@midscene/core 0.24.2-beta-20250801024655.0 → 0.24.2-beta-20250805024613.0

package/dist/es/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-d836fa73.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, a7 as BaseAgentParserOpt, C as CallAIFn, aB as CodeGenerationChunk, a6 as Color, H as DumpMeta, N as ElementById, B as EnsureObject, aa as ExecutionRecorderItem, as as ExecutionTaskAction, ar as ExecutionTaskActionApply, ad as ExecutionTaskHitBy, aq as ExecutionTaskInsightAssertion, ap as ExecutionTaskInsightAssertionApply, ao as ExecutionTaskInsightAssertionParam, ah as ExecutionTaskInsightDumpLog, aj as ExecutionTaskInsightLocate, ai as ExecutionTaskInsightLocateApply, ag as ExecutionTaskInsightLocateOutput, af as ExecutionTaskInsightLocateParam, an as ExecutionTaskInsightQuery, am as ExecutionTaskInsightQueryApply, al as ExecutionTaskInsightQueryOutput, ak as ExecutionTaskInsightQueryParam, au as ExecutionTaskLog, at as ExecutionTaskLogApply, aw as ExecutionTaskPlanning, av as ExecutionTaskPlanningApply, ae as ExecutionTaskReturn, ab as ExecutionTaskType, ac as ExecutorContext, a$ as FreeFn, ax as GroupedActionDump, J as InsightDump, F as InsightExtractParam, K as LiteUISection, aD as LocateOption, G as LocateResultElement, y as LocateValidatorResult, x as LocatorValidatorOption, b2 as MidsceneYamlConfig, b3 as MidsceneYamlConfigOutput, l as MidsceneYamlConfigResult, j as MidsceneYamlFlowItem, aK as MidsceneYamlFlowItemAIAction, aP as MidsceneYamlFlowItemAIAsk, aL as MidsceneYamlFlowItemAIAssert, aQ as MidsceneYamlFlowItemAIBoolean, aU as MidsceneYamlFlowItemAIHover, aV as MidsceneYamlFlowItemAIInput, aW as MidsceneYamlFlowItemAIKeyboardPress, aR as MidsceneYamlFlowItemAILocate, aN as MidsceneYamlFlowItemAINumber, aM as MidsceneYamlFlowItemAIQuery, k as MidsceneYamlFlowItemAIRightClick, aX as MidsceneYamlFlowItemAIScroll, aO as MidsceneYamlFlowItemAIString, aT as MidsceneYamlFlowItemAITap, aS as MidsceneYamlFlowItemAIWaitFor, aY as MidsceneYamlFlowItemEvaluateJavaScript, a_ as MidsceneYamlFlowItemLogScreenshot, aZ as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aI as MidsceneYamlScriptAndroidEnv, aJ as MidsceneYamlScriptEnv, aG as MidsceneYamlScriptEnvBase, aH as MidsceneYamlScriptWebEnv, i as MidsceneYamlTask, O as OnTaskStartTip, ay as PageType, P as PartialInsightDumpFromSDK, W as PlanningAIResponse, V as PlanningAction, a4 as PlanningActionParamAndroidLongPress, a5 as PlanningActionParamAndroidPull, 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, a9 as PlaywrightParserOpt, a8 as PuppeteerParserOpt, aE as ReferenceImage, R as ReportDumpWithAttributes, b1 as ScriptPlayerStatusValue, b0 as ScriptPlayerTaskStatus, aC as StreamingAIResponse, aA as StreamingCallback, az as StreamingCodeGenerationOptions, aF as scrollParam } from './types-d836fa73.js';
-import { c as callAiFn } from './llm-planning-d7096b0d.js';
-export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-d7096b0d.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, T as TMultimodalPrompt, i as TUserPrompt, j as InsightAssertionResponse, A as AIDescribeElementResponse } from './types-512d3687.js';
+export { y as AIAssertionResponse, w as AIDataExtractionResponse, u as AIElementCoordinatesResponse, t as AIElementLocatorResponse, v as AIElementResponse, p as AIResponseFormat, x as AISectionLocatorResponse, s as AISingleElementResponse, q as AISingleElementResponseById, r as AISingleElementResponseByPosition, o as AIUsageInfo, V as AgentAssertOpt, C as AgentDescribeElementAtPointResult, S as AgentWaitForOpt, a9 as BaseAgentParserOpt, F as CallAIFn, aD as CodeGenerationChunk, a8 as Color, J as DumpMeta, O as ElementById, G as EnsureObject, ac as ExecutionRecorderItem, au as ExecutionTaskAction, at as ExecutionTaskActionApply, af as ExecutionTaskHitBy, as as ExecutionTaskInsightAssertion, ar as ExecutionTaskInsightAssertionApply, aq as ExecutionTaskInsightAssertionParam, aj as ExecutionTaskInsightDumpLog, al as ExecutionTaskInsightLocate, ak as ExecutionTaskInsightLocateApply, ai as ExecutionTaskInsightLocateOutput, ah as ExecutionTaskInsightLocateParam, ap as ExecutionTaskInsightQuery, ao as ExecutionTaskInsightQueryApply, an as ExecutionTaskInsightQueryOutput, am as ExecutionTaskInsightQueryParam, aw as ExecutionTaskLog, av as ExecutionTaskLogApply, ay as ExecutionTaskPlanning, ax as ExecutionTaskPlanningApply, ag as ExecutionTaskReturn, ad as ExecutionTaskType, ae as ExecutorContext, b1 as FreeFn, az as GroupedActionDump, K as InsightDump, N as LiteUISection, aF as LocateOption, H as LocateResultElement, B as LocateValidatorResult, z as LocatorValidatorOption, b4 as MidsceneYamlConfig, b5 as MidsceneYamlConfigOutput, n as MidsceneYamlConfigResult, l as MidsceneYamlFlowItem, aM as MidsceneYamlFlowItemAIAction, aR as MidsceneYamlFlowItemAIAsk, aN as MidsceneYamlFlowItemAIAssert, aS as MidsceneYamlFlowItemAIBoolean, aW as MidsceneYamlFlowItemAIHover, aX as MidsceneYamlFlowItemAIInput, aY as MidsceneYamlFlowItemAIKeyboardPress, aT as MidsceneYamlFlowItemAILocate, aP as MidsceneYamlFlowItemAINumber, aO as MidsceneYamlFlowItemAIQuery, m as MidsceneYamlFlowItemAIRightClick, aZ as MidsceneYamlFlowItemAIScroll, aQ as MidsceneYamlFlowItemAIString, aV as MidsceneYamlFlowItemAITap, aU as MidsceneYamlFlowItemAIWaitFor, a_ as MidsceneYamlFlowItemEvaluateJavaScript, b0 as MidsceneYamlFlowItemLogScreenshot, a$ as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aK as MidsceneYamlScriptAndroidEnv, aL as MidsceneYamlScriptEnv, aI as MidsceneYamlScriptEnvBase, aJ as MidsceneYamlScriptWebEnv, k as MidsceneYamlTask, Q as OnTaskStartTip, aA as PageType, P as PartialInsightDumpFromSDK, Y as PlanningAIResponse, X as PlanningAction, a6 as PlanningActionParamAndroidLongPress, a7 as PlanningActionParamAndroidPull, a2 as PlanningActionParamAssert, a4 as PlanningActionParamError, _ as PlanningActionParamHover, a0 as PlanningActionParamInputOrKeyPress, $ as PlanningActionParamRightClick, a1 as PlanningActionParamScroll, a3 as PlanningActionParamSleep, Z as PlanningActionParamTap, a5 as PlanningActionParamWaitFor, W as PlanningLocateParam, ab as PlaywrightParserOpt, aa as PuppeteerParserOpt, aG as ReferenceImage, R as ReportDumpWithAttributes, b3 as ScriptPlayerStatusValue, b2 as ScriptPlayerTaskStatus, aE as StreamingAIResponse, aC as StreamingCallback, aB as StreamingCodeGenerationOptions, aH as scrollParam } from './types-512d3687.js';
+import { c as callAiFn } from './llm-planning-877248da.js';
+export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-877248da.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>;
-    assert(assertion: string): Promise<InsightAssertionResponse>;
+    extract<T>(dataDemand: InsightExtractParam, opt?: InsightExtractOption, multimodalPrompt?: TMultimodalPrompt): Promise<any>;
+    assert(assertion: TUserPrompt): 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, TMultimodalPrompt, TUserPrompt, UIContext, Insight as default };

package/dist/es/index.js

@@ -1,6 +1,6 @@
 import {
   getVersion
-} from "./chunk-FMBJ3CM2.js";
+} from "./chunk-NZFWY3M5.js";
 import {
   AiAssert,
   AiExtractElementInfo,
@@ -11,7 +11,7 @@ import {
   describeUserPage,
   expandSearchArea,
   plan
-} from "./chunk-5HH6E7M4.js";
+} from "./chunk-JS4CT3XV.js";
 
 // src/ai-model/action-executor.ts
 import {
@@ -186,20 +186,44 @@ ${this.latestErrorTask()?.errorStack}`
 // src/ai-model/prompt/describe.ts
 import { getPreferredLanguage } from "@midscene/shared/env";
 var elementDescriberInstruction = () => {
-  return `Describe the element in the red rectangle for precise identification. Use ${getPreferredLanguage()}.
+  return `
+Describe the element in the red rectangle for precise identification. Use ${getPreferredLanguage()}.
 
-Rules:
-1. Start with element type (button, input, link, etc.)    
-2. Include key identifiers:
-   - Text content: "with text 'Submit'"
-   - Visual features: "blue background", "icon only"
-   - Position: "top-right", "below search bar"
-3. Keep description under 20 words
-4. Don't mention the red rectangle
+CRITICAL REQUIREMENTS:
+1. UNIQUENESS: The description must uniquely identify this element on the current page
+2. UNIVERSALITY: Use generic, reusable selectors that work across different contexts
+3. PRECISION: Be specific enough to distinguish from similar elements
+
+DESCRIPTION STRUCTURE:
+1. Element type (button, input, link, div, etc.)
+2. Primary identifier (in order of preference):
+   - Unique text content: "with text 'Login'"
+   - Unique attribute: "with aria-label 'Search'"
+   - Unique class/ID: "with class 'primary-button'"
+   - Unique position: "in header navigation"
+3. Secondary identifiers (if needed for uniqueness):
+   - Visual features: "blue background", "with icon"
+   - Relative position: "below search bar", "in sidebar"
+   - Parent context: "in login form", "in main menu"
+
+GUIDELINES:
+- Keep description under 25 words
+- Prioritize semantic identifiers over visual ones
+- Use consistent terminology across similar elements
+- Avoid page-specific or temporary content
+- Don't mention the red rectangle or selection box
+- Focus on stable, reusable characteristics
+
+EXAMPLES:
+- "Login button with text 'Sign In'"
+- "Search input with placeholder 'Enter keywords'"
+- "Navigation link with text 'Home' in header"
+- "Submit button in contact form"
+- "Menu icon with aria-label 'Open menu'"
 
 Return JSON:
 {
-  "description": "brief element type with key identifiers",
+  "description": "unique element identifier",
   "error"?: "error message if any"
 }`;
 };
@@ -379,7 +403,7 @@ ${parseResult.errors.join("\n")}`;
       rect
     };
   }
-  async extract(dataDemand, opt) {
+  async extract(dataDemand, opt, multimodalPrompt) {
     assert2(
       typeof dataDemand === "object" || typeof dataDemand === "string",
       `dataDemand should be object or string, but get ${typeof dataDemand}`
@@ -391,6 +415,7 @@ ${parseResult.errors.join("\n")}`;
     const { parseResult, usage } = await AiExtractElementInfo({
       context,
       dataQuery: dataDemand,
+      multimodalPrompt,
       extractOption: opt
     });
     const timeCost = Date.now() - startTime;
@@ -431,11 +456,6 @@ ${parseResult.errors.join("\n")}`;
     };
   }
   async assert(assertion) {
-    if (typeof assertion !== "string") {
-      throw new Error(
-        "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."
-      );
-    }
     const dumpSubscriber = this.onceDumpUpdatedFn;
     this.onceDumpUpdatedFn = void 0;
     const context = await this.contextRetrieverFn("assert");

package/dist/es/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,QAAQ;AACb,aAAK,eACH,GAAG,YAAY,OAAO,MAAM,WAAW,IAAI;AAC7C,aAAK,aAAa,EAAE;AAEpB,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF;AAAA,IACF;AAGA,aAAS,IAAI,YAAY,GAAG,IAAI,KAAK,MAAM,QAAQ,KAAK;AACtD,WAAK,MAAM,CAAC,EAAE,SAAS;AAAA,IACzB;AAEA,QAAI,uBAAuB;AACzB,WAAK,SAAS;AAAA,IAChB,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AAEA,QAAI,KAAK,MAAM,QAAQ;AAErB,YAAM,cAAc,KAAK,IAAI,WAAW,KAAK,MAAM,SAAS,CAAC;AAC7D,aAAO,KAAK,MAAM,WAAW,EAAE;AAAA,IACjC;AAAA,EACF;AAAA,EAEA,iBAA0B;AACxB,WAAO,KAAK,WAAW;AAAA,EACzB;AAAA,EAEA,kBAAwC;AACtC,QAAI,KAAK,WAAW,SAAS;AAC3B,aAAO;AAAA,IACT;AACA,UAAM,iBAAiB,KAAK,MAAM;AAAA,MAChC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,kBAAkB,GAAG;AACvB,aAAO,KAAK,MAAM,cAAc;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AAAA,EAEA,OAAsB;AACpB,QAAI,mBAAmB;AAEvB,QAAI,aAAa,GAAG;AAClB,YAAM,iBAAiB,mBAAmB;AAC1C,UAAI,gBAAgB;AAClB,2BAAmB,WAAW,cAAc;AAAA,MAC9C,OAAO;AACL,2BAAmB,GAAG,aAAa,CAAC;AAAA,MACtC;AAAA,IACF;AACA,UAAM,WAA0B;AAAA,MAC9B,YAAY,WAAW;AAAA,MACvB,YAAY,YAAY,mBAAmB,KAAK;AAAA,MAChD,mBAAmB;AAAA,MACnB,SAAS,KAAK,IAAI;AAAA,MAClB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACF;;;AC9NA,SAAS,4BAA4B;AAE9B,IAAM,8BAA8B,MAAM;AAC/C,SAAO,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;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,IAAI,MAAM,gBAAgB;AAAA,MACxB,QAAQ,UAAU,KAAK;AAAA,MACvB;AAAA,MACA,0BAA0B;AAAA,MAC1B,cAAc;AAAA,IAChB,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,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,UACzB;AAAA,QACF;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;;;AEjaA,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: ${task.type}, will try to execute it directly`,\n          );\n          returnValue = await task.executor(param, executorContext);\n        }\n\n        Object.assign(task, returnValue);\n        task.status = 'finished';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        taskIndex++;\n      } catch (e: any) {\n        successfullyCompleted = false;\n        task.error = e;\n        task.errorMessage =\n          e?.message || (typeof e === 'string' ? e : 'error-without-message');\n        task.errorStack = e.stack;\n\n        task.status = 'failed';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        break;\n      }\n    }\n\n    // set all remaining tasks as cancelled\n    for (let i = taskIndex + 1; i < this.tasks.length; i++) {\n      this.tasks[i].status = 'cancelled';\n    }\n\n    if (successfullyCompleted) {\n      this.status = 'completed';\n    } else {\n      this.status = 'error';\n    }\n\n    if (this.tasks.length) {\n      // return the last output\n      const outputIndex = Math.min(taskIndex, this.tasks.length - 1);\n      return this.tasks[outputIndex].output;\n    }\n  }\n\n  isInErrorState(): boolean {\n    return this.status === 'error';\n  }\n\n  latestErrorTask(): ExecutionTask | null {\n    if (this.status !== 'error') {\n      return null;\n    }\n    const errorTaskIndex = this.tasks.findIndex(\n      (task) => task.status === 'failed',\n    );\n    if (errorTaskIndex >= 0) {\n      return this.tasks[errorTaskIndex];\n    }\n    return null;\n  }\n\n  dump(): ExecutionDump {\n    let modelDescription = '';\n\n    if (vlLocateMode()) {\n      const uiTarsModelVer = uiTarsModelVersion();\n      if (uiTarsModelVer) {\n        modelDescription = `UI-TARS=${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 ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const {\n      parseResult,\n      rect,\n      elementById,\n      rawResponse,\n      usage,\n      isOrderSensitive,\n    } = await AiLocateElement({\n      callAI: callAI || this.aiVendorFn,\n      context,\n      targetElementDescription: queryPrompt,\n      searchConfig: searchAreaResponse,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI model failed to locate: \\n${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          isOrderSensitive,\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,QAAQ;AACb,aAAK,eACH,GAAG,YAAY,OAAO,MAAM,WAAW,IAAI;AAC7C,aAAK,aAAa,EAAE;AAEpB,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF;AAAA,IACF;AAGA,aAAS,IAAI,YAAY,GAAG,IAAI,KAAK,MAAM,QAAQ,KAAK;AACtD,WAAK,MAAM,CAAC,EAAE,SAAS;AAAA,IACzB;AAEA,QAAI,uBAAuB;AACzB,WAAK,SAAS;AAAA,IAChB,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AAEA,QAAI,KAAK,MAAM,QAAQ;AAErB,YAAM,cAAc,KAAK,IAAI,WAAW,KAAK,MAAM,SAAS,CAAC;AAC7D,aAAO,KAAK,MAAM,WAAW,EAAE;AAAA,IACjC;AAAA,EACF;AAAA,EAEA,iBAA0B;AACxB,WAAO,KAAK,WAAW;AAAA,EACzB;AAAA,EAEA,kBAAwC;AACtC,QAAI,KAAK,WAAW,SAAS;AAC3B,aAAO;AAAA,IACT;AACA,UAAM,iBAAiB,KAAK,MAAM;AAAA,MAChC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,kBAAkB,GAAG;AACvB,aAAO,KAAK,MAAM,cAAc;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AAAA,EAEA,OAAsB;AACpB,QAAI,mBAAmB;AAEvB,QAAI,aAAa,GAAG;AAClB,YAAM,iBAAiB,mBAAmB;AAC1C,UAAI,gBAAgB;AAClB,2BAAmB,WAAW,cAAc;AAAA,MAC9C,OAAO;AACL,2BAAmB,GAAG,aAAa,CAAC;AAAA,MACtC;AAAA,IACF;AACA,UAAM,WAA0B;AAAA,MAC9B,YAAY,WAAW;AAAA,MACvB,YAAY,YAAY,mBAAmB,KAAK;AAAA,MAChD,mBAAmB;AAAA,MACnB,SAAS,KAAK,IAAI;AAAA,MAClB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACF;;;AC9NA,SAAS,4BAA4B;AAE9B,IAAM,8BAA8B,MAAM;AAC/C,SAAO;AAAA,4EACmE,qBAAqB,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAuClG;;;ACVA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA,gBAAAA;AAAA,OACK;AACP,SAAS,yBAAyB,kBAAkB;AACpD,SAAS,gBAAgB;AACzB,SAAS,UAAAC,eAAc;;;AClCvB,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;;;AD2BA,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;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,IAAI,MAAM,gBAAgB;AAAA,MACxB,QAAQ,UAAU,KAAK;AAAA,MACvB;AAAA,MACA,0BAA0B;AAAA,MAC1B,cAAc;AAAA,IAChB,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,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,UACzB;AAAA,QACF;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,WAAO;AAAA,MACL,SAAS;AAAA,MACT;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,QACJ,YACA,KACA,kBACc;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,WAA2D;AACtE,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;;;AErZA,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: ${task.type}, will try to execute it directly`,\n          );\n          returnValue = await task.executor(param, executorContext);\n        }\n\n        Object.assign(task, returnValue);\n        task.status = 'finished';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        taskIndex++;\n      } catch (e: any) {\n        successfullyCompleted = false;\n        task.error = e;\n        task.errorMessage =\n          e?.message || (typeof e === 'string' ? e : 'error-without-message');\n        task.errorStack = e.stack;\n\n        task.status = 'failed';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        break;\n      }\n    }\n\n    // set all remaining tasks as cancelled\n    for (let i = taskIndex + 1; i < this.tasks.length; i++) {\n      this.tasks[i].status = 'cancelled';\n    }\n\n    if (successfullyCompleted) {\n      this.status = 'completed';\n    } else {\n      this.status = 'error';\n    }\n\n    if (this.tasks.length) {\n      // return the last output\n      const outputIndex = Math.min(taskIndex, this.tasks.length - 1);\n      return this.tasks[outputIndex].output;\n    }\n  }\n\n  isInErrorState(): boolean {\n    return this.status === 'error';\n  }\n\n  latestErrorTask(): ExecutionTask | null {\n    if (this.status !== 'error') {\n      return null;\n    }\n    const errorTaskIndex = this.tasks.findIndex(\n      (task) => task.status === 'failed',\n    );\n    if (errorTaskIndex >= 0) {\n      return this.tasks[errorTaskIndex];\n    }\n    return null;\n  }\n\n  dump(): ExecutionDump {\n    let modelDescription = '';\n\n    if (vlLocateMode()) {\n      const uiTarsModelVer = uiTarsModelVersion();\n      if (uiTarsModelVer) {\n        modelDescription = `UI-TARS=${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 `\nDescribe the element in the red rectangle for precise identification. Use ${getPreferredLanguage()}.\n\nCRITICAL REQUIREMENTS:\n1. UNIQUENESS: The description must uniquely identify this element on the current page\n2. UNIVERSALITY: Use generic, reusable selectors that work across different contexts\n3. PRECISION: Be specific enough to distinguish from similar elements\n\nDESCRIPTION STRUCTURE:\n1. Element type (button, input, link, div, etc.)\n2. Primary identifier (in order of preference):\n   - Unique text content: \"with text 'Login'\"\n   - Unique attribute: \"with aria-label 'Search'\"\n   - Unique class/ID: \"with class 'primary-button'\"\n   - Unique position: \"in header navigation\"\n3. Secondary identifiers (if needed for uniqueness):\n   - Visual features: \"blue background\", \"with icon\"\n   - Relative position: \"below search bar\", \"in sidebar\"\n   - Parent context: \"in login form\", \"in main menu\"\n\nGUIDELINES:\n- Keep description under 25 words\n- Prioritize semantic identifiers over visual ones\n- Use consistent terminology across similar elements\n- Avoid page-specific or temporary content\n- Don't mention the red rectangle or selection box\n- Focus on stable, reusable characteristics\n\nEXAMPLES:\n- \"Login button with text 'Sign In'\"\n- \"Search input with placeholder 'Enter keywords'\"\n- \"Navigation link with text 'Home' in header\"\n- \"Submit button in contact form\"\n- \"Menu icon with aria-label 'Open menu'\"\n\nReturn JSON:\n{\n  \"description\": \"unique element identifier\",\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  TMultimodalPrompt,\n  TUserPrompt,\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 ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const {\n      parseResult,\n      rect,\n      elementById,\n      rawResponse,\n      usage,\n      isOrderSensitive,\n    } = await AiLocateElement({\n      callAI: callAI || this.aiVendorFn,\n      context,\n      targetElementDescription: queryPrompt,\n      searchConfig: searchAreaResponse,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI model failed to locate: \\n${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          isOrderSensitive,\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    multimodalPrompt?: TMultimodalPrompt,\n  ): Promise<any> {\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const context = await this.contextRetrieverFn('extract');\n\n    const startTime = Date.now();\n    const { parseResult, usage } = await AiExtractElementInfo<T>({\n      context,\n      dataQuery: dataDemand,\n      multimodalPrompt,\n      extractOption: opt,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(parseResult),\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${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: TUserPrompt): Promise<InsightAssertionResponse> {\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/es/{llm-planning-d7096b0d.d.ts → llm-planning-877248da.d.ts}

@@ -1,10 +1,10 @@
-import { m as AIUsageInfo, U as UIContext, aE as ReferenceImage, r as AIElementLocatorResponse, N as ElementById, g as InsightExtractOption, u as AIDataExtractionResponse, w as AIAssertionResponse, ay as PageType, W as PlanningAIResponse } from './types-d836fa73.js';
+import { o as AIUsageInfo, U as UIContext, i as TUserPrompt, aG as ReferenceImage, t as AIElementLocatorResponse, O as ElementById, T as TMultimodalPrompt, h as InsightExtractOption, w as AIDataExtractionResponse, y as AIAssertionResponse, aA as PageType, Y as PlanningAIResponse } from './types-512d3687.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,
@@ -53,7 +53,7 @@ declare function describeUserPage<ElementType extends BaseElement = BaseElement>
 
 declare function AiLocateElement<ElementType extends BaseElement = BaseElement>(options: {
     context: UIContext<ElementType>;
-    targetElementDescription: string;
+    targetElementDescription: TUserPrompt;
     referenceImage?: ReferenceImage;
     callAI?: typeof callAiFn<AIElementResponse | [number, number]>;
     searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;
@@ -67,7 +67,7 @@ declare function AiLocateElement<ElementType extends BaseElement = BaseElement>(
 }>;
 declare function AiLocateSection(options: {
     context: UIContext<BaseElement>;
-    sectionDescription: string;
+    sectionDescription: TUserPrompt;
     callAI?: typeof callAiFn<AISectionLocatorResponse>;
 }): Promise<{
     rect?: Rect;
@@ -78,6 +78,7 @@ declare function AiLocateSection(options: {
 }>;
 declare function AiExtractElementInfo<T, ElementType extends BaseElement = BaseElement>(options: {
     dataQuery: string | Record<string, string>;
+    multimodalPrompt?: TMultimodalPrompt;
     context: UIContext<ElementType>;
     extractOption?: InsightExtractOption;
 }): Promise<{
@@ -86,7 +87,7 @@ declare function AiExtractElementInfo<T, ElementType extends BaseElement = BaseE
     usage: AIUsageInfo | undefined;
 }>;
 declare function AiAssert<ElementType extends BaseElement = BaseElement>(options: {
-    assertion: string;
+    assertion: TUserPrompt;
     context: UIContext<ElementType>;
 }): Promise<{
     content: AIAssertionResponse;
@@ -101,4 +102,4 @@ declare function plan(userInstruction: string, opts: {
     actionContext?: string;
 }): Promise<PlanningAIResponse>;
 
-export { AiLocateElement as A, AiAssert as a, AIActionType as b, callAiFn as c, describeUserPage as d, elementByPositionWithElementInfo as e, AiExtractElementInfo as f, AiLocateSection as g, adaptBboxToRect as h, plan as p };
+export { AiLocateElement as A, AiAssert as a, AIActionType as b, callAiFn as c, describeUserPage as d, type AIArgs as e, elementByPositionWithElementInfo as f, AiExtractElementInfo as g, AiLocateSection as h, adaptBboxToRect as i, plan as p };

package/dist/es/{types-d836fa73.d.ts → types-512d3687.d.ts}

@@ -3,7 +3,6 @@ import { Rect, BaseElement, ElementTreeNode, Size } from '@midscene/shared/types
 import { ChatCompletionMessageParam } from 'openai/resources';
 
 interface LocateOption {
-    prompt?: string;
     deepThink?: boolean;
     cacheable?: boolean;
     xpath?: string;
@@ -17,7 +16,7 @@ interface ReferenceImage {
     rect?: Rect;
 }
 interface DetailedLocateParam extends LocateOption {
-    prompt: string;
+    prompt: TUserPrompt;
     referenceImage?: ReferenceImage;
 }
 interface scrollParam {
@@ -102,25 +101,25 @@ interface MidsceneYamlFlowItemAIWaitFor {
     timeout?: number;
 }
 interface MidsceneYamlFlowItemAITap extends LocateOption {
-    aiTap: string;
+    aiTap: TUserPrompt;
 }
 interface MidsceneYamlFlowItemAIRightClick extends LocateOption {
-    aiRightClick: string;
+    aiRightClick: TUserPrompt;
 }
 interface MidsceneYamlFlowItemAIHover extends LocateOption {
-    aiHover: string;
+    aiHover: TUserPrompt;
 }
 interface MidsceneYamlFlowItemAIInput extends LocateOption {
     aiInput: string;
-    locate: string;
+    locate: TUserPrompt;
 }
 interface MidsceneYamlFlowItemAIKeyboardPress extends LocateOption {
     aiKeyboardPress: string;
-    locate?: string;
+    locate?: TUserPrompt;
 }
 interface MidsceneYamlFlowItemAIScroll extends LocateOption, PlanningActionParamScroll {
     aiScroll: null;
-    locate?: string;
+    locate?: TUserPrompt;
 }
 interface MidsceneYamlFlowItemEvaluateJavaScript {
     javascript: string;
@@ -310,9 +309,9 @@ interface InsightDump extends DumpMeta {
     type: 'locate' | 'extract' | 'assert';
     logId: string;
     userQuery: {
-        element?: string;
+        element?: TUserPrompt;
         dataDemand?: InsightExtractParam;
-        assertion?: string;
+        assertion?: TUserPrompt;
     };
     matchedElement: BaseElement[];
     matchedRect?: Rect;
@@ -382,7 +381,7 @@ interface PlanningActionParamInputOrKeyPress {
 }
 type PlanningActionParamScroll = scrollParam;
 interface PlanningActionParamAssert {
-    assertion: string;
+    assertion: TUserPrompt;
 }
 interface PlanningActionParamSleep {
     timeMs: number;
@@ -542,5 +541,25 @@ interface StreamingAIResponse {
     /** Whether the response was streamed */
     isStreamed: boolean;
 }
+type TMultimodalPrompt = {
+    /**
+     * Support use image to inspect elements.
+     * The "images" field is an object that uses image name as key and image url as value.
+     * The image url can be a local path, a http link , or a base64 string.
+     */
+    images?: {
+        name: string;
+        url: string;
+    }[];
+    /**
+     * By default, the image url in the "images" filed starts with `https://` or `http://` will be directly sent to the LLM.
+     * In case the images are not accessible to the LLM (One common case is that image url is internal network only.), you can enable this option.
+     * Then image will be download and convert to base64 format.
+     */
+    convertHttpImage2Base64?: boolean;
+};
+type TUserPrompt = string | ({
+    prompt: string;
+} & Partial<TMultimodalPrompt>);
 
-export { type PlanningActionParamScroll as $, type AIDescribeElementResponse as A, type EnsureObject as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type InsightExtractParam as F, type LocateResultElement as G, type DumpMeta as H, type InsightAction as I, type InsightDump as J, type LiteUISection as K, type LocateResult as L, type MidsceneYamlScript as M, type ElementById as N, type OnTaskStartTip as O, type PartialInsightDumpFromSDK as P, type AgentWaitForOpt as Q, type ReportDumpWithAttributes as R, type AgentAssertOpt as S, type PlanningLocateParam as T, UIContext as U, type PlanningAction as V, type PlanningAIResponse as W, type PlanningActionParamTap as X, type PlanningActionParamHover as Y, type PlanningActionParamRightClick as Z, type PlanningActionParamInputOrKeyPress as _, type ExecutionTaskProgressOptions as a, type FreeFn as a$, type PlanningActionParamAssert as a0, type PlanningActionParamSleep as a1, type PlanningActionParamError as a2, type PlanningActionParamWaitFor as a3, type PlanningActionParamAndroidLongPress as a4, type PlanningActionParamAndroidPull as a5, type Color as a6, type BaseAgentParserOpt as a7, type PuppeteerParserOpt as a8, type PlaywrightParserOpt as a9, type StreamingCallback as aA, type CodeGenerationChunk as aB, type StreamingAIResponse as aC, type LocateOption as aD, type ReferenceImage as aE, type scrollParam as aF, type MidsceneYamlScriptEnvBase as aG, type MidsceneYamlScriptWebEnv as aH, type MidsceneYamlScriptAndroidEnv as aI, type MidsceneYamlScriptEnv as aJ, type MidsceneYamlFlowItemAIAction as aK, type MidsceneYamlFlowItemAIAssert as aL, type MidsceneYamlFlowItemAIQuery as aM, type MidsceneYamlFlowItemAINumber as aN, type MidsceneYamlFlowItemAIString as aO, type MidsceneYamlFlowItemAIAsk as aP, type MidsceneYamlFlowItemAIBoolean as aQ, type MidsceneYamlFlowItemAILocate as aR, type MidsceneYamlFlowItemAIWaitFor as aS, type MidsceneYamlFlowItemAITap as aT, type MidsceneYamlFlowItemAIHover as aU, type MidsceneYamlFlowItemAIInput as aV, type MidsceneYamlFlowItemAIKeyboardPress as aW, type MidsceneYamlFlowItemAIScroll as aX, type MidsceneYamlFlowItemEvaluateJavaScript as aY, type MidsceneYamlFlowItemSleep as aZ, type MidsceneYamlFlowItemLogScreenshot as a_, type ExecutionRecorderItem as aa, type ExecutionTaskType as ab, type ExecutorContext as ac, type ExecutionTaskHitBy as ad, type ExecutionTaskReturn as ae, type ExecutionTaskInsightLocateParam as af, type ExecutionTaskInsightLocateOutput as ag, type ExecutionTaskInsightDumpLog as ah, type ExecutionTaskInsightLocateApply as ai, type ExecutionTaskInsightLocate as aj, type ExecutionTaskInsightQueryParam as ak, type ExecutionTaskInsightQueryOutput as al, type ExecutionTaskInsightQueryApply as am, type ExecutionTaskInsightQuery as an, type ExecutionTaskInsightAssertionParam as ao, type ExecutionTaskInsightAssertionApply as ap, type ExecutionTaskInsightAssertion as aq, type ExecutionTaskActionApply as ar, type ExecutionTaskAction as as, type ExecutionTaskLogApply as at, type ExecutionTaskLog as au, type ExecutionTaskPlanningApply as av, type ExecutionTaskPlanning as aw, type GroupedActionDump as ax, type PageType as ay, type StreamingCodeGenerationOptions as az, type ExecutionTaskApply as b, type ScriptPlayerTaskStatus as b0, type ScriptPlayerStatusValue as b1, type MidsceneYamlConfig as b2, type MidsceneYamlConfigOutput as b3, 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 PlanningActionParamRightClick as $, type AIDescribeElementResponse as A, type LocateValidatorResult as B, type AgentDescribeElementAtPointResult as C, type DumpSubscriber as D, type ExecutionTask as E, type CallAIFn as F, type EnsureObject as G, type LocateResultElement as H, type InsightAction as I, type DumpMeta as J, type InsightDump as K, type LocateResult as L, type MidsceneYamlScript as M, type LiteUISection as N, type ElementById as O, type PartialInsightDumpFromSDK as P, type OnTaskStartTip as Q, type ReportDumpWithAttributes as R, type AgentWaitForOpt as S, type TMultimodalPrompt as T, UIContext as U, type AgentAssertOpt as V, type PlanningLocateParam as W, type PlanningAction as X, type PlanningAIResponse as Y, type PlanningActionParamTap as Z, type PlanningActionParamHover as _, type ExecutionTaskProgressOptions as a, type MidsceneYamlFlowItemSleep as a$, type PlanningActionParamInputOrKeyPress as a0, type PlanningActionParamScroll as a1, type PlanningActionParamAssert as a2, type PlanningActionParamSleep as a3, type PlanningActionParamError as a4, type PlanningActionParamWaitFor as a5, type PlanningActionParamAndroidLongPress as a6, type PlanningActionParamAndroidPull as a7, type Color as a8, type BaseAgentParserOpt as a9, type PageType as aA, type StreamingCodeGenerationOptions as aB, type StreamingCallback as aC, type CodeGenerationChunk as aD, type StreamingAIResponse as aE, type LocateOption as aF, type ReferenceImage as aG, type scrollParam as aH, type MidsceneYamlScriptEnvBase as aI, type MidsceneYamlScriptWebEnv as aJ, type MidsceneYamlScriptAndroidEnv as aK, type MidsceneYamlScriptEnv as aL, type MidsceneYamlFlowItemAIAction as aM, type MidsceneYamlFlowItemAIAssert as aN, type MidsceneYamlFlowItemAIQuery as aO, type MidsceneYamlFlowItemAINumber as aP, type MidsceneYamlFlowItemAIString as aQ, type MidsceneYamlFlowItemAIAsk as aR, type MidsceneYamlFlowItemAIBoolean as aS, type MidsceneYamlFlowItemAILocate as aT, type MidsceneYamlFlowItemAIWaitFor as aU, type MidsceneYamlFlowItemAITap as aV, type MidsceneYamlFlowItemAIHover as aW, type MidsceneYamlFlowItemAIInput as aX, type MidsceneYamlFlowItemAIKeyboardPress as aY, type MidsceneYamlFlowItemAIScroll as aZ, type MidsceneYamlFlowItemEvaluateJavaScript as a_, type PuppeteerParserOpt as aa, type PlaywrightParserOpt as ab, type ExecutionRecorderItem as ac, type ExecutionTaskType as ad, type ExecutorContext as ae, type ExecutionTaskHitBy as af, type ExecutionTaskReturn as ag, type ExecutionTaskInsightLocateParam as ah, type ExecutionTaskInsightLocateOutput as ai, type ExecutionTaskInsightDumpLog as aj, type ExecutionTaskInsightLocateApply as ak, type ExecutionTaskInsightLocate as al, type ExecutionTaskInsightQueryParam as am, type ExecutionTaskInsightQueryOutput as an, type ExecutionTaskInsightQueryApply as ao, type ExecutionTaskInsightQuery as ap, type ExecutionTaskInsightAssertionParam as aq, type ExecutionTaskInsightAssertionApply as ar, type ExecutionTaskInsightAssertion as as, type ExecutionTaskActionApply as at, type ExecutionTaskAction as au, type ExecutionTaskLogApply as av, type ExecutionTaskLog as aw, type ExecutionTaskPlanningApply as ax, type ExecutionTaskPlanning as ay, type GroupedActionDump as az, type ExecutionTaskApply as b, type MidsceneYamlFlowItemLogScreenshot as b0, type FreeFn as b1, type ScriptPlayerTaskStatus as b2, type ScriptPlayerStatusValue as b3, type MidsceneYamlConfig as b4, type MidsceneYamlConfigOutput as b5, 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 TUserPrompt as i, type InsightAssertionResponse as j, type MidsceneYamlTask as k, type MidsceneYamlFlowItem as l, type MidsceneYamlFlowItemAIRightClick as m, type MidsceneYamlConfigResult as n, type AIUsageInfo as o, AIResponseFormat as p, type AISingleElementResponseById as q, type AISingleElementResponseByPosition as r, type AISingleElementResponse as s, type AIElementLocatorResponse as t, type AIElementCoordinatesResponse as u, type AIElementResponse as v, type AIDataExtractionResponse as w, type AISectionLocatorResponse as x, type AIAssertionResponse as y, type LocatorValidatorOption as z };

package/dist/es/utils.d.ts

@@ -1,4 +1,4 @@
-import { R as ReportDumpWithAttributes } from './types-d836fa73.js';
+import { R as ReportDumpWithAttributes } from './types-512d3687.js';
 import { Rect } from '@midscene/shared/types';
 import '@midscene/shared/constants';
 import 'openai/resources';

package/dist/es/utils.js

@@ -12,7 +12,7 @@ import {
   uploadTestInfoToServer,
   writeDumpReport,
   writeLogFile
-} from "./chunk-FMBJ3CM2.js";
+} from "./chunk-NZFWY3M5.js";
 export {
   getTmpDir,
   getTmpFile,

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

@@ -1,9 +1,9 @@
-import { aA as StreamingCallback, m as AIUsageInfo, az as StreamingCodeGenerationOptions, aC as StreamingAIResponse, V as PlanningAction, j as MidsceneYamlFlowItem } from './types-d836fa73.js';
+import { aC as StreamingCallback, o as AIUsageInfo, aB as StreamingCodeGenerationOptions, aE as StreamingAIResponse, X as PlanningAction, l as MidsceneYamlFlowItem } from './types-512d3687.js';
 import OpenAI from 'openai';
 import { ChatCompletionMessageParam } from 'openai/resources';
 export { ChatCompletionMessageParam } from 'openai/resources';
-import { b as AIActionType } from './llm-planning-d7096b0d.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-d7096b0d.js';
+import { b as AIActionType, e as AIArgs } from './llm-planning-877248da.js';
+export { a as AiAssert, g as AiExtractElementInfo, A as AiLocateElement, h as AiLocateSection, i as adaptBboxToRect, c as callAiFn, d as describeUserPage, f as elementByPositionWithElementInfo, p as plan } from './llm-planning-877248da.js';
 import { vlLocateMode } from '@midscene/shared/env';
 import { actionParser } from '@ui-tars/action-parser';
 import { Size } from '@midscene/shared/types';
@@ -21,6 +21,10 @@ declare function callToGetJSONObject<T>(messages: ChatCompletionMessageParam[],
     content: T;
     usage?: AIUsageInfo;
 }>;
+declare function callAiFnWithStringResponse<T>(msgs: AIArgs, AIActionTypeValue: AIActionType): Promise<{
+    content: string;
+    usage?: AIUsageInfo;
+}>;
 
 declare function systemPromptToLocateElement(vlMode: ReturnType<typeof vlLocateMode>): string;
 
@@ -92,4 +96,4 @@ declare function vlmPlanning(options: {
 }>;
 declare function resizeImageForUiTars(imageBase64: string, size: Size): Promise<string>;
 
-export { AIActionType, call as callAi, callToGetJSONObject, generatePlaywrightTest, generatePlaywrightTestStream, generateYamlTest, generateYamlTestStream, resizeImageForUiTars, systemPromptToLocateElement, vlmPlanning };
+export { AIActionType, AIArgs, call as callAi, callAiFnWithStringResponse, callToGetJSONObject, generatePlaywrightTest, generatePlaywrightTestStream, generateYamlTest, generateYamlTestStream, resizeImageForUiTars, systemPromptToLocateElement, vlmPlanning };

package/dist/lib/ai-model.js

@@ -18,8 +18,8 @@
 
 
 
-var _chunk5HH6E7M4js = require('./chunk-5HH6E7M4.js');
 
+var _chunkJS4CT3XVjs = require('./chunk-JS4CT3XV.js');
 
 
 
@@ -39,4 +39,6 @@ var _chunk5HH6E7M4js = require('./chunk-5HH6E7M4.js');
 
 
 
-exports.AIActionType = _chunk5HH6E7M4js.AIActionType; exports.AiAssert = _chunk5HH6E7M4js.AiAssert; exports.AiExtractElementInfo = _chunk5HH6E7M4js.AiExtractElementInfo; exports.AiLocateElement = _chunk5HH6E7M4js.AiLocateElement; exports.AiLocateSection = _chunk5HH6E7M4js.AiLocateSection; exports.adaptBboxToRect = _chunk5HH6E7M4js.adaptBboxToRect; exports.callAi = _chunk5HH6E7M4js.call; exports.callAiFn = _chunk5HH6E7M4js.callAiFn; exports.callToGetJSONObject = _chunk5HH6E7M4js.callToGetJSONObject; exports.describeUserPage = _chunk5HH6E7M4js.describeUserPage; exports.elementByPositionWithElementInfo = _chunk5HH6E7M4js.elementByPositionWithElementInfo; exports.generatePlaywrightTest = _chunk5HH6E7M4js.generatePlaywrightTest; exports.generatePlaywrightTestStream = _chunk5HH6E7M4js.generatePlaywrightTestStream; exports.generateYamlTest = _chunk5HH6E7M4js.generateYamlTest; exports.generateYamlTestStream = _chunk5HH6E7M4js.generateYamlTestStream; exports.plan = _chunk5HH6E7M4js.plan; exports.resizeImageForUiTars = _chunk5HH6E7M4js.resizeImageForUiTars; exports.systemPromptToLocateElement = _chunk5HH6E7M4js.systemPromptToLocateElement; exports.vlmPlanning = _chunk5HH6E7M4js.vlmPlanning;
+
+
+exports.AIActionType = _chunkJS4CT3XVjs.AIActionType; exports.AiAssert = _chunkJS4CT3XVjs.AiAssert; exports.AiExtractElementInfo = _chunkJS4CT3XVjs.AiExtractElementInfo; exports.AiLocateElement = _chunkJS4CT3XVjs.AiLocateElement; exports.AiLocateSection = _chunkJS4CT3XVjs.AiLocateSection; exports.adaptBboxToRect = _chunkJS4CT3XVjs.adaptBboxToRect; exports.callAi = _chunkJS4CT3XVjs.call; exports.callAiFn = _chunkJS4CT3XVjs.callAiFn; exports.callAiFnWithStringResponse = _chunkJS4CT3XVjs.callAiFnWithStringResponse; exports.callToGetJSONObject = _chunkJS4CT3XVjs.callToGetJSONObject; exports.describeUserPage = _chunkJS4CT3XVjs.describeUserPage; exports.elementByPositionWithElementInfo = _chunkJS4CT3XVjs.elementByPositionWithElementInfo; exports.generatePlaywrightTest = _chunkJS4CT3XVjs.generatePlaywrightTest; exports.generatePlaywrightTestStream = _chunkJS4CT3XVjs.generatePlaywrightTestStream; exports.generateYamlTest = _chunkJS4CT3XVjs.generateYamlTest; exports.generateYamlTestStream = _chunkJS4CT3XVjs.generateYamlTestStream; exports.plan = _chunkJS4CT3XVjs.plan; exports.resizeImageForUiTars = _chunkJS4CT3XVjs.resizeImageForUiTars; exports.systemPromptToLocateElement = _chunkJS4CT3XVjs.systemPromptToLocateElement; exports.vlmPlanning = _chunkJS4CT3XVjs.vlmPlanning;