 +143,7 @@ declare abstract class BaseElement {
     abstract rect: Rect;
     abstract center: [number, number];
     abstract locator?: string;
+    abstract xpaths?: string[];
 }
 interface ElementTreeNode<ElementType extends BaseElement = BaseElement> {
     node: ElementType | null;
@@ -165,6 +166,7 @@ type AISingleElementResponseById = {
     id: string;
     reason?: string;
     text?: string;
+    xpaths?: string[];
 };
 type AISingleElementResponseByPosition = {
     position?: {
@@ -181,6 +183,7 @@ interface AIElementLocatorResponse {
         id: string;
         reason?: string;
         text?: string;
+        xpaths?: string[];
     }[];
     bbox?: [number, number, number, number];
     errors?: string[];
@@ -230,6 +233,11 @@ type LocateResultElement = {
     indexId?: number;
     center: [number, number];
     rect: Rect;
+    xpaths: string[];
+    attributes: {
+        nodeType: NodeType;
+        [key: string]: string;
+    };
 };
 interface LocateResult {
     element: LocateResultElement | null;
@@ -262,7 +270,6 @@ interface InsightDump extends DumpMeta {
         dataDemand?: InsightExtractParam;
         assertion?: string;
     };
-    quickAnswer?: Partial<AISingleElementResponse> | null;
     matchedElement: BaseElement[];
     matchedRect?: Rect;
     deepThink?: boolean;
@@ -319,6 +326,8 @@ interface PlanningAIResponse {
     error?: string;
     usage?: AIUsageInfo;
     rawResponse?: string;
+    yamlFlow?: MidsceneYamlFlowItem[];
+    yamlString?: string;
 }
 type PlanningActionParamTap = null;
 type PlanningActionParamHover = null;
@@ -376,7 +385,6 @@ interface ExecutionTaskApply<Type extends ExecutionTaskType = any, TaskParam = a
     param?: TaskParam;
     thought?: string;
     locate?: PlanningLocateParam | null;
-    quickAnswer?: AISingleElementResponse | null;
     pageContext?: UIContext;
     executor: (param: TaskParam, context: ExecutorContext) => Promise<ExecutionTaskReturn<TaskOutput, TaskLog> | undefined | void> | undefined | void;
 }
@@ -439,4 +447,4 @@ interface GroupedActionDump {
 }
 type PageType = 'puppeteer' | 'playwright' | 'static' | 'chrome-extension-proxy' | 'android';
 
-export { type PlanningActionParamError as $, type AISingleElementResponse as A, BaseElement as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type PartialInsightDumpFromSDK as F, type LiteUISection as G, type ElementById as H, type InsightAction as I, type AgentWaitForOpt as J, type AgentAssertOpt as K, type LocateResult as L, type MidsceneYamlScript as M, type PlanningLocateParam as N, type OnTaskStartTip as O, type Point as P, type PlanningAction as Q, type Rect as R, type Size as S, type PlanningAIResponse as T, UIContext as U, type PlanningActionParamTap as V, type PlanningActionParamHover as W, type PlanningActionParamInputOrKeyPress as X, type PlanningActionParamScroll as Y, type PlanningActionParamAssert as Z, type PlanningActionParamSleep as _, type ExecutionTaskProgressOptions as a, type PlanningActionParamWaitFor as a0, type Color as a1, type BaseAgentParserOpt as a2, type PuppeteerParserOpt as a3, type PlaywrightParserOpt as a4, type ExecutionRecorderItem as a5, type ExecutionTaskType as a6, type ExecutorContext as a7, type TaskCacheInfo as a8, type ExecutionTaskReturn as a9, type MidsceneYamlFlowItemAIQuery as aA, type MidsceneYamlFlowItemAINumber as aB, type MidsceneYamlFlowItemAINString as aC, type MidsceneYamlFlowItemAIBoolean as aD, type MidsceneYamlFlowItemAILocate as aE, type MidsceneYamlFlowItemAIWaitFor as aF, type MidsceneYamlFlowItemAITap as aG, type MidsceneYamlFlowItemAIHover as aH, type MidsceneYamlFlowItemAIInput as aI, type MidsceneYamlFlowItemAIKeyboardPress as aJ, type MidsceneYamlFlowItemAIScroll as aK, type MidsceneYamlFlowItemEvaluateJavaScript as aL, type MidsceneYamlFlowItemSleep as aM, type FreeFn as aN, type ScriptPlayerTaskStatus as aO, type ScriptPlayerStatusValue as aP, type ExecutionTaskInsightLocateParam as aa, type ExecutionTaskInsightLocateOutput as ab, type ExecutionTaskInsightDumpLog as ac, type ExecutionTaskInsightLocateApply as ad, type ExecutionTaskInsightLocate as ae, type ExecutionTaskInsightQueryParam as af, type ExecutionTaskInsightQueryOutput as ag, type ExecutionTaskInsightQueryApply as ah, type ExecutionTaskInsightQuery as ai, type ExecutionTaskInsightAssertionParam as aj, type ExecutionTaskInsightAssertionApply as ak, type ExecutionTaskInsightAssertion as al, type ExecutionTaskActionApply as am, type ExecutionTaskAction as an, type ExecutionTaskPlanningApply as ao, type ExecutionTaskPlanning as ap, type GroupedActionDump as aq, type PageType as ar, type LocateOption as as, type scrollParam as at, type MidsceneYamlScriptEnvBase as au, type MidsceneYamlScriptWebEnv as av, type MidsceneYamlScriptAndroidEnv as aw, type MidsceneYamlScriptEnv as ax, type MidsceneYamlFlowItemAIAction as ay, type MidsceneYamlFlowItemAIAssert 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 AIElementLocatorResponse as o, type AIElementCoordinatesResponse as p, type AIElementResponse as q, type AIDataExtractionResponse as r, type AISectionLocatorResponse as s, type AIAssertionResponse as t, type EnsureObject as u, type InsightExtractParam as v, type LocateResultElement as w, type DumpMeta as x, type ReportDumpWithAttributes as y, type InsightDump as z };
+export { type PlanningActionParamError as $, type AIUsageInfo as A, BaseElement as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type PartialInsightDumpFromSDK as F, type LiteUISection as G, type ElementById as H, type InsightAction as I, type AgentWaitForOpt as J, type AgentAssertOpt as K, type LocateResult as L, type MidsceneYamlScript as M, type PlanningLocateParam as N, type OnTaskStartTip as O, type Point as P, type PlanningAction as Q, type Rect as R, type Size as S, type PlanningAIResponse as T, UIContext as U, type PlanningActionParamTap as V, type PlanningActionParamHover as W, type PlanningActionParamInputOrKeyPress as X, type PlanningActionParamScroll as Y, type PlanningActionParamAssert as Z, type PlanningActionParamSleep as _, type ExecutionTaskProgressOptions as a, type PlanningActionParamWaitFor as a0, type Color as a1, type BaseAgentParserOpt as a2, type PuppeteerParserOpt as a3, type PlaywrightParserOpt as a4, type ExecutionRecorderItem as a5, type ExecutionTaskType as a6, type ExecutorContext as a7, type TaskCacheInfo as a8, type ExecutionTaskReturn as a9, type MidsceneYamlFlowItemAIQuery as aA, type MidsceneYamlFlowItemAINumber as aB, type MidsceneYamlFlowItemAINString as aC, type MidsceneYamlFlowItemAIBoolean as aD, type MidsceneYamlFlowItemAILocate as aE, type MidsceneYamlFlowItemAIWaitFor as aF, type MidsceneYamlFlowItemAITap as aG, type MidsceneYamlFlowItemAIHover as aH, type MidsceneYamlFlowItemAIInput as aI, type MidsceneYamlFlowItemAIKeyboardPress as aJ, type MidsceneYamlFlowItemAIScroll as aK, type MidsceneYamlFlowItemEvaluateJavaScript as aL, type MidsceneYamlFlowItemSleep as aM, type FreeFn as aN, type ScriptPlayerTaskStatus as aO, type ScriptPlayerStatusValue as aP, type ExecutionTaskInsightLocateParam as aa, type ExecutionTaskInsightLocateOutput as ab, type ExecutionTaskInsightDumpLog as ac, type ExecutionTaskInsightLocateApply as ad, type ExecutionTaskInsightLocate as ae, type ExecutionTaskInsightQueryParam as af, type ExecutionTaskInsightQueryOutput as ag, type ExecutionTaskInsightQueryApply as ah, type ExecutionTaskInsightQuery as ai, type ExecutionTaskInsightAssertionParam as aj, type ExecutionTaskInsightAssertionApply as ak, type ExecutionTaskInsightAssertion as al, type ExecutionTaskActionApply as am, type ExecutionTaskAction as an, type ExecutionTaskPlanningApply as ao, type ExecutionTaskPlanning as ap, type GroupedActionDump as aq, type PageType as ar, type LocateOption as as, type scrollParam as at, type MidsceneYamlScriptEnvBase as au, type MidsceneYamlScriptWebEnv as av, type MidsceneYamlScriptAndroidEnv as aw, type MidsceneYamlScriptEnv as ax, type MidsceneYamlFlowItemAIAction as ay, type MidsceneYamlFlowItemAIAssert 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, AIResponseFormat as k, type AISingleElementResponseById as l, type AISingleElementResponseByPosition as m, type AISingleElementResponse as n, type AIElementLocatorResponse as o, type AIElementCoordinatesResponse as p, type AIElementResponse as q, type AIDataExtractionResponse as r, type AISectionLocatorResponse as s, type AIAssertionResponse as t, type EnsureObject as u, type InsightExtractParam as v, type LocateResultElement as w, type DumpMeta as x, type ReportDumpWithAttributes as y, type InsightDump as z };

package/dist/lib/utils.d.ts

@@ -1,4 +1,4 @@
-import { y as ReportDumpWithAttributes, R as Rect } from './types-8c83481c.js';
+import { y as ReportDumpWithAttributes, R as Rect } from './types-cbcbeb4e.js';
 import '@midscene/shared/constants';
 import 'openai/resources';
 

package/dist/lib/utils.js

@@ -13,7 +13,7 @@
 
 
 
-var _chunkHLRMNAQKjs = require('./chunk-HLRMNAQK.js');
+var _chunkTOTKWQNQjs = require('./chunk-TOTKWQNQ.js');
 
 
 
@@ -29,4 +29,4 @@ var _chunkHLRMNAQKjs = require('./chunk-HLRMNAQK.js');
 
 
 
-exports.getLogDir = _chunkHLRMNAQKjs.getLogDir; exports.getTmpDir = _chunkHLRMNAQKjs.getTmpDir; exports.getTmpFile = _chunkHLRMNAQKjs.getTmpFile; exports.getVersion = _chunkHLRMNAQKjs.getVersion; exports.groupedActionDumpFileExt = _chunkHLRMNAQKjs.groupedActionDumpFileExt; exports.overlapped = _chunkHLRMNAQKjs.overlapped; exports.replaceStringWithFirstAppearance = _chunkHLRMNAQKjs.replaceStringWithFirstAppearance; exports.replacerForPageObject = _chunkHLRMNAQKjs.replacerForPageObject; exports.reportHTMLContent = _chunkHLRMNAQKjs.reportHTMLContent; exports.sleep = _chunkHLRMNAQKjs.sleep; exports.stringifyDumpData = _chunkHLRMNAQKjs.stringifyDumpData; exports.uploadTestInfoToServer = _chunkHLRMNAQKjs.uploadTestInfoToServer; exports.writeDumpReport = _chunkHLRMNAQKjs.writeDumpReport; exports.writeLogFile = _chunkHLRMNAQKjs.writeLogFile;
+exports.getLogDir = _chunkTOTKWQNQjs.getLogDir; exports.getTmpDir = _chunkTOTKWQNQjs.getTmpDir; exports.getTmpFile = _chunkTOTKWQNQjs.getTmpFile; exports.getVersion = _chunkTOTKWQNQjs.getVersion; exports.groupedActionDumpFileExt = _chunkTOTKWQNQjs.groupedActionDumpFileExt; exports.overlapped = _chunkTOTKWQNQjs.overlapped; exports.replaceStringWithFirstAppearance = _chunkTOTKWQNQjs.replaceStringWithFirstAppearance; exports.replacerForPageObject = _chunkTOTKWQNQjs.replacerForPageObject; exports.reportHTMLContent = _chunkTOTKWQNQjs.reportHTMLContent; exports.sleep = _chunkTOTKWQNQjs.sleep; exports.stringifyDumpData = _chunkTOTKWQNQjs.stringifyDumpData; exports.uploadTestInfoToServer = _chunkTOTKWQNQjs.uploadTestInfoToServer; exports.writeDumpReport = _chunkTOTKWQNQjs.writeDumpReport; exports.writeLogFile = _chunkTOTKWQNQjs.writeLogFile;

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

@@ -1,8 +1,8 @@
-import { k as AIUsageInfo, Q as PlanningAction } from './types-8c83481c.js';
+import { A as AIUsageInfo, Q as PlanningAction, i as MidsceneYamlFlowItem, S as Size } from './types-cbcbeb4e.js';
 import { ChatCompletionMessageParam } from 'openai/resources';
 export { ChatCompletionMessageParam } from 'openai/resources';
-import { b as AIActionType } from './llm-planning-e16d1814.js';
-export { a as AiAssert, e as AiExtractElementInfo, A as AiLocateElement, f as AiLocateSection, g as adaptBboxToRect, c as callAiFn, d as describeUserPage, p as plan } from './llm-planning-e16d1814.js';
+import { b as AIActionType } from './llm-planning-26db5c81.js';
+export { a as AiAssert, f as AiExtractElementInfo, A as AiLocateElement, g as AiLocateSection, h as adaptBboxToRect, c as callAiFn, d as describeUserPage, e as elementByPositionWithElementInfo, p as plan } from './llm-planning-26db5c81.js';
 import { vlLocateMode } from '@midscene/shared/env';
 import { actionParser } from '@ui-tars/action-parser';
 import '@midscene/shared/constants';
@@ -23,8 +23,10 @@ declare function vlmPlanning(options: {
     };
 }): Promise<{
     actions: PlanningAction<any>[];
-    realActions: ReturnType<typeof actionParser>['parsed'];
+    actionsFromModel: ReturnType<typeof actionParser>['parsed'];
     action_summary: string;
+    yamlFlow: MidsceneYamlFlowItem[];
 }>;
+declare function resizeImageForUiTars(imageBase64: string, size: Size): Promise<string>;
 
-export { callToGetJSONObject, systemPromptToLocateElement, vlmPlanning };
+export { callToGetJSONObject, resizeImageForUiTars, systemPromptToLocateElement, vlmPlanning };

package/dist/types/index.d.ts

@@ -1,7 +1,7 @@
-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, A as AISingleElementResponse } from './types-8c83481c.js';
-export { t as AIAssertionResponse, r as AIDataExtractionResponse, p as AIElementCoordinatesResponse, o as AIElementLocatorResponse, q as AIElementResponse, l as AIResponseFormat, s as AISectionLocatorResponse, m as AISingleElementResponseById, n as AISingleElementResponseByPosition, k as AIUsageInfo, K as AgentAssertOpt, J as AgentWaitForOpt, a2 as BaseAgentParserOpt, C as CallAIFn, a1 as Color, x as DumpMeta, H as ElementById, j as ElementTreeNode, u as EnsureObject, a5 as ExecutionRecorderItem, an as ExecutionTaskAction, am as ExecutionTaskActionApply, al as ExecutionTaskInsightAssertion, ak as ExecutionTaskInsightAssertionApply, aj as ExecutionTaskInsightAssertionParam, ac as ExecutionTaskInsightDumpLog, ae as ExecutionTaskInsightLocate, ad as ExecutionTaskInsightLocateApply, ab as ExecutionTaskInsightLocateOutput, aa as ExecutionTaskInsightLocateParam, ai as ExecutionTaskInsightQuery, ah as ExecutionTaskInsightQueryApply, ag as ExecutionTaskInsightQueryOutput, af as ExecutionTaskInsightQueryParam, ap as ExecutionTaskPlanning, ao as ExecutionTaskPlanningApply, a9 as ExecutionTaskReturn, a6 as ExecutionTaskType, a7 as ExecutorContext, aN as FreeFn, aq as GroupedActionDump, z as InsightDump, v as InsightExtractParam, G as LiteUISection, as as LocateOption, w as LocateResultElement, i as MidsceneYamlFlowItem, ay as MidsceneYamlFlowItemAIAction, az as MidsceneYamlFlowItemAIAssert, aD as MidsceneYamlFlowItemAIBoolean, aH as MidsceneYamlFlowItemAIHover, aI as MidsceneYamlFlowItemAIInput, aJ as MidsceneYamlFlowItemAIKeyboardPress, aE as MidsceneYamlFlowItemAILocate, aC as MidsceneYamlFlowItemAINString, aB as MidsceneYamlFlowItemAINumber, aA as MidsceneYamlFlowItemAIQuery, aK as MidsceneYamlFlowItemAIScroll, aG as MidsceneYamlFlowItemAITap, aF as MidsceneYamlFlowItemAIWaitFor, aL as MidsceneYamlFlowItemEvaluateJavaScript, aM as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aw as MidsceneYamlScriptAndroidEnv, ax as MidsceneYamlScriptEnv, au as MidsceneYamlScriptEnvBase, av as MidsceneYamlScriptWebEnv, h as MidsceneYamlTask, O as OnTaskStartTip, ar as PageType, F as PartialInsightDumpFromSDK, T as PlanningAIResponse, Q as PlanningAction, Z as PlanningActionParamAssert, $ as PlanningActionParamError, W as PlanningActionParamHover, X as PlanningActionParamInputOrKeyPress, Y as PlanningActionParamScroll, _ as PlanningActionParamSleep, V as PlanningActionParamTap, a0 as PlanningActionParamWaitFor, N as PlanningLocateParam, a4 as PlaywrightParserOpt, P as Point, a3 as PuppeteerParserOpt, R as Rect, y as ReportDumpWithAttributes, aP as ScriptPlayerStatusValue, aO as ScriptPlayerTaskStatus, S as Size, a8 as TaskCacheInfo, at as scrollParam } from './types-8c83481c.js';
-import { c as callAiFn } from './llm-planning-e16d1814.js';
-export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-e16d1814.js';
+import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, B as BaseElement, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightAssertionResponse } from './types-cbcbeb4e.js';
+export { t as AIAssertionResponse, r as AIDataExtractionResponse, p as AIElementCoordinatesResponse, o as AIElementLocatorResponse, q as AIElementResponse, k as AIResponseFormat, s as AISectionLocatorResponse, n as AISingleElementResponse, l as AISingleElementResponseById, m as AISingleElementResponseByPosition, A as AIUsageInfo, K as AgentAssertOpt, J as AgentWaitForOpt, a2 as BaseAgentParserOpt, C as CallAIFn, a1 as Color, x as DumpMeta, H as ElementById, j as ElementTreeNode, u as EnsureObject, a5 as ExecutionRecorderItem, an as ExecutionTaskAction, am as ExecutionTaskActionApply, al as ExecutionTaskInsightAssertion, ak as ExecutionTaskInsightAssertionApply, aj as ExecutionTaskInsightAssertionParam, ac as ExecutionTaskInsightDumpLog, ae as ExecutionTaskInsightLocate, ad as ExecutionTaskInsightLocateApply, ab as ExecutionTaskInsightLocateOutput, aa as ExecutionTaskInsightLocateParam, ai as ExecutionTaskInsightQuery, ah as ExecutionTaskInsightQueryApply, ag as ExecutionTaskInsightQueryOutput, af as ExecutionTaskInsightQueryParam, ap as ExecutionTaskPlanning, ao as ExecutionTaskPlanningApply, a9 as ExecutionTaskReturn, a6 as ExecutionTaskType, a7 as ExecutorContext, aN as FreeFn, aq as GroupedActionDump, z as InsightDump, v as InsightExtractParam, G as LiteUISection, as as LocateOption, w as LocateResultElement, i as MidsceneYamlFlowItem, ay as MidsceneYamlFlowItemAIAction, az as MidsceneYamlFlowItemAIAssert, aD as MidsceneYamlFlowItemAIBoolean, aH as MidsceneYamlFlowItemAIHover, aI as MidsceneYamlFlowItemAIInput, aJ as MidsceneYamlFlowItemAIKeyboardPress, aE as MidsceneYamlFlowItemAILocate, aC as MidsceneYamlFlowItemAINString, aB as MidsceneYamlFlowItemAINumber, aA as MidsceneYamlFlowItemAIQuery, aK as MidsceneYamlFlowItemAIScroll, aG as MidsceneYamlFlowItemAITap, aF as MidsceneYamlFlowItemAIWaitFor, aL as MidsceneYamlFlowItemEvaluateJavaScript, aM as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aw as MidsceneYamlScriptAndroidEnv, ax as MidsceneYamlScriptEnv, au as MidsceneYamlScriptEnvBase, av as MidsceneYamlScriptWebEnv, h as MidsceneYamlTask, O as OnTaskStartTip, ar as PageType, F as PartialInsightDumpFromSDK, T as PlanningAIResponse, Q as PlanningAction, Z as PlanningActionParamAssert, $ as PlanningActionParamError, W as PlanningActionParamHover, X as PlanningActionParamInputOrKeyPress, Y as PlanningActionParamScroll, _ as PlanningActionParamSleep, V as PlanningActionParamTap, a0 as PlanningActionParamWaitFor, N as PlanningLocateParam, a4 as PlaywrightParserOpt, P as Point, a3 as PuppeteerParserOpt, R as Rect, y as ReportDumpWithAttributes, aP as ScriptPlayerStatusValue, aO as ScriptPlayerTaskStatus, S as Size, a8 as TaskCacheInfo, at as scrollParam } from './types-cbcbeb4e.js';
+import { c as callAiFn } from './llm-planning-26db5c81.js';
+export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-26db5c81.js';
 export { getVersion } from './utils.js';
 export { MIDSCENE_MODEL_NAME, getAIConfig } from '@midscene/shared/env';
 import '@midscene/shared/constants';
@@ -24,8 +24,8 @@ declare class Executor {
 }
 
 interface LocateOpts {
+    context?: UIContext<BaseElement>;
     callAI?: typeof callAiFn<AIElementResponse>;
-    quickAnswer?: Partial<AISingleElementResponse>;
 }
 declare class Insight<ElementType extends BaseElement = BaseElement, ContextType extends UIContext<ElementType> = UIContext<ElementType>> {
     contextRetrieverFn: (action: InsightAction) => Promise<ContextType> | ContextType;
@@ -40,4 +40,4 @@ declare class Insight<ElementType extends BaseElement = BaseElement, ContextType
     assert(assertion: string): Promise<InsightAssertionResponse>;
 }
 
-export { AISingleElementResponse, BaseElement, DetailedLocateParam, DumpSubscriber, ExecutionDump, ExecutionTask, ExecutionTaskApply, ExecutionTaskProgressOptions, Executor, Insight, InsightAction, InsightAssertionResponse, InsightOptions, InsightTaskInfo, LocateResult, UIContext, Insight as default };
+export { BaseElement, DetailedLocateParam, DumpSubscriber, ExecutionDump, ExecutionTask, ExecutionTaskApply, ExecutionTaskProgressOptions, Executor, Insight, InsightAction, InsightAssertionResponse, InsightOptions, InsightTaskInfo, LocateResult, UIContext, Insight as default };

package/dist/types/{llm-planning-e16d1814.d.ts → llm-planning-26db5c81.d.ts}

@@ -1,4 +1,4 @@
-import { k as AIUsageInfo, R as Rect, B as BaseElement, U as UIContext, A as AISingleElementResponse, n as AISingleElementResponseByPosition, o as AIElementLocatorResponse, H as ElementById, r as AIDataExtractionResponse, t as AIAssertionResponse, ar as PageType, T as PlanningAIResponse } from './types-8c83481c.js';
+import { A as AIUsageInfo, R as Rect, j as ElementTreeNode, B as BaseElement, U as UIContext, o as AIElementLocatorResponse, H as ElementById, r as AIDataExtractionResponse, t as AIAssertionResponse, ar as PageType, T as PlanningAIResponse } from './types-cbcbeb4e.js';
 import { ChatCompletionSystemMessageParam, ChatCompletionUserMessageParam } from 'openai/resources';
 
 type AIArgs = [
@@ -17,6 +17,13 @@ declare function callAiFn<T>(msgs: AIArgs, AIActionTypeValue: AIActionType): Pro
 }>;
 declare function adaptBboxToRect(bbox: number[], width: number, height: number, offsetX?: number, offsetY?: number): Rect;
 
+declare function elementByPositionWithElementInfo(treeRoot: ElementTreeNode<BaseElement>, position: {
+    x: number;
+    y: number;
+}, options?: {
+    requireStrictDistance?: boolean;
+    filterPositionElements?: boolean;
+}): BaseElement | undefined;
 declare function describeUserPage<ElementType extends BaseElement = BaseElement>(context: Omit<UIContext<ElementType>, 'describer'>, opt?: {
     truncateTextLength?: number;
     filterNonTextContent?: boolean;
@@ -44,7 +51,6 @@ declare function AiLocateElement<ElementType extends BaseElement = BaseElement>(
     context: UIContext<ElementType>;
     targetElementDescription: string;
     callAI?: typeof callAiFn<AIElementResponse | [number, number]>;
-    quickAnswer?: Partial<AISingleElementResponse | AISingleElementResponseByPosition>;
     searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;
 }): Promise<{
     parseResult: AIElementLocatorResponse;
@@ -88,4 +94,4 @@ declare function plan(userInstruction: string, opts: {
     actionContext?: string;
 }): Promise<PlanningAIResponse>;
 
-export { AiLocateElement as A, AiAssert as a, AIActionType as b, callAiFn as c, describeUserPage as d, AiExtractElementInfo as e, AiLocateSection as f, adaptBboxToRect as g, plan as p };
+export { AiLocateElement as A, AiAssert as a, AIActionType as b, callAiFn as c, describeUserPage as d, elementByPositionWithElementInfo as e, AiExtractElementInfo as f, AiLocateSection as g, adaptBboxToRect as h, plan as p };

package/dist/types/tree.d.ts

@@ -1,5 +1,5 @@
 import * as _midscene_shared_constants from '@midscene/shared/constants';
-import { B as BaseElement, j as ElementTreeNode } from './types-8c83481c.js';
+import { B as BaseElement, j as ElementTreeNode } from './types-cbcbeb4e.js';
 import 'openai/resources';
 
 declare function truncateText(text: string | number | object | undefined, maxLength?: number): string;