@donggui/core 1.5.13 → 1.6.0

package/dist/es/service/index.mjs

@@ -38,7 +38,6 @@ class Service {
         const context = opt?.context || await this.contextRetrieverFn();
         let searchArea;
         let searchAreaRawResponse;
-        let searchAreaUsage;
         let searchAreaResponse;
         if (searchAreaPrompt) {
             searchAreaResponse = await AiLocateSection({
@@ -49,7 +48,7 @@ class Service {
             });
             assert(searchAreaResponse.rect, `cannot find search area for "${searchAreaPrompt}"${searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''}`);
             searchAreaRawResponse = searchAreaResponse.rawResponse;
-            searchAreaUsage = searchAreaResponse.usage;
+            searchAreaResponse.usage;
             searchArea = searchAreaResponse.rect;
         }
         const startTime = Date.now();
@@ -69,7 +68,6 @@ class Service {
             usage,
             searchArea,
             searchAreaRawResponse,
-            searchAreaUsage,
             reasoning_content
         };
         let errorLog;
@@ -82,9 +80,9 @@ class Service {
             matchedElement: [],
             matchedRect: rect,
             data: null,
-            taskInfo,
-            deepLocate: !!searchArea,
-            error: errorLog
+            deepLocate: !!searchArea || !!searchAreaPrompt,
+            error: errorLog,
+            taskInfo
         };
         const elements = parseResult.elements || [];
         const dump = createServiceDump({

package/dist/es/service/index.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"service/index.mjs","sources":["../../../src/service/index.ts"],"sourcesContent":["import { isAutoGLM, isUITars } from '@/ai-model/auto-glm/util';\nimport {\n  AIResponseParseError,\n  AiExtractElementInfo,\n  AiLocateElement,\n  callAIWithObjectResponse,\n} from '@/ai-model/index';\nimport { AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport { type AIArgs, expandSearchArea } from '@/common';\nimport type {\n  AIDescribeElementResponse,\n  AIUsageInfo,\n  DetailedLocateParam,\n  LocateResultWithDump,\n  PartialServiceDumpFromSDK,\n  Rect,\n  ServiceExtractOption,\n  ServiceExtractParam,\n  ServiceExtractResult,\n  ServiceTaskInfo,\n  UIContext,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TMultimodalPrompt } from '../common';\nimport { createServiceDump } from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\ninterface ServiceOptions {\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n}\n\nconst debug = getDebug('ai:service');\nexport default class Service {\n  contextRetrieverFn: () => Promise<UIContext> | UIContext;\n\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n\n  constructor(\n    context: UIContext | (() => Promise<UIContext> | UIContext),\n    opt?: ServiceOptions,\n  ) {\n    assert(context, 'context is required for Service');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\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    modelConfig: IModelConfig,\n    abortSignal?: AbortSignal,\n  ): Promise<LocateResultWithDump> {\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    let searchAreaPrompt;\n    if (query.deepLocate) {\n      searchAreaPrompt = query.prompt;\n    }\n\n    const { modelFamily } = modelConfig;\n\n    if (searchAreaPrompt && !modelFamily) {\n      console.warn(\n        'The \"deepLocate\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/model-config',\n      );\n      searchAreaPrompt = undefined;\n    }\n\n    if (searchAreaPrompt && isAutoGLM(modelFamily)) {\n      console.warn('The \"deepLocate\" feature is not supported with AutoGLM.');\n      searchAreaPrompt = undefined;\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn());\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        modelConfig,\n        abortSignal,\n      });\n      assert(\n        searchAreaResponse.rect,\n        `cannot find search area for \"${searchAreaPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const { parseResult, rect, rawResponse, usage, reasoning_content } =\n      await AiLocateElement({\n        context,\n        targetElementDescription: queryPrompt,\n        searchConfig: searchAreaResponse,\n        modelConfig,\n        abortSignal,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `failed to locate element: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      matchedElement: [],\n      matchedRect: rect,\n      data: null,\n      taskInfo,\n      deepLocate: !!searchArea,\n      error: errorLog,\n    };\n\n    const elements = parseResult.elements || [];\n\n    const dump = createServiceDump({\n      ...dumpData,\n      matchedElement: elements,\n    });\n\n    if (errorLog) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    if (elements.length > 1) {\n      throw new ServiceError(\n        `locate: multiple elements found, length = ${elements.length}`,\n        dump,\n      );\n    }\n\n    if (elements.length === 1) {\n      return {\n        element: {\n          center: elements[0]!.center,\n          rect: elements[0]!.rect,\n          description: elements[0]!.description,\n        },\n        rect,\n        dump,\n      };\n    }\n\n    return {\n      element: null,\n      rect,\n      dump,\n    };\n  }\n\n  async extract<T>(\n    dataDemand: ServiceExtractParam,\n    modelConfig: IModelConfig,\n    opt?: ServiceExtractOption,\n    pageDescription?: string,\n    multimodalPrompt?: TMultimodalPrompt,\n    context?: UIContext,\n  ): Promise<ServiceExtractResult<T>> {\n    assert(context, 'context is required for extract');\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n\n    const startTime = Date.now();\n\n    let parseResult: Awaited<\n      ReturnType<typeof AiExtractElementInfo<T>>\n    >['parseResult'];\n    let rawResponse: string;\n    let usage: Awaited<ReturnType<typeof AiExtractElementInfo<T>>>['usage'];\n    let reasoning_content: string | undefined;\n\n    try {\n      const result = await AiExtractElementInfo<T>({\n        context,\n        dataQuery: dataDemand,\n        multimodalPrompt,\n        extractOption: opt,\n        modelConfig,\n        pageDescription,\n      });\n      parseResult = result.parseResult;\n      rawResponse = result.rawResponse;\n      usage = result.usage;\n      reasoning_content = result.reasoning_content;\n    } catch (error) {\n      if (error instanceof AIResponseParseError) {\n        // Create dump with usage and rawResponse from the error\n        const timeCost = Date.now() - startTime;\n        const taskInfo: ServiceTaskInfo = {\n          ...(this.taskInfo ? this.taskInfo : {}),\n          durationMs: timeCost,\n          rawResponse: error.rawResponse,\n          usage: error.usage,\n        };\n        const dump = createServiceDump({\n          type: 'extract',\n          userQuery: { dataDemand },\n          matchedElement: [],\n          data: null,\n          taskInfo,\n          error: error.message,\n        });\n        throw new ServiceError(error.message, dump);\n      }\n      throw error;\n    }\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse,\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\n      type: 'extract',\n      userQuery: {\n        dataDemand,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      error: errorLog,\n    };\n\n    const { data, thought } = parseResult || {};\n\n    const dump = createServiceDump({\n      ...dumpData,\n      data,\n    });\n\n    if (errorLog && !data) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    return {\n      data,\n      thought,\n      usage,\n      reasoning_content,\n      dump,\n    };\n  }\n\n  async describe(\n    target: Rect | [number, number],\n    modelConfig: IModelConfig,\n    opt?: {\n      deepLocate?: boolean;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for service.describe');\n    const context = await this.contextRetrieverFn();\n    const { shotSize } = context;\n    const screenshotBase64 = context.screenshot.base64;\n    assert(screenshotBase64, 'screenshot is required for service.describe');\n    // The result of the \"describe\" function will be used for positioning, so essentially it is a form of grounding.\n    const { modelFamily } = modelConfig;\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: shotSize,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepLocate) {\n      const searchArea = expandSearchArea(targetRect, shotSize);\n      // Always crop in describe mode. Unlike locate's deepLocate (where\n      // cropping too small loses context for finding elements), describe's\n      // deepLocate intentionally zooms in so the model produces a more\n      // precise description from a focused view. expandSearchArea already\n      // guarantees a minimum 400x400 area with surrounding context.\n      debug('describe: cropping to searchArea', searchArea);\n      const croppedResult = await cropByRect(\n        imagePayload,\n        searchArea,\n        modelFamily === 'qwen2.5-vl',\n      );\n      imagePayload = croppedResult.imageBase64;\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 res = await callAIWithObjectResponse<AIDescribeElementResponse>(\n      msgs,\n      modelConfig,\n    );\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"],"names":["debug","getDebug","Service","query","opt","modelConfig","abortSignal","queryPrompt","assert","searchAreaPrompt","modelFamily","console","undefined","isAutoGLM","context","searchArea","searchAreaRawResponse","searchAreaUsage","searchAreaResponse","AiLocateSection","startTime","Date","parseResult","rect","rawResponse","usage","reasoning_content","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","elements","dump","createServiceDump","ServiceError","dataDemand","pageDescription","multimodalPrompt","result","AiExtractElementInfo","error","AIResponseParseError","data","thought","target","shotSize","screenshotBase64","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","croppedResult","cropByRect","msgs","res","callAIWithObjectResponse","content","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;;AA2CA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IAqBnB,MAAM,OACJC,KAA0B,EAC1BC,GAAe,EACfC,WAAyB,EACzBC,WAAyB,EACM;QAC/B,MAAMC,cAAc,AAAiB,YAAjB,OAAOJ,QAAqBA,QAAQA,MAAM,MAAM;QACpEK,OAAOD,aAAa;QAEpBC,OAAO,AAAiB,YAAjB,OAAOL,OAAoB;QAElC,IAAIM;QACJ,IAAIN,MAAM,UAAU,EAClBM,mBAAmBN,MAAM,MAAM;QAGjC,MAAM,EAAEO,WAAW,EAAE,GAAGL;QAExB,IAAII,oBAAoB,CAACC,aAAa;YACpCC,QAAQ,IAAI,CACV;YAEFF,mBAAmBG;QACrB;QAEA,IAAIH,oBAAoBI,UAAUH,cAAc;YAC9CC,QAAQ,IAAI,CAAC;YACbF,mBAAmBG;QACrB;QAEA,MAAME,UAAUV,KAAK,WAAY,MAAM,IAAI,CAAC,kBAAkB;QAE9D,IAAIW;QACJ,IAAIC;QACJ,IAAIC;QACJ,IAAIC;QAGJ,IAAIT,kBAAkB;YACpBS,qBAAqB,MAAMC,gBAAgB;gBACzCL;gBACA,oBAAoBL;gBACpBJ;gBACAC;YACF;YACAE,OACEU,mBAAmB,IAAI,EACvB,CAAC,6BAA6B,EAAET,iBAAiB,CAAC,EAChDS,mBAAmB,KAAK,GAAG,CAAC,EAAE,EAAEA,mBAAmB,KAAK,EAAE,GAAG,IAC7D;YAEJF,wBAAwBE,mBAAmB,WAAW;YACtDD,kBAAkBC,mBAAmB,KAAK;YAC1CH,aAAaG,mBAAmB,IAAI;QACtC;QAEA,MAAME,YAAYC,KAAK,GAAG;QAC1B,MAAM,EAAEC,WAAW,EAAEC,IAAI,EAAEC,WAAW,EAAEC,KAAK,EAAEC,iBAAiB,EAAE,GAChE,MAAMC,gBAAgB;YACpBb;YACA,0BAA0BP;YAC1B,cAAcW;YACdb;YACAC;QACF;QAEF,MAAMsB,WAAWP,KAAK,GAAG,KAAKD;QAC9B,MAAMS,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACN;YAC5B,gBAAgBM,KAAK,SAAS,CAACR;YAC/BG;YACAV;YACAC;YACAC;YACAS;QACF;QAEA,IAAIK;QACJ,IAAIT,YAAY,MAAM,EAAE,QACtBS,WAAW,CAAC,4BAA4B,EAAET,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG3E,MAAMU,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAASzB;YACX;YACA,gBAAgB,EAAE;YAClB,aAAagB;YACb,MAAM;YACNM;YACA,YAAY,CAAC,CAACd;YACd,OAAOgB;QACT;QAEA,MAAME,WAAWX,YAAY,QAAQ,IAAI,EAAE;QAE3C,MAAMY,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACX,gBAAgBC;QAClB;QAEA,IAAIF,UACF,MAAM,IAAIK,aAAaL,UAAUG;QAGnC,IAAID,SAAS,MAAM,GAAG,GACpB,MAAM,IAAIG,aACR,CAAC,0CAA0C,EAAEH,SAAS,MAAM,EAAE,EAC9DC;QAIJ,IAAID,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,SAAS;gBACP,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM;gBAC3B,MAAMA,QAAQ,CAAC,EAAE,CAAE,IAAI;gBACvB,aAAaA,QAAQ,CAAC,EAAE,CAAE,WAAW;YACvC;YACAV;YACAW;QACF;QAGF,OAAO;YACL,SAAS;YACTX;YACAW;QACF;IACF;IAEA,MAAM,QACJG,UAA+B,EAC/BhC,WAAyB,EACzBD,GAA0B,EAC1BkC,eAAwB,EACxBC,gBAAoC,EACpCzB,OAAmB,EACe;QAClCN,OAAOM,SAAS;QAChBN,OACE,AAAsB,YAAtB,OAAO6B,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAGvE,MAAMjB,YAAYC,KAAK,GAAG;QAE1B,IAAIC;QAGJ,IAAIE;QACJ,IAAIC;QACJ,IAAIC;QAEJ,IAAI;YACF,MAAMc,SAAS,MAAMC,qBAAwB;gBAC3C3B;gBACA,WAAWuB;gBACXE;gBACA,eAAenC;gBACfC;gBACAiC;YACF;YACAhB,cAAckB,OAAO,WAAW;YAChChB,cAAcgB,OAAO,WAAW;YAChCf,QAAQe,OAAO,KAAK;YACpBd,oBAAoBc,OAAO,iBAAiB;QAC9C,EAAE,OAAOE,OAAO;YACd,IAAIA,iBAAiBC,sBAAsB;gBAEzC,MAAMf,WAAWP,KAAK,GAAG,KAAKD;gBAC9B,MAAMS,WAA4B;oBAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;oBACtC,YAAYD;oBACZ,aAAac,MAAM,WAAW;oBAC9B,OAAOA,MAAM,KAAK;gBACpB;gBACA,MAAMR,OAAOC,kBAAkB;oBAC7B,MAAM;oBACN,WAAW;wBAAEE;oBAAW;oBACxB,gBAAgB,EAAE;oBAClB,MAAM;oBACNR;oBACA,OAAOa,MAAM,OAAO;gBACtB;gBACA,MAAM,IAAIN,aAAaM,MAAM,OAAO,EAAER;YACxC;YACA,MAAMQ;QACR;QAEA,MAAMd,WAAWP,KAAK,GAAG,KAAKD;QAC9B,MAAMS,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZJ;YACA,gBAAgBM,KAAK,SAAS,CAACR;YAC/BG;YACAC;QACF;QAEA,IAAIK;QACJ,IAAIT,YAAY,MAAM,EAAE,QACtBS,WAAW,CAAC,qBAAqB,EAAET,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMU,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTK;YACF;YACA,gBAAgB,EAAE;YAClB,MAAM;YACNR;YACA,OAAOE;QACT;QAEA,MAAM,EAAEa,IAAI,EAAEC,OAAO,EAAE,GAAGvB,eAAe,CAAC;QAE1C,MAAMY,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACXY;QACF;QAEA,IAAIb,YAAY,CAACa,MACf,MAAM,IAAIR,aAAaL,UAAUG;QAGnC,OAAO;YACLU;YACAC;YACApB;YACAC;YACAQ;QACF;IACF;IAEA,MAAM,SACJY,MAA+B,EAC/BzC,WAAyB,EACzBD,GAEC,EACwD;QACzDI,OAAOsC,QAAQ;QACf,MAAMhC,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAC7C,MAAM,EAAEiC,QAAQ,EAAE,GAAGjC;QACrB,MAAMkC,mBAAmBlC,QAAQ,UAAU,CAAC,MAAM;QAClDN,OAAOwC,kBAAkB;QAEzB,MAAM,EAAEtC,WAAW,EAAE,GAAGL;QACxB,MAAM4C,eAAeC;QAGrB,MAAMC,kBAAkB;QACxB,MAAMC,aAAmBC,MAAM,OAAO,CAACP,UACnC;YACE,MAAMQ,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC/C,KAAKG,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC9C,OAAOA;YACP,QAAQA;QACV,IACAL;QAEJ,IAAIS,eAAe,MAAMC,wBAAwB;YAC/C,gBAAgBR;YAChB,MAAMD;YACN,sBAAsB;gBACpB;oBACE,MAAMK;gBACR;aACD;YACD,iBAAiB;QACnB;QAEA,IAAIhD,KAAK,YAAY;YACnB,MAAMW,aAAa0C,iBAAiBL,YAAYL;YAMhD/C,MAAM,oCAAoCe;YAC1C,MAAM2C,gBAAgB,MAAMC,WAC1BJ,cACAxC,YACAL,AAAgB,iBAAhBA;YAEF6C,eAAeG,cAAc,WAAW;QAC1C;QAEA,MAAME,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,MAAM,MAAMC,yBAChBF,MACAvD;QAGF,MAAM,EAAE0D,OAAO,EAAE,GAAGF;QACpBrD,OAAO,CAACuD,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1DvD,OAAOuD,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IA9UA,YACEjD,OAA2D,EAC3DV,GAAoB,CACpB;QAPF;QAEA;QAMEI,OAAOM,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMkD,QAAQ,OAAO,CAAClD;QAGlD,IAAI,AAAyB,WAAlBV,KAAK,UACd,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AAiUF"}
+{"version":3,"file":"service/index.mjs","sources":["../../../src/service/index.ts"],"sourcesContent":["import { isAutoGLM, isUITars } from '@/ai-model/auto-glm/util';\nimport {\n  AIResponseParseError,\n  AiExtractElementInfo,\n  AiLocateElement,\n  callAIWithObjectResponse,\n} from '@/ai-model/index';\nimport { AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport { type AIArgs, expandSearchArea } from '@/common';\nimport type {\n  AIDescribeElementResponse,\n  AIUsageInfo,\n  DetailedLocateParam,\n  LocateResultWithDump,\n  PartialServiceDumpFromSDK,\n  Rect,\n  ServiceExtractOption,\n  ServiceExtractParam,\n  ServiceExtractResult,\n  ServiceTaskInfo,\n  UIContext,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TMultimodalPrompt } from '../common';\nimport { createServiceDump } from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\ninterface ServiceOptions {\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n}\n\nconst debug = getDebug('ai:service');\nexport default class Service {\n  contextRetrieverFn: () => Promise<UIContext> | UIContext;\n\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n\n  constructor(\n    context: UIContext | (() => Promise<UIContext> | UIContext),\n    opt?: ServiceOptions,\n  ) {\n    assert(context, 'context is required for Service');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\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    modelConfig: IModelConfig,\n    abortSignal?: AbortSignal,\n  ): Promise<LocateResultWithDump> {\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    let searchAreaPrompt;\n    if (query.deepLocate) {\n      searchAreaPrompt = query.prompt;\n    }\n\n    const { modelFamily } = modelConfig;\n\n    if (searchAreaPrompt && !modelFamily) {\n      console.warn(\n        'The \"deepLocate\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/model-config',\n      );\n      searchAreaPrompt = undefined;\n    }\n\n    if (searchAreaPrompt && isAutoGLM(modelFamily)) {\n      console.warn('The \"deepLocate\" feature is not supported with AutoGLM.');\n      searchAreaPrompt = undefined;\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn());\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        modelConfig,\n        abortSignal,\n      });\n      assert(\n        searchAreaResponse.rect,\n        `cannot find search area for \"${searchAreaPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const { parseResult, rect, rawResponse, usage, reasoning_content } =\n      await AiLocateElement({\n        context,\n        targetElementDescription: queryPrompt,\n        searchConfig: searchAreaResponse,\n        modelConfig,\n        abortSignal,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `failed to locate element: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      matchedElement: [],\n      matchedRect: rect,\n      data: null,\n      deepLocate: !!searchArea || !!searchAreaPrompt,\n      error: errorLog,\n      taskInfo,\n    };\n\n    const elements = parseResult.elements || [];\n\n    const dump = createServiceDump({\n      ...dumpData,\n      matchedElement: elements,\n    });\n\n    if (errorLog) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    if (elements.length > 1) {\n      throw new ServiceError(\n        `locate: multiple elements found, length = ${elements.length}`,\n        dump,\n      );\n    }\n\n    if (elements.length === 1) {\n      return {\n        element: {\n          center: elements[0]!.center,\n          rect: elements[0]!.rect,\n          description: elements[0]!.description,\n        },\n        rect,\n        dump,\n      };\n    }\n\n    return {\n      element: null,\n      rect,\n      dump,\n    };\n  }\n\n  async extract<T>(\n    dataDemand: ServiceExtractParam,\n    modelConfig: IModelConfig,\n    opt?: ServiceExtractOption,\n    pageDescription?: string,\n    multimodalPrompt?: TMultimodalPrompt,\n    context?: UIContext,\n  ): Promise<ServiceExtractResult<T>> {\n    assert(context, 'context is required for extract');\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n\n    const startTime = Date.now();\n\n    let parseResult: Awaited<\n      ReturnType<typeof AiExtractElementInfo<T>>\n    >['parseResult'];\n    let rawResponse: string;\n    let usage: Awaited<ReturnType<typeof AiExtractElementInfo<T>>>['usage'];\n    let reasoning_content: string | undefined;\n\n    try {\n      const result = await AiExtractElementInfo<T>({\n        context,\n        dataQuery: dataDemand,\n        multimodalPrompt,\n        extractOption: opt,\n        modelConfig,\n        pageDescription,\n      });\n      parseResult = result.parseResult;\n      rawResponse = result.rawResponse;\n      usage = result.usage;\n      reasoning_content = result.reasoning_content;\n    } catch (error) {\n      if (error instanceof AIResponseParseError) {\n        // Create dump with usage and rawResponse from the error\n        const timeCost = Date.now() - startTime;\n        const taskInfo: ServiceTaskInfo = {\n          ...(this.taskInfo ? this.taskInfo : {}),\n          durationMs: timeCost,\n          rawResponse: error.rawResponse,\n          usage: error.usage,\n        };\n        const dump = createServiceDump({\n          type: 'extract',\n          userQuery: { dataDemand },\n          matchedElement: [],\n          data: null,\n          taskInfo,\n          error: error.message,\n        });\n        throw new ServiceError(error.message, dump);\n      }\n      throw error;\n    }\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse,\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\n      type: 'extract',\n      userQuery: {\n        dataDemand,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      error: errorLog,\n    };\n\n    const { data, thought } = parseResult || {};\n\n    const dump = createServiceDump({\n      ...dumpData,\n      data,\n    });\n\n    if (errorLog && !data) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    return {\n      data,\n      thought,\n      usage,\n      reasoning_content,\n      dump,\n    };\n  }\n\n  async describe(\n    target: Rect | [number, number],\n    modelConfig: IModelConfig,\n    opt?: {\n      deepLocate?: boolean;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for service.describe');\n    const context = await this.contextRetrieverFn();\n    const { shotSize } = context;\n    const screenshotBase64 = context.screenshot.base64;\n    assert(screenshotBase64, 'screenshot is required for service.describe');\n    // The result of the \"describe\" function will be used for positioning, so essentially it is a form of grounding.\n    const { modelFamily } = modelConfig;\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: shotSize,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepLocate) {\n      const searchArea = expandSearchArea(targetRect, shotSize);\n      // Always crop in describe mode. Unlike locate's deepLocate (where\n      // cropping too small loses context for finding elements), describe's\n      // deepLocate intentionally zooms in so the model produces a more\n      // precise description from a focused view. expandSearchArea already\n      // guarantees a minimum 400x400 area with surrounding context.\n      debug('describe: cropping to searchArea', searchArea);\n      const croppedResult = await cropByRect(\n        imagePayload,\n        searchArea,\n        modelFamily === 'qwen2.5-vl',\n      );\n      imagePayload = croppedResult.imageBase64;\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 res = await callAIWithObjectResponse<AIDescribeElementResponse>(\n      msgs,\n      modelConfig,\n    );\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"],"names":["debug","getDebug","Service","query","opt","modelConfig","abortSignal","queryPrompt","assert","searchAreaPrompt","modelFamily","console","undefined","isAutoGLM","context","searchArea","searchAreaRawResponse","searchAreaResponse","AiLocateSection","startTime","Date","parseResult","rect","rawResponse","usage","reasoning_content","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","elements","dump","createServiceDump","ServiceError","dataDemand","pageDescription","multimodalPrompt","result","AiExtractElementInfo","error","AIResponseParseError","data","thought","target","shotSize","screenshotBase64","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","croppedResult","cropByRect","msgs","res","callAIWithObjectResponse","content","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;;AA2CA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IAqBnB,MAAM,OACJC,KAA0B,EAC1BC,GAAe,EACfC,WAAyB,EACzBC,WAAyB,EACM;QAC/B,MAAMC,cAAc,AAAiB,YAAjB,OAAOJ,QAAqBA,QAAQA,MAAM,MAAM;QACpEK,OAAOD,aAAa;QAEpBC,OAAO,AAAiB,YAAjB,OAAOL,OAAoB;QAElC,IAAIM;QACJ,IAAIN,MAAM,UAAU,EAClBM,mBAAmBN,MAAM,MAAM;QAGjC,MAAM,EAAEO,WAAW,EAAE,GAAGL;QAExB,IAAII,oBAAoB,CAACC,aAAa;YACpCC,QAAQ,IAAI,CACV;YAEFF,mBAAmBG;QACrB;QAEA,IAAIH,oBAAoBI,UAAUH,cAAc;YAC9CC,QAAQ,IAAI,CAAC;YACbF,mBAAmBG;QACrB;QAEA,MAAME,UAAUV,KAAK,WAAY,MAAM,IAAI,CAAC,kBAAkB;QAE9D,IAAIW;QACJ,IAAIC;QAEJ,IAAIC;QAGJ,IAAIR,kBAAkB;YACpBQ,qBAAqB,MAAMC,gBAAgB;gBACzCJ;gBACA,oBAAoBL;gBACpBJ;gBACAC;YACF;YACAE,OACES,mBAAmB,IAAI,EACvB,CAAC,6BAA6B,EAAER,iBAAiB,CAAC,EAChDQ,mBAAmB,KAAK,GAAG,CAAC,EAAE,EAAEA,mBAAmB,KAAK,EAAE,GAAG,IAC7D;YAEJD,wBAAwBC,mBAAmB,WAAW;YACpCA,mBAAmB,KAAK;YAC1CF,aAAaE,mBAAmB,IAAI;QACtC;QAEA,MAAME,YAAYC,KAAK,GAAG;QAC1B,MAAM,EAAEC,WAAW,EAAEC,IAAI,EAAEC,WAAW,EAAEC,KAAK,EAAEC,iBAAiB,EAAE,GAChE,MAAMC,gBAAgB;YACpBZ;YACA,0BAA0BP;YAC1B,cAAcU;YACdZ;YACAC;QACF;QAEF,MAAMqB,WAAWP,KAAK,GAAG,KAAKD;QAC9B,MAAMS,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACN;YAC5B,gBAAgBM,KAAK,SAAS,CAACR;YAC/BG;YACAT;YACAC;YACAS;QACF;QAEA,IAAIK;QACJ,IAAIT,YAAY,MAAM,EAAE,QACtBS,WAAW,CAAC,4BAA4B,EAAET,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG3E,MAAMU,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAASxB;YACX;YACA,gBAAgB,EAAE;YAClB,aAAae;YACb,MAAM;YACN,YAAY,CAAC,CAACP,cAAc,CAAC,CAACN;YAC9B,OAAOqB;YACPF;QACF;QAEA,MAAMI,WAAWX,YAAY,QAAQ,IAAI,EAAE;QAE3C,MAAMY,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACX,gBAAgBC;QAClB;QAEA,IAAIF,UACF,MAAM,IAAIK,aAAaL,UAAUG;QAGnC,IAAID,SAAS,MAAM,GAAG,GACpB,MAAM,IAAIG,aACR,CAAC,0CAA0C,EAAEH,SAAS,MAAM,EAAE,EAC9DC;QAIJ,IAAID,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,SAAS;gBACP,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM;gBAC3B,MAAMA,QAAQ,CAAC,EAAE,CAAE,IAAI;gBACvB,aAAaA,QAAQ,CAAC,EAAE,CAAE,WAAW;YACvC;YACAV;YACAW;QACF;QAGF,OAAO;YACL,SAAS;YACTX;YACAW;QACF;IACF;IAEA,MAAM,QACJG,UAA+B,EAC/B/B,WAAyB,EACzBD,GAA0B,EAC1BiC,eAAwB,EACxBC,gBAAoC,EACpCxB,OAAmB,EACe;QAClCN,OAAOM,SAAS;QAChBN,OACE,AAAsB,YAAtB,OAAO4B,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAGvE,MAAMjB,YAAYC,KAAK,GAAG;QAE1B,IAAIC;QAGJ,IAAIE;QACJ,IAAIC;QACJ,IAAIC;QAEJ,IAAI;YACF,MAAMc,SAAS,MAAMC,qBAAwB;gBAC3C1B;gBACA,WAAWsB;gBACXE;gBACA,eAAelC;gBACfC;gBACAgC;YACF;YACAhB,cAAckB,OAAO,WAAW;YAChChB,cAAcgB,OAAO,WAAW;YAChCf,QAAQe,OAAO,KAAK;YACpBd,oBAAoBc,OAAO,iBAAiB;QAC9C,EAAE,OAAOE,OAAO;YACd,IAAIA,iBAAiBC,sBAAsB;gBAEzC,MAAMf,WAAWP,KAAK,GAAG,KAAKD;gBAC9B,MAAMS,WAA4B;oBAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;oBACtC,YAAYD;oBACZ,aAAac,MAAM,WAAW;oBAC9B,OAAOA,MAAM,KAAK;gBACpB;gBACA,MAAMR,OAAOC,kBAAkB;oBAC7B,MAAM;oBACN,WAAW;wBAAEE;oBAAW;oBACxB,gBAAgB,EAAE;oBAClB,MAAM;oBACNR;oBACA,OAAOa,MAAM,OAAO;gBACtB;gBACA,MAAM,IAAIN,aAAaM,MAAM,OAAO,EAAER;YACxC;YACA,MAAMQ;QACR;QAEA,MAAMd,WAAWP,KAAK,GAAG,KAAKD;QAC9B,MAAMS,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZJ;YACA,gBAAgBM,KAAK,SAAS,CAACR;YAC/BG;YACAC;QACF;QAEA,IAAIK;QACJ,IAAIT,YAAY,MAAM,EAAE,QACtBS,WAAW,CAAC,qBAAqB,EAAET,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMU,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTK;YACF;YACA,gBAAgB,EAAE;YAClB,MAAM;YACNR;YACA,OAAOE;QACT;QAEA,MAAM,EAAEa,IAAI,EAAEC,OAAO,EAAE,GAAGvB,eAAe,CAAC;QAE1C,MAAMY,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACXY;QACF;QAEA,IAAIb,YAAY,CAACa,MACf,MAAM,IAAIR,aAAaL,UAAUG;QAGnC,OAAO;YACLU;YACAC;YACApB;YACAC;YACAQ;QACF;IACF;IAEA,MAAM,SACJY,MAA+B,EAC/BxC,WAAyB,EACzBD,GAEC,EACwD;QACzDI,OAAOqC,QAAQ;QACf,MAAM/B,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAC7C,MAAM,EAAEgC,QAAQ,EAAE,GAAGhC;QACrB,MAAMiC,mBAAmBjC,QAAQ,UAAU,CAAC,MAAM;QAClDN,OAAOuC,kBAAkB;QAEzB,MAAM,EAAErC,WAAW,EAAE,GAAGL;QACxB,MAAM2C,eAAeC;QAGrB,MAAMC,kBAAkB;QACxB,MAAMC,aAAmBC,MAAM,OAAO,CAACP,UACnC;YACE,MAAMQ,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC/C,KAAKG,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC9C,OAAOA;YACP,QAAQA;QACV,IACAL;QAEJ,IAAIS,eAAe,MAAMC,wBAAwB;YAC/C,gBAAgBR;YAChB,MAAMD;YACN,sBAAsB;gBACpB;oBACE,MAAMK;gBACR;aACD;YACD,iBAAiB;QACnB;QAEA,IAAI/C,KAAK,YAAY;YACnB,MAAMW,aAAayC,iBAAiBL,YAAYL;YAMhD9C,MAAM,oCAAoCe;YAC1C,MAAM0C,gBAAgB,MAAMC,WAC1BJ,cACAvC,YACAL,AAAgB,iBAAhBA;YAEF4C,eAAeG,cAAc,WAAW;QAC1C;QAEA,MAAME,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,MAAM,MAAMC,yBAChBF,MACAtD;QAGF,MAAM,EAAEyD,OAAO,EAAE,GAAGF;QACpBpD,OAAO,CAACsD,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1DtD,OAAOsD,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IA7UA,YACEhD,OAA2D,EAC3DV,GAAoB,CACpB;QAPF;QAEA;QAMEI,OAAOM,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMiD,QAAQ,OAAO,CAACjD;QAGlD,IAAI,AAAyB,WAAlBV,KAAK,UACd,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AAgUF"}

package/dist/es/types.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"types.mjs","sources":["../../src/types.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n\nimport {\n  existsSync,\n  mkdirSync,\n  readFileSync,\n  rmSync,\n  writeFileSync,\n} from 'node:fs';\nimport { join } from 'node:path';\nimport type { NodeType } from '@midscene/shared/constants';\nimport type { CreateOpenAIClientFn, TModelConfig } from '@midscene/shared/env';\nimport type {\n  BaseElement,\n  LocateResultElement,\n  Rect,\n  Size,\n} from '@midscene/shared/types';\nimport type { z } from 'zod';\nimport type { TUserPrompt } from './common';\nimport { restoreImageReferences } from './dump/image-restoration';\nimport { ScreenshotItem } from './screenshot-item';\nimport type {\n  DetailedLocateParam,\n  MidsceneYamlFlowItem,\n  ServiceExtractOption,\n} from './yaml';\n\nexport type {\n  ElementTreeNode,\n  BaseElement,\n  Rect,\n  Size,\n  Point,\n} from '@midscene/shared/types';\nexport * from './yaml';\n\nexport type AIUsageInfo = Record<string, any> & {\n  prompt_tokens: number | undefined;\n  completion_tokens: number | undefined;\n  total_tokens: number | undefined;\n  cached_input: number | undefined;\n  time_cost: number | undefined;\n  model_name: string | undefined;\n  model_description: string | undefined;\n  intent: string | undefined;\n  request_id: string | undefined;\n};\n\nexport type { LocateResultElement };\n\nexport type AISingleElementResponseByPosition = {\n  position?: {\n    x: number;\n    y: number;\n  };\n  bbox?: [number, number, number, number];\n  reason: string;\n  text: string;\n};\n\nexport interface AIElementCoordinatesResponse {\n  bbox: [number, number, number, number];\n  errors?: string[];\n}\n\nexport type AIElementResponse = AIElementCoordinatesResponse;\n\nexport interface AIDataExtractionResponse<DataDemand> {\n  data: DataDemand;\n  errors?: string[];\n  thought?: string;\n}\n\nexport interface AISectionLocatorResponse {\n  bbox: [number, number, number, number];\n  references_bbox?: [number, number, number, number][];\n  error?: string;\n}\n\nexport interface AIAssertionResponse {\n  pass: boolean;\n  thought: string;\n}\n\nexport interface AIDescribeElementResponse {\n  description: string;\n  error?: string;\n}\n\nexport interface LocatorValidatorOption {\n  centerDistanceThreshold?: number;\n}\n\nexport interface LocateValidatorResult {\n  pass: boolean;\n  rect: Rect;\n  center: [number, number];\n  centerDistance?: number;\n}\n\nexport interface AgentDescribeElementAtPointResult {\n  prompt: string;\n  deepLocate: boolean;\n  verifyResult?: LocateValidatorResult;\n}\n\n/**\n * context\n */\n\nexport abstract class UIContext {\n  /**\n   * screenshot of the current UI state. which size is shotSize(be shrunk by screenshotShrinkFactor),\n   */\n  abstract screenshot: ScreenshotItem;\n\n  /**\n   * screenshot size after shrinking\n   */\n  abstract shotSize: Size;\n\n  /**\n   * The ratio for converting shrunk screenshot coordinates to logical coordinates.\n   *\n   * Example:\n   * - Physical screen width: 3000px, dpr=6\n   * - Logical width: 500px\n   * - User-defined screenshotShrinkFactor: 2\n   * - Actual shrunk screenshot width: 3000 / 2 = 1500px\n   * - shrunkShotToLogicalRatio: dpr / screenshotShrinkFactor = 6 / 2 = 3\n   * - To map back to logical coordinates: 1500 / shrunkShotToLogicalRatio = 500px\n   */\n  abstract shrunkShotToLogicalRatio: number;\n\n  abstract _isFrozen?: boolean;\n\n  // @deprecated - backward compatibility for aiLocate\n  abstract deprecatedDpr?: number;\n}\n\nexport type EnsureObject<T> = { [K in keyof T]: any };\n\nexport type ServiceAction = 'locate' | 'extract' | 'assert' | 'describe';\n\nexport type ServiceExtractParam = string | Record<string, string>;\n\nexport type ElementCacheFeature = Record<string, unknown>;\n\nexport interface LocateResult {\n  element: LocateResultElement | null;\n  rect?: Rect;\n}\n\nexport type ThinkingLevel = 'off' | 'medium' | 'high';\n\nexport type DeepThinkOption = 'unset' | true | false;\n\nexport interface ServiceTaskInfo {\n  durationMs: number;\n  formatResponse?: string;\n  rawResponse?: string;\n  usage?: AIUsageInfo;\n  searchArea?: Rect;\n  searchAreaRawResponse?: string;\n  searchAreaUsage?: AIUsageInfo;\n  reasoning_content?: string;\n}\n\nexport interface DumpMeta {\n  logTime: number;\n}\n\nexport interface ReportDumpWithAttributes {\n  dumpString: string;\n  attributes?: Record<string, any>;\n}\n\nexport interface ServiceDump extends DumpMeta {\n  type: 'locate' | 'extract' | 'assert';\n  logId: string;\n  userQuery: {\n    element?: TUserPrompt;\n    dataDemand?: ServiceExtractParam;\n    assertion?: TUserPrompt;\n  };\n  matchedElement: LocateResultElement[];\n  matchedRect?: Rect;\n  deepLocate?: boolean;\n  data: any;\n  assertionPass?: boolean;\n  assertionThought?: string;\n  taskInfo: ServiceTaskInfo;\n  error?: string;\n  output?: any;\n}\n\nexport type PartialServiceDumpFromSDK = Omit<\n  ServiceDump,\n  'logTime' | 'logId' | 'model_name'\n>;\n\nexport interface ServiceResultBase {\n  dump: ServiceDump;\n}\n\nexport type LocateResultWithDump = LocateResult & ServiceResultBase;\n\nexport interface ServiceExtractResult<T> extends ServiceResultBase {\n  data: T;\n  thought?: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}\n\nexport class ServiceError extends Error {\n  dump: ServiceDump;\n\n  constructor(message: string, dump: ServiceDump) {\n    super(message);\n    this.name = 'ServiceError';\n    this.dump = dump;\n  }\n}\n\n// intermediate variables to optimize the return value by AI\nexport interface LiteUISection {\n  name: string;\n  description: string;\n  sectionCharacteristics: string;\n  textIds: string[];\n}\n\nexport type ElementById = (id: string) => BaseElement | null;\n\nexport type ServiceAssertionResponse = AIAssertionResponse & {\n  usage?: AIUsageInfo;\n};\n\n/**\n * agent\n */\n\nexport type OnTaskStartTip = (tip: string) => Promise<void> | void;\n\nexport interface AgentWaitForOpt extends ServiceExtractOption {\n  checkIntervalMs?: number;\n  timeoutMs?: number;\n}\n\nexport interface AgentAssertOpt {\n  keepRawResponse?: boolean;\n}\n\n/**\n * planning\n *\n */\n\nexport interface PlanningLocateParam extends DetailedLocateParam {\n  bbox?: [number, number, number, number];\n}\n\nexport interface PlanningAction<ParamType = any> {\n  thought?: string;\n  log?: string; // a brief preamble to the user explaining what you’re about to do\n  type: string;\n  param: ParamType;\n}\n\nexport type SubGoalStatus = 'pending' | 'running' | 'finished';\n\nexport interface SubGoal {\n  index: number;\n  status: SubGoalStatus;\n  description: string;\n  logs?: string[];\n}\n\nexport interface RawResponsePlanningAIResponse {\n  action: PlanningAction;\n  thought?: string;\n  log: string;\n  memory?: string;\n  error?: string;\n  finalizeMessage?: string;\n  finalizeSuccess?: boolean;\n  updateSubGoals?: SubGoal[];\n  markFinishedIndexes?: number[];\n}\n\nexport interface PlanningAIResponse\n  extends Omit<RawResponsePlanningAIResponse, 'action'> {\n  actions?: PlanningAction[];\n  usage?: AIUsageInfo;\n  rawResponse?: string;\n  yamlFlow?: MidsceneYamlFlowItem[];\n  yamlString?: string;\n  error?: string;\n  reasoning_content?: string;\n  shouldContinuePlanning: boolean;\n  output?: string; // Output message from <complete> tag (same as finalizeMessage)\n}\n\nexport interface PlanningActionParamSleep {\n  timeMs: number;\n}\n\nexport interface PlanningActionParamError {\n  thought: string;\n}\n\nexport type PlanningActionParamWaitFor = AgentWaitForOpt & {};\n\nexport interface LongPressParam {\n  duration?: number;\n}\n\nexport interface PullParam {\n  direction: 'up' | 'down';\n  distance?: number;\n  duration?: number;\n}\n/**\n * misc\n */\n\nexport interface Color {\n  name: string;\n  hex: string;\n}\n\nexport interface BaseAgentParserOpt {\n  selector?: string;\n}\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PuppeteerParserOpt extends BaseAgentParserOpt {}\n\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PlaywrightParserOpt extends BaseAgentParserOpt {}\n\n/*\naction\n*/\nexport interface ExecutionTaskProgressOptions {\n  onTaskStart?: (task: ExecutionTask) => Promise<void> | void;\n}\n\nexport interface ExecutionRecorderItem {\n  type: 'screenshot';\n  ts: number;\n  screenshot?: ScreenshotItem;\n  timing?: string;\n}\n\nexport type ExecutionTaskType = 'Planning' | 'Insight' | 'Action Space' | 'Log';\n\nexport interface ExecutorContext {\n  task: ExecutionTask;\n  element?: LocateResultElement | null;\n  uiContext?: UIContext;\n}\n\nexport interface ExecutionTaskApply<\n  Type extends ExecutionTaskType = any,\n  TaskParam = any,\n  TaskOutput = any,\n  TaskLog = any,\n> {\n  type: Type;\n  subType?: string;\n  param?: TaskParam;\n  thought?: string;\n  uiContext?: UIContext;\n  executor: (\n    param: TaskParam,\n    context: ExecutorContext,\n  ) => // biome-ignore lint/suspicious/noConfusingVoidType: void is intentionally allowed as some executors may not return a value\n    | Promise<ExecutionTaskReturn<TaskOutput, TaskLog> | undefined | void>\n    | undefined\n    | void;\n}\n\nexport interface ExecutionTaskHitBy {\n  from: string;\n  context: Record<string, any>;\n}\n\nexport interface ExecutionTaskReturn<TaskOutput = unknown, TaskLog = unknown> {\n  output?: TaskOutput;\n  log?: TaskLog;\n  recorder?: ExecutionRecorderItem[];\n  hitBy?: ExecutionTaskHitBy;\n}\n\nexport type ExecutionTask<\n  E extends ExecutionTaskApply<any, any, any> = ExecutionTaskApply<\n    any,\n    any,\n    any\n  >,\n> = E &\n  ExecutionTaskReturn<\n    E extends ExecutionTaskApply<any, any, infer TaskOutput, any>\n      ? TaskOutput\n      : unknown,\n    E extends ExecutionTaskApply<any, any, any, infer TaskLog>\n      ? TaskLog\n      : unknown\n  > & {\n    taskId: string;\n    status: 'pending' | 'running' | 'finished' | 'failed' | 'cancelled';\n    error?: Error;\n    errorMessage?: string;\n    errorStack?: string;\n    timing?: {\n      start: number;\n      getUiContextStart?: number;\n      getUiContextEnd?: number;\n      callAiStart?: number;\n      callAiEnd?: number;\n      beforeInvokeActionHookStart?: number;\n      beforeInvokeActionHookEnd?: number;\n      callActionStart?: number;\n      callActionEnd?: number;\n      afterInvokeActionHookStart?: number;\n      afterInvokeActionHookEnd?: number;\n      captureAfterCallingSnapshotStart?: number;\n      captureAfterCallingSnapshotEnd?: number;\n      end?: number;\n      cost?: number;\n    };\n    usage?: AIUsageInfo;\n    searchAreaUsage?: AIUsageInfo;\n    reasoning_content?: string;\n  };\n\nexport interface IExecutionDump extends DumpMeta {\n  name: string;\n  description?: string;\n  tasks: ExecutionTask[];\n  aiActContext?: string;\n}\n\n/**\n * Replacer function for JSON serialization that handles Page, Browser objects and ScreenshotItem\n */\nfunction replacerForDumpSerialization(_key: string, value: any): any {\n  if (value && value.constructor?.name === 'Page') {\n    return '[Page object]';\n  }\n  if (value && value.constructor?.name === 'Browser') {\n    return '[Browser object]';\n  }\n  // Handle ScreenshotItem serialization\n  if (value && typeof value.toSerializable === 'function') {\n    return value.toSerializable();\n  }\n  return value;\n}\n\n/**\n * Reviver function for JSON deserialization that handles ScreenshotItem formats.\n *\n * BEHAVIOR:\n * - For { $screenshot: \"id\" } format: Left as-is (plain object)\n *   Consumer must use imageMap to restore base64 data\n * - For { base64: \"...\" } format: Creates ScreenshotItem from base64 data\n *\n * @param key - JSON key being processed\n * @param value - JSON value being processed\n * @returns Restored value\n */\nfunction reviverForDumpDeserialization(key: string, value: any): any {\n  // Only process screenshot fields\n  if (key !== 'screenshot' || typeof value !== 'object' || value === null) {\n    return value;\n  }\n\n  // Handle serialized format: { $screenshot: \"id\" }\n  // Leave as plain object — consumer uses imageMap to restore\n  if (ScreenshotItem.isSerialized(value)) {\n    return value;\n  }\n\n  // Handle inline base64 format: { base64: \"...\" }\n  if ('base64' in value && typeof value.base64 === 'string') {\n    return value;\n  }\n\n  return value;\n}\n\n/**\n * ExecutionDump class for serializing and deserializing execution dumps\n */\nexport class ExecutionDump implements IExecutionDump {\n  logTime: number;\n  name: string;\n  description?: string;\n  tasks: ExecutionTask[];\n  aiActContext?: string;\n\n  constructor(data: IExecutionDump) {\n    this.logTime = data.logTime;\n    this.name = data.name;\n    this.description = data.description;\n    this.tasks = data.tasks;\n    this.aiActContext = data.aiActContext;\n  }\n\n  /**\n   * Serialize the ExecutionDump to a JSON string\n   */\n  serialize(indents?: number): string {\n    return JSON.stringify(this.toJSON(), replacerForDumpSerialization, indents);\n  }\n\n  /**\n   * Convert to a plain object for JSON serialization\n   */\n  toJSON(): IExecutionDump {\n    return {\n      logTime: this.logTime,\n      name: this.name,\n      description: this.description,\n      tasks: this.tasks.map((task) => ({\n        ...task,\n        recorder: task.recorder || [],\n      })),\n      aiActContext: this.aiActContext,\n    };\n  }\n\n  /**\n   * Create an ExecutionDump instance from a serialized JSON string\n   */\n  static fromSerializedString(serialized: string): ExecutionDump {\n    const parsed = JSON.parse(\n      serialized,\n      reviverForDumpDeserialization,\n    ) as IExecutionDump;\n    return new ExecutionDump(parsed);\n  }\n\n  /**\n   * Create an ExecutionDump instance from a plain object\n   */\n  static fromJSON(data: IExecutionDump): ExecutionDump {\n    return new ExecutionDump(data);\n  }\n\n  /**\n   * Collect all ScreenshotItem instances from tasks.\n   * Scans through uiContext and recorder items to find screenshots.\n   *\n   * @returns Array of ScreenshotItem instances\n   */\n  collectScreenshots(): ScreenshotItem[] {\n    const screenshots: ScreenshotItem[] = [];\n\n    for (const task of this.tasks) {\n      // Collect uiContext.screenshot if present\n      if (task.uiContext?.screenshot instanceof ScreenshotItem) {\n        screenshots.push(task.uiContext.screenshot);\n      }\n\n      // Collect recorder screenshots\n      if (task.recorder) {\n        for (const record of task.recorder) {\n          if (record.screenshot instanceof ScreenshotItem) {\n            screenshots.push(record.screenshot);\n          }\n        }\n      }\n    }\n\n    return screenshots;\n  }\n}\n\n/*\ntask - service-locate\n*/\nexport type ExecutionTaskInsightLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskInsightLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport type ExecutionTaskInsightDump = ServiceDump;\n\nexport type ExecutionTaskInsightLocateApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightLocateParam,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightLocate =\n  ExecutionTask<ExecutionTaskInsightLocateApply>;\n\n/*\ntask - service-query\n*/\nexport interface ExecutionTaskInsightQueryParam {\n  dataDemand: ServiceExtractParam;\n}\n\nexport interface ExecutionTaskInsightQueryOutput {\n  data: any;\n}\n\nexport type ExecutionTaskInsightQueryApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightQueryParam,\n  any,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightQuery =\n  ExecutionTask<ExecutionTaskInsightQueryApply>;\n\n/*\ntask - assertion\n*/\nexport interface ExecutionTaskInsightAssertionParam {\n  assertion: string;\n}\n\nexport type ExecutionTaskInsightAssertionApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightAssertionParam,\n  ServiceAssertionResponse,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightAssertion =\n  ExecutionTask<ExecutionTaskInsightAssertionApply>;\n\n/*\ntask - action (i.e. interact) \n*/\nexport type ExecutionTaskActionApply<ActionParam = any> = ExecutionTaskApply<\n  'Action Space',\n  ActionParam,\n  void,\n  void\n>;\n\nexport type ExecutionTaskAction = ExecutionTask<ExecutionTaskActionApply>;\n\n/*\ntask - Log\n*/\n\nexport type ExecutionTaskLogApply<\n  LogParam = {\n    content: string;\n  },\n> = ExecutionTaskApply<'Log', LogParam, void, void>;\n\nexport type ExecutionTaskLog = ExecutionTask<ExecutionTaskLogApply>;\n\n/*\ntask - planning\n*/\n\nexport type ExecutionTaskPlanningApply = ExecutionTaskApply<\n  'Planning',\n  {\n    userInstruction: string;\n    aiActContext?: string;\n  },\n  PlanningAIResponse\n>;\n\nexport type ExecutionTaskPlanning = ExecutionTask<ExecutionTaskPlanningApply>;\n\n/*\ntask - planning-locate\n*/\nexport type ExecutionTaskPlanningLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskPlanningLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport type ExecutionTaskPlanningDump = ServiceDump;\n\nexport type ExecutionTaskPlanningLocateApply = ExecutionTaskApply<\n  'Planning',\n  ExecutionTaskPlanningLocateParam,\n  ExecutionTaskPlanningLocateOutput,\n  ExecutionTaskPlanningDump\n>;\n\nexport type ExecutionTaskPlanningLocate =\n  ExecutionTask<ExecutionTaskPlanningLocateApply>;\n\n/*\nGrouped dump\n*/\nexport interface IGroupedActionDump {\n  sdkVersion: string;\n  groupName: string;\n  groupDescription?: string;\n  modelBriefs: string[];\n  executions: IExecutionDump[];\n  deviceType?: string;\n}\n\n/**\n * GroupedActionDump class for serializing and deserializing grouped action dumps\n */\nexport class GroupedActionDump implements IGroupedActionDump {\n  sdkVersion: string;\n  groupName: string;\n  groupDescription?: string;\n  modelBriefs: string[];\n  executions: ExecutionDump[];\n  deviceType?: string;\n\n  constructor(data: IGroupedActionDump) {\n    this.sdkVersion = data.sdkVersion;\n    this.groupName = data.groupName;\n    this.groupDescription = data.groupDescription;\n    this.modelBriefs = data.modelBriefs;\n    this.executions = data.executions.map((exec) =>\n      exec instanceof ExecutionDump ? exec : ExecutionDump.fromJSON(exec),\n    );\n    this.deviceType = data.deviceType;\n  }\n\n  /**\n   * Serialize the GroupedActionDump to a JSON string\n   * Uses compact { $screenshot: id } format\n   */\n  serialize(indents?: number): string {\n    return JSON.stringify(this.toJSON(), replacerForDumpSerialization, indents);\n  }\n\n  /**\n   * Serialize the GroupedActionDump with inline screenshots to a JSON string.\n   * Each ScreenshotItem is replaced with { base64: \"...\", capturedAt }.\n   */\n  serializeWithInlineScreenshots(indents?: number): string {\n    const processValue = (obj: unknown): unknown => {\n      if (obj instanceof ScreenshotItem) {\n        return { base64: obj.base64, capturedAt: obj.capturedAt };\n      }\n      if (Array.isArray(obj)) {\n        return obj.map(processValue);\n      }\n      if (obj && typeof obj === 'object') {\n        const entries = Object.entries(obj).map(([key, value]) => [\n          key,\n          processValue(value),\n        ]);\n        return Object.fromEntries(entries);\n      }\n      return obj;\n    };\n\n    const data = processValue(this.toJSON());\n    return JSON.stringify(data, null, indents);\n  }\n\n  /**\n   * Convert to a plain object for JSON serialization\n   */\n  toJSON(): IGroupedActionDump {\n    return {\n      sdkVersion: this.sdkVersion,\n      groupName: this.groupName,\n      groupDescription: this.groupDescription,\n      modelBriefs: this.modelBriefs,\n      executions: this.executions.map((exec) => exec.toJSON()),\n      deviceType: this.deviceType,\n    };\n  }\n\n  /**\n   * Create a GroupedActionDump instance from a serialized JSON string\n   */\n  static fromSerializedString(serialized: string): GroupedActionDump {\n    const parsed = JSON.parse(\n      serialized,\n      reviverForDumpDeserialization,\n    ) as IGroupedActionDump;\n    return new GroupedActionDump(parsed);\n  }\n\n  /**\n   * Create a GroupedActionDump instance from a plain object\n   */\n  static fromJSON(data: IGroupedActionDump): GroupedActionDump {\n    return new GroupedActionDump(data);\n  }\n\n  /**\n   * Collect all ScreenshotItem instances from all executions.\n   *\n   * @returns Array of all ScreenshotItem instances across all executions\n   */\n  collectAllScreenshots(): ScreenshotItem[] {\n    const screenshots: ScreenshotItem[] = [];\n    for (const execution of this.executions) {\n      screenshots.push(...execution.collectScreenshots());\n    }\n    return screenshots;\n  }\n\n  /**\n   * Serialize the dump to files with screenshots as separate PNG files.\n   * Creates:\n   * - {basePath} - dump JSON with { $screenshot: id } references\n   * - {basePath}.screenshots/ - PNG files\n   * - {basePath}.screenshots.json - ID to path mapping\n   *\n   * @param basePath - Base path for the dump file\n   */\n  serializeToFiles(basePath: string): void {\n    const screenshotsDir = `${basePath}.screenshots`;\n    if (!existsSync(screenshotsDir)) {\n      mkdirSync(screenshotsDir, { recursive: true });\n    }\n\n    // Write screenshots to separate files\n    const screenshotMap: Record<string, string> = {};\n    const screenshots = this.collectAllScreenshots();\n\n    for (const screenshot of screenshots) {\n      if (screenshot.hasBase64()) {\n        const imagePath = join(\n          screenshotsDir,\n          `${screenshot.id}.${screenshot.extension}`,\n        );\n        const rawBase64 = screenshot.rawBase64;\n        writeFileSync(imagePath, Buffer.from(rawBase64, 'base64'));\n        screenshotMap[screenshot.id] = imagePath;\n      }\n    }\n\n    // Write screenshot map file\n    writeFileSync(\n      `${basePath}.screenshots.json`,\n      JSON.stringify(screenshotMap),\n      'utf-8',\n    );\n\n    // Write dump JSON with references\n    writeFileSync(basePath, this.serialize(), 'utf-8');\n  }\n\n  /**\n   * Read dump from files and return JSON string with inline screenshots.\n   * Reads the dump JSON and screenshot files, then inlines the base64 data.\n   *\n   * @param basePath - Base path for the dump file\n   * @returns JSON string with inline screenshots ({ base64: \"...\" } format)\n   */\n  static fromFilesAsInlineJson(basePath: string): string {\n    const dumpString = readFileSync(basePath, 'utf-8');\n    const screenshotsMapPath = `${basePath}.screenshots.json`;\n\n    if (!existsSync(screenshotsMapPath)) {\n      return dumpString;\n    }\n\n    // Read screenshot map and build imageMap from files\n    const screenshotMap: Record<string, string> = JSON.parse(\n      readFileSync(screenshotsMapPath, 'utf-8'),\n    );\n\n    const imageMap: Record<string, string> = {};\n    for (const [id, filePath] of Object.entries(screenshotMap)) {\n      if (existsSync(filePath)) {\n        const data = readFileSync(filePath);\n        const mime =\n          filePath.endsWith('.jpeg') || filePath.endsWith('.jpg')\n            ? 'jpeg'\n            : 'png';\n        imageMap[id] = `data:image/${mime};base64,${data.toString('base64')}`;\n      }\n    }\n\n    // Restore image references\n    const dumpData = JSON.parse(dumpString);\n    const processedData = restoreImageReferences(\n      dumpData,\n      (id) => imageMap[id] ?? '',\n    );\n    return JSON.stringify(processedData);\n  }\n\n  /**\n   * Clean up all files associated with a serialized dump.\n   *\n   * @param basePath - Base path for the dump file\n   */\n  static cleanupFiles(basePath: string): void {\n    const filesToClean = [\n      basePath,\n      `${basePath}.screenshots.json`,\n      `${basePath}.screenshots`,\n    ];\n\n    for (const filePath of filesToClean) {\n      try {\n        rmSync(filePath, { force: true, recursive: true });\n      } catch {\n        // Ignore errors - file may already be deleted\n      }\n    }\n  }\n\n  /**\n   * Get all file paths associated with a serialized dump.\n   *\n   * @param basePath - Base path for the dump file\n   * @returns Array of all associated file paths\n   */\n  static getFilePaths(basePath: string): string[] {\n    return [\n      basePath,\n      `${basePath}.screenshots.json`,\n      `${basePath}.screenshots`,\n    ];\n  }\n}\n\nexport type InterfaceType =\n  | 'puppeteer'\n  | 'playwright'\n  | 'static'\n  | 'chrome-extension-proxy'\n  | 'android'\n  | string;\n\nexport interface StreamingCodeGenerationOptions {\n  /** Whether to enable streaming output */\n  stream?: boolean;\n  /** Callback function to handle streaming chunks */\n  onChunk?: StreamingCallback;\n  /** Callback function to handle streaming completion */\n  onComplete?: (finalCode: string) => void;\n  /** Callback function to handle streaming errors */\n  onError?: (error: Error) => void;\n}\n\nexport type StreamingCallback = (chunk: CodeGenerationChunk) => void;\n\nexport interface CodeGenerationChunk {\n  /** The incremental content chunk */\n  content: string;\n  /** The reasoning content */\n  reasoning_content: string;\n  /** The accumulated content so far */\n  accumulated: string;\n  /** Whether this is the final chunk */\n  isComplete: boolean;\n  /** Token usage information if available */\n  usage?: AIUsageInfo;\n}\n\nexport interface StreamingAIResponse {\n  /** The final accumulated content */\n  content: string;\n  /** Token usage information */\n  usage?: AIUsageInfo;\n  /** Whether the response was streamed */\n  isStreamed: boolean;\n}\n\nexport interface DeviceAction<TParam = any, TReturn = any> {\n  name: string;\n  description?: string;\n  interfaceAlias?: string;\n  paramSchema?: z.ZodType<TParam>;\n  call: (param: TParam, context: ExecutorContext) => Promise<TReturn> | TReturn;\n  delayAfterRunner?: number;\n  /**\n   * An example param object for this action.\n   * Locate fields with { prompt } will automatically get bbox injected when needed.\n   */\n  sample?: { [K in keyof TParam]?: any };\n}\n\n/**\n * Type utilities for extracting types from DeviceAction definitions\n */\n\n/**\n * Extract parameter type from a DeviceAction\n */\nexport type ActionParam<Action extends DeviceAction<any, any>> =\n  Action extends DeviceAction<infer P, any> ? P : never;\n\n/**\n * Extract return type from a DeviceAction\n */\nexport type ActionReturn<Action extends DeviceAction<any, any>> =\n  Action extends DeviceAction<any, infer R> ? R : never;\n\n/**\n * Web-specific types\n */\nexport interface WebElementInfo extends BaseElement {\n  id: string;\n  attributes: {\n    nodeType: NodeType;\n    [key: string]: string;\n  };\n}\n\n/**\n * Agent\n */\n\nexport type CacheConfig = {\n  strategy?: 'read-only' | 'read-write' | 'write-only';\n  id: string;\n  cacheAdapter?: import('./agent/cache-adapter').CacheAdapter;\n};\n\nexport type Cache =\n  | false // No read, no write\n  | true // Will throw error at runtime - deprecated\n  | CacheConfig; // Object configuration (requires explicit id)\n\nexport interface AgentOpt {\n  testId?: string;\n  // @deprecated\n  cacheId?: string; // Keep backward compatibility, but marked as deprecated\n  groupName?: string;\n  groupDescription?: string;\n  /* if auto generate report, default true */\n  generateReport?: boolean;\n  /* if auto print report msg, default true */\n  autoPrintReportMsg?: boolean;\n\n  /**\n   * Use directory-based report format with separate image files.\n   *\n   * When enabled:\n   * - Screenshots are saved as PNG files in a `screenshots/` subdirectory\n   * - Report is generated as `index.html` with relative image paths\n   * - Reduces memory usage and report file size\n   *\n   * IMPORTANT: 'html-and-external-assets' reports must be served via HTTP server\n   * (e.g., `npx serve ./report-dir`). The file:// protocol will not\n   * work due to browser CORS restrictions.\n   *\n   * @default 'single-html'\n   */\n  outputFormat?: 'single-html' | 'html-and-external-assets';\n\n  onTaskStartTip?: OnTaskStartTip;\n  aiActContext?: string;\n  aiActionContext?: string;\n  /* custom report file name */\n  reportFileName?: string;\n  modelConfig?: TModelConfig;\n  cache?: Cache;\n  /**\n   * Maximum number of replanning cycles for aiAct.\n   * Defaults to 20 (40 for `vlm-ui-tars`) when not provided.\n   * If omitted, the agent will also read `MIDSCENE_REPLANNING_CYCLE_LIMIT` for backward compatibility.\n   */\n  replanningCycleLimit?: number;\n\n  /**\n   * Wait time in milliseconds after each action execution.\n   * This allows the UI to settle and stabilize before the next action.\n   * Defaults to 300ms when not provided.\n   */\n  waitAfterAction?: number;\n\n  /**\n   * When set to true, Midscene will use the target device's time (Android/iOS)\n   * instead of the system time. Useful when the device time differs from the\n   * host machine. Default: false\n   */\n  useDeviceTimestamp?: boolean;\n\n  /**\n   * Custom screenshot shrink factor to reduce AI token usage.\n   * When set, the screenshot will be scaled down by this factor from the physical resolution.\n   *\n   * Example:\n   * - Physical screen width: 3000px, dpr=6\n   * - Logical width: 500px\n   * - screenshotShrinkFactor: 2\n   * - Actual shrunk screenshot width: 3000 / 2 = 1500px\n   * - AI analyzes the 1500px screenshot\n   * - Coordinates are transformed back to logical (500px) before actions execute\n   *\n   * Benefits:\n   * - Reduces token usage for high-resolution screenshots\n   * - Maintains accuracy by scaling coordinates appropriately\n   *\n   * Must be >= 1 (shrinking only, enlarging is not supported).\n   *\n   * @default 1 (no shrinking, uses original physical screenshot)\n   */\n  screenshotShrinkFactor?: number;\n\n  /**\n   * Custom OpenAI client factory function\n   *\n   * If provided, this function will be called to create OpenAI client instances\n   * for each AI call, allowing you to:\n   * - Wrap clients with observability tools (langsmith, langfuse)\n   * - Use custom OpenAI-compatible clients\n   * - Apply different configurations based on intent\n   *\n   * @param config - Resolved model configuration\n   * @returns OpenAI client instance (original or wrapped)\n   *\n   * @example\n   * ```typescript\n   * createOpenAIClient: async (openai, opts) => {\n   *   // Wrap with langsmith for planning tasks\n   *   if (opts.baseURL?.includes('planning')) {\n   *     return wrapOpenAI(openai, { metadata: { task: 'planning' } });\n   *   }\n   *\n   *   return openai;\n   * }\n   * ```\n   */\n  createOpenAIClient?: CreateOpenAIClientFn;\n}\n\nexport type TestStatus =\n  | 'passed'\n  | 'failed'\n  | 'timedOut'\n  | 'skipped'\n  | 'interrupted';\n\nexport interface ReportFileWithAttributes {\n  reportFilePath: string;\n  reportAttributes: {\n    testDuration: number;\n    testStatus: TestStatus;\n    testTitle: string;\n    testId: string;\n    testDescription: string;\n  };\n}\n"],"names":["UIContext","ServiceError","Error","message","dump","replacerForDumpSerialization","_key","value","reviverForDumpDeserialization","key","ScreenshotItem","ExecutionDump","indents","JSON","task","serialized","parsed","data","screenshots","record","GroupedActionDump","processValue","obj","Array","entries","Object","exec","execution","basePath","screenshotsDir","existsSync","mkdirSync","screenshotMap","screenshot","imagePath","join","rawBase64","writeFileSync","Buffer","dumpString","readFileSync","screenshotsMapPath","imageMap","id","filePath","mime","dumpData","processedData","restoreImageReferences","filesToClean","rmSync"],"mappings":";;;;;AAAqD;;;;;;;;;;AA+G9C,MAAeA;AA4BtB;AA4EO,MAAMC,qBAAqBC;IAGhC,YAAYC,OAAe,EAAEC,IAAiB,CAAE;QAC9C,KAAK,CAACD,UAHR;QAIE,IAAI,CAAC,IAAI,GAAG;QACZ,IAAI,CAAC,IAAI,GAAGC;IACd;AACF;AAgOA,SAASC,6BAA6BC,IAAY,EAAEC,KAAU;IAC5D,IAAIA,SAASA,MAAM,WAAW,EAAE,SAAS,QACvC,OAAO;IAET,IAAIA,SAASA,MAAM,WAAW,EAAE,SAAS,WACvC,OAAO;IAGT,IAAIA,SAAS,AAAgC,cAAhC,OAAOA,MAAM,cAAc,EACtC,OAAOA,MAAM,cAAc;IAE7B,OAAOA;AACT;AAcA,SAASC,8BAA8BC,GAAW,EAAEF,KAAU;IAE5D,IAAIE,AAAQ,iBAARA,OAAwB,AAAiB,YAAjB,OAAOF,SAAsBA,AAAU,SAAVA,OACvD,OAAOA;IAKT,IAAIG,eAAe,YAAY,CAACH,QAC9B,OAAOA;IAIL,YAAYA,SAAgBA,MAAM,MAAM;IAI5C,OAAOA;AACT;AAKO,MAAMI;IAkBX,UAAUC,OAAgB,EAAU;QAClC,OAAOC,KAAK,SAAS,CAAC,IAAI,CAAC,MAAM,IAAIR,8BAA8BO;IACrE;IAKA,SAAyB;QACvB,OAAO;YACL,SAAS,IAAI,CAAC,OAAO;YACrB,MAAM,IAAI,CAAC,IAAI;YACf,aAAa,IAAI,CAAC,WAAW;YAC7B,OAAO,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,CAACE,OAAU;oBAC/B,GAAGA,IAAI;oBACP,UAAUA,KAAK,QAAQ,IAAI,EAAE;gBAC/B;YACA,cAAc,IAAI,CAAC,YAAY;QACjC;IACF;IAKA,OAAO,qBAAqBC,UAAkB,EAAiB;QAC7D,MAAMC,SAASH,KAAK,KAAK,CACvBE,YACAP;QAEF,OAAO,IAAIG,cAAcK;IAC3B;IAKA,OAAO,SAASC,IAAoB,EAAiB;QACnD,OAAO,IAAIN,cAAcM;IAC3B;IAQA,qBAAuC;QACrC,MAAMC,cAAgC,EAAE;QAExC,KAAK,MAAMJ,QAAQ,IAAI,CAAC,KAAK,CAAE;YAE7B,IAAIA,KAAK,SAAS,EAAE,sBAAsBJ,gBACxCQ,YAAY,IAAI,CAACJ,KAAK,SAAS,CAAC,UAAU;YAI5C,IAAIA,KAAK,QAAQ,EACf;gBAAA,KAAK,MAAMK,UAAUL,KAAK,QAAQ,CAChC,IAAIK,OAAO,UAAU,YAAYT,gBAC/BQ,YAAY,IAAI,CAACC,OAAO,UAAU;YAEtC;QAEJ;QAEA,OAAOD;IACT;IA3EA,YAAYD,IAAoB,CAAE;QANlC;QACA;QACA;QACA;QACA;QAGE,IAAI,CAAC,OAAO,GAAGA,KAAK,OAAO;QAC3B,IAAI,CAAC,IAAI,GAAGA,KAAK,IAAI;QACrB,IAAI,CAAC,WAAW,GAAGA,KAAK,WAAW;QACnC,IAAI,CAAC,KAAK,GAAGA,KAAK,KAAK;QACvB,IAAI,CAAC,YAAY,GAAGA,KAAK,YAAY;IACvC;AAsEF;AAwIO,MAAMG;IAuBX,UAAUR,OAAgB,EAAU;QAClC,OAAOC,KAAK,SAAS,CAAC,IAAI,CAAC,MAAM,IAAIR,8BAA8BO;IACrE;IAMA,+BAA+BA,OAAgB,EAAU;QACvD,MAAMS,eAAe,CAACC;YACpB,IAAIA,eAAeZ,gBACjB,OAAO;gBAAE,QAAQY,IAAI,MAAM;gBAAE,YAAYA,IAAI,UAAU;YAAC;YAE1D,IAAIC,MAAM,OAAO,CAACD,MAChB,OAAOA,IAAI,GAAG,CAACD;YAEjB,IAAIC,OAAO,AAAe,YAAf,OAAOA,KAAkB;gBAClC,MAAME,UAAUC,OAAO,OAAO,CAACH,KAAK,GAAG,CAAC,CAAC,CAACb,KAAKF,MAAM,GAAK;wBACxDE;wBACAY,aAAad;qBACd;gBACD,OAAOkB,OAAO,WAAW,CAACD;YAC5B;YACA,OAAOF;QACT;QAEA,MAAML,OAAOI,aAAa,IAAI,CAAC,MAAM;QACrC,OAAOR,KAAK,SAAS,CAACI,MAAM,MAAML;IACpC;IAKA,SAA6B;QAC3B,OAAO;YACL,YAAY,IAAI,CAAC,UAAU;YAC3B,WAAW,IAAI,CAAC,SAAS;YACzB,kBAAkB,IAAI,CAAC,gBAAgB;YACvC,aAAa,IAAI,CAAC,WAAW;YAC7B,YAAY,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,CAACc,OAASA,KAAK,MAAM;YACrD,YAAY,IAAI,CAAC,UAAU;QAC7B;IACF;IAKA,OAAO,qBAAqBX,UAAkB,EAAqB;QACjE,MAAMC,SAASH,KAAK,KAAK,CACvBE,YACAP;QAEF,OAAO,IAAIY,kBAAkBJ;IAC/B;IAKA,OAAO,SAASC,IAAwB,EAAqB;QAC3D,OAAO,IAAIG,kBAAkBH;IAC/B;IAOA,wBAA0C;QACxC,MAAMC,cAAgC,EAAE;QACxC,KAAK,MAAMS,aAAa,IAAI,CAAC,UAAU,CACrCT,YAAY,IAAI,IAAIS,UAAU,kBAAkB;QAElD,OAAOT;IACT;IAWA,iBAAiBU,QAAgB,EAAQ;QACvC,MAAMC,iBAAiB,GAAGD,SAAS,YAAY,CAAC;QAChD,IAAI,CAACE,WAAWD,iBACdE,UAAUF,gBAAgB;YAAE,WAAW;QAAK;QAI9C,MAAMG,gBAAwC,CAAC;QAC/C,MAAMd,cAAc,IAAI,CAAC,qBAAqB;QAE9C,KAAK,MAAMe,cAAcf,YACvB,IAAIe,WAAW,SAAS,IAAI;YAC1B,MAAMC,YAAYC,KAChBN,gBACA,GAAGI,WAAW,EAAE,CAAC,CAAC,EAAEA,WAAW,SAAS,EAAE;YAE5C,MAAMG,YAAYH,WAAW,SAAS;YACtCI,cAAcH,WAAWI,OAAO,IAAI,CAACF,WAAW;YAChDJ,aAAa,CAACC,WAAW,EAAE,CAAC,GAAGC;QACjC;QAIFG,cACE,GAAGT,SAAS,iBAAiB,CAAC,EAC9Bf,KAAK,SAAS,CAACmB,gBACf;QAIFK,cAAcT,UAAU,IAAI,CAAC,SAAS,IAAI;IAC5C;IASA,OAAO,sBAAsBA,QAAgB,EAAU;QACrD,MAAMW,aAAaC,aAAaZ,UAAU;QAC1C,MAAMa,qBAAqB,GAAGb,SAAS,iBAAiB,CAAC;QAEzD,IAAI,CAACE,WAAWW,qBACd,OAAOF;QAIT,MAAMP,gBAAwCnB,KAAK,KAAK,CACtD2B,aAAaC,oBAAoB;QAGnC,MAAMC,WAAmC,CAAC;QAC1C,KAAK,MAAM,CAACC,IAAIC,SAAS,IAAInB,OAAO,OAAO,CAACO,eAC1C,IAAIF,WAAWc,WAAW;YACxB,MAAM3B,OAAOuB,aAAaI;YAC1B,MAAMC,OACJD,SAAS,QAAQ,CAAC,YAAYA,SAAS,QAAQ,CAAC,UAC5C,SACA;YACNF,QAAQ,CAACC,GAAG,GAAG,CAAC,WAAW,EAAEE,KAAK,QAAQ,EAAE5B,KAAK,QAAQ,CAAC,WAAW;QACvE;QAIF,MAAM6B,WAAWjC,KAAK,KAAK,CAAC0B;QAC5B,MAAMQ,gBAAgBC,uBACpBF,UACA,CAACH,KAAOD,QAAQ,CAACC,GAAG,IAAI;QAE1B,OAAO9B,KAAK,SAAS,CAACkC;IACxB;IAOA,OAAO,aAAanB,QAAgB,EAAQ;QAC1C,MAAMqB,eAAe;YACnBrB;YACA,GAAGA,SAAS,iBAAiB,CAAC;YAC9B,GAAGA,SAAS,YAAY,CAAC;SAC1B;QAED,KAAK,MAAMgB,YAAYK,aACrB,IAAI;YACFC,OAAON,UAAU;gBAAE,OAAO;gBAAM,WAAW;YAAK;QAClD,EAAE,OAAM,CAER;IAEJ;IAQA,OAAO,aAAahB,QAAgB,EAAY;QAC9C,OAAO;YACLA;YACA,GAAGA,SAAS,iBAAiB,CAAC;YAC9B,GAAGA,SAAS,YAAY,CAAC;SAC1B;IACH;IA9MA,YAAYX,IAAwB,CAAE;QAPtC;QACA;QACA;QACA;QACA;QACA;QAGE,IAAI,CAAC,UAAU,GAAGA,KAAK,UAAU;QACjC,IAAI,CAAC,SAAS,GAAGA,KAAK,SAAS;QAC/B,IAAI,CAAC,gBAAgB,GAAGA,KAAK,gBAAgB;QAC7C,IAAI,CAAC,WAAW,GAAGA,KAAK,WAAW;QACnC,IAAI,CAAC,UAAU,GAAGA,KAAK,UAAU,CAAC,GAAG,CAAC,CAACS,OACrCA,gBAAgBf,gBAAgBe,OAAOf,cAAc,QAAQ,CAACe;QAEhE,IAAI,CAAC,UAAU,GAAGT,KAAK,UAAU;IACnC;AAsMF"}
+{"version":3,"file":"types.mjs","sources":["../../src/types.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n\nimport {\n  existsSync,\n  mkdirSync,\n  readFileSync,\n  rmSync,\n  writeFileSync,\n} from 'node:fs';\nimport { join } from 'node:path';\nimport type { NodeType } from '@midscene/shared/constants';\nimport type { CreateOpenAIClientFn, TModelConfig } from '@midscene/shared/env';\nimport type {\n  BaseElement,\n  LocateResultElement,\n  Rect,\n  Size,\n} from '@midscene/shared/types';\nimport type { z } from 'zod';\nimport type { TUserPrompt } from './common';\nimport { restoreImageReferences } from './dump/image-restoration';\nimport { ScreenshotItem } from './screenshot-item';\nimport type {\n  DetailedLocateParam,\n  MidsceneYamlFlowItem,\n  ServiceExtractOption,\n} from './yaml';\n\nexport type {\n  ElementTreeNode,\n  BaseElement,\n  Rect,\n  Size,\n  Point,\n} from '@midscene/shared/types';\nexport * from './yaml';\n\nexport type AIUsageInfo = Record<string, any> & {\n  prompt_tokens: number | undefined;\n  completion_tokens: number | undefined;\n  total_tokens: number | undefined;\n  cached_input: number | undefined;\n  time_cost: number | undefined;\n  model_name: string | undefined;\n  model_description: string | undefined;\n  intent: string | undefined;\n  request_id: string | undefined;\n};\n\nexport type { LocateResultElement };\n\nexport type AISingleElementResponseByPosition = {\n  position?: {\n    x: number;\n    y: number;\n  };\n  bbox?: [number, number, number, number];\n  reason: string;\n  text: string;\n};\n\nexport interface AIElementCoordinatesResponse {\n  bbox: [number, number, number, number];\n  errors?: string[];\n}\n\nexport type AIElementResponse = AIElementCoordinatesResponse;\n\nexport interface AIDataExtractionResponse<DataDemand> {\n  data: DataDemand;\n  errors?: string[];\n  thought?: string;\n}\n\nexport interface AISectionLocatorResponse {\n  bbox: [number, number, number, number];\n  references_bbox?: [number, number, number, number][];\n  error?: string;\n}\n\nexport interface AIAssertionResponse {\n  pass: boolean;\n  thought: string;\n}\n\nexport interface AIDescribeElementResponse {\n  description: string;\n  error?: string;\n}\n\nexport interface LocatorValidatorOption {\n  centerDistanceThreshold?: number;\n}\n\nexport interface LocateValidatorResult {\n  pass: boolean;\n  rect: Rect;\n  center: [number, number];\n  centerDistance?: number;\n}\n\nexport interface AgentDescribeElementAtPointResult {\n  prompt: string;\n  deepLocate: boolean;\n  verifyResult?: LocateValidatorResult;\n}\n\n/**\n * context\n */\n\nexport abstract class UIContext {\n  /**\n   * screenshot of the current UI state. which size is shotSize(be shrunk by screenshotShrinkFactor),\n   */\n  abstract screenshot: ScreenshotItem;\n\n  /**\n   * screenshot size after shrinking\n   */\n  abstract shotSize: Size;\n\n  /**\n   * The ratio for converting shrunk screenshot coordinates to logical coordinates.\n   *\n   * Example:\n   * - Physical screen width: 3000px, dpr=6\n   * - Logical width: 500px\n   * - User-defined screenshotShrinkFactor: 2\n   * - Actual shrunk screenshot width: 3000 / 2 = 1500px\n   * - shrunkShotToLogicalRatio: dpr / screenshotShrinkFactor = 6 / 2 = 3\n   * - To map back to logical coordinates: 1500 / shrunkShotToLogicalRatio = 500px\n   */\n  abstract shrunkShotToLogicalRatio: number;\n\n  abstract _isFrozen?: boolean;\n\n  // @deprecated - backward compatibility for aiLocate\n  abstract deprecatedDpr?: number;\n}\n\nexport type EnsureObject<T> = { [K in keyof T]: any };\n\nexport type ServiceAction = 'locate' | 'extract' | 'assert' | 'describe';\n\nexport type ServiceExtractParam = string | Record<string, string>;\n\nexport type ElementCacheFeature = Record<string, unknown>;\n\nexport interface LocateResult {\n  element: LocateResultElement | null;\n  rect?: Rect;\n}\n\nexport type ThinkingLevel = 'off' | 'medium' | 'high';\n\nexport type DeepThinkOption = 'unset' | true | false;\n\nexport interface ServiceTaskInfo {\n  durationMs: number;\n  formatResponse?: string;\n  rawResponse?: string;\n  usage?: AIUsageInfo;\n  searchArea?: Rect;\n  searchAreaRawResponse?: string;\n  searchAreaUsage?: AIUsageInfo;\n  reasoning_content?: string;\n}\n\nexport interface DumpMeta {\n  logTime: number;\n}\n\nexport interface ReportDumpWithAttributes {\n  dumpString: string;\n  attributes?: Record<string, any>;\n}\n\nexport interface ServiceDump extends DumpMeta {\n  type: 'locate' | 'extract' | 'assert';\n  logId: string;\n  userQuery: {\n    element?: TUserPrompt;\n    dataDemand?: ServiceExtractParam;\n    assertion?: TUserPrompt;\n  };\n  matchedElement: LocateResultElement[];\n  matchedRect?: Rect;\n  deepLocate?: boolean;\n  data: any;\n  assertionPass?: boolean;\n  assertionThought?: string;\n  assertionReason?: string;\n  beforeScreenshot?: string;\n  systemCheckResults?: SystemCheckResults;\n  taskInfo: ServiceTaskInfo;\n  error?: string;\n  output?: any;\n}\n\nexport type PartialServiceDumpFromSDK = Omit<\n  ServiceDump,\n  'logTime' | 'logId' | 'model_name'\n>;\n\nexport interface ServiceResultBase {\n  dump: ServiceDump;\n}\n\nexport type LocateResultWithDump = LocateResult & ServiceResultBase;\n\nexport interface ServiceExtractResult<T> extends ServiceResultBase {\n  data: T;\n  thought?: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}\n\nexport class ServiceError extends Error {\n  dump: ServiceDump;\n\n  constructor(message: string, dump: ServiceDump) {\n    super(message);\n    this.name = 'ServiceError';\n    this.dump = dump;\n  }\n}\n\n// intermediate variables to optimize the return value by AI\nexport interface LiteUISection {\n  name: string;\n  description: string;\n  sectionCharacteristics: string;\n  textIds: string[];\n}\n\nexport type ElementById = (id: string) => BaseElement | null;\n\nexport type ServiceAssertionResponse = AIAssertionResponse & {\n  usage?: AIUsageInfo;\n};\n\n/**\n * agent\n */\n\nexport type OnTaskStartTip = (tip: string) => Promise<void> | void;\n\nexport interface AgentWaitForOpt extends ServiceExtractOption {\n  checkIntervalMs?: number;\n  timeoutMs?: number;\n}\n\nexport interface ReferenceImage {\n  name: string;\n  url: string;\n}\n\nexport interface IgnoreRegion {\n  x: number;\n  y: number;\n  width: number;\n  height: number;\n}\n\nexport interface VideoInput {\n  /** 视频帧列表（Base64 编码的图片），最多 150 帧（5秒 * 30fps） */\n  frames: string[];\n  /** 视频时长（秒），最大 5 秒 */\n  duration?: number;\n  /** 帧率，默认 30 */\n  fps?: number;\n  /** 视频格式，默认 'frames' */\n  format?: 'frames' | 'base64' | 'url';\n  /** 视频 URL（当 format 为 'url' 时使用） */\n  url?: string;\n}\n\nexport interface VideoAssertionOptions {\n  /** 是否验证动画流畅度，默认 true */\n  checkSmoothness?: boolean;\n  /** 流畅度阈值（0-100），默认 60 */\n  smoothnessThreshold?: number;\n  /** 是否验证动画时长，默认 false */\n  checkDuration?: boolean;\n  /** 预期动画时长范围（秒） */\n  expectedDuration?: {\n    min?: number;\n    max?: number;\n  };\n  /** 是否验证动画完整性，默认 true */\n  checkCompleteness?: boolean;\n  /** 关键帧验证配置 */\n  keyframes?: {\n    /** 关键帧时间点（秒） */\n    timestamps: number[];\n    /** 关键帧描述 */\n    descriptions?: string[];\n  };\n  /** 是否忽略背景变化，默认 false */\n  ignoreBackgroundChanges?: boolean;\n  /** 动画类型，用于优化分析 */\n  animationType?: 'transition' | 'loading' | 'interaction' | 'auto';\n}\n\nexport interface VideoDetails {\n  /** 动画流畅度评分（0-100） */\n  smoothnessScore: number;\n  /** 动画时长（秒） */\n  duration: number;\n  /** 动画是否完整 */\n  isComplete: boolean;\n  /** 关键帧匹配结果 */\n  keyframeMatches?: {\n    timestamp: number;\n    matched: boolean;\n    description?: string;\n  }[];\n  /** 动画类型识别 */\n  detectedAnimationType: string;\n  /** 动画质量评估 */\n  qualityAssessment: {\n    /** 是否有卡顿 */\n    hasStuttering: boolean;\n    /** 是否有跳帧 */\n    hasFrameDropping: boolean;\n    /** 平均帧间隔（ms） */\n    averageFrameInterval: number;\n    /** 帧间隔标准差 */\n    frameIntervalStdDev: number;\n  };\n  /** 可接受的差异列表 */\n  acceptableDifferences: string[];\n  /** 不可接受的差异列表 */\n  unacceptableDifferences: string[];\n}\n\nexport interface RetryOptions {\n  maxRetries: number;\n  retryInterval: number;\n}\n\nexport interface DiffDetails {\n  layoutMatch: boolean;\n  colorMatch: boolean;\n  elementPositionMatch: boolean;\n  textContentMatch: boolean;\n  resourceMatch: boolean;\n  acceptableDifferences: string[];\n  unacceptableDifferences: string[];\n}\n\nexport interface AssertSnapshot {\n  beforeScreenshot: string;\n  afterScreenshot: string;\n  timestamp: number;\n  assertion: string;\n}\n\nexport interface AgentAssertOpt {\n  keepRawResponse?: boolean;\n  /** 启用系统级校验（白屏、遮挡等），默认 true */\n  enableSystemCheck?: boolean;\n  /** 自定义系统校验规则 */\n  customSystemCheckRules?: string;\n  /** 业务知识上下文 */\n  businessContext?: string;\n  /** 手动传入操作前截图（Base64） */\n  beforeScreenshot?: string;\n  /** 手动传入操作后截图（Base64） */\n  afterScreenshot?: string;\n  /**\n   * 截图模式\n   * - 'auto': 自动判断（默认）\n   * - 'lastAction': 使用最后一次 action 的前后截图\n   * - 'flow': 使用整个流程的开始和结束截图\n   * - 'manual': 使用手动传入的截图\n   * - 'currentOnly': 仅使用当前截图（无操作前截图）\n   * - 'diff': 与基准图片进行 Diff 对比\n   * - 'video': 与基准视频进行动画对比\n   */\n  screenshotMode?:\n    | 'auto'\n    | 'lastAction'\n    | 'flow'\n    | 'manual'\n    | 'currentOnly'\n    | 'diff'\n    | 'video';\n  /** 基准图片列表（仅 Diff 模式） */\n  referenceImages?: ReferenceImage[];\n  /** Diff 差异阈值（0-1），默认 0.1 */\n  diffThreshold?: number;\n  /** 忽略对比的区域 */\n  ignoreRegions?: IgnoreRegion[];\n  /** 是否忽略动态内容，默认 false */\n  ignoreDynamicContent?: boolean;\n  /** 严格模式，任何差异都视为失败，默认 false */\n  strictMode?: boolean;\n  /** 视频断言配置（仅 Video 模式） */\n  videoOptions?: VideoAssertionOptions;\n  /** 当前视频输入（仅 Video 模式），用户传入的视频帧或视频 URL */\n  currentVideo?: VideoInput;\n  /** 断言重试配置 */\n  retryOptions?: RetryOptions;\n  /** 是否保存快照，默认 false */\n  saveSnapshot?: boolean;\n  /** 快照保存路径 */\n  snapshotPath?: string;\n}\n\nexport interface ActionScreenshotContext {\n  beforeScreenshot: string;\n  afterScreenshot: string;\n  actionType: string;\n  actionParam?: any;\n  timestamp: number;\n}\n\nexport interface SystemCheckResults {\n  whiteScreen?: boolean;\n  layoutBlocked?: boolean;\n  loadingContent?: boolean;\n  errorPrompt?: boolean;\n  backendError?: boolean;\n}\n\n/**\n * planning\n *\n */\n\nexport interface PlanningLocateParam extends DetailedLocateParam {\n  bbox?: [number, number, number, number];\n}\n\nexport interface PlanningAction<ParamType = any> {\n  thought?: string;\n  log?: string; // a brief preamble to the user explaining what you’re about to do\n  type: string;\n  param: ParamType;\n}\n\nexport type SubGoalStatus = 'pending' | 'running' | 'finished';\n\nexport interface SubGoal {\n  index: number;\n  status: SubGoalStatus;\n  description: string;\n  logs?: string[];\n}\n\nexport interface RawResponsePlanningAIResponse {\n  action: PlanningAction;\n  thought?: string;\n  log: string;\n  memory?: string;\n  error?: string;\n  finalizeMessage?: string;\n  finalizeSuccess?: boolean;\n  updateSubGoals?: SubGoal[];\n  markFinishedIndexes?: number[];\n}\n\nexport interface PlanningAIResponse\n  extends Omit<RawResponsePlanningAIResponse, 'action'> {\n  actions?: PlanningAction[];\n  usage?: AIUsageInfo;\n  rawResponse?: string;\n  yamlFlow?: MidsceneYamlFlowItem[];\n  yamlString?: string;\n  error?: string;\n  reasoning_content?: string;\n  shouldContinuePlanning: boolean;\n  output?: string; // Output message from <complete> tag (same as finalizeMessage)\n}\n\nexport interface PlanningActionParamSleep {\n  timeMs: number;\n}\n\nexport interface PlanningActionParamError {\n  thought: string;\n}\n\nexport type PlanningActionParamWaitFor = AgentWaitForOpt & {};\n\nexport interface LongPressParam {\n  duration?: number;\n}\n\nexport interface PullParam {\n  direction: 'up' | 'down';\n  distance?: number;\n  duration?: number;\n}\n/**\n * misc\n */\n\nexport interface Color {\n  name: string;\n  hex: string;\n}\n\nexport interface BaseAgentParserOpt {\n  selector?: string;\n}\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PuppeteerParserOpt extends BaseAgentParserOpt {}\n\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PlaywrightParserOpt extends BaseAgentParserOpt {}\n\n/*\naction\n*/\nexport interface ExecutionTaskProgressOptions {\n  onTaskStart?: (task: ExecutionTask) => Promise<void> | void;\n}\n\nexport interface ExecutionRecorderItem {\n  type: 'screenshot';\n  ts: number;\n  screenshot?: ScreenshotItem;\n  timing?: string;\n}\n\nexport type ExecutionTaskType = 'Planning' | 'Insight' | 'Action Space' | 'Log';\n\nexport interface ExecutorContext {\n  task: ExecutionTask;\n  element?: LocateResultElement | null;\n  uiContext?: UIContext;\n}\n\nexport interface ExecutionTaskApply<\n  Type extends ExecutionTaskType = any,\n  TaskParam = any,\n  TaskOutput = any,\n  TaskLog = any,\n> {\n  type: Type;\n  subType?: string;\n  param?: TaskParam;\n  thought?: string;\n  uiContext?: UIContext;\n  executor: (\n    param: TaskParam,\n    context: ExecutorContext,\n  ) => // biome-ignore lint/suspicious/noConfusingVoidType: void is intentionally allowed as some executors may not return a value\n    | Promise<ExecutionTaskReturn<TaskOutput, TaskLog> | undefined | void>\n    | undefined\n    | void;\n}\n\nexport interface ExecutionTaskHitBy {\n  from: string;\n  context: Record<string, any>;\n}\n\nexport interface ExecutionTaskReturn<TaskOutput = unknown, TaskLog = unknown> {\n  output?: TaskOutput;\n  log?: TaskLog;\n  recorder?: ExecutionRecorderItem[];\n  hitBy?: ExecutionTaskHitBy;\n}\n\nexport type ExecutionTask<\n  E extends ExecutionTaskApply<any, any, any> = ExecutionTaskApply<\n    any,\n    any,\n    any\n  >,\n> = E &\n  ExecutionTaskReturn<\n    E extends ExecutionTaskApply<any, any, infer TaskOutput, any>\n      ? TaskOutput\n      : unknown,\n    E extends ExecutionTaskApply<any, any, any, infer TaskLog>\n      ? TaskLog\n      : unknown\n  > & {\n    taskId: string;\n    status: 'pending' | 'running' | 'finished' | 'failed' | 'cancelled';\n    error?: Error;\n    errorMessage?: string;\n    errorStack?: string;\n    timing?: {\n      start: number;\n      getUiContextStart?: number;\n      getUiContextEnd?: number;\n      callAiStart?: number;\n      callAiEnd?: number;\n      beforeInvokeActionHookStart?: number;\n      beforeInvokeActionHookEnd?: number;\n      callActionStart?: number;\n      callActionEnd?: number;\n      afterInvokeActionHookStart?: number;\n      afterInvokeActionHookEnd?: number;\n      captureAfterCallingSnapshotStart?: number;\n      captureAfterCallingSnapshotEnd?: number;\n      end?: number;\n      cost?: number;\n    };\n    usage?: AIUsageInfo;\n    searchAreaUsage?: AIUsageInfo;\n    reasoning_content?: string;\n  };\n\nexport interface IExecutionDump extends DumpMeta {\n  name: string;\n  description?: string;\n  tasks: ExecutionTask[];\n  aiActContext?: string;\n}\n\n/**\n * Replacer function for JSON serialization that handles Page, Browser objects and ScreenshotItem\n */\nfunction replacerForDumpSerialization(_key: string, value: any): any {\n  if (value && value.constructor?.name === 'Page') {\n    return '[Page object]';\n  }\n  if (value && value.constructor?.name === 'Browser') {\n    return '[Browser object]';\n  }\n  // Handle ScreenshotItem serialization\n  if (value && typeof value.toSerializable === 'function') {\n    return value.toSerializable();\n  }\n  return value;\n}\n\n/**\n * Reviver function for JSON deserialization that handles ScreenshotItem formats.\n *\n * BEHAVIOR:\n * - For { $screenshot: \"id\" } format: Left as-is (plain object)\n *   Consumer must use imageMap to restore base64 data\n * - For { base64: \"...\" } format: Creates ScreenshotItem from base64 data\n *\n * @param key - JSON key being processed\n * @param value - JSON value being processed\n * @returns Restored value\n */\nfunction reviverForDumpDeserialization(key: string, value: any): any {\n  // Only process screenshot fields\n  if (key !== 'screenshot' || typeof value !== 'object' || value === null) {\n    return value;\n  }\n\n  // Handle serialized format: { $screenshot: \"id\" }\n  // Leave as plain object — consumer uses imageMap to restore\n  if (ScreenshotItem.isSerialized(value)) {\n    return value;\n  }\n\n  // Handle inline base64 format: { base64: \"...\" }\n  if ('base64' in value && typeof value.base64 === 'string') {\n    return value;\n  }\n\n  return value;\n}\n\n/**\n * ExecutionDump class for serializing and deserializing execution dumps\n */\nexport class ExecutionDump implements IExecutionDump {\n  logTime: number;\n  name: string;\n  description?: string;\n  tasks: ExecutionTask[];\n  aiActContext?: string;\n\n  constructor(data: IExecutionDump) {\n    this.logTime = data.logTime;\n    this.name = data.name;\n    this.description = data.description;\n    this.tasks = data.tasks;\n    this.aiActContext = data.aiActContext;\n  }\n\n  /**\n   * Serialize the ExecutionDump to a JSON string\n   */\n  serialize(indents?: number): string {\n    return JSON.stringify(this.toJSON(), replacerForDumpSerialization, indents);\n  }\n\n  /**\n   * Convert to a plain object for JSON serialization\n   */\n  toJSON(): IExecutionDump {\n    return {\n      logTime: this.logTime,\n      name: this.name,\n      description: this.description,\n      tasks: this.tasks.map((task) => ({\n        ...task,\n        recorder: task.recorder || [],\n      })),\n      aiActContext: this.aiActContext,\n    };\n  }\n\n  /**\n   * Create an ExecutionDump instance from a serialized JSON string\n   */\n  static fromSerializedString(serialized: string): ExecutionDump {\n    const parsed = JSON.parse(\n      serialized,\n      reviverForDumpDeserialization,\n    ) as IExecutionDump;\n    return new ExecutionDump(parsed);\n  }\n\n  /**\n   * Create an ExecutionDump instance from a plain object\n   */\n  static fromJSON(data: IExecutionDump): ExecutionDump {\n    return new ExecutionDump(data);\n  }\n\n  /**\n   * Collect all ScreenshotItem instances from tasks.\n   * Scans through uiContext and recorder items to find screenshots.\n   *\n   * @returns Array of ScreenshotItem instances\n   */\n  collectScreenshots(): ScreenshotItem[] {\n    const screenshots: ScreenshotItem[] = [];\n\n    for (const task of this.tasks) {\n      // Collect uiContext.screenshot if present\n      if (task.uiContext?.screenshot instanceof ScreenshotItem) {\n        screenshots.push(task.uiContext.screenshot);\n      }\n\n      // Collect recorder screenshots\n      if (task.recorder) {\n        for (const record of task.recorder) {\n          if (record.screenshot instanceof ScreenshotItem) {\n            screenshots.push(record.screenshot);\n          }\n        }\n      }\n    }\n\n    return screenshots;\n  }\n}\n\n/*\ntask - service-locate\n*/\nexport type ExecutionTaskInsightLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskInsightLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport type ExecutionTaskInsightDump = ServiceDump;\n\nexport type ExecutionTaskInsightLocateApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightLocateParam,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightLocate =\n  ExecutionTask<ExecutionTaskInsightLocateApply>;\n\n/*\ntask - service-query\n*/\nexport interface ExecutionTaskInsightQueryParam {\n  dataDemand: ServiceExtractParam;\n}\n\nexport interface ExecutionTaskInsightQueryOutput {\n  data: any;\n}\n\nexport type ExecutionTaskInsightQueryApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightQueryParam,\n  any,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightQuery =\n  ExecutionTask<ExecutionTaskInsightQueryApply>;\n\n/*\ntask - assertion\n*/\nexport interface ExecutionTaskInsightAssertionParam {\n  assertion: string;\n}\n\nexport type ExecutionTaskInsightAssertionApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightAssertionParam,\n  ServiceAssertionResponse,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightAssertion =\n  ExecutionTask<ExecutionTaskInsightAssertionApply>;\n\n/*\ntask - action (i.e. interact) \n*/\nexport type ExecutionTaskActionApply<ActionParam = any> = ExecutionTaskApply<\n  'Action Space',\n  ActionParam,\n  void,\n  void\n>;\n\nexport type ExecutionTaskAction = ExecutionTask<ExecutionTaskActionApply>;\n\n/*\ntask - Log\n*/\n\nexport type ExecutionTaskLogApply<\n  LogParam = {\n    content: string;\n  },\n> = ExecutionTaskApply<'Log', LogParam, void, void>;\n\nexport type ExecutionTaskLog = ExecutionTask<ExecutionTaskLogApply>;\n\n/*\ntask - planning\n*/\n\nexport type ExecutionTaskPlanningApply = ExecutionTaskApply<\n  'Planning',\n  {\n    userInstruction: string;\n    aiActContext?: string;\n  },\n  PlanningAIResponse\n>;\n\nexport type ExecutionTaskPlanning = ExecutionTask<ExecutionTaskPlanningApply>;\n\n/*\ntask - planning-locate\n*/\nexport type ExecutionTaskPlanningLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskPlanningLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport type ExecutionTaskPlanningDump = ServiceDump;\n\nexport type ExecutionTaskPlanningLocateApply = ExecutionTaskApply<\n  'Planning',\n  ExecutionTaskPlanningLocateParam,\n  ExecutionTaskPlanningLocateOutput,\n  ExecutionTaskPlanningDump\n>;\n\nexport type ExecutionTaskPlanningLocate =\n  ExecutionTask<ExecutionTaskPlanningLocateApply>;\n\n/*\nGrouped dump\n*/\nexport interface IGroupedActionDump {\n  sdkVersion: string;\n  groupName: string;\n  groupDescription?: string;\n  modelBriefs: string[];\n  executions: IExecutionDump[];\n  deviceType?: string;\n}\n\n/**\n * GroupedActionDump class for serializing and deserializing grouped action dumps\n */\nexport class GroupedActionDump implements IGroupedActionDump {\n  sdkVersion: string;\n  groupName: string;\n  groupDescription?: string;\n  modelBriefs: string[];\n  executions: ExecutionDump[];\n  deviceType?: string;\n\n  constructor(data: IGroupedActionDump) {\n    this.sdkVersion = data.sdkVersion;\n    this.groupName = data.groupName;\n    this.groupDescription = data.groupDescription;\n    this.modelBriefs = data.modelBriefs;\n    this.executions = data.executions.map((exec) =>\n      exec instanceof ExecutionDump ? exec : ExecutionDump.fromJSON(exec),\n    );\n    this.deviceType = data.deviceType;\n  }\n\n  /**\n   * Serialize the GroupedActionDump to a JSON string\n   * Uses compact { $screenshot: id } format\n   */\n  serialize(indents?: number): string {\n    return JSON.stringify(this.toJSON(), replacerForDumpSerialization, indents);\n  }\n\n  /**\n   * Serialize the GroupedActionDump with inline screenshots to a JSON string.\n   * Each ScreenshotItem is replaced with { base64: \"...\", capturedAt }.\n   */\n  serializeWithInlineScreenshots(indents?: number): string {\n    const processValue = (obj: unknown): unknown => {\n      if (obj instanceof ScreenshotItem) {\n        return { base64: obj.base64, capturedAt: obj.capturedAt };\n      }\n      if (Array.isArray(obj)) {\n        return obj.map(processValue);\n      }\n      if (obj && typeof obj === 'object') {\n        const entries = Object.entries(obj).map(([key, value]) => [\n          key,\n          processValue(value),\n        ]);\n        return Object.fromEntries(entries);\n      }\n      return obj;\n    };\n\n    const data = processValue(this.toJSON());\n    return JSON.stringify(data, null, indents);\n  }\n\n  /**\n   * Convert to a plain object for JSON serialization\n   */\n  toJSON(): IGroupedActionDump {\n    return {\n      sdkVersion: this.sdkVersion,\n      groupName: this.groupName,\n      groupDescription: this.groupDescription,\n      modelBriefs: this.modelBriefs,\n      executions: this.executions.map((exec) => exec.toJSON()),\n      deviceType: this.deviceType,\n    };\n  }\n\n  /**\n   * Create a GroupedActionDump instance from a serialized JSON string\n   */\n  static fromSerializedString(serialized: string): GroupedActionDump {\n    const parsed = JSON.parse(\n      serialized,\n      reviverForDumpDeserialization,\n    ) as IGroupedActionDump;\n    return new GroupedActionDump(parsed);\n  }\n\n  /**\n   * Create a GroupedActionDump instance from a plain object\n   */\n  static fromJSON(data: IGroupedActionDump): GroupedActionDump {\n    return new GroupedActionDump(data);\n  }\n\n  /**\n   * Collect all ScreenshotItem instances from all executions.\n   *\n   * @returns Array of all ScreenshotItem instances across all executions\n   */\n  collectAllScreenshots(): ScreenshotItem[] {\n    const screenshots: ScreenshotItem[] = [];\n    for (const execution of this.executions) {\n      screenshots.push(...execution.collectScreenshots());\n    }\n    return screenshots;\n  }\n\n  /**\n   * Serialize the dump to files with screenshots as separate PNG files.\n   * Creates:\n   * - {basePath} - dump JSON with { $screenshot: id } references\n   * - {basePath}.screenshots/ - PNG files\n   * - {basePath}.screenshots.json - ID to path mapping\n   *\n   * @param basePath - Base path for the dump file\n   */\n  serializeToFiles(basePath: string): void {\n    const screenshotsDir = `${basePath}.screenshots`;\n    if (!existsSync(screenshotsDir)) {\n      mkdirSync(screenshotsDir, { recursive: true });\n    }\n\n    // Write screenshots to separate files\n    const screenshotMap: Record<string, string> = {};\n    const screenshots = this.collectAllScreenshots();\n\n    for (const screenshot of screenshots) {\n      if (screenshot.hasBase64()) {\n        const imagePath = join(\n          screenshotsDir,\n          `${screenshot.id}.${screenshot.extension}`,\n        );\n        const rawBase64 = screenshot.rawBase64;\n        writeFileSync(imagePath, Buffer.from(rawBase64, 'base64'));\n        screenshotMap[screenshot.id] = imagePath;\n      }\n    }\n\n    // Write screenshot map file\n    writeFileSync(\n      `${basePath}.screenshots.json`,\n      JSON.stringify(screenshotMap),\n      'utf-8',\n    );\n\n    // Write dump JSON with references\n    writeFileSync(basePath, this.serialize(), 'utf-8');\n  }\n\n  /**\n   * Read dump from files and return JSON string with inline screenshots.\n   * Reads the dump JSON and screenshot files, then inlines the base64 data.\n   *\n   * @param basePath - Base path for the dump file\n   * @returns JSON string with inline screenshots ({ base64: \"...\" } format)\n   */\n  static fromFilesAsInlineJson(basePath: string): string {\n    const dumpString = readFileSync(basePath, 'utf-8');\n    const screenshotsMapPath = `${basePath}.screenshots.json`;\n\n    if (!existsSync(screenshotsMapPath)) {\n      return dumpString;\n    }\n\n    // Read screenshot map and build imageMap from files\n    const screenshotMap: Record<string, string> = JSON.parse(\n      readFileSync(screenshotsMapPath, 'utf-8'),\n    );\n\n    const imageMap: Record<string, string> = {};\n    for (const [id, filePath] of Object.entries(screenshotMap)) {\n      if (existsSync(filePath)) {\n        const data = readFileSync(filePath);\n        const mime =\n          filePath.endsWith('.jpeg') || filePath.endsWith('.jpg')\n            ? 'jpeg'\n            : 'png';\n        imageMap[id] = `data:image/${mime};base64,${data.toString('base64')}`;\n      }\n    }\n\n    // Restore image references\n    const dumpData = JSON.parse(dumpString);\n    const processedData = restoreImageReferences(\n      dumpData,\n      (id) => imageMap[id] ?? '',\n    );\n    return JSON.stringify(processedData);\n  }\n\n  /**\n   * Clean up all files associated with a serialized dump.\n   *\n   * @param basePath - Base path for the dump file\n   */\n  static cleanupFiles(basePath: string): void {\n    const filesToClean = [\n      basePath,\n      `${basePath}.screenshots.json`,\n      `${basePath}.screenshots`,\n    ];\n\n    for (const filePath of filesToClean) {\n      try {\n        rmSync(filePath, { force: true, recursive: true });\n      } catch {\n        // Ignore errors - file may already be deleted\n      }\n    }\n  }\n\n  /**\n   * Get all file paths associated with a serialized dump.\n   *\n   * @param basePath - Base path for the dump file\n   * @returns Array of all associated file paths\n   */\n  static getFilePaths(basePath: string): string[] {\n    return [\n      basePath,\n      `${basePath}.screenshots.json`,\n      `${basePath}.screenshots`,\n    ];\n  }\n}\n\nexport type InterfaceType =\n  | 'puppeteer'\n  | 'playwright'\n  | 'static'\n  | 'chrome-extension-proxy'\n  | 'android'\n  | string;\n\nexport interface StreamingCodeGenerationOptions {\n  /** Whether to enable streaming output */\n  stream?: boolean;\n  /** Callback function to handle streaming chunks */\n  onChunk?: StreamingCallback;\n  /** Callback function to handle streaming completion */\n  onComplete?: (finalCode: string) => void;\n  /** Callback function to handle streaming errors */\n  onError?: (error: Error) => void;\n}\n\nexport type StreamingCallback = (chunk: CodeGenerationChunk) => void;\n\nexport interface CodeGenerationChunk {\n  /** The incremental content chunk */\n  content: string;\n  /** The reasoning content */\n  reasoning_content: string;\n  /** The accumulated content so far */\n  accumulated: string;\n  /** Whether this is the final chunk */\n  isComplete: boolean;\n  /** Token usage information if available */\n  usage?: AIUsageInfo;\n}\n\nexport interface StreamingAIResponse {\n  /** The final accumulated content */\n  content: string;\n  /** Token usage information */\n  usage?: AIUsageInfo;\n  /** Whether the response was streamed */\n  isStreamed: boolean;\n}\n\nexport interface DeviceAction<TParam = any, TReturn = any> {\n  name: string;\n  description?: string;\n  interfaceAlias?: string;\n  paramSchema?: z.ZodType<TParam>;\n  call: (param: TParam, context: ExecutorContext) => Promise<TReturn> | TReturn;\n  delayAfterRunner?: number;\n  /**\n   * An example param object for this action.\n   * Locate fields with { prompt } will automatically get bbox injected when needed.\n   */\n  sample?: { [K in keyof TParam]?: any };\n}\n\n/**\n * Type utilities for extracting types from DeviceAction definitions\n */\n\n/**\n * Extract parameter type from a DeviceAction\n */\nexport type ActionParam<Action extends DeviceAction<any, any>> =\n  Action extends DeviceAction<infer P, any> ? P : never;\n\n/**\n * Extract return type from a DeviceAction\n */\nexport type ActionReturn<Action extends DeviceAction<any, any>> =\n  Action extends DeviceAction<any, infer R> ? R : never;\n\n/**\n * Web-specific types\n */\nexport interface WebElementInfo extends BaseElement {\n  id: string;\n  attributes: {\n    nodeType: NodeType;\n    [key: string]: string;\n  };\n}\n\n/**\n * Agent\n */\n\nexport type CacheConfig = {\n  strategy?: 'read-only' | 'read-write' | 'write-only';\n  id: string;\n  cacheAdapter?: import('./agent/cache-adapter').CacheAdapter;\n};\n\nexport type Cache =\n  | false // No read, no write\n  | true // Will throw error at runtime - deprecated\n  | CacheConfig; // Object configuration (requires explicit id)\n\nexport interface AgentOpt {\n  testId?: string;\n  // @deprecated\n  cacheId?: string; // Keep backward compatibility, but marked as deprecated\n  groupName?: string;\n  groupDescription?: string;\n  /* if auto generate report, default true */\n  generateReport?: boolean;\n  /* if auto print report msg, default true */\n  autoPrintReportMsg?: boolean;\n\n  /**\n   * Use directory-based report format with separate image files.\n   *\n   * When enabled:\n   * - Screenshots are saved as PNG files in a `screenshots/` subdirectory\n   * - Report is generated as `index.html` with relative image paths\n   * - Reduces memory usage and report file size\n   *\n   * IMPORTANT: 'html-and-external-assets' reports must be served via HTTP server\n   * (e.g., `npx serve ./report-dir`). The file:// protocol will not\n   * work due to browser CORS restrictions.\n   *\n   * @default 'single-html'\n   */\n  outputFormat?: 'single-html' | 'html-and-external-assets';\n\n  onTaskStartTip?: OnTaskStartTip;\n  aiActContext?: string;\n  aiActionContext?: string;\n  /* custom report file name */\n  reportFileName?: string;\n  modelConfig?: TModelConfig;\n  cache?: Cache;\n  /**\n   * Maximum number of replanning cycles for aiAct.\n   * Defaults to 20 (40 for `vlm-ui-tars`) when not provided.\n   * If omitted, the agent will also read `MIDSCENE_REPLANNING_CYCLE_LIMIT` for backward compatibility.\n   */\n  replanningCycleLimit?: number;\n\n  /**\n   * Wait time in milliseconds after each action execution.\n   * This allows the UI to settle and stabilize before the next action.\n   * Defaults to 300ms when not provided.\n   */\n  waitAfterAction?: number;\n\n  /**\n   * When set to true, Midscene will use the target device's time (Android/iOS)\n   * instead of the system time. Useful when the device time differs from the\n   * host machine. Default: false\n   */\n  useDeviceTimestamp?: boolean;\n\n  /**\n   * Custom screenshot shrink factor to reduce AI token usage.\n   * When set, the screenshot will be scaled down by this factor from the physical resolution.\n   *\n   * Example:\n   * - Physical screen width: 3000px, dpr=6\n   * - Logical width: 500px\n   * - screenshotShrinkFactor: 2\n   * - Actual shrunk screenshot width: 3000 / 2 = 1500px\n   * - AI analyzes the 1500px screenshot\n   * - Coordinates are transformed back to logical (500px) before actions execute\n   *\n   * Benefits:\n   * - Reduces token usage for high-resolution screenshots\n   * - Maintains accuracy by scaling coordinates appropriately\n   *\n   * Must be >= 1 (shrinking only, enlarging is not supported).\n   *\n   * @default 1 (no shrinking, uses original physical screenshot)\n   */\n  screenshotShrinkFactor?: number;\n\n  /**\n   * Custom OpenAI client factory function\n   *\n   * If provided, this function will be called to create OpenAI client instances\n   * for each AI call, allowing you to:\n   * - Wrap clients with observability tools (langsmith, langfuse)\n   * - Use custom OpenAI-compatible clients\n   * - Apply different configurations based on intent\n   *\n   * @param config - Resolved model configuration\n   * @returns OpenAI client instance (original or wrapped)\n   *\n   * @example\n   * ```typescript\n   * createOpenAIClient: async (openai, opts) => {\n   *   // Wrap with langsmith for planning tasks\n   *   if (opts.baseURL?.includes('planning')) {\n   *     return wrapOpenAI(openai, { metadata: { task: 'planning' } });\n   *   }\n   *\n   *   return openai;\n   * }\n   * ```\n   */\n  createOpenAIClient?: CreateOpenAIClientFn;\n}\n\nexport type TestStatus =\n  | 'passed'\n  | 'failed'\n  | 'timedOut'\n  | 'skipped'\n  | 'interrupted';\n\nexport interface ReportFileWithAttributes {\n  reportFilePath: string;\n  reportAttributes: {\n    testDuration: number;\n    testStatus: TestStatus;\n    testTitle: string;\n    testId: string;\n    testDescription: string;\n  };\n}\n"],"names":["UIContext","ServiceError","Error","message","dump","replacerForDumpSerialization","_key","value","reviverForDumpDeserialization","key","ScreenshotItem","ExecutionDump","indents","JSON","task","serialized","parsed","data","screenshots","record","GroupedActionDump","processValue","obj","Array","entries","Object","exec","execution","basePath","screenshotsDir","existsSync","mkdirSync","screenshotMap","screenshot","imagePath","join","rawBase64","writeFileSync","Buffer","dumpString","readFileSync","screenshotsMapPath","imageMap","id","filePath","mime","dumpData","processedData","restoreImageReferences","filesToClean","rmSync"],"mappings":";;;;;AAAqD;;;;;;;;;;AA+G9C,MAAeA;AA4BtB;AA+EO,MAAMC,qBAAqBC;IAGhC,YAAYC,OAAe,EAAEC,IAAiB,CAAE;QAC9C,KAAK,CAACD,UAHR;QAIE,IAAI,CAAC,IAAI,GAAG;QACZ,IAAI,CAAC,IAAI,GAAGC;IACd;AACF;AA0YA,SAASC,6BAA6BC,IAAY,EAAEC,KAAU;IAC5D,IAAIA,SAASA,MAAM,WAAW,EAAE,SAAS,QACvC,OAAO;IAET,IAAIA,SAASA,MAAM,WAAW,EAAE,SAAS,WACvC,OAAO;IAGT,IAAIA,SAAS,AAAgC,cAAhC,OAAOA,MAAM,cAAc,EACtC,OAAOA,MAAM,cAAc;IAE7B,OAAOA;AACT;AAcA,SAASC,8BAA8BC,GAAW,EAAEF,KAAU;IAE5D,IAAIE,AAAQ,iBAARA,OAAwB,AAAiB,YAAjB,OAAOF,SAAsBA,AAAU,SAAVA,OACvD,OAAOA;IAKT,IAAIG,eAAe,YAAY,CAACH,QAC9B,OAAOA;IAIL,YAAYA,SAAgBA,MAAM,MAAM;IAI5C,OAAOA;AACT;AAKO,MAAMI;IAkBX,UAAUC,OAAgB,EAAU;QAClC,OAAOC,KAAK,SAAS,CAAC,IAAI,CAAC,MAAM,IAAIR,8BAA8BO;IACrE;IAKA,SAAyB;QACvB,OAAO;YACL,SAAS,IAAI,CAAC,OAAO;YACrB,MAAM,IAAI,CAAC,IAAI;YACf,aAAa,IAAI,CAAC,WAAW;YAC7B,OAAO,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,CAACE,OAAU;oBAC/B,GAAGA,IAAI;oBACP,UAAUA,KAAK,QAAQ,IAAI,EAAE;gBAC/B;YACA,cAAc,IAAI,CAAC,YAAY;QACjC;IACF;IAKA,OAAO,qBAAqBC,UAAkB,EAAiB;QAC7D,MAAMC,SAASH,KAAK,KAAK,CACvBE,YACAP;QAEF,OAAO,IAAIG,cAAcK;IAC3B;IAKA,OAAO,SAASC,IAAoB,EAAiB;QACnD,OAAO,IAAIN,cAAcM;IAC3B;IAQA,qBAAuC;QACrC,MAAMC,cAAgC,EAAE;QAExC,KAAK,MAAMJ,QAAQ,IAAI,CAAC,KAAK,CAAE;YAE7B,IAAIA,KAAK,SAAS,EAAE,sBAAsBJ,gBACxCQ,YAAY,IAAI,CAACJ,KAAK,SAAS,CAAC,UAAU;YAI5C,IAAIA,KAAK,QAAQ,EACf;gBAAA,KAAK,MAAMK,UAAUL,KAAK,QAAQ,CAChC,IAAIK,OAAO,UAAU,YAAYT,gBAC/BQ,YAAY,IAAI,CAACC,OAAO,UAAU;YAEtC;QAEJ;QAEA,OAAOD;IACT;IA3EA,YAAYD,IAAoB,CAAE;QANlC;QACA;QACA;QACA;QACA;QAGE,IAAI,CAAC,OAAO,GAAGA,KAAK,OAAO;QAC3B,IAAI,CAAC,IAAI,GAAGA,KAAK,IAAI;QACrB,IAAI,CAAC,WAAW,GAAGA,KAAK,WAAW;QACnC,IAAI,CAAC,KAAK,GAAGA,KAAK,KAAK;QACvB,IAAI,CAAC,YAAY,GAAGA,KAAK,YAAY;IACvC;AAsEF;AAwIO,MAAMG;IAuBX,UAAUR,OAAgB,EAAU;QAClC,OAAOC,KAAK,SAAS,CAAC,IAAI,CAAC,MAAM,IAAIR,8BAA8BO;IACrE;IAMA,+BAA+BA,OAAgB,EAAU;QACvD,MAAMS,eAAe,CAACC;YACpB,IAAIA,eAAeZ,gBACjB,OAAO;gBAAE,QAAQY,IAAI,MAAM;gBAAE,YAAYA,IAAI,UAAU;YAAC;YAE1D,IAAIC,MAAM,OAAO,CAACD,MAChB,OAAOA,IAAI,GAAG,CAACD;YAEjB,IAAIC,OAAO,AAAe,YAAf,OAAOA,KAAkB;gBAClC,MAAME,UAAUC,OAAO,OAAO,CAACH,KAAK,GAAG,CAAC,CAAC,CAACb,KAAKF,MAAM,GAAK;wBACxDE;wBACAY,aAAad;qBACd;gBACD,OAAOkB,OAAO,WAAW,CAACD;YAC5B;YACA,OAAOF;QACT;QAEA,MAAML,OAAOI,aAAa,IAAI,CAAC,MAAM;QACrC,OAAOR,KAAK,SAAS,CAACI,MAAM,MAAML;IACpC;IAKA,SAA6B;QAC3B,OAAO;YACL,YAAY,IAAI,CAAC,UAAU;YAC3B,WAAW,IAAI,CAAC,SAAS;YACzB,kBAAkB,IAAI,CAAC,gBAAgB;YACvC,aAAa,IAAI,CAAC,WAAW;YAC7B,YAAY,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,CAACc,OAASA,KAAK,MAAM;YACrD,YAAY,IAAI,CAAC,UAAU;QAC7B;IACF;IAKA,OAAO,qBAAqBX,UAAkB,EAAqB;QACjE,MAAMC,SAASH,KAAK,KAAK,CACvBE,YACAP;QAEF,OAAO,IAAIY,kBAAkBJ;IAC/B;IAKA,OAAO,SAASC,IAAwB,EAAqB;QAC3D,OAAO,IAAIG,kBAAkBH;IAC/B;IAOA,wBAA0C;QACxC,MAAMC,cAAgC,EAAE;QACxC,KAAK,MAAMS,aAAa,IAAI,CAAC,UAAU,CACrCT,YAAY,IAAI,IAAIS,UAAU,kBAAkB;QAElD,OAAOT;IACT;IAWA,iBAAiBU,QAAgB,EAAQ;QACvC,MAAMC,iBAAiB,GAAGD,SAAS,YAAY,CAAC;QAChD,IAAI,CAACE,WAAWD,iBACdE,UAAUF,gBAAgB;YAAE,WAAW;QAAK;QAI9C,MAAMG,gBAAwC,CAAC;QAC/C,MAAMd,cAAc,IAAI,CAAC,qBAAqB;QAE9C,KAAK,MAAMe,cAAcf,YACvB,IAAIe,WAAW,SAAS,IAAI;YAC1B,MAAMC,YAAYC,KAChBN,gBACA,GAAGI,WAAW,EAAE,CAAC,CAAC,EAAEA,WAAW,SAAS,EAAE;YAE5C,MAAMG,YAAYH,WAAW,SAAS;YACtCI,cAAcH,WAAWI,OAAO,IAAI,CAACF,WAAW;YAChDJ,aAAa,CAACC,WAAW,EAAE,CAAC,GAAGC;QACjC;QAIFG,cACE,GAAGT,SAAS,iBAAiB,CAAC,EAC9Bf,KAAK,SAAS,CAACmB,gBACf;QAIFK,cAAcT,UAAU,IAAI,CAAC,SAAS,IAAI;IAC5C;IASA,OAAO,sBAAsBA,QAAgB,EAAU;QACrD,MAAMW,aAAaC,aAAaZ,UAAU;QAC1C,MAAMa,qBAAqB,GAAGb,SAAS,iBAAiB,CAAC;QAEzD,IAAI,CAACE,WAAWW,qBACd,OAAOF;QAIT,MAAMP,gBAAwCnB,KAAK,KAAK,CACtD2B,aAAaC,oBAAoB;QAGnC,MAAMC,WAAmC,CAAC;QAC1C,KAAK,MAAM,CAACC,IAAIC,SAAS,IAAInB,OAAO,OAAO,CAACO,eAC1C,IAAIF,WAAWc,WAAW;YACxB,MAAM3B,OAAOuB,aAAaI;YAC1B,MAAMC,OACJD,SAAS,QAAQ,CAAC,YAAYA,SAAS,QAAQ,CAAC,UAC5C,SACA;YACNF,QAAQ,CAACC,GAAG,GAAG,CAAC,WAAW,EAAEE,KAAK,QAAQ,EAAE5B,KAAK,QAAQ,CAAC,WAAW;QACvE;QAIF,MAAM6B,WAAWjC,KAAK,KAAK,CAAC0B;QAC5B,MAAMQ,gBAAgBC,uBACpBF,UACA,CAACH,KAAOD,QAAQ,CAACC,GAAG,IAAI;QAE1B,OAAO9B,KAAK,SAAS,CAACkC;IACxB;IAOA,OAAO,aAAanB,QAAgB,EAAQ;QAC1C,MAAMqB,eAAe;YACnBrB;YACA,GAAGA,SAAS,iBAAiB,CAAC;YAC9B,GAAGA,SAAS,YAAY,CAAC;SAC1B;QAED,KAAK,MAAMgB,YAAYK,aACrB,IAAI;YACFC,OAAON,UAAU;gBAAE,OAAO;gBAAM,WAAW;YAAK;QAClD,EAAE,OAAM,CAER;IAEJ;IAQA,OAAO,aAAahB,QAAgB,EAAY;QAC9C,OAAO;YACLA;YACA,GAAGA,SAAS,iBAAiB,CAAC;YAC9B,GAAGA,SAAS,YAAY,CAAC;SAC1B;IACH;IA9MA,YAAYX,IAAwB,CAAE;QAPtC;QACA;QACA;QACA;QACA;QACA;QAGE,IAAI,CAAC,UAAU,GAAGA,KAAK,UAAU;QACjC,IAAI,CAAC,SAAS,GAAGA,KAAK,SAAS;QAC/B,IAAI,CAAC,gBAAgB,GAAGA,KAAK,gBAAgB;QAC7C,IAAI,CAAC,WAAW,GAAGA,KAAK,WAAW;QACnC,IAAI,CAAC,UAAU,GAAGA,KAAK,UAAU,CAAC,GAAG,CAAC,CAACS,OACrCA,gBAAgBf,gBAAgBe,OAAOf,cAAc,QAAQ,CAACe;QAEhE,IAAI,CAAC,UAAU,GAAGT,KAAK,UAAU;IACnC;AAsMF"}