@midscene/core 0.23.5-beta-20250728060950.0 → 0.23.5-beta-20250728070606.0

package/dist/lib/index.d.ts

@@ -1,7 +1,7 @@
-import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightExtractOption, h as InsightAssertionResponse, A as AIDescribeElementResponse } from './types-4d9c9a85.js';
-export { w as AIAssertionResponse, u as AIDataExtractionResponse, s as AIElementCoordinatesResponse, r as AIElementLocatorResponse, t as AIElementResponse, n as AIResponseFormat, v as AISectionLocatorResponse, q as AISingleElementResponse, o as AISingleElementResponseById, p as AISingleElementResponseByPosition, m as AIUsageInfo, S as AgentAssertOpt, z as AgentDescribeElementAtPointResult, Q as AgentWaitForOpt, a5 as BaseAgentParserOpt, C as CallAIFn, az as CodeGenerationChunk, a4 as Color, H as DumpMeta, N as ElementById, B as EnsureObject, a8 as ExecutionRecorderItem, aq as ExecutionTaskAction, ap as ExecutionTaskActionApply, ab as ExecutionTaskHitBy, ao as ExecutionTaskInsightAssertion, an as ExecutionTaskInsightAssertionApply, am as ExecutionTaskInsightAssertionParam, af as ExecutionTaskInsightDumpLog, ah as ExecutionTaskInsightLocate, ag as ExecutionTaskInsightLocateApply, ae as ExecutionTaskInsightLocateOutput, ad as ExecutionTaskInsightLocateParam, al as ExecutionTaskInsightQuery, ak as ExecutionTaskInsightQueryApply, aj as ExecutionTaskInsightQueryOutput, ai as ExecutionTaskInsightQueryParam, as as ExecutionTaskLog, ar as ExecutionTaskLogApply, au as ExecutionTaskPlanning, at as ExecutionTaskPlanningApply, ac as ExecutionTaskReturn, a9 as ExecutionTaskType, aa as ExecutorContext, aZ as FreeFn, av as GroupedActionDump, J as InsightDump, F as InsightExtractParam, K as LiteUISection, aB as LocateOption, G as LocateResultElement, y as LocateValidatorResult, x as LocatorValidatorOption, b0 as MidsceneYamlConfig, b1 as MidsceneYamlConfigOutput, l as MidsceneYamlConfigResult, j as MidsceneYamlFlowItem, aI as MidsceneYamlFlowItemAIAction, aN as MidsceneYamlFlowItemAIAsk, aJ as MidsceneYamlFlowItemAIAssert, aO as MidsceneYamlFlowItemAIBoolean, aS as MidsceneYamlFlowItemAIHover, aT as MidsceneYamlFlowItemAIInput, aU as MidsceneYamlFlowItemAIKeyboardPress, aP as MidsceneYamlFlowItemAILocate, aL as MidsceneYamlFlowItemAINumber, aK as MidsceneYamlFlowItemAIQuery, k as MidsceneYamlFlowItemAIRightClick, aV as MidsceneYamlFlowItemAIScroll, aM as MidsceneYamlFlowItemAIString, aR as MidsceneYamlFlowItemAITap, aQ as MidsceneYamlFlowItemAIWaitFor, aW as MidsceneYamlFlowItemEvaluateJavaScript, aY as MidsceneYamlFlowItemLogScreenshot, aX as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aG as MidsceneYamlScriptAndroidEnv, aH as MidsceneYamlScriptEnv, aE as MidsceneYamlScriptEnvBase, aF as MidsceneYamlScriptWebEnv, i as MidsceneYamlTask, O as OnTaskStartTip, aw as PageType, P as PartialInsightDumpFromSDK, W as PlanningAIResponse, V as PlanningAction, a0 as PlanningActionParamAssert, a2 as PlanningActionParamError, Y as PlanningActionParamHover, _ as PlanningActionParamInputOrKeyPress, Z as PlanningActionParamRightClick, $ as PlanningActionParamScroll, a1 as PlanningActionParamSleep, X as PlanningActionParamTap, a3 as PlanningActionParamWaitFor, T as PlanningLocateParam, a7 as PlaywrightParserOpt, a6 as PuppeteerParserOpt, aC as ReferenceImage, R as ReportDumpWithAttributes, a$ as ScriptPlayerStatusValue, a_ as ScriptPlayerTaskStatus, aA as StreamingAIResponse, ay as StreamingCallback, ax as StreamingCodeGenerationOptions, aD as scrollParam } from './types-4d9c9a85.js';
-import { c as callAiFn } from './llm-planning-3407bd42.js';
-export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-3407bd42.js';
+import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightExtractParam, h as InsightExtractOption, i as InsightAssertionResponse, A as AIDescribeElementResponse } from './types-27dc17c9.js';
+export { x as AIAssertionResponse, v as AIDataExtractionResponse, t as AIElementCoordinatesResponse, s as AIElementLocatorResponse, u as AIElementResponse, o as AIResponseFormat, w as AISectionLocatorResponse, r as AISingleElementResponse, p as AISingleElementResponseById, q as AISingleElementResponseByPosition, n as AIUsageInfo, S as AgentAssertOpt, B as AgentDescribeElementAtPointResult, Q as AgentWaitForOpt, a5 as BaseAgentParserOpt, C as CallAIFn, az as CodeGenerationChunk, a4 as Color, H as DumpMeta, N as ElementById, F as EnsureObject, a8 as ExecutionRecorderItem, aq as ExecutionTaskAction, ap as ExecutionTaskActionApply, ab as ExecutionTaskHitBy, ao as ExecutionTaskInsightAssertion, an as ExecutionTaskInsightAssertionApply, am as ExecutionTaskInsightAssertionParam, af as ExecutionTaskInsightDumpLog, ah as ExecutionTaskInsightLocate, ag as ExecutionTaskInsightLocateApply, ae as ExecutionTaskInsightLocateOutput, ad as ExecutionTaskInsightLocateParam, al as ExecutionTaskInsightQuery, ak as ExecutionTaskInsightQueryApply, aj as ExecutionTaskInsightQueryOutput, ai as ExecutionTaskInsightQueryParam, as as ExecutionTaskLog, ar as ExecutionTaskLogApply, au as ExecutionTaskPlanning, at as ExecutionTaskPlanningApply, ac as ExecutionTaskReturn, a9 as ExecutionTaskType, aa as ExecutorContext, a_ as FreeFn, av as GroupedActionDump, J as InsightDump, K as LiteUISection, aC as LocateOption, G as LocateResultElement, z as LocateValidatorResult, y as LocatorValidatorOption, b1 as MidsceneYamlConfig, b2 as MidsceneYamlConfigOutput, m as MidsceneYamlConfigResult, k as MidsceneYamlFlowItem, aJ as MidsceneYamlFlowItemAIAction, aO as MidsceneYamlFlowItemAIAsk, aK as MidsceneYamlFlowItemAIAssert, aP as MidsceneYamlFlowItemAIBoolean, aT as MidsceneYamlFlowItemAIHover, aU as MidsceneYamlFlowItemAIInput, aV as MidsceneYamlFlowItemAIKeyboardPress, aQ as MidsceneYamlFlowItemAILocate, aM as MidsceneYamlFlowItemAINumber, aL as MidsceneYamlFlowItemAIQuery, l as MidsceneYamlFlowItemAIRightClick, aW as MidsceneYamlFlowItemAIScroll, aN as MidsceneYamlFlowItemAIString, aS as MidsceneYamlFlowItemAITap, aR as MidsceneYamlFlowItemAIWaitFor, aX as MidsceneYamlFlowItemEvaluateJavaScript, aZ as MidsceneYamlFlowItemLogScreenshot, aY as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aH as MidsceneYamlScriptAndroidEnv, aI as MidsceneYamlScriptEnv, aF as MidsceneYamlScriptEnvBase, aG as MidsceneYamlScriptWebEnv, j as MidsceneYamlTask, O as OnTaskStartTip, aw as PageType, P as PartialInsightDumpFromSDK, W as PlanningAIResponse, V as PlanningAction, a0 as PlanningActionParamAssert, a2 as PlanningActionParamError, Y as PlanningActionParamHover, _ as PlanningActionParamInputOrKeyPress, Z as PlanningActionParamRightClick, $ as PlanningActionParamScroll, a1 as PlanningActionParamSleep, X as PlanningActionParamTap, a3 as PlanningActionParamWaitFor, T as PlanningLocateParam, a7 as PlaywrightParserOpt, a6 as PuppeteerParserOpt, aD as ReferenceImage, R as ReportDumpWithAttributes, b0 as ScriptPlayerStatusValue, a$ as ScriptPlayerTaskStatus, aA as StreamingAIResponse, ay as StreamingCallback, ax as StreamingCodeGenerationOptions, aB as TUserPrompt, aE as scrollParam } from './types-27dc17c9.js';
+import { c as callAiFn } from './llm-planning-4ac580dc.js';
+export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-4ac580dc.js';
 import { BaseElement, Rect } from '@midscene/shared/types';
 export { BaseElement, ElementTreeNode, Point, Rect, Size } from '@midscene/shared/types';
 export { getVersion } from './utils.js';
@@ -36,13 +36,11 @@ declare class Insight<ElementType extends BaseElement = BaseElement, ContextType
     taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;
     constructor(context: ContextType | ((action: InsightAction) => Promise<ContextType> | ContextType), opt?: InsightOptions);
     locate(query: DetailedLocateParam, opt?: LocateOpts): Promise<LocateResult>;
-    extract<T = any>(input: string, opt?: InsightExtractOption): Promise<T>;
-    extract<T extends Record<string, string>>(input: T, opt?: InsightExtractOption): Promise<Record<keyof T, any>>;
-    extract<T extends object>(input: Record<keyof T, string>, opt?: InsightExtractOption): Promise<T>;
+    extract<T>(dataDemand: InsightExtractParam, opt?: InsightExtractOption, promptImages?: Record<string, string>): Promise<any>;
     assert(assertion: string): Promise<InsightAssertionResponse>;
     describe(target: Rect | [number, number], opt?: {
         deepThink?: boolean;
     }): Promise<Pick<AIDescribeElementResponse, 'description'>>;
 }
 
-export { AIDescribeElementResponse, DetailedLocateParam, DumpSubscriber, ExecutionDump, ExecutionTask, ExecutionTaskApply, ExecutionTaskProgressOptions, Executor, Insight, InsightAction, InsightAssertionResponse, InsightExtractOption, InsightOptions, InsightTaskInfo, LocateResult, UIContext, Insight as default };
+export { AIDescribeElementResponse, DetailedLocateParam, DumpSubscriber, ExecutionDump, ExecutionTask, ExecutionTaskApply, ExecutionTaskProgressOptions, Executor, Insight, InsightAction, InsightAssertionResponse, InsightExtractOption, InsightExtractParam, 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 _chunk73GOZ45Rjs = require('./chunk-73GOZ45R.js');
+var _chunkQSWMWTO4js = require('./chunk-QSWMWTO4.js');
 
 
 
@@ -11,7 +11,7 @@ var _chunk73GOZ45Rjs = require('./chunk-73GOZ45R.js');
 
 
 
-var _chunkTV47VKQ7js = require('./chunk-TV47VKQ7.js');
+var _chunkQFXN2AP7js = require('./chunk-QFXN2AP7.js');
 
 // src/ai-model/action-executor.ts
 
@@ -171,7 +171,7 @@ ${_optionalChain([this, 'access', _7 => _7.latestErrorTask, 'call', _8 => _8(),
       }
     }
     const dumpData = {
-      sdkVersion: _chunk73GOZ45Rjs.getVersion.call(void 0, ),
+      sdkVersion: _chunkQSWMWTO4js.getVersion.call(void 0, ),
       model_name: _env.getAIConfig.call(void 0, _env.MIDSCENE_MODEL_NAME) || "",
       model_description: modelDescription,
       logTime: Date.now(),
@@ -219,7 +219,7 @@ var _logger = require('@midscene/shared/logger');
 
 function emitInsightDump(data, dumpSubscriber) {
   const baseData = {
-    sdkVersion: _chunk73GOZ45Rjs.getVersion.call(void 0, ),
+    sdkVersion: _chunkQSWMWTO4js.getVersion.call(void 0, ),
     logTime: Date.now(),
     model_name: _env.getAIConfig.call(void 0, _env.MIDSCENE_MODEL_NAME) || ""
   };
@@ -235,7 +235,7 @@ function emitInsightDump(data, dumpSubscriber) {
 var debug = _logger.getDebug.call(void 0, "ai:insight");
 var Insight = class {
   constructor(context, opt) {
-    this.aiVendorFn = _chunkTV47VKQ7js.callAiFn;
+    this.aiVendorFn = _chunkQFXN2AP7js.callAiFn;
     _utils.assert.call(void 0, context, "context is required for Insight");
     if (typeof context === "function") {
       this.contextRetrieverFn = context;
@@ -278,7 +278,7 @@ var Insight = class {
     let searchAreaUsage = void 0;
     let searchAreaResponse = void 0;
     if (searchAreaPrompt) {
-      searchAreaResponse = await _chunkTV47VKQ7js.AiLocateSection.call(void 0, {
+      searchAreaResponse = await _chunkQFXN2AP7js.AiLocateSection.call(void 0, {
         context,
         sectionDescription: searchAreaPrompt
       });
@@ -291,7 +291,7 @@ var Insight = class {
       searchArea = searchAreaResponse.rect;
     }
     const startTime = Date.now();
-    const { parseResult, rect, elementById, rawResponse, usage } = await _chunkTV47VKQ7js.AiLocateElement.call(void 0, {
+    const { parseResult, rect, elementById, rawResponse, usage } = await _chunkQFXN2AP7js.AiLocateElement.call(void 0, {
       callAI: callAI || this.aiVendorFn,
       context,
       targetElementDescription: queryPrompt,
@@ -370,7 +370,7 @@ ${parseResult.errors.join("\n")}`;
       rect
     };
   }
-  async extract(dataDemand, opt) {
+  async extract(dataDemand, opt, promptImages) {
     _utils.assert.call(void 0, 
       typeof dataDemand === "object" || typeof dataDemand === "string",
       `dataDemand should be object or string, but get ${typeof dataDemand}`
@@ -379,9 +379,10 @@ ${parseResult.errors.join("\n")}`;
     this.onceDumpUpdatedFn = void 0;
     const context = await this.contextRetrieverFn("extract");
     const startTime = Date.now();
-    const { parseResult, usage } = await _chunkTV47VKQ7js.AiExtractElementInfo.call(void 0, {
+    const { parseResult, usage } = await _chunkQFXN2AP7js.AiExtractElementInfo.call(void 0, {
       context,
       dataQuery: dataDemand,
+      promptImages,
       extractOption: opt
     });
     const timeCost = Date.now() - startTime;
@@ -431,7 +432,7 @@ ${parseResult.errors.join("\n")}`;
     this.onceDumpUpdatedFn = void 0;
     const context = await this.contextRetrieverFn("assert");
     const startTime = Date.now();
-    const assertResult = await _chunkTV47VKQ7js.AiAssert.call(void 0, {
+    const assertResult = await _chunkQFXN2AP7js.AiAssert.call(void 0, {
       assertion,
       context
     });
@@ -485,7 +486,7 @@ ${parseResult.errors.join("\n")}`;
       borderThickness: 3
     });
     if (_optionalChain([opt, 'optionalAccess', _22 => _22.deepThink])) {
-      const searchArea = _chunkTV47VKQ7js.expandSearchArea.call(void 0, targetRect, context.size);
+      const searchArea = _chunkQFXN2AP7js.expandSearchArea.call(void 0, targetRect, context.size);
       debug("describe: set searchArea", searchArea);
       imagePayload = await _img.cropByRect.call(void 0, 
         imagePayload,
@@ -508,7 +509,7 @@ ${parseResult.errors.join("\n")}`;
         ]
       }
     ];
-    const callAIFn = this.aiVendorFn || _chunkTV47VKQ7js.callToGetJSONObject;
+    const callAIFn = this.aiVendorFn || _chunkQFXN2AP7js.callToGetJSONObject;
     const res = await callAIFn(msgs, 4 /* DESCRIBE_ELEMENT */);
     const { content } = res;
     _utils.assert.call(void 0, !content.error, `describe failed: ${content.error}`);
@@ -531,6 +532,6 @@ var src_default = Insight;
 
 
 
-exports.AiAssert = _chunkTV47VKQ7js.AiAssert; exports.AiLocateElement = _chunkTV47VKQ7js.AiLocateElement; exports.Executor = Executor; exports.Insight = Insight; exports.MIDSCENE_MODEL_NAME = _env.MIDSCENE_MODEL_NAME; exports.default = src_default; exports.describeUserPage = _chunkTV47VKQ7js.describeUserPage; exports.getAIConfig = _env.getAIConfig; exports.getVersion = _chunk73GOZ45Rjs.getVersion; exports.plan = _chunkTV47VKQ7js.plan;
+exports.AiAssert = _chunkQFXN2AP7js.AiAssert; exports.AiLocateElement = _chunkQFXN2AP7js.AiLocateElement; exports.Executor = Executor; exports.Insight = Insight; exports.MIDSCENE_MODEL_NAME = _env.MIDSCENE_MODEL_NAME; exports.default = src_default; exports.describeUserPage = _chunkQFXN2AP7js.describeUserPage; exports.getAIConfig = _env.getAIConfig; exports.getVersion = _chunkQSWMWTO4js.getVersion; exports.plan = _chunkQFXN2AP7js.plan;
 
 //# sourceMappingURL=index.js.map

package/dist/lib/index.js.map

@@ -1 +1 @@
-{"version":3,"mappings":";;;;;;;;;;;;;;;;AAUA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,cAAc;AAEhB,IAAM,WAAN,MAAe;AAAA,EAUpB,YACE,MACA,SAGA;AACA,SAAK,SACH,SAAS,SAAS,QAAQ,MAAM,SAAS,IAAI,YAAY;AAC3D,SAAK,OAAO;AACZ,SAAK,SAAS,SAAS,SAAS,CAAC,GAAG;AAAA,MAAI,CAAC,SACvC,KAAK,kBAAkB,IAAI;AAAA,IAC7B;AACA,SAAK,cAAc,SAAS;AAAA,EAC9B;AAAA,EAEQ,kBAAkB,MAAyC;AACjE,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,MAAgE;AAC3E;AAAA,MACE,KAAK,WAAW;AAAA,MAChB;AAAA,QAAyD,KAAK,gBAAgB,GAAG,KAAK;AAAA,EAAK,KAAK,gBAAgB,GAAG,UAAU;AAAA,IAC/H;AACA,QAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,WAAK,MAAM,KAAK,GAAG,KAAK,IAAI,CAAC,SAAS,KAAK,kBAAkB,IAAI,CAAC,CAAC;AAAA,IACrE,OAAO;AACL,WAAK,MAAM,KAAK,KAAK,kBAAkB,IAAI,CAAC;AAAA,IAC9C;AACA,QAAI,KAAK,WAAW,WAAW;AAC7B,WAAK,SAAS;AAAA,IAChB;AAAA,EACF;AAAA,EAEA,MAAM,QAAsB;AAC1B,QAAI,KAAK,WAAW,UAAU,KAAK,MAAM,SAAS,GAAG;AACnD,cAAQ;AAAA,QACN;AAAA,MACF;AAAA,IACF;AAEA,WAAO,KAAK,WAAW,WAAW,6BAA6B;AAC/D,WAAO,KAAK,WAAW,aAAa,+BAA+B;AACnE,WAAO,KAAK,WAAW,SAAS,4BAA4B;AAE5D,UAAM,mBAAmB,KAAK,MAAM;AAAA,MAClC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,mBAAmB,GAAG;AAExB;AAAA,IACF;AAEA,SAAK,SAAS;AACd,QAAI,YAAY;AAChB,QAAI,wBAAwB;AAE5B,QAAI;AAEJ,WAAO,YAAY,KAAK,MAAM,QAAQ;AACpC,YAAM,OAAO,KAAK,MAAM,SAAS;AACjC;AAAA,QACE,KAAK,WAAW;AAAA,QAChB,2CAA2C,KAAK,MAAM;AAAA,MACxD;AACA,WAAK,SAAS;AAAA,QACZ,OAAO,KAAK,IAAI;AAAA,MAClB;AACA,UAAI;AACF,aAAK,SAAS;AACd,YAAI;AACF,cAAI,KAAK,aAAa;AACpB,kBAAM,KAAK,YAAY,IAAI;AAAA,UAC7B;AAAA,QACF,SAAS,GAAG;AACV,kBAAQ,MAAM,wBAAwB,CAAC;AAAA,QACzC;AACA;AAAA,UACE,CAAC,WAAW,UAAU,UAAU,EAAE,QAAQ,KAAK,IAAI,KAAK;AAAA,UACxD,0BAA0B,KAAK,IAAI;AAAA,QACrC;AAEA,cAAM,EAAE,UAAU,MAAM,IAAI;AAC5B,eAAO,UAAU,uCAAuC,KAAK,IAAI,EAAE;AAEnE,YAAI;AACJ,cAAM,kBAAmC;AAAA,UACvC;AAAA,UACA,SAAS,oBAAoB;AAAA,QAC/B;AAEA,YAAI,KAAK,SAAS,WAAW;AAC3B;AAAA,YACE,KAAK,YAAY,YACf,KAAK,YAAY,WACjB,KAAK,YAAY,YACjB,KAAK,YAAY,aACjB,KAAK,YAAY,YACjB,KAAK,YAAY;AAAA,YACnB,gCAAgC,KAAK,OAAO;AAAA,UAC9C;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AACxD,cAAI,KAAK,YAAY,UAAU;AAC7B,iCACE,aACC;AAAA,UACL;AAAA,QACF,WAAW,KAAK,SAAS,YAAY,KAAK,SAAS,YAAY;AAC7D,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D,OAAO;AACL,kBAAQ;AAAA,YACN,0BAA0B,KAAK,IAAI;AAAA,UACrC;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D;AAEA,eAAO,OAAO,MAAM,WAAW;AAC/B,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF,SAAS,GAAQ;AACf,gCAAwB;AACxB,aAAK,QACH,GAAG,YAAY,OAAO,MAAM,WAAW,IAAI;AAC7C,aAAK,aAAa,EAAE;AAEpB,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF;AAAA,IACF;AAGA,aAAS,IAAI,YAAY,GAAG,IAAI,KAAK,MAAM,QAAQ,KAAK;AACtD,WAAK,MAAM,CAAC,EAAE,SAAS;AAAA,IACzB;AAEA,QAAI,uBAAuB;AACzB,WAAK,SAAS;AAAA,IAChB,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AAEA,QAAI,KAAK,MAAM,QAAQ;AAErB,YAAM,cAAc,KAAK,IAAI,WAAW,KAAK,MAAM,SAAS,CAAC;AAC7D,aAAO,KAAK,MAAM,WAAW,EAAE;AAAA,IACjC;AAAA,EACF;AAAA,EAEA,iBAA0B;AACxB,WAAO,KAAK,WAAW;AAAA,EACzB;AAAA,EAEA,kBAAwC;AACtC,QAAI,KAAK,WAAW,SAAS;AAC3B,aAAO;AAAA,IACT;AACA,UAAM,iBAAiB,KAAK,MAAM;AAAA,MAChC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,kBAAkB,GAAG;AACvB,aAAO,KAAK,MAAM,cAAc;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AAAA,EAEA,OAAsB;AACpB,QAAI,mBAAmB;AAEvB,QAAI,aAAa,GAAG;AAClB,YAAM,iBAAiB,mBAAmB;AAC1C,UAAI,gBAAgB;AAClB,2BAAmB,WAAW,cAAc;AAAA,MAC9C,OAAO;AACL,2BAAmB,GAAG,aAAa,CAAC;AAAA,MACtC;AAAA,IACF;AACA,UAAM,WAA0B;AAAA,MAC9B,YAAY,WAAW;AAAA,MACvB,YAAY,YAAY,mBAAmB,KAAK;AAAA,MAChD,mBAAmB;AAAA,MACnB,SAAS,KAAK,IAAI;AAAA,MAClB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACF;;;AC7NA,SAAS,4BAA4B;AAE9B,IAAM,8BAA8B,MAAM;AAC/C,SAAO,6EAA6E,qBAAqB,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAgB5G;;;ACYA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA,gBAAAA;AAAA,OACK;AACP,SAAS,yBAAyB,kBAAkB;AACpD,SAAS,gBAAgB;AACzB,SAAS,UAAAC,eAAc;;;AChCvB,SAAS,uBAAAC,sBAAqB,eAAAC,oBAAmB;AACjD,SAAS,YAAY;AAEd,SAAS,gBACd,MACA,gBACA;AACA,QAAM,WAAqB;AAAA,IACzB,YAAY,WAAW;AAAA,IACvB,SAAS,KAAK,IAAI;AAAA,IAClB,YAAYA,aAAYD,oBAAmB,KAAK;AAAA,EAClD;AACA,QAAM,YAAyB;AAAA,IAC7B,OAAO,KAAK;AAAA,IACZ,GAAG;AAAA,IACH,GAAG;AAAA,EACL;AAEA,mBAAiB,SAAS;AAC5B;;;ADyBA,IAAM,QAAQ,SAAS,YAAY;AACnC,IAAqB,UAArB,MAGE;AAAA,EAWA,YACE,SAGA,KACA;AAXF,sBAAoD;AAYlD,IAAAD,QAAO,SAAS,iCAAiC;AACjD,QAAI,OAAO,YAAY,YAAY;AACjC,WAAK,qBAAqB;AAAA,IAC5B,OAAO;AACL,WAAK,qBAAqB,MAAM,QAAQ,QAAQ,OAAO;AAAA,IACzD;AAEA,QAAI,OAAO,KAAK,eAAe,aAAa;AAC1C,WAAK,aAAa,IAAI;AAAA,IACxB;AACA,QAAI,OAAO,KAAK,aAAa,aAAa;AACxC,WAAK,WAAW,IAAI;AAAA,IACtB;AAAA,EACF;AAAA,EAEA,MAAM,OACJ,OACA,KACuB;AACvB,UAAM,EAAE,OAAO,IAAI,OAAO,CAAC;AAC3B,UAAM,cAAc,OAAO,UAAU,WAAW,QAAQ,MAAM;AAC9D,IAAAA,QAAO,aAAa,8BAA8B;AAClD,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,IAAAA,QAAO,OAAO,UAAU,UAAU,sCAAsC;AAExE,UAAM,wBAAwB;AAAA,MAC5B;AAAA,IACF;AACA,QAAI,uBAAuB;AACzB,YAAM,yBAAyB,qBAAqB;AAAA,IACtD;AACA,QAAI;AACJ,QAAI,MAAM,aAAa,uBAAuB;AAC5C,yBAAmB,MAAM;AAAA,IAC3B;AAEA,QAAI,oBAAoB,CAACD,cAAa,GAAG;AACvC,cAAQ;AAAA,QACN;AAAA,MACF;AACA,yBAAmB;AAAA,IACrB;AAEA,UAAM,UAAU,KAAK,WAAY,MAAM,KAAK,mBAAmB,QAAQ;AAEvE,QAAI,aAA+B;AACnC,QAAI,wBAA4C;AAChD,QAAI,kBAA2C;AAC/C,QAAI,qBAEY;AAChB,QAAI,kBAAkB;AACpB,2BAAqB,MAAM,gBAAgB;AAAA,QACzC;AAAA,QACA,oBAAoB;AAAA,MACtB,CAAC;AACD,MAAAC;AAAA,QACE,mBAAmB;AAAA,QACnB,gCAAgC,gBAAgB,IAC9C,mBAAmB,QAAQ,KAAK,mBAAmB,KAAK,KAAK,EAC/D;AAAA,MACF;AACA,8BAAwB,mBAAmB;AAC3C,wBAAkB,mBAAmB;AACrC,mBAAa,mBAAmB;AAAA,IAClC;AAEA,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,aAAa,aAAa,MAAM,IACzD,MAAM,gBAAgB;AAAA,MACpB,QAAQ,UAAU,KAAK;AAAA,MACvB;AAAA,MACA,0BAA0B;AAAA,MAC1B,cAAc;AAAA,IAChB,CAAC;AAEH,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,MACvC,gBAAgB,KAAK,UAAU,WAAW;AAAA,MAC1C;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAgC,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAC1E;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT,SAAS;AAAA,MACX;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,aAAa;AAAA,MACb,MAAM;AAAA,MACN;AAAA,MACA,WAAW,CAAC,CAAC;AAAA,MACb,OAAO;AAAA,IACT;AAEA,UAAM,WAA0B,CAAC;AACjC,KAAC,YAAY,YAAY,CAAC,GAAG,QAAQ,CAAC,SAAS;AAC7C,UAAI,QAAQ,MAAM;AAChB,cAAM,UAAU,YAAY,MAAM,EAAE;AAEpC,YAAI,CAAC,SAAS;AACZ,kBAAQ;AAAA,YACN,kCAAkC,KAAK,EAAE;AAAA,UAC3C;AACA;AAAA,QACF;AACA,iBAAS,KAAK,OAAO;AAAA,MACvB;AAAA,IACF,CAAC;AAED;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,IACF;AAEA,QAAI,UAAU;AACZ,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,IAAAA;AAAA,MACE,SAAS,UAAU;AAAA,MACnB,6CAA6C,SAAS,MAAM;AAAA,IAC9D;AAEA,QAAI,SAAS,WAAW,GAAG;AACzB,aAAO;AAAA,QACL,SAAS;AAAA,UACP,IAAI,SAAS,CAAC,EAAG;AAAA,UACjB,SAAS,SAAS,CAAC,EAAG;AAAA,UACtB,QAAQ,SAAS,CAAC,EAAG;AAAA,UACrB,MAAM,SAAS,CAAC,EAAG;AAAA,UACnB,QAAQ,SAAS,CAAC,EAAG,UAAU,CAAC;AAAA,UAChC,YAAY,SAAS,CAAC,EAAG;AAAA,QAC3B;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,WAAO;AAAA,MACL,SAAS;AAAA,MACT;AAAA,IACF;AAAA,EACF;AAAA,EAYA,MAAM,QACJ,YACA,KACc;AACd,IAAAA;AAAA,MACE,OAAO,eAAe,YAAY,OAAO,eAAe;AAAA,MACxD,kDAAkD,OAAO,UAAU;AAAA,IACrE;AACA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,SAAS;AAEvD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,IAAI,MAAM,qBAAwB;AAAA,MAC3D;AAAA,MACA,WAAW;AAAA,MACX,eAAe;AAAA,IACjB,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,IACzC;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAwB,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAClE;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,OAAO;AAAA,IACT;AAEA,UAAM,EAAE,KAAK,IAAI,eAAe,CAAC;AAGjC;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH;AAAA,MACF;AAAA,MACA;AAAA,IACF;AAEA,QAAI,YAAY,CAAC,MAAM;AACrB,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,WAAO;AAAA,MACL;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,WAAsD;AACjE,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,QAAQ;AACtD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,eAAe,MAAM,SAAS;AAAA,MAClC;AAAA,MACA;AAAA,IACF,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,aAAa,OAAO;AAAA,IAClD;AAEA,UAAM,EAAE,SAAS,KAAK,IAAI,aAAa;AACvC,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,eAAe;AAAA,MACf,kBAAkB;AAAA,MAClB,OAAO,OAAO,SAAY;AAAA,IAC5B;AACA,oBAAgB,UAAU,cAAc;AAExC,WAAO;AAAA,MACL;AAAA,MACA;AAAA,MACA,OAAO,aAAa;AAAA,IACtB;AAAA,EACF;AAAA,EACA,MAAM,SACJ,QACA,KAGyD;AACzD,IAAAA,QAAO,QAAQ,yCAAyC;AACxD,UAAM,UAAU,MAAM,KAAK,mBAAmB,UAAU;AACxD,UAAM,EAAE,kBAAkB,KAAK,IAAI;AACnC,IAAAA,QAAO,kBAAkB,6CAA6C;AAEtE,UAAM,eAAe,4BAA4B;AAGjD,UAAM,kBAAkB;AACxB,UAAM,aAAmB,MAAM,QAAQ,MAAM,IACzC;AAAA,MACE,MAAM,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAChD,KAAK,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAC/C,OAAO;AAAA,MACP,QAAQ;AAAA,IACV,IACA;AAEJ,QAAI,eAAe,MAAM,wBAAwB;AAAA,MAC/C,gBAAgB;AAAA,MAChB;AAAA,MACA,sBAAsB;AAAA,QACpB;AAAA,UACE,MAAM;AAAA,QACR;AAAA,MACF;AAAA,MACA,iBAAiB;AAAA,IACnB,CAAC;AAED,QAAI,KAAK,WAAW;AAClB,YAAM,aAAa,iBAAiB,YAAY,QAAQ,IAAI;AAC5D,YAAM,4BAA4B,UAAU;AAC5C,qBAAe,MAAM;AAAA,QACnB;AAAA,QACA;AAAA,QACA,qBAAqB,oBAAoB;AAAA,MAC3C;AAAA,IACF;AAEA,UAAM,OAAe;AAAA,MACnB,EAAE,MAAM,UAAU,SAAS,aAAa;AAAA,MACxC;AAAA,QACE,MAAM;AAAA,QACN,SAAS;AAAA,UACP;AAAA,YACE,MAAM;AAAA,YACN,WAAW;AAAA,cACT,KAAK;AAAA,cACL,QAAQ;AAAA,YACV;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,WACJ,KAAK,cAAc;AAErB,UAAM,MAAM,MAAM,SAAS,8BAAmC;AAE9D,UAAM,EAAE,QAAQ,IAAI;AACpB,IAAAA,QAAO,CAAC,QAAQ,OAAO,oBAAoB,QAAQ,KAAK,EAAE;AAC1D,IAAAA,QAAO,QAAQ,aAAa,gCAAgC;AAC5D,WAAO;AAAA,EACT;AACF;;;AE1ZA,SAAS,eAAAE,cAAa,uBAAAD,4BAA2B;AAGjD,IAAO,cAAQ","names":["vlLocateMode","assert","MIDSCENE_MODEL_NAME","getAIConfig"],"ignoreList":[],"sources":["../../src/ai-model/action-executor.ts","../../src/ai-model/prompt/describe.ts","../../src/insight/index.ts","../../src/insight/utils.ts","../../src/index.ts"],"sourcesContent":["import type {\n  ExecutionDump,\n  ExecutionTask,\n  ExecutionTaskApply,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskProgressOptions,\n  ExecutionTaskReturn,\n  ExecutorContext,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n  MIDSCENE_MODEL_NAME,\n  getAIConfig,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\n\nexport class Executor {\n  name: string;\n\n  tasks: ExecutionTask[];\n\n  // status of executor\n  status: 'init' | 'pending' | 'running' | 'completed' | 'error';\n\n  onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n\n  constructor(\n    name: string,\n    options?: ExecutionTaskProgressOptions & {\n      tasks?: ExecutionTaskApply[];\n    },\n  ) {\n    this.status =\n      options?.tasks && options.tasks.length > 0 ? 'pending' : 'init';\n    this.name = name;\n    this.tasks = (options?.tasks || []).map((item) =>\n      this.markTaskAsPending(item),\n    );\n    this.onTaskStart = options?.onTaskStart;\n  }\n\n  private markTaskAsPending(task: ExecutionTaskApply): ExecutionTask {\n    return {\n      status: 'pending',\n      ...task,\n    };\n  }\n\n  async append(task: ExecutionTaskApply[] | ExecutionTaskApply): Promise<void> {\n    assert(\n      this.status !== 'error',\n      `executor is in error state, cannot append task\\nerror=${this.latestErrorTask()?.error}\\n${this.latestErrorTask()?.errorStack}`,\n    );\n    if (Array.isArray(task)) {\n      this.tasks.push(...task.map((item) => this.markTaskAsPending(item)));\n    } else {\n      this.tasks.push(this.markTaskAsPending(task));\n    }\n    if (this.status !== 'running') {\n      this.status = 'pending';\n    }\n  }\n\n  async flush(): Promise<any> {\n    if (this.status === 'init' && this.tasks.length > 0) {\n      console.warn(\n        'illegal state for executor, status is init but tasks are not empty',\n      );\n    }\n\n    assert(this.status !== 'running', 'executor is already running');\n    assert(this.status !== 'completed', 'executor is already completed');\n    assert(this.status !== 'error', 'executor is in error state');\n\n    const nextPendingIndex = this.tasks.findIndex(\n      (task) => task.status === 'pending',\n    );\n    if (nextPendingIndex < 0) {\n      // all tasks are completed\n      return;\n    }\n\n    this.status = 'running';\n    let taskIndex = nextPendingIndex;\n    let successfullyCompleted = true;\n\n    let previousFindOutput: ExecutionTaskInsightLocateOutput | undefined;\n\n    while (taskIndex < this.tasks.length) {\n      const task = this.tasks[taskIndex];\n      assert(\n        task.status === 'pending',\n        `task status should be pending, but got: ${task.status}`,\n      );\n      task.timing = {\n        start: Date.now(),\n      };\n      try {\n        task.status = 'running';\n        try {\n          if (this.onTaskStart) {\n            await this.onTaskStart(task);\n          }\n        } catch (e) {\n          console.error('error in onTaskStart', e);\n        }\n        assert(\n          ['Insight', 'Action', 'Planning'].indexOf(task.type) >= 0,\n          `unsupported task type: ${task.type}`,\n        );\n\n        const { executor, param } = task;\n        assert(executor, `executor is required for task type: ${task.type}`);\n\n        let returnValue;\n        const executorContext: ExecutorContext = {\n          task,\n          element: previousFindOutput?.element,\n        };\n\n        if (task.type === 'Insight') {\n          assert(\n            task.subType === 'Locate' ||\n              task.subType === 'Query' ||\n              task.subType === 'Assert' ||\n              task.subType === 'Boolean' ||\n              task.subType === 'Number' ||\n              task.subType === 'String',\n            `unsupported insight subType: ${task.subType}`,\n          );\n          returnValue = await task.executor(param, executorContext);\n          if (task.subType === 'Locate') {\n            previousFindOutput = (\n              returnValue as ExecutionTaskReturn<ExecutionTaskInsightLocateOutput>\n            )?.output;\n          }\n        } else if (task.type === 'Action' || task.type === 'Planning') {\n          returnValue = await task.executor(param, executorContext);\n        } else {\n          console.warn(\n            `unsupported task type: ${task.type}, will try to execute it directly`,\n          );\n          returnValue = await task.executor(param, executorContext);\n        }\n\n        Object.assign(task, returnValue);\n        task.status = 'finished';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        taskIndex++;\n      } catch (e: any) {\n        successfullyCompleted = false;\n        task.error =\n          e?.message || (typeof e === 'string' ? e : 'error-without-message');\n        task.errorStack = e.stack;\n\n        task.status = 'failed';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        break;\n      }\n    }\n\n    // set all remaining tasks as cancelled\n    for (let i = taskIndex + 1; i < this.tasks.length; i++) {\n      this.tasks[i].status = 'cancelled';\n    }\n\n    if (successfullyCompleted) {\n      this.status = 'completed';\n    } else {\n      this.status = 'error';\n    }\n\n    if (this.tasks.length) {\n      // return the last output\n      const outputIndex = Math.min(taskIndex, this.tasks.length - 1);\n      return this.tasks[outputIndex].output;\n    }\n  }\n\n  isInErrorState(): boolean {\n    return this.status === 'error';\n  }\n\n  latestErrorTask(): ExecutionTask | null {\n    if (this.status !== 'error') {\n      return null;\n    }\n    const errorTaskIndex = this.tasks.findIndex(\n      (task) => task.status === 'failed',\n    );\n    if (errorTaskIndex >= 0) {\n      return this.tasks[errorTaskIndex];\n    }\n    return null;\n  }\n\n  dump(): ExecutionDump {\n    let modelDescription = '';\n\n    if (vlLocateMode()) {\n      const uiTarsModelVer = uiTarsModelVersion();\n      if (uiTarsModelVer) {\n        modelDescription = `UI-TARS=${uiTarsModelVer}`;\n      } else {\n        modelDescription = `${vlLocateMode()} mode`;\n      }\n    }\n    const dumpData: ExecutionDump = {\n      sdkVersion: getVersion(),\n      model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n      model_description: modelDescription,\n      logTime: Date.now(),\n      name: this.name,\n      tasks: this.tasks,\n    };\n    return dumpData;\n  }\n}\n","import { getPreferredLanguage } from '@midscene/shared/env';\n\nexport const elementDescriberInstruction = () => {\n  return `Describe the element in the red rectangle for precise identification. Use ${getPreferredLanguage()}.\n\nRules:\n1. Start with element type (button, input, link, etc.)    \n2. Include key identifiers:\n   - Text content: \"with text 'Submit'\"\n   - Visual features: \"blue background\", \"icon only\"\n   - Position: \"top-right\", \"below search bar\"\n3. Keep description under 20 words\n4. Don't mention the red rectangle\n\nReturn JSON:\n{\n  \"description\": \"brief element type with key identifiers\",\n  \"error\"?: \"error message if any\"\n}`;\n};\n","import {\n  AIActionType,\n  type AIArgs,\n  callAiFn,\n  expandSearchArea,\n} from '@/ai-model/common';\nimport {\n  AiExtractElementInfo,\n  AiLocateElement,\n  callToGetJSONObject,\n} from '@/ai-model/index';\nimport { AiAssert, AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n  AIDescribeElementResponse,\n  AIElementResponse,\n  AIUsageInfo,\n  BaseElement,\n  DetailedLocateParam,\n  DumpSubscriber,\n  InsightAction,\n  InsightAssertionResponse,\n  InsightExtractOption,\n  InsightExtractParam,\n  InsightOptions,\n  InsightTaskInfo,\n  LocateResult,\n  PartialInsightDumpFromSDK,\n  Rect,\n  UIContext,\n} from '@/types';\nimport {\n  MIDSCENE_FORCE_DEEP_THINK,\n  MIDSCENE_USE_QWEN_VL,\n  getAIConfigInBoolean,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext<BaseElement>;\n  callAI?: typeof callAiFn<AIElementResponse>;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\nconst debug = getDebug('ai:insight');\nexport default class Insight<\n  ElementType extends BaseElement = BaseElement,\n  ContextType extends UIContext<ElementType> = UIContext<ElementType>,\n> {\n  contextRetrieverFn: (\n    action: InsightAction,\n  ) => Promise<ContextType> | ContextType;\n\n  aiVendorFn: (...args: Array<any>) => Promise<any> = callAiFn;\n\n  onceDumpUpdatedFn?: DumpSubscriber;\n\n  taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n  constructor(\n    context:\n      | ContextType\n      | ((action: InsightAction) => Promise<ContextType> | ContextType),\n    opt?: InsightOptions,\n  ) {\n    assert(context, 'context is required for Insight');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\n\n    if (typeof opt?.aiVendorFn !== 'undefined') {\n      this.aiVendorFn = opt.aiVendorFn;\n    }\n    if (typeof opt?.taskInfo !== 'undefined') {\n      this.taskInfo = opt.taskInfo;\n    }\n  }\n\n  async locate(\n    query: DetailedLocateParam,\n    opt?: LocateOpts,\n  ): Promise<LocateResult> {\n    const { callAI } = opt || {};\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    const globalDeepThinkSwitch = getAIConfigInBoolean(\n      MIDSCENE_FORCE_DEEP_THINK,\n    );\n    if (globalDeepThinkSwitch) {\n      debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n    }\n    let searchAreaPrompt;\n    if (query.deepThink || globalDeepThinkSwitch) {\n      searchAreaPrompt = query.prompt;\n    }\n\n    if (searchAreaPrompt && !vlLocateMode()) {\n      console.warn(\n        'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n      );\n      searchAreaPrompt = undefined;\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn('locate'));\n\n    let searchArea: Rect | undefined = undefined;\n    let searchAreaRawResponse: string | undefined = undefined;\n    let searchAreaUsage: AIUsageInfo | undefined = undefined;\n    let searchAreaResponse:\n      | Awaited<ReturnType<typeof AiLocateSection>>\n      | undefined = undefined;\n    if (searchAreaPrompt) {\n      searchAreaResponse = await AiLocateSection({\n        context,\n        sectionDescription: searchAreaPrompt,\n      });\n      assert(\n        searchAreaResponse.rect,\n        `cannot find search area for \"${searchAreaPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const { parseResult, rect, elementById, rawResponse, usage } =\n      await AiLocateElement({\n        callAI: callAI || this.aiVendorFn,\n        context,\n        targetElementDescription: queryPrompt,\n        searchConfig: searchAreaResponse,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      matchedElement: [],\n      matchedRect: rect,\n      data: null,\n      taskInfo,\n      deepThink: !!searchArea,\n      error: errorLog,\n    };\n\n    const elements: BaseElement[] = [];\n    (parseResult.elements || []).forEach((item) => {\n      if ('id' in item) {\n        const element = elementById(item?.id);\n\n        if (!element) {\n          console.warn(\n            `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\n          );\n          return;\n        }\n        elements.push(element);\n      }\n    });\n\n    emitInsightDump(\n      {\n        ...dumpData,\n        matchedElement: elements,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog) {\n      throw new Error(errorLog);\n    }\n\n    assert(\n      elements.length <= 1,\n      `locate: multiple elements found, length = ${elements.length}`,\n    );\n\n    if (elements.length === 1) {\n      return {\n        element: {\n          id: elements[0]!.id,\n          indexId: elements[0]!.indexId,\n          center: elements[0]!.center,\n          rect: elements[0]!.rect,\n          xpaths: elements[0]!.xpaths || [],\n          attributes: elements[0]!.attributes,\n        },\n        rect,\n      };\n    }\n    return {\n      element: null,\n      rect,\n    };\n  }\n\n  async extract<T = any>(input: string, opt?: InsightExtractOption): Promise<T>;\n  async extract<T extends Record<string, string>>(\n    input: T,\n    opt?: InsightExtractOption,\n  ): Promise<Record<keyof T, any>>;\n  async extract<T extends object>(\n    input: Record<keyof T, string>,\n    opt?: InsightExtractOption,\n  ): Promise<T>;\n\n  async extract<T>(\n    dataDemand: InsightExtractParam,\n    opt?: InsightExtractOption,\n  ): Promise<any> {\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const context = await this.contextRetrieverFn('extract');\n\n    const startTime = Date.now();\n    const { parseResult, usage } = await AiExtractElementInfo<T>({\n      context,\n      dataQuery: dataDemand,\n      extractOption: opt,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(parseResult),\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'extract',\n      userQuery: {\n        dataDemand,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      error: errorLog,\n    };\n\n    const { data } = parseResult || {};\n\n    // 4\n    emitInsightDump(\n      {\n        ...dumpData,\n        data,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog && !data) {\n      throw new Error(errorLog);\n    }\n\n    return {\n      data,\n      usage,\n    };\n  }\n\n  async assert(assertion: string): Promise<InsightAssertionResponse> {\n    if (typeof assertion !== 'string') {\n      throw new Error(\n        'This is the assert method for Midscene, the first argument should be a string. If you want to use the assert method from Node.js, please import it from the Node.js assert module.',\n      );\n    }\n\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const context = await this.contextRetrieverFn('assert');\n    const startTime = Date.now();\n    const assertResult = await AiAssert({\n      assertion,\n      context,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(assertResult.content),\n    };\n\n    const { thought, pass } = assertResult.content;\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'assert',\n      userQuery: {\n        assertion,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      assertionPass: pass,\n      assertionThought: thought,\n      error: pass ? undefined : thought,\n    };\n    emitInsightDump(dumpData, dumpSubscriber);\n\n    return {\n      pass,\n      thought,\n      usage: assertResult.usage,\n    };\n  }\n  async describe(\n    target: Rect | [number, number],\n    opt?: {\n      deepThink?: boolean;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for insight.describe');\n    const context = await this.contextRetrieverFn('describe');\n    const { screenshotBase64, size } = context;\n    assert(screenshotBase64, 'screenshot is required for insight.describe');\n\n    const systemPrompt = elementDescriberInstruction();\n\n    // Convert [x,y] center point to Rect if needed\n    const defaultRectSize = 30;\n    const targetRect: Rect = Array.isArray(target)\n      ? {\n          left: Math.floor(target[0] - defaultRectSize / 2),\n          top: Math.floor(target[1] - defaultRectSize / 2),\n          width: defaultRectSize,\n          height: defaultRectSize,\n        }\n      : target;\n\n    let imagePayload = await compositeElementInfoImg({\n      inputImgBase64: screenshotBase64,\n      size,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepThink) {\n      const searchArea = expandSearchArea(targetRect, context.size);\n      debug('describe: set searchArea', searchArea);\n      imagePayload = await cropByRect(\n        imagePayload,\n        searchArea,\n        getAIConfigInBoolean(MIDSCENE_USE_QWEN_VL),\n      );\n    }\n\n    const msgs: AIArgs = [\n      { role: 'system', content: systemPrompt },\n      {\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: imagePayload,\n              detail: 'high',\n            },\n          },\n        ],\n      },\n    ];\n\n    const callAIFn =\n      this.aiVendorFn || callToGetJSONObject<AIDescribeElementResponse>;\n\n    const res = await callAIFn(msgs, AIActionType.DESCRIBE_ELEMENT);\n\n    const { content } = res;\n    assert(!content.error, `describe failed: ${content.error}`);\n    assert(content.description, 'failed to describe the element');\n    return content;\n  }\n}\n","import type {\n  DumpMeta,\n  DumpSubscriber,\n  InsightDump,\n  PartialInsightDumpFromSDK,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport { MIDSCENE_MODEL_NAME, getAIConfig } from '@midscene/shared/env';\nimport { uuid } from '@midscene/shared/utils';\n\nexport function emitInsightDump(\n  data: PartialInsightDumpFromSDK,\n  dumpSubscriber?: DumpSubscriber,\n) {\n  const baseData: DumpMeta = {\n    sdkVersion: getVersion(),\n    logTime: Date.now(),\n    model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n  };\n  const finalData: InsightDump = {\n    logId: uuid(),\n    ...baseData,\n    ...data,\n  };\n\n  dumpSubscriber?.(finalData);\n}\n","import { Executor } from './ai-model/action-executor';\nimport Insight from './insight/index';\nimport { getVersion } from './utils';\n\nexport {\n  plan,\n  describeUserPage,\n  AiLocateElement,\n  AiAssert,\n} from './ai-model/index';\n\nexport { getAIConfig, MIDSCENE_MODEL_NAME } from '@midscene/shared/env';\n\nexport type * from './types';\nexport default Insight;\nexport { Executor, Insight, getVersion };\n\nexport type {\n  MidsceneYamlScript,\n  MidsceneYamlTask,\n  MidsceneYamlFlowItem,\n  MidsceneYamlFlowItemAIRightClick,\n  MidsceneYamlConfigResult,\n} from './yaml';\n"]}
+{"version":3,"mappings":";;;;;;;;;;;;;;;;AAUA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,cAAc;AAEhB,IAAM,WAAN,MAAe;AAAA,EAUpB,YACE,MACA,SAGA;AACA,SAAK,SACH,SAAS,SAAS,QAAQ,MAAM,SAAS,IAAI,YAAY;AAC3D,SAAK,OAAO;AACZ,SAAK,SAAS,SAAS,SAAS,CAAC,GAAG;AAAA,MAAI,CAAC,SACvC,KAAK,kBAAkB,IAAI;AAAA,IAC7B;AACA,SAAK,cAAc,SAAS;AAAA,EAC9B;AAAA,EAEQ,kBAAkB,MAAyC;AACjE,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,MAAgE;AAC3E;AAAA,MACE,KAAK,WAAW;AAAA,MAChB;AAAA,QAAyD,KAAK,gBAAgB,GAAG,KAAK;AAAA,EAAK,KAAK,gBAAgB,GAAG,UAAU;AAAA,IAC/H;AACA,QAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,WAAK,MAAM,KAAK,GAAG,KAAK,IAAI,CAAC,SAAS,KAAK,kBAAkB,IAAI,CAAC,CAAC;AAAA,IACrE,OAAO;AACL,WAAK,MAAM,KAAK,KAAK,kBAAkB,IAAI,CAAC;AAAA,IAC9C;AACA,QAAI,KAAK,WAAW,WAAW;AAC7B,WAAK,SAAS;AAAA,IAChB;AAAA,EACF;AAAA,EAEA,MAAM,QAAsB;AAC1B,QAAI,KAAK,WAAW,UAAU,KAAK,MAAM,SAAS,GAAG;AACnD,cAAQ;AAAA,QACN;AAAA,MACF;AAAA,IACF;AAEA,WAAO,KAAK,WAAW,WAAW,6BAA6B;AAC/D,WAAO,KAAK,WAAW,aAAa,+BAA+B;AACnE,WAAO,KAAK,WAAW,SAAS,4BAA4B;AAE5D,UAAM,mBAAmB,KAAK,MAAM;AAAA,MAClC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,mBAAmB,GAAG;AAExB;AAAA,IACF;AAEA,SAAK,SAAS;AACd,QAAI,YAAY;AAChB,QAAI,wBAAwB;AAE5B,QAAI;AAEJ,WAAO,YAAY,KAAK,MAAM,QAAQ;AACpC,YAAM,OAAO,KAAK,MAAM,SAAS;AACjC;AAAA,QACE,KAAK,WAAW;AAAA,QAChB,2CAA2C,KAAK,MAAM;AAAA,MACxD;AACA,WAAK,SAAS;AAAA,QACZ,OAAO,KAAK,IAAI;AAAA,MAClB;AACA,UAAI;AACF,aAAK,SAAS;AACd,YAAI;AACF,cAAI,KAAK,aAAa;AACpB,kBAAM,KAAK,YAAY,IAAI;AAAA,UAC7B;AAAA,QACF,SAAS,GAAG;AACV,kBAAQ,MAAM,wBAAwB,CAAC;AAAA,QACzC;AACA;AAAA,UACE,CAAC,WAAW,UAAU,UAAU,EAAE,QAAQ,KAAK,IAAI,KAAK;AAAA,UACxD,0BAA0B,KAAK,IAAI;AAAA,QACrC;AAEA,cAAM,EAAE,UAAU,MAAM,IAAI;AAC5B,eAAO,UAAU,uCAAuC,KAAK,IAAI,EAAE;AAEnE,YAAI;AACJ,cAAM,kBAAmC;AAAA,UACvC;AAAA,UACA,SAAS,oBAAoB;AAAA,QAC/B;AAEA,YAAI,KAAK,SAAS,WAAW;AAC3B;AAAA,YACE,KAAK,YAAY,YACf,KAAK,YAAY,WACjB,KAAK,YAAY,YACjB,KAAK,YAAY,aACjB,KAAK,YAAY,YACjB,KAAK,YAAY;AAAA,YACnB,gCAAgC,KAAK,OAAO;AAAA,UAC9C;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AACxD,cAAI,KAAK,YAAY,UAAU;AAC7B,iCACE,aACC;AAAA,UACL;AAAA,QACF,WAAW,KAAK,SAAS,YAAY,KAAK,SAAS,YAAY;AAC7D,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D,OAAO;AACL,kBAAQ;AAAA,YACN,0BAA0B,KAAK,IAAI;AAAA,UACrC;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D;AAEA,eAAO,OAAO,MAAM,WAAW;AAC/B,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF,SAAS,GAAQ;AACf,gCAAwB;AACxB,aAAK,QACH,GAAG,YAAY,OAAO,MAAM,WAAW,IAAI;AAC7C,aAAK,aAAa,EAAE;AAEpB,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF;AAAA,IACF;AAGA,aAAS,IAAI,YAAY,GAAG,IAAI,KAAK,MAAM,QAAQ,KAAK;AACtD,WAAK,MAAM,CAAC,EAAE,SAAS;AAAA,IACzB;AAEA,QAAI,uBAAuB;AACzB,WAAK,SAAS;AAAA,IAChB,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AAEA,QAAI,KAAK,MAAM,QAAQ;AAErB,YAAM,cAAc,KAAK,IAAI,WAAW,KAAK,MAAM,SAAS,CAAC;AAC7D,aAAO,KAAK,MAAM,WAAW,EAAE;AAAA,IACjC;AAAA,EACF;AAAA,EAEA,iBAA0B;AACxB,WAAO,KAAK,WAAW;AAAA,EACzB;AAAA,EAEA,kBAAwC;AACtC,QAAI,KAAK,WAAW,SAAS;AAC3B,aAAO;AAAA,IACT;AACA,UAAM,iBAAiB,KAAK,MAAM;AAAA,MAChC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,kBAAkB,GAAG;AACvB,aAAO,KAAK,MAAM,cAAc;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AAAA,EAEA,OAAsB;AACpB,QAAI,mBAAmB;AAEvB,QAAI,aAAa,GAAG;AAClB,YAAM,iBAAiB,mBAAmB;AAC1C,UAAI,gBAAgB;AAClB,2BAAmB,WAAW,cAAc;AAAA,MAC9C,OAAO;AACL,2BAAmB,GAAG,aAAa,CAAC;AAAA,MACtC;AAAA,IACF;AACA,UAAM,WAA0B;AAAA,MAC9B,YAAY,WAAW;AAAA,MACvB,YAAY,YAAY,mBAAmB,KAAK;AAAA,MAChD,mBAAmB;AAAA,MACnB,SAAS,KAAK,IAAI;AAAA,MAClB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACF;;;AC7NA,SAAS,4BAA4B;AAE9B,IAAM,8BAA8B,MAAM;AAC/C,SAAO,6EAA6E,qBAAqB,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAgB5G;;;ACYA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA,gBAAAA;AAAA,OACK;AACP,SAAS,yBAAyB,kBAAkB;AACpD,SAAS,gBAAgB;AACzB,SAAS,UAAAC,eAAc;;;AChCvB,SAAS,uBAAAC,sBAAqB,eAAAC,oBAAmB;AACjD,SAAS,YAAY;AAEd,SAAS,gBACd,MACA,gBACA;AACA,QAAM,WAAqB;AAAA,IACzB,YAAY,WAAW;AAAA,IACvB,SAAS,KAAK,IAAI;AAAA,IAClB,YAAYA,aAAYD,oBAAmB,KAAK;AAAA,EAClD;AACA,QAAM,YAAyB;AAAA,IAC7B,OAAO,KAAK;AAAA,IACZ,GAAG;AAAA,IACH,GAAG;AAAA,EACL;AAEA,mBAAiB,SAAS;AAC5B;;;ADyBA,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,EAEA,MAAM,QACJ,YACA,KACA,cACc;AACd,IAAAA;AAAA,MACE,OAAO,eAAe,YAAY,OAAO,eAAe;AAAA,MACxD,kDAAkD,OAAO,UAAU;AAAA,IACrE;AACA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,SAAS;AAEvD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,IAAI,MAAM,qBAAwB;AAAA,MAC3D;AAAA,MACA,WAAW;AAAA,MACX;AAAA,MACA,eAAe;AAAA,IACjB,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,IACzC;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAwB,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAClE;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,OAAO;AAAA,IACT;AAEA,UAAM,EAAE,KAAK,IAAI,eAAe,CAAC;AAGjC;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH;AAAA,MACF;AAAA,MACA;AAAA,IACF;AAEA,QAAI,YAAY,CAAC,MAAM;AACrB,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,WAAO;AAAA,MACL;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,WAAsD;AACjE,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,QAAQ;AACtD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,eAAe,MAAM,SAAS;AAAA,MAClC;AAAA,MACA;AAAA,IACF,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,aAAa,OAAO;AAAA,IAClD;AAEA,UAAM,EAAE,SAAS,KAAK,IAAI,aAAa;AACvC,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,eAAe;AAAA,MACf,kBAAkB;AAAA,MAClB,OAAO,OAAO,SAAY;AAAA,IAC5B;AACA,oBAAgB,UAAU,cAAc;AAExC,WAAO;AAAA,MACL;AAAA,MACA;AAAA,MACA,OAAO,aAAa;AAAA,IACtB;AAAA,EACF;AAAA,EACA,MAAM,SACJ,QACA,KAGyD;AACzD,IAAAA,QAAO,QAAQ,yCAAyC;AACxD,UAAM,UAAU,MAAM,KAAK,mBAAmB,UAAU;AACxD,UAAM,EAAE,kBAAkB,KAAK,IAAI;AACnC,IAAAA,QAAO,kBAAkB,6CAA6C;AAEtE,UAAM,eAAe,4BAA4B;AAGjD,UAAM,kBAAkB;AACxB,UAAM,aAAmB,MAAM,QAAQ,MAAM,IACzC;AAAA,MACE,MAAM,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAChD,KAAK,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAC/C,OAAO;AAAA,MACP,QAAQ;AAAA,IACV,IACA;AAEJ,QAAI,eAAe,MAAM,wBAAwB;AAAA,MAC/C,gBAAgB;AAAA,MAChB;AAAA,MACA,sBAAsB;AAAA,QACpB;AAAA,UACE,MAAM;AAAA,QACR;AAAA,MACF;AAAA,MACA,iBAAiB;AAAA,IACnB,CAAC;AAED,QAAI,KAAK,WAAW;AAClB,YAAM,aAAa,iBAAiB,YAAY,QAAQ,IAAI;AAC5D,YAAM,4BAA4B,UAAU;AAC5C,qBAAe,MAAM;AAAA,QACnB;AAAA,QACA;AAAA,QACA,qBAAqB,oBAAoB;AAAA,MAC3C;AAAA,IACF;AAEA,UAAM,OAAe;AAAA,MACnB,EAAE,MAAM,UAAU,SAAS,aAAa;AAAA,MACxC;AAAA,QACE,MAAM;AAAA,QACN,SAAS;AAAA,UACP;AAAA,YACE,MAAM;AAAA,YACN,WAAW;AAAA,cACT,KAAK;AAAA,cACL,QAAQ;AAAA,YACV;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,WACJ,KAAK,cAAc;AAErB,UAAM,MAAM,MAAM,SAAS,8BAAmC;AAE9D,UAAM,EAAE,QAAQ,IAAI;AACpB,IAAAA,QAAO,CAAC,QAAQ,OAAO,oBAAoB,QAAQ,KAAK,EAAE;AAC1D,IAAAA,QAAO,QAAQ,aAAa,gCAAgC;AAC5D,WAAO;AAAA,EACT;AACF;;;AElZA,SAAS,eAAAE,cAAa,uBAAAD,4BAA2B;AAGjD,IAAO,cAAQ","names":["vlLocateMode","assert","MIDSCENE_MODEL_NAME","getAIConfig"],"ignoreList":[],"sources":["../../src/ai-model/action-executor.ts","../../src/ai-model/prompt/describe.ts","../../src/insight/index.ts","../../src/insight/utils.ts","../../src/index.ts"],"sourcesContent":["import type {\n  ExecutionDump,\n  ExecutionTask,\n  ExecutionTaskApply,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskProgressOptions,\n  ExecutionTaskReturn,\n  ExecutorContext,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n  MIDSCENE_MODEL_NAME,\n  getAIConfig,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\n\nexport class Executor {\n  name: string;\n\n  tasks: ExecutionTask[];\n\n  // status of executor\n  status: 'init' | 'pending' | 'running' | 'completed' | 'error';\n\n  onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n\n  constructor(\n    name: string,\n    options?: ExecutionTaskProgressOptions & {\n      tasks?: ExecutionTaskApply[];\n    },\n  ) {\n    this.status =\n      options?.tasks && options.tasks.length > 0 ? 'pending' : 'init';\n    this.name = name;\n    this.tasks = (options?.tasks || []).map((item) =>\n      this.markTaskAsPending(item),\n    );\n    this.onTaskStart = options?.onTaskStart;\n  }\n\n  private markTaskAsPending(task: ExecutionTaskApply): ExecutionTask {\n    return {\n      status: 'pending',\n      ...task,\n    };\n  }\n\n  async append(task: ExecutionTaskApply[] | ExecutionTaskApply): Promise<void> {\n    assert(\n      this.status !== 'error',\n      `executor is in error state, cannot append task\\nerror=${this.latestErrorTask()?.error}\\n${this.latestErrorTask()?.errorStack}`,\n    );\n    if (Array.isArray(task)) {\n      this.tasks.push(...task.map((item) => this.markTaskAsPending(item)));\n    } else {\n      this.tasks.push(this.markTaskAsPending(task));\n    }\n    if (this.status !== 'running') {\n      this.status = 'pending';\n    }\n  }\n\n  async flush(): Promise<any> {\n    if (this.status === 'init' && this.tasks.length > 0) {\n      console.warn(\n        'illegal state for executor, status is init but tasks are not empty',\n      );\n    }\n\n    assert(this.status !== 'running', 'executor is already running');\n    assert(this.status !== 'completed', 'executor is already completed');\n    assert(this.status !== 'error', 'executor is in error state');\n\n    const nextPendingIndex = this.tasks.findIndex(\n      (task) => task.status === 'pending',\n    );\n    if (nextPendingIndex < 0) {\n      // all tasks are completed\n      return;\n    }\n\n    this.status = 'running';\n    let taskIndex = nextPendingIndex;\n    let successfullyCompleted = true;\n\n    let previousFindOutput: ExecutionTaskInsightLocateOutput | undefined;\n\n    while (taskIndex < this.tasks.length) {\n      const task = this.tasks[taskIndex];\n      assert(\n        task.status === 'pending',\n        `task status should be pending, but got: ${task.status}`,\n      );\n      task.timing = {\n        start: Date.now(),\n      };\n      try {\n        task.status = 'running';\n        try {\n          if (this.onTaskStart) {\n            await this.onTaskStart(task);\n          }\n        } catch (e) {\n          console.error('error in onTaskStart', e);\n        }\n        assert(\n          ['Insight', 'Action', 'Planning'].indexOf(task.type) >= 0,\n          `unsupported task type: ${task.type}`,\n        );\n\n        const { executor, param } = task;\n        assert(executor, `executor is required for task type: ${task.type}`);\n\n        let returnValue;\n        const executorContext: ExecutorContext = {\n          task,\n          element: previousFindOutput?.element,\n        };\n\n        if (task.type === 'Insight') {\n          assert(\n            task.subType === 'Locate' ||\n              task.subType === 'Query' ||\n              task.subType === 'Assert' ||\n              task.subType === 'Boolean' ||\n              task.subType === 'Number' ||\n              task.subType === 'String',\n            `unsupported insight subType: ${task.subType}`,\n          );\n          returnValue = await task.executor(param, executorContext);\n          if (task.subType === 'Locate') {\n            previousFindOutput = (\n              returnValue as ExecutionTaskReturn<ExecutionTaskInsightLocateOutput>\n            )?.output;\n          }\n        } else if (task.type === 'Action' || task.type === 'Planning') {\n          returnValue = await task.executor(param, executorContext);\n        } else {\n          console.warn(\n            `unsupported task type: ${task.type}, will try to execute it directly`,\n          );\n          returnValue = await task.executor(param, executorContext);\n        }\n\n        Object.assign(task, returnValue);\n        task.status = 'finished';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        taskIndex++;\n      } catch (e: any) {\n        successfullyCompleted = false;\n        task.error =\n          e?.message || (typeof e === 'string' ? e : 'error-without-message');\n        task.errorStack = e.stack;\n\n        task.status = 'failed';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        break;\n      }\n    }\n\n    // set all remaining tasks as cancelled\n    for (let i = taskIndex + 1; i < this.tasks.length; i++) {\n      this.tasks[i].status = 'cancelled';\n    }\n\n    if (successfullyCompleted) {\n      this.status = 'completed';\n    } else {\n      this.status = 'error';\n    }\n\n    if (this.tasks.length) {\n      // return the last output\n      const outputIndex = Math.min(taskIndex, this.tasks.length - 1);\n      return this.tasks[outputIndex].output;\n    }\n  }\n\n  isInErrorState(): boolean {\n    return this.status === 'error';\n  }\n\n  latestErrorTask(): ExecutionTask | null {\n    if (this.status !== 'error') {\n      return null;\n    }\n    const errorTaskIndex = this.tasks.findIndex(\n      (task) => task.status === 'failed',\n    );\n    if (errorTaskIndex >= 0) {\n      return this.tasks[errorTaskIndex];\n    }\n    return null;\n  }\n\n  dump(): ExecutionDump {\n    let modelDescription = '';\n\n    if (vlLocateMode()) {\n      const uiTarsModelVer = uiTarsModelVersion();\n      if (uiTarsModelVer) {\n        modelDescription = `UI-TARS=${uiTarsModelVer}`;\n      } else {\n        modelDescription = `${vlLocateMode()} mode`;\n      }\n    }\n    const dumpData: ExecutionDump = {\n      sdkVersion: getVersion(),\n      model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n      model_description: modelDescription,\n      logTime: Date.now(),\n      name: this.name,\n      tasks: this.tasks,\n    };\n    return dumpData;\n  }\n}\n","import { getPreferredLanguage } from '@midscene/shared/env';\n\nexport const elementDescriberInstruction = () => {\n  return `Describe the element in the red rectangle for precise identification. Use ${getPreferredLanguage()}.\n\nRules:\n1. Start with element type (button, input, link, etc.)    \n2. Include key identifiers:\n   - Text content: \"with text 'Submit'\"\n   - Visual features: \"blue background\", \"icon only\"\n   - Position: \"top-right\", \"below search bar\"\n3. Keep description under 20 words\n4. Don't mention the red rectangle\n\nReturn JSON:\n{\n  \"description\": \"brief element type with key identifiers\",\n  \"error\"?: \"error message if any\"\n}`;\n};\n","import {\n  AIActionType,\n  type AIArgs,\n  callAiFn,\n  expandSearchArea,\n} from '@/ai-model/common';\nimport {\n  AiExtractElementInfo,\n  AiLocateElement,\n  callToGetJSONObject,\n} from '@/ai-model/index';\nimport { AiAssert, AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n  AIDescribeElementResponse,\n  AIElementResponse,\n  AIUsageInfo,\n  BaseElement,\n  DetailedLocateParam,\n  DumpSubscriber,\n  InsightAction,\n  InsightAssertionResponse,\n  InsightExtractOption,\n  InsightExtractParam,\n  InsightOptions,\n  InsightTaskInfo,\n  LocateResult,\n  PartialInsightDumpFromSDK,\n  Rect,\n  UIContext,\n} from '@/types';\nimport {\n  MIDSCENE_FORCE_DEEP_THINK,\n  MIDSCENE_USE_QWEN_VL,\n  getAIConfigInBoolean,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext<BaseElement>;\n  callAI?: typeof callAiFn<AIElementResponse>;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\nconst debug = getDebug('ai:insight');\nexport default class Insight<\n  ElementType extends BaseElement = BaseElement,\n  ContextType extends UIContext<ElementType> = UIContext<ElementType>,\n> {\n  contextRetrieverFn: (\n    action: InsightAction,\n  ) => Promise<ContextType> | ContextType;\n\n  aiVendorFn: (...args: Array<any>) => Promise<any> = callAiFn;\n\n  onceDumpUpdatedFn?: DumpSubscriber;\n\n  taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n  constructor(\n    context:\n      | ContextType\n      | ((action: InsightAction) => Promise<ContextType> | ContextType),\n    opt?: InsightOptions,\n  ) {\n    assert(context, 'context is required for Insight');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\n\n    if (typeof opt?.aiVendorFn !== 'undefined') {\n      this.aiVendorFn = opt.aiVendorFn;\n    }\n    if (typeof opt?.taskInfo !== 'undefined') {\n      this.taskInfo = opt.taskInfo;\n    }\n  }\n\n  async locate(\n    query: DetailedLocateParam,\n    opt?: LocateOpts,\n  ): Promise<LocateResult> {\n    const { callAI } = opt || {};\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    const globalDeepThinkSwitch = getAIConfigInBoolean(\n      MIDSCENE_FORCE_DEEP_THINK,\n    );\n    if (globalDeepThinkSwitch) {\n      debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n    }\n    let searchAreaPrompt;\n    if (query.deepThink || globalDeepThinkSwitch) {\n      searchAreaPrompt = query.prompt;\n    }\n\n    if (searchAreaPrompt && !vlLocateMode()) {\n      console.warn(\n        'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n      );\n      searchAreaPrompt = undefined;\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn('locate'));\n\n    let searchArea: Rect | undefined = undefined;\n    let searchAreaRawResponse: string | undefined = undefined;\n    let searchAreaUsage: AIUsageInfo | undefined = undefined;\n    let searchAreaResponse:\n      | Awaited<ReturnType<typeof AiLocateSection>>\n      | undefined = undefined;\n    if (searchAreaPrompt) {\n      searchAreaResponse = await AiLocateSection({\n        context,\n        sectionDescription: searchAreaPrompt,\n      });\n      assert(\n        searchAreaResponse.rect,\n        `cannot find search area for \"${searchAreaPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const { parseResult, rect, elementById, rawResponse, usage } =\n      await AiLocateElement({\n        callAI: callAI || this.aiVendorFn,\n        context,\n        targetElementDescription: queryPrompt,\n        searchConfig: searchAreaResponse,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      matchedElement: [],\n      matchedRect: rect,\n      data: null,\n      taskInfo,\n      deepThink: !!searchArea,\n      error: errorLog,\n    };\n\n    const elements: BaseElement[] = [];\n    (parseResult.elements || []).forEach((item) => {\n      if ('id' in item) {\n        const element = elementById(item?.id);\n\n        if (!element) {\n          console.warn(\n            `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\n          );\n          return;\n        }\n        elements.push(element);\n      }\n    });\n\n    emitInsightDump(\n      {\n        ...dumpData,\n        matchedElement: elements,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog) {\n      throw new Error(errorLog);\n    }\n\n    assert(\n      elements.length <= 1,\n      `locate: multiple elements found, length = ${elements.length}`,\n    );\n\n    if (elements.length === 1) {\n      return {\n        element: {\n          id: elements[0]!.id,\n          indexId: elements[0]!.indexId,\n          center: elements[0]!.center,\n          rect: elements[0]!.rect,\n          xpaths: elements[0]!.xpaths || [],\n          attributes: elements[0]!.attributes,\n        },\n        rect,\n      };\n    }\n    return {\n      element: null,\n      rect,\n    };\n  }\n\n  async extract<T>(\n    dataDemand: InsightExtractParam,\n    opt?: InsightExtractOption,\n    promptImages?: Record<string, string>,\n  ): Promise<any> {\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const context = await this.contextRetrieverFn('extract');\n\n    const startTime = Date.now();\n    const { parseResult, usage } = await AiExtractElementInfo<T>({\n      context,\n      dataQuery: dataDemand,\n      promptImages,\n      extractOption: opt,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(parseResult),\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'extract',\n      userQuery: {\n        dataDemand,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      error: errorLog,\n    };\n\n    const { data } = parseResult || {};\n\n    // 4\n    emitInsightDump(\n      {\n        ...dumpData,\n        data,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog && !data) {\n      throw new Error(errorLog);\n    }\n\n    return {\n      data,\n      usage,\n    };\n  }\n\n  async assert(assertion: string): Promise<InsightAssertionResponse> {\n    if (typeof assertion !== 'string') {\n      throw new Error(\n        'This is the assert method for Midscene, the first argument should be a string. If you want to use the assert method from Node.js, please import it from the Node.js assert module.',\n      );\n    }\n\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const context = await this.contextRetrieverFn('assert');\n    const startTime = Date.now();\n    const assertResult = await AiAssert({\n      assertion,\n      context,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(assertResult.content),\n    };\n\n    const { thought, pass } = assertResult.content;\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'assert',\n      userQuery: {\n        assertion,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      assertionPass: pass,\n      assertionThought: thought,\n      error: pass ? undefined : thought,\n    };\n    emitInsightDump(dumpData, dumpSubscriber);\n\n    return {\n      pass,\n      thought,\n      usage: assertResult.usage,\n    };\n  }\n  async describe(\n    target: Rect | [number, number],\n    opt?: {\n      deepThink?: boolean;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for insight.describe');\n    const context = await this.contextRetrieverFn('describe');\n    const { screenshotBase64, size } = context;\n    assert(screenshotBase64, 'screenshot is required for insight.describe');\n\n    const systemPrompt = elementDescriberInstruction();\n\n    // Convert [x,y] center point to Rect if needed\n    const defaultRectSize = 30;\n    const targetRect: Rect = Array.isArray(target)\n      ? {\n          left: Math.floor(target[0] - defaultRectSize / 2),\n          top: Math.floor(target[1] - defaultRectSize / 2),\n          width: defaultRectSize,\n          height: defaultRectSize,\n        }\n      : target;\n\n    let imagePayload = await compositeElementInfoImg({\n      inputImgBase64: screenshotBase64,\n      size,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepThink) {\n      const searchArea = expandSearchArea(targetRect, context.size);\n      debug('describe: set searchArea', searchArea);\n      imagePayload = await cropByRect(\n        imagePayload,\n        searchArea,\n        getAIConfigInBoolean(MIDSCENE_USE_QWEN_VL),\n      );\n    }\n\n    const msgs: AIArgs = [\n      { role: 'system', content: systemPrompt },\n      {\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: imagePayload,\n              detail: 'high',\n            },\n          },\n        ],\n      },\n    ];\n\n    const callAIFn =\n      this.aiVendorFn || callToGetJSONObject<AIDescribeElementResponse>;\n\n    const res = await callAIFn(msgs, AIActionType.DESCRIBE_ELEMENT);\n\n    const { content } = res;\n    assert(!content.error, `describe failed: ${content.error}`);\n    assert(content.description, 'failed to describe the element');\n    return content;\n  }\n}\n","import type {\n  DumpMeta,\n  DumpSubscriber,\n  InsightDump,\n  PartialInsightDumpFromSDK,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport { MIDSCENE_MODEL_NAME, getAIConfig } from '@midscene/shared/env';\nimport { uuid } from '@midscene/shared/utils';\n\nexport function emitInsightDump(\n  data: PartialInsightDumpFromSDK,\n  dumpSubscriber?: DumpSubscriber,\n) {\n  const baseData: DumpMeta = {\n    sdkVersion: getVersion(),\n    logTime: Date.now(),\n    model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n  };\n  const finalData: InsightDump = {\n    logId: uuid(),\n    ...baseData,\n    ...data,\n  };\n\n  dumpSubscriber?.(finalData);\n}\n","import { Executor } from './ai-model/action-executor';\nimport Insight from './insight/index';\nimport { getVersion } from './utils';\n\nexport {\n  plan,\n  describeUserPage,\n  AiLocateElement,\n  AiAssert,\n} from './ai-model/index';\n\nexport { getAIConfig, MIDSCENE_MODEL_NAME } from '@midscene/shared/env';\n\nexport type * from './types';\nexport default Insight;\nexport { Executor, Insight, getVersion };\n\nexport type {\n  MidsceneYamlScript,\n  MidsceneYamlTask,\n  MidsceneYamlFlowItem,\n  MidsceneYamlFlowItemAIRightClick,\n  MidsceneYamlConfigResult,\n} from './yaml';\n"]}

package/dist/lib/{llm-planning-3407bd42.d.ts → llm-planning-4ac580dc.d.ts}

@@ -1,10 +1,10 @@
-import { m as AIUsageInfo, U as UIContext, aC as ReferenceImage, r as AIElementLocatorResponse, N as ElementById, g as InsightExtractOption, u as AIDataExtractionResponse, w as AIAssertionResponse, aw as PageType, W as PlanningAIResponse } from './types-4d9c9a85.js';
+import { n as AIUsageInfo, U as UIContext, aD as ReferenceImage, s as AIElementLocatorResponse, N as ElementById, h as InsightExtractOption, v as AIDataExtractionResponse, x as AIAssertionResponse, aw as PageType, W as PlanningAIResponse } from './types-27dc17c9.js';
 import { Rect, ElementTreeNode, BaseElement } from '@midscene/shared/types';
 import { ChatCompletionSystemMessageParam, ChatCompletionUserMessageParam } from 'openai/resources';
 
 type AIArgs = [
     ChatCompletionSystemMessageParam,
-    ChatCompletionUserMessageParam
+    ...ChatCompletionUserMessageParam[]
 ];
 declare enum AIActionType {
     ASSERT = 0,
@@ -77,6 +77,7 @@ declare function AiLocateSection(options: {
 }>;
 declare function AiExtractElementInfo<T, ElementType extends BaseElement = BaseElement>(options: {
     dataQuery: string | Record<string, string>;
+    promptImages?: Record<string, string>;
     context: UIContext<ElementType>;
     extractOption?: InsightExtractOption;
 }): Promise<{

package/dist/lib/{types-4d9c9a85.d.ts → types-27dc17c9.d.ts}

@@ -524,5 +524,10 @@ interface StreamingAIResponse {
     /** Whether the response was streamed */
     isStreamed: boolean;
 }
+type TUserPrompt = string | {
+    text: string;
+    /** let use image to inspect elements */
+    images: Record<string, string>;
+};
 
-export { type PlanningActionParamScroll as $, type AIDescribeElementResponse as A, type EnsureObject as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type InsightExtractParam as F, type LocateResultElement as G, type DumpMeta as H, type InsightAction as I, type InsightDump as J, type LiteUISection as K, type LocateResult as L, type MidsceneYamlScript as M, type ElementById as N, type OnTaskStartTip as O, type PartialInsightDumpFromSDK as P, type AgentWaitForOpt as Q, type ReportDumpWithAttributes as R, type AgentAssertOpt as S, type PlanningLocateParam as T, UIContext as U, type PlanningAction as V, type PlanningAIResponse as W, type PlanningActionParamTap as X, type PlanningActionParamHover as Y, type PlanningActionParamRightClick as Z, type PlanningActionParamInputOrKeyPress as _, type ExecutionTaskProgressOptions as a, type ScriptPlayerStatusValue as a$, type PlanningActionParamAssert as a0, type PlanningActionParamSleep as a1, type PlanningActionParamError as a2, type PlanningActionParamWaitFor as a3, type Color as a4, type BaseAgentParserOpt as a5, type PuppeteerParserOpt as a6, type PlaywrightParserOpt as a7, type ExecutionRecorderItem as a8, type ExecutionTaskType as a9, type StreamingAIResponse as aA, type LocateOption as aB, type ReferenceImage as aC, type scrollParam as aD, type MidsceneYamlScriptEnvBase as aE, type MidsceneYamlScriptWebEnv as aF, type MidsceneYamlScriptAndroidEnv as aG, type MidsceneYamlScriptEnv as aH, type MidsceneYamlFlowItemAIAction as aI, type MidsceneYamlFlowItemAIAssert as aJ, type MidsceneYamlFlowItemAIQuery as aK, type MidsceneYamlFlowItemAINumber as aL, type MidsceneYamlFlowItemAIString as aM, type MidsceneYamlFlowItemAIAsk as aN, type MidsceneYamlFlowItemAIBoolean as aO, type MidsceneYamlFlowItemAILocate as aP, type MidsceneYamlFlowItemAIWaitFor as aQ, type MidsceneYamlFlowItemAITap as aR, type MidsceneYamlFlowItemAIHover as aS, type MidsceneYamlFlowItemAIInput as aT, type MidsceneYamlFlowItemAIKeyboardPress as aU, type MidsceneYamlFlowItemAIScroll as aV, type MidsceneYamlFlowItemEvaluateJavaScript as aW, type MidsceneYamlFlowItemSleep as aX, type MidsceneYamlFlowItemLogScreenshot as aY, type FreeFn as aZ, type ScriptPlayerTaskStatus as a_, type ExecutorContext as aa, type ExecutionTaskHitBy as ab, type ExecutionTaskReturn as ac, type ExecutionTaskInsightLocateParam as ad, type ExecutionTaskInsightLocateOutput as ae, type ExecutionTaskInsightDumpLog as af, type ExecutionTaskInsightLocateApply as ag, type ExecutionTaskInsightLocate as ah, type ExecutionTaskInsightQueryParam as ai, type ExecutionTaskInsightQueryOutput as aj, type ExecutionTaskInsightQueryApply as ak, type ExecutionTaskInsightQuery as al, type ExecutionTaskInsightAssertionParam as am, type ExecutionTaskInsightAssertionApply as an, type ExecutionTaskInsightAssertion as ao, type ExecutionTaskActionApply as ap, type ExecutionTaskAction as aq, type ExecutionTaskLogApply as ar, type ExecutionTaskLog as as, type ExecutionTaskPlanningApply as at, type ExecutionTaskPlanning as au, type GroupedActionDump as av, type PageType as aw, type StreamingCodeGenerationOptions as ax, type StreamingCallback as ay, type CodeGenerationChunk as az, type ExecutionTaskApply as b, type MidsceneYamlConfig as b0, type MidsceneYamlConfigOutput as b1, type ExecutionDump as c, type InsightTaskInfo as d, type InsightOptions as e, type DetailedLocateParam as f, type InsightExtractOption as g, type InsightAssertionResponse as h, type MidsceneYamlTask as i, type MidsceneYamlFlowItem as j, type MidsceneYamlFlowItemAIRightClick as k, type MidsceneYamlConfigResult as l, type AIUsageInfo as m, AIResponseFormat as n, type AISingleElementResponseById as o, type AISingleElementResponseByPosition as p, type AISingleElementResponse as q, type AIElementLocatorResponse as r, type AIElementCoordinatesResponse as s, type AIElementResponse as t, type AIDataExtractionResponse as u, type AISectionLocatorResponse as v, type AIAssertionResponse as w, type LocatorValidatorOption as x, type LocateValidatorResult as y, type AgentDescribeElementAtPointResult as z };
+export { type PlanningActionParamScroll as $, type AIDescribeElementResponse as A, type AgentDescribeElementAtPointResult as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type EnsureObject as F, type LocateResultElement as G, type DumpMeta as H, type InsightAction as I, type InsightDump as J, type LiteUISection as K, type LocateResult as L, type MidsceneYamlScript as M, type ElementById as N, type OnTaskStartTip as O, type PartialInsightDumpFromSDK as P, type AgentWaitForOpt as Q, type ReportDumpWithAttributes as R, type AgentAssertOpt as S, type PlanningLocateParam as T, UIContext as U, type PlanningAction as V, type PlanningAIResponse as W, type PlanningActionParamTap as X, type PlanningActionParamHover as Y, type PlanningActionParamRightClick as Z, type PlanningActionParamInputOrKeyPress as _, type ExecutionTaskProgressOptions as a, type ScriptPlayerTaskStatus as a$, type PlanningActionParamAssert as a0, type PlanningActionParamSleep as a1, type PlanningActionParamError as a2, type PlanningActionParamWaitFor as a3, type Color as a4, type BaseAgentParserOpt as a5, type PuppeteerParserOpt as a6, type PlaywrightParserOpt as a7, type ExecutionRecorderItem as a8, type ExecutionTaskType as a9, type StreamingAIResponse as aA, type TUserPrompt as aB, type LocateOption as aC, type ReferenceImage as aD, type scrollParam as aE, type MidsceneYamlScriptEnvBase as aF, type MidsceneYamlScriptWebEnv as aG, type MidsceneYamlScriptAndroidEnv as aH, type MidsceneYamlScriptEnv as aI, type MidsceneYamlFlowItemAIAction as aJ, type MidsceneYamlFlowItemAIAssert as aK, type MidsceneYamlFlowItemAIQuery as aL, type MidsceneYamlFlowItemAINumber as aM, type MidsceneYamlFlowItemAIString as aN, type MidsceneYamlFlowItemAIAsk as aO, type MidsceneYamlFlowItemAIBoolean as aP, type MidsceneYamlFlowItemAILocate as aQ, type MidsceneYamlFlowItemAIWaitFor as aR, type MidsceneYamlFlowItemAITap as aS, type MidsceneYamlFlowItemAIHover as aT, type MidsceneYamlFlowItemAIInput as aU, type MidsceneYamlFlowItemAIKeyboardPress as aV, type MidsceneYamlFlowItemAIScroll as aW, type MidsceneYamlFlowItemEvaluateJavaScript as aX, type MidsceneYamlFlowItemSleep as aY, type MidsceneYamlFlowItemLogScreenshot as aZ, type FreeFn as a_, type ExecutorContext as aa, type ExecutionTaskHitBy as ab, type ExecutionTaskReturn as ac, type ExecutionTaskInsightLocateParam as ad, type ExecutionTaskInsightLocateOutput as ae, type ExecutionTaskInsightDumpLog as af, type ExecutionTaskInsightLocateApply as ag, type ExecutionTaskInsightLocate as ah, type ExecutionTaskInsightQueryParam as ai, type ExecutionTaskInsightQueryOutput as aj, type ExecutionTaskInsightQueryApply as ak, type ExecutionTaskInsightQuery as al, type ExecutionTaskInsightAssertionParam as am, type ExecutionTaskInsightAssertionApply as an, type ExecutionTaskInsightAssertion as ao, type ExecutionTaskActionApply as ap, type ExecutionTaskAction as aq, type ExecutionTaskLogApply as ar, type ExecutionTaskLog as as, type ExecutionTaskPlanningApply as at, type ExecutionTaskPlanning as au, type GroupedActionDump as av, type PageType as aw, type StreamingCodeGenerationOptions as ax, type StreamingCallback as ay, type CodeGenerationChunk as az, type ExecutionTaskApply as b, type ScriptPlayerStatusValue as b0, type MidsceneYamlConfig as b1, type MidsceneYamlConfigOutput as b2, type ExecutionDump as c, type InsightTaskInfo as d, type InsightOptions as e, type DetailedLocateParam as f, type InsightExtractParam as g, type InsightExtractOption as h, type InsightAssertionResponse as i, type MidsceneYamlTask as j, type MidsceneYamlFlowItem as k, type MidsceneYamlFlowItemAIRightClick as l, type MidsceneYamlConfigResult as m, type AIUsageInfo as n, AIResponseFormat as o, type AISingleElementResponseById as p, type AISingleElementResponseByPosition as q, type AISingleElementResponse as r, type AIElementLocatorResponse as s, type AIElementCoordinatesResponse as t, type AIElementResponse as u, type AIDataExtractionResponse as v, type AISectionLocatorResponse as w, type AIAssertionResponse as x, type LocatorValidatorOption as y, type LocateValidatorResult as z };

package/dist/lib/utils.d.ts

@@ -1,4 +1,4 @@
-import { R as ReportDumpWithAttributes } from './types-4d9c9a85.js';
+import { R as ReportDumpWithAttributes } from './types-27dc17c9.js';
 import { Rect } from '@midscene/shared/types';
 import '@midscene/shared/constants';
 import 'openai/resources';

package/dist/lib/utils.js

@@ -12,7 +12,7 @@
 
 
 
-var _chunk73GOZ45Rjs = require('./chunk-73GOZ45R.js');
+var _chunkQSWMWTO4js = require('./chunk-QSWMWTO4.js');
 
 
 
@@ -27,4 +27,4 @@ var _chunk73GOZ45Rjs = require('./chunk-73GOZ45R.js');
 
 
 
-exports.getTmpDir = _chunk73GOZ45Rjs.getTmpDir; exports.getTmpFile = _chunk73GOZ45Rjs.getTmpFile; exports.getVersion = _chunk73GOZ45Rjs.getVersion; exports.groupedActionDumpFileExt = _chunk73GOZ45Rjs.groupedActionDumpFileExt; exports.insertScriptBeforeClosingHtml = _chunk73GOZ45Rjs.insertScriptBeforeClosingHtml; exports.overlapped = _chunk73GOZ45Rjs.overlapped; exports.replacerForPageObject = _chunk73GOZ45Rjs.replacerForPageObject; exports.reportHTMLContent = _chunk73GOZ45Rjs.reportHTMLContent; exports.sleep = _chunk73GOZ45Rjs.sleep; exports.stringifyDumpData = _chunk73GOZ45Rjs.stringifyDumpData; exports.uploadTestInfoToServer = _chunk73GOZ45Rjs.uploadTestInfoToServer; exports.writeDumpReport = _chunk73GOZ45Rjs.writeDumpReport; exports.writeLogFile = _chunk73GOZ45Rjs.writeLogFile;
+exports.getTmpDir = _chunkQSWMWTO4js.getTmpDir; exports.getTmpFile = _chunkQSWMWTO4js.getTmpFile; exports.getVersion = _chunkQSWMWTO4js.getVersion; exports.groupedActionDumpFileExt = _chunkQSWMWTO4js.groupedActionDumpFileExt; exports.insertScriptBeforeClosingHtml = _chunkQSWMWTO4js.insertScriptBeforeClosingHtml; exports.overlapped = _chunkQSWMWTO4js.overlapped; exports.replacerForPageObject = _chunkQSWMWTO4js.replacerForPageObject; exports.reportHTMLContent = _chunkQSWMWTO4js.reportHTMLContent; exports.sleep = _chunkQSWMWTO4js.sleep; exports.stringifyDumpData = _chunkQSWMWTO4js.stringifyDumpData; exports.uploadTestInfoToServer = _chunkQSWMWTO4js.uploadTestInfoToServer; exports.writeDumpReport = _chunkQSWMWTO4js.writeDumpReport; exports.writeLogFile = _chunkQSWMWTO4js.writeLogFile;

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

@@ -1,9 +1,9 @@
-import { ay as StreamingCallback, m as AIUsageInfo, ax as StreamingCodeGenerationOptions, aA as StreamingAIResponse, V as PlanningAction, j as MidsceneYamlFlowItem } from './types-4d9c9a85.js';
+import { ay as StreamingCallback, n as AIUsageInfo, ax as StreamingCodeGenerationOptions, aA as StreamingAIResponse, V as PlanningAction, k as MidsceneYamlFlowItem } from './types-27dc17c9.js';
 import OpenAI from 'openai';
 import { ChatCompletionMessageParam } from 'openai/resources';
 export { ChatCompletionMessageParam } from 'openai/resources';
-import { b as AIActionType } from './llm-planning-3407bd42.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-3407bd42.js';
+import { b as AIActionType } from './llm-planning-4ac580dc.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-4ac580dc.js';
 import { vlLocateMode } from '@midscene/shared/env';
 import { actionParser } from '@ui-tars/action-parser';
 import { Size } from '@midscene/shared/types';

package/dist/types/index.d.ts

@@ -1,7 +1,7 @@
-import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightExtractOption, h as InsightAssertionResponse, A as AIDescribeElementResponse } from './types-4d9c9a85.js';
-export { w as AIAssertionResponse, u as AIDataExtractionResponse, s as AIElementCoordinatesResponse, r as AIElementLocatorResponse, t as AIElementResponse, n as AIResponseFormat, v as AISectionLocatorResponse, q as AISingleElementResponse, o as AISingleElementResponseById, p as AISingleElementResponseByPosition, m as AIUsageInfo, S as AgentAssertOpt, z as AgentDescribeElementAtPointResult, Q as AgentWaitForOpt, a5 as BaseAgentParserOpt, C as CallAIFn, az as CodeGenerationChunk, a4 as Color, H as DumpMeta, N as ElementById, B as EnsureObject, a8 as ExecutionRecorderItem, aq as ExecutionTaskAction, ap as ExecutionTaskActionApply, ab as ExecutionTaskHitBy, ao as ExecutionTaskInsightAssertion, an as ExecutionTaskInsightAssertionApply, am as ExecutionTaskInsightAssertionParam, af as ExecutionTaskInsightDumpLog, ah as ExecutionTaskInsightLocate, ag as ExecutionTaskInsightLocateApply, ae as ExecutionTaskInsightLocateOutput, ad as ExecutionTaskInsightLocateParam, al as ExecutionTaskInsightQuery, ak as ExecutionTaskInsightQueryApply, aj as ExecutionTaskInsightQueryOutput, ai as ExecutionTaskInsightQueryParam, as as ExecutionTaskLog, ar as ExecutionTaskLogApply, au as ExecutionTaskPlanning, at as ExecutionTaskPlanningApply, ac as ExecutionTaskReturn, a9 as ExecutionTaskType, aa as ExecutorContext, aZ as FreeFn, av as GroupedActionDump, J as InsightDump, F as InsightExtractParam, K as LiteUISection, aB as LocateOption, G as LocateResultElement, y as LocateValidatorResult, x as LocatorValidatorOption, b0 as MidsceneYamlConfig, b1 as MidsceneYamlConfigOutput, l as MidsceneYamlConfigResult, j as MidsceneYamlFlowItem, aI as MidsceneYamlFlowItemAIAction, aN as MidsceneYamlFlowItemAIAsk, aJ as MidsceneYamlFlowItemAIAssert, aO as MidsceneYamlFlowItemAIBoolean, aS as MidsceneYamlFlowItemAIHover, aT as MidsceneYamlFlowItemAIInput, aU as MidsceneYamlFlowItemAIKeyboardPress, aP as MidsceneYamlFlowItemAILocate, aL as MidsceneYamlFlowItemAINumber, aK as MidsceneYamlFlowItemAIQuery, k as MidsceneYamlFlowItemAIRightClick, aV as MidsceneYamlFlowItemAIScroll, aM as MidsceneYamlFlowItemAIString, aR as MidsceneYamlFlowItemAITap, aQ as MidsceneYamlFlowItemAIWaitFor, aW as MidsceneYamlFlowItemEvaluateJavaScript, aY as MidsceneYamlFlowItemLogScreenshot, aX as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aG as MidsceneYamlScriptAndroidEnv, aH as MidsceneYamlScriptEnv, aE as MidsceneYamlScriptEnvBase, aF as MidsceneYamlScriptWebEnv, i as MidsceneYamlTask, O as OnTaskStartTip, aw as PageType, P as PartialInsightDumpFromSDK, W as PlanningAIResponse, V as PlanningAction, a0 as PlanningActionParamAssert, a2 as PlanningActionParamError, Y as PlanningActionParamHover, _ as PlanningActionParamInputOrKeyPress, Z as PlanningActionParamRightClick, $ as PlanningActionParamScroll, a1 as PlanningActionParamSleep, X as PlanningActionParamTap, a3 as PlanningActionParamWaitFor, T as PlanningLocateParam, a7 as PlaywrightParserOpt, a6 as PuppeteerParserOpt, aC as ReferenceImage, R as ReportDumpWithAttributes, a$ as ScriptPlayerStatusValue, a_ as ScriptPlayerTaskStatus, aA as StreamingAIResponse, ay as StreamingCallback, ax as StreamingCodeGenerationOptions, aD as scrollParam } from './types-4d9c9a85.js';
-import { c as callAiFn } from './llm-planning-3407bd42.js';
-export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-3407bd42.js';
+import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightExtractParam, h as InsightExtractOption, i as InsightAssertionResponse, A as AIDescribeElementResponse } from './types-27dc17c9.js';
+export { x as AIAssertionResponse, v as AIDataExtractionResponse, t as AIElementCoordinatesResponse, s as AIElementLocatorResponse, u as AIElementResponse, o as AIResponseFormat, w as AISectionLocatorResponse, r as AISingleElementResponse, p as AISingleElementResponseById, q as AISingleElementResponseByPosition, n as AIUsageInfo, S as AgentAssertOpt, B as AgentDescribeElementAtPointResult, Q as AgentWaitForOpt, a5 as BaseAgentParserOpt, C as CallAIFn, az as CodeGenerationChunk, a4 as Color, H as DumpMeta, N as ElementById, F as EnsureObject, a8 as ExecutionRecorderItem, aq as ExecutionTaskAction, ap as ExecutionTaskActionApply, ab as ExecutionTaskHitBy, ao as ExecutionTaskInsightAssertion, an as ExecutionTaskInsightAssertionApply, am as ExecutionTaskInsightAssertionParam, af as ExecutionTaskInsightDumpLog, ah as ExecutionTaskInsightLocate, ag as ExecutionTaskInsightLocateApply, ae as ExecutionTaskInsightLocateOutput, ad as ExecutionTaskInsightLocateParam, al as ExecutionTaskInsightQuery, ak as ExecutionTaskInsightQueryApply, aj as ExecutionTaskInsightQueryOutput, ai as ExecutionTaskInsightQueryParam, as as ExecutionTaskLog, ar as ExecutionTaskLogApply, au as ExecutionTaskPlanning, at as ExecutionTaskPlanningApply, ac as ExecutionTaskReturn, a9 as ExecutionTaskType, aa as ExecutorContext, a_ as FreeFn, av as GroupedActionDump, J as InsightDump, K as LiteUISection, aC as LocateOption, G as LocateResultElement, z as LocateValidatorResult, y as LocatorValidatorOption, b1 as MidsceneYamlConfig, b2 as MidsceneYamlConfigOutput, m as MidsceneYamlConfigResult, k as MidsceneYamlFlowItem, aJ as MidsceneYamlFlowItemAIAction, aO as MidsceneYamlFlowItemAIAsk, aK as MidsceneYamlFlowItemAIAssert, aP as MidsceneYamlFlowItemAIBoolean, aT as MidsceneYamlFlowItemAIHover, aU as MidsceneYamlFlowItemAIInput, aV as MidsceneYamlFlowItemAIKeyboardPress, aQ as MidsceneYamlFlowItemAILocate, aM as MidsceneYamlFlowItemAINumber, aL as MidsceneYamlFlowItemAIQuery, l as MidsceneYamlFlowItemAIRightClick, aW as MidsceneYamlFlowItemAIScroll, aN as MidsceneYamlFlowItemAIString, aS as MidsceneYamlFlowItemAITap, aR as MidsceneYamlFlowItemAIWaitFor, aX as MidsceneYamlFlowItemEvaluateJavaScript, aZ as MidsceneYamlFlowItemLogScreenshot, aY as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aH as MidsceneYamlScriptAndroidEnv, aI as MidsceneYamlScriptEnv, aF as MidsceneYamlScriptEnvBase, aG as MidsceneYamlScriptWebEnv, j as MidsceneYamlTask, O as OnTaskStartTip, aw as PageType, P as PartialInsightDumpFromSDK, W as PlanningAIResponse, V as PlanningAction, a0 as PlanningActionParamAssert, a2 as PlanningActionParamError, Y as PlanningActionParamHover, _ as PlanningActionParamInputOrKeyPress, Z as PlanningActionParamRightClick, $ as PlanningActionParamScroll, a1 as PlanningActionParamSleep, X as PlanningActionParamTap, a3 as PlanningActionParamWaitFor, T as PlanningLocateParam, a7 as PlaywrightParserOpt, a6 as PuppeteerParserOpt, aD as ReferenceImage, R as ReportDumpWithAttributes, b0 as ScriptPlayerStatusValue, a$ as ScriptPlayerTaskStatus, aA as StreamingAIResponse, ay as StreamingCallback, ax as StreamingCodeGenerationOptions, aB as TUserPrompt, aE as scrollParam } from './types-27dc17c9.js';
+import { c as callAiFn } from './llm-planning-4ac580dc.js';
+export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-4ac580dc.js';
 import { BaseElement, Rect } from '@midscene/shared/types';
 export { BaseElement, ElementTreeNode, Point, Rect, Size } from '@midscene/shared/types';
 export { getVersion } from './utils.js';
@@ -36,13 +36,11 @@ declare class Insight<ElementType extends BaseElement = BaseElement, ContextType
     taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;
     constructor(context: ContextType | ((action: InsightAction) => Promise<ContextType> | ContextType), opt?: InsightOptions);
     locate(query: DetailedLocateParam, opt?: LocateOpts): Promise<LocateResult>;
-    extract<T = any>(input: string, opt?: InsightExtractOption): Promise<T>;
-    extract<T extends Record<string, string>>(input: T, opt?: InsightExtractOption): Promise<Record<keyof T, any>>;
-    extract<T extends object>(input: Record<keyof T, string>, opt?: InsightExtractOption): Promise<T>;
+    extract<T>(dataDemand: InsightExtractParam, opt?: InsightExtractOption, promptImages?: Record<string, string>): Promise<any>;
     assert(assertion: string): Promise<InsightAssertionResponse>;
     describe(target: Rect | [number, number], opt?: {
         deepThink?: boolean;
     }): Promise<Pick<AIDescribeElementResponse, 'description'>>;
 }
 
-export { AIDescribeElementResponse, DetailedLocateParam, DumpSubscriber, ExecutionDump, ExecutionTask, ExecutionTaskApply, ExecutionTaskProgressOptions, Executor, Insight, InsightAction, InsightAssertionResponse, InsightExtractOption, InsightOptions, InsightTaskInfo, LocateResult, UIContext, Insight as default };
+export { AIDescribeElementResponse, DetailedLocateParam, DumpSubscriber, ExecutionDump, ExecutionTask, ExecutionTaskApply, ExecutionTaskProgressOptions, Executor, Insight, InsightAction, InsightAssertionResponse, InsightExtractOption, InsightExtractParam, InsightOptions, InsightTaskInfo, LocateResult, UIContext, Insight as default };

package/dist/types/{llm-planning-3407bd42.d.ts → llm-planning-4ac580dc.d.ts}

@@ -1,10 +1,10 @@
-import { m as AIUsageInfo, U as UIContext, aC as ReferenceImage, r as AIElementLocatorResponse, N as ElementById, g as InsightExtractOption, u as AIDataExtractionResponse, w as AIAssertionResponse, aw as PageType, W as PlanningAIResponse } from './types-4d9c9a85.js';
+import { n as AIUsageInfo, U as UIContext, aD as ReferenceImage, s as AIElementLocatorResponse, N as ElementById, h as InsightExtractOption, v as AIDataExtractionResponse, x as AIAssertionResponse, aw as PageType, W as PlanningAIResponse } from './types-27dc17c9.js';
 import { Rect, ElementTreeNode, BaseElement } from '@midscene/shared/types';
 import { ChatCompletionSystemMessageParam, ChatCompletionUserMessageParam } from 'openai/resources';
 
 type AIArgs = [
     ChatCompletionSystemMessageParam,
-    ChatCompletionUserMessageParam
+    ...ChatCompletionUserMessageParam[]
 ];
 declare enum AIActionType {
     ASSERT = 0,
@@ -77,6 +77,7 @@ declare function AiLocateSection(options: {
 }>;
 declare function AiExtractElementInfo<T, ElementType extends BaseElement = BaseElement>(options: {
     dataQuery: string | Record<string, string>;
+    promptImages?: Record<string, string>;
     context: UIContext<ElementType>;
     extractOption?: InsightExtractOption;
 }): Promise<{

package/dist/types/{types-4d9c9a85.d.ts → types-27dc17c9.d.ts}

@@ -524,5 +524,10 @@ interface StreamingAIResponse {
     /** Whether the response was streamed */
     isStreamed: boolean;
 }
+type TUserPrompt = string | {
+    text: string;
+    /** let use image to inspect elements */
+    images: Record<string, string>;
+};
 
-export { type PlanningActionParamScroll as $, type AIDescribeElementResponse as A, type EnsureObject as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type InsightExtractParam as F, type LocateResultElement as G, type DumpMeta as H, type InsightAction as I, type InsightDump as J, type LiteUISection as K, type LocateResult as L, type MidsceneYamlScript as M, type ElementById as N, type OnTaskStartTip as O, type PartialInsightDumpFromSDK as P, type AgentWaitForOpt as Q, type ReportDumpWithAttributes as R, type AgentAssertOpt as S, type PlanningLocateParam as T, UIContext as U, type PlanningAction as V, type PlanningAIResponse as W, type PlanningActionParamTap as X, type PlanningActionParamHover as Y, type PlanningActionParamRightClick as Z, type PlanningActionParamInputOrKeyPress as _, type ExecutionTaskProgressOptions as a, type ScriptPlayerStatusValue as a$, type PlanningActionParamAssert as a0, type PlanningActionParamSleep as a1, type PlanningActionParamError as a2, type PlanningActionParamWaitFor as a3, type Color as a4, type BaseAgentParserOpt as a5, type PuppeteerParserOpt as a6, type PlaywrightParserOpt as a7, type ExecutionRecorderItem as a8, type ExecutionTaskType as a9, type StreamingAIResponse as aA, type LocateOption as aB, type ReferenceImage as aC, type scrollParam as aD, type MidsceneYamlScriptEnvBase as aE, type MidsceneYamlScriptWebEnv as aF, type MidsceneYamlScriptAndroidEnv as aG, type MidsceneYamlScriptEnv as aH, type MidsceneYamlFlowItemAIAction as aI, type MidsceneYamlFlowItemAIAssert as aJ, type MidsceneYamlFlowItemAIQuery as aK, type MidsceneYamlFlowItemAINumber as aL, type MidsceneYamlFlowItemAIString as aM, type MidsceneYamlFlowItemAIAsk as aN, type MidsceneYamlFlowItemAIBoolean as aO, type MidsceneYamlFlowItemAILocate as aP, type MidsceneYamlFlowItemAIWaitFor as aQ, type MidsceneYamlFlowItemAITap as aR, type MidsceneYamlFlowItemAIHover as aS, type MidsceneYamlFlowItemAIInput as aT, type MidsceneYamlFlowItemAIKeyboardPress as aU, type MidsceneYamlFlowItemAIScroll as aV, type MidsceneYamlFlowItemEvaluateJavaScript as aW, type MidsceneYamlFlowItemSleep as aX, type MidsceneYamlFlowItemLogScreenshot as aY, type FreeFn as aZ, type ScriptPlayerTaskStatus as a_, type ExecutorContext as aa, type ExecutionTaskHitBy as ab, type ExecutionTaskReturn as ac, type ExecutionTaskInsightLocateParam as ad, type ExecutionTaskInsightLocateOutput as ae, type ExecutionTaskInsightDumpLog as af, type ExecutionTaskInsightLocateApply as ag, type ExecutionTaskInsightLocate as ah, type ExecutionTaskInsightQueryParam as ai, type ExecutionTaskInsightQueryOutput as aj, type ExecutionTaskInsightQueryApply as ak, type ExecutionTaskInsightQuery as al, type ExecutionTaskInsightAssertionParam as am, type ExecutionTaskInsightAssertionApply as an, type ExecutionTaskInsightAssertion as ao, type ExecutionTaskActionApply as ap, type ExecutionTaskAction as aq, type ExecutionTaskLogApply as ar, type ExecutionTaskLog as as, type ExecutionTaskPlanningApply as at, type ExecutionTaskPlanning as au, type GroupedActionDump as av, type PageType as aw, type StreamingCodeGenerationOptions as ax, type StreamingCallback as ay, type CodeGenerationChunk as az, type ExecutionTaskApply as b, type MidsceneYamlConfig as b0, type MidsceneYamlConfigOutput as b1, type ExecutionDump as c, type InsightTaskInfo as d, type InsightOptions as e, type DetailedLocateParam as f, type InsightExtractOption as g, type InsightAssertionResponse as h, type MidsceneYamlTask as i, type MidsceneYamlFlowItem as j, type MidsceneYamlFlowItemAIRightClick as k, type MidsceneYamlConfigResult as l, type AIUsageInfo as m, AIResponseFormat as n, type AISingleElementResponseById as o, type AISingleElementResponseByPosition as p, type AISingleElementResponse as q, type AIElementLocatorResponse as r, type AIElementCoordinatesResponse as s, type AIElementResponse as t, type AIDataExtractionResponse as u, type AISectionLocatorResponse as v, type AIAssertionResponse as w, type LocatorValidatorOption as x, type LocateValidatorResult as y, type AgentDescribeElementAtPointResult as z };
+export { type PlanningActionParamScroll as $, type AIDescribeElementResponse as A, type AgentDescribeElementAtPointResult as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type EnsureObject as F, type LocateResultElement as G, type DumpMeta as H, type InsightAction as I, type InsightDump as J, type LiteUISection as K, type LocateResult as L, type MidsceneYamlScript as M, type ElementById as N, type OnTaskStartTip as O, type PartialInsightDumpFromSDK as P, type AgentWaitForOpt as Q, type ReportDumpWithAttributes as R, type AgentAssertOpt as S, type PlanningLocateParam as T, UIContext as U, type PlanningAction as V, type PlanningAIResponse as W, type PlanningActionParamTap as X, type PlanningActionParamHover as Y, type PlanningActionParamRightClick as Z, type PlanningActionParamInputOrKeyPress as _, type ExecutionTaskProgressOptions as a, type ScriptPlayerTaskStatus as a$, type PlanningActionParamAssert as a0, type PlanningActionParamSleep as a1, type PlanningActionParamError as a2, type PlanningActionParamWaitFor as a3, type Color as a4, type BaseAgentParserOpt as a5, type PuppeteerParserOpt as a6, type PlaywrightParserOpt as a7, type ExecutionRecorderItem as a8, type ExecutionTaskType as a9, type StreamingAIResponse as aA, type TUserPrompt as aB, type LocateOption as aC, type ReferenceImage as aD, type scrollParam as aE, type MidsceneYamlScriptEnvBase as aF, type MidsceneYamlScriptWebEnv as aG, type MidsceneYamlScriptAndroidEnv as aH, type MidsceneYamlScriptEnv as aI, type MidsceneYamlFlowItemAIAction as aJ, type MidsceneYamlFlowItemAIAssert as aK, type MidsceneYamlFlowItemAIQuery as aL, type MidsceneYamlFlowItemAINumber as aM, type MidsceneYamlFlowItemAIString as aN, type MidsceneYamlFlowItemAIAsk as aO, type MidsceneYamlFlowItemAIBoolean as aP, type MidsceneYamlFlowItemAILocate as aQ, type MidsceneYamlFlowItemAIWaitFor as aR, type MidsceneYamlFlowItemAITap as aS, type MidsceneYamlFlowItemAIHover as aT, type MidsceneYamlFlowItemAIInput as aU, type MidsceneYamlFlowItemAIKeyboardPress as aV, type MidsceneYamlFlowItemAIScroll as aW, type MidsceneYamlFlowItemEvaluateJavaScript as aX, type MidsceneYamlFlowItemSleep as aY, type MidsceneYamlFlowItemLogScreenshot as aZ, type FreeFn as a_, type ExecutorContext as aa, type ExecutionTaskHitBy as ab, type ExecutionTaskReturn as ac, type ExecutionTaskInsightLocateParam as ad, type ExecutionTaskInsightLocateOutput as ae, type ExecutionTaskInsightDumpLog as af, type ExecutionTaskInsightLocateApply as ag, type ExecutionTaskInsightLocate as ah, type ExecutionTaskInsightQueryParam as ai, type ExecutionTaskInsightQueryOutput as aj, type ExecutionTaskInsightQueryApply as ak, type ExecutionTaskInsightQuery as al, type ExecutionTaskInsightAssertionParam as am, type ExecutionTaskInsightAssertionApply as an, type ExecutionTaskInsightAssertion as ao, type ExecutionTaskActionApply as ap, type ExecutionTaskAction as aq, type ExecutionTaskLogApply as ar, type ExecutionTaskLog as as, type ExecutionTaskPlanningApply as at, type ExecutionTaskPlanning as au, type GroupedActionDump as av, type PageType as aw, type StreamingCodeGenerationOptions as ax, type StreamingCallback as ay, type CodeGenerationChunk as az, type ExecutionTaskApply as b, type ScriptPlayerStatusValue as b0, type MidsceneYamlConfig as b1, type MidsceneYamlConfigOutput as b2, type ExecutionDump as c, type InsightTaskInfo as d, type InsightOptions as e, type DetailedLocateParam as f, type InsightExtractParam as g, type InsightExtractOption as h, type InsightAssertionResponse as i, type MidsceneYamlTask as j, type MidsceneYamlFlowItem as k, type MidsceneYamlFlowItemAIRightClick as l, type MidsceneYamlConfigResult as m, type AIUsageInfo as n, AIResponseFormat as o, type AISingleElementResponseById as p, type AISingleElementResponseByPosition as q, type AISingleElementResponse as r, type AIElementLocatorResponse as s, type AIElementCoordinatesResponse as t, type AIElementResponse as u, type AIDataExtractionResponse as v, type AISectionLocatorResponse as w, type AIAssertionResponse as x, type LocatorValidatorOption as y, type LocateValidatorResult as z };

package/dist/types/utils.d.ts

@@ -1,4 +1,4 @@
-import { R as ReportDumpWithAttributes } from './types-4d9c9a85.js';
+import { R as ReportDumpWithAttributes } from './types-27dc17c9.js';
 import { Rect } from '@midscene/shared/types';
 import '@midscene/shared/constants';
 import 'openai/resources';

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@midscene/core",
   "description": "Automate browser actions, extract data, and perform assertions using AI. It offers JavaScript SDK, Chrome extension, and support for scripting in YAML. See https://midscenejs.com/ for details.",
-  "version": "0.23.5-beta-20250728060950.0",
+  "version": "0.23.5-beta-20250728070606.0",
   "repository": "https://github.com/web-infra-dev/midscene",
   "homepage": "https://midscenejs.com/",
   "jsnext:source": "./src/index.ts",
@@ -44,8 +44,8 @@
     "langsmith": "0.3.7",
     "openai": "4.81.0",
     "socks-proxy-agent": "8.0.4",
-    "@midscene/recorder": "0.23.5-beta-20250728060950.0",
-    "@midscene/shared": "0.23.5-beta-20250728060950.0"
+    "@midscene/recorder": "0.23.5-beta-20250728070606.0",
+    "@midscene/shared": "0.23.5-beta-20250728070606.0"
   },
   "devDependencies": {
     "@modern-js/module-tools": "2.60.6",
@@ -65,7 +65,7 @@
   "scripts": {
     "dev": "npm run build:watch",
     "build": "modern build",
-    "build:watch": "modern build -w",
+    "build:watch": "modern build -w --no-clear",
     "new": "modern new",
     "upgrade": "modern upgrade",
     "test": "vitest --run",