@midscene/core 1.9.7 → 1.9.8-beta-20260618014851.0

package/dist/es/agent/utils.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"agent/utils.mjs","sources":["../../../src/agent/utils.ts"],"sourcesContent":["import { pixelBboxToRect } from '@/ai-model/workflows/inspect/locate-result-rect';\nimport type { TMultimodalPrompt, TUserPrompt } from '@/common';\nimport type { AbstractInterface } from '@/device';\nimport { ScreenshotItem } from '@/screenshot-item';\nimport type {\n  ElementCacheFeature,\n  LocateResultElement,\n  PixelBbox,\n  PlanningLocateParam,\n  PlanningLocateParamWithLocatedPixelBbox,\n  Rect,\n  Size,\n  UIContext,\n} from '@/types';\nimport { uploadTestInfoToServer } from '@/utils';\nimport {\n  MIDSCENE_REPORT_QUIET,\n  MIDSCENE_REPORT_TAG_NAME,\n  globalConfigManager,\n} from '@midscene/shared/env';\nimport { generateElementByRect } from '@midscene/shared/extractor';\nimport {\n  imageInfoOfBase64,\n  normalizeBase64Image,\n  resizeImgBase64,\n} from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { assert, logMsg, uuid } from '@midscene/shared/utils';\nimport dayjs from 'dayjs';\nimport type { TaskCache } from './task-cache';\nimport { debug as cacheDebug } from './task-cache';\n\nconst agentDebug = getDebug('agent');\n\nexport async function commonContextParser(\n  interfaceInstance: AbstractInterface,\n  _opt: {\n    uploadServerUrl?: string;\n    screenshotShrinkFactor?: number;\n  },\n): Promise<UIContext> {\n  const debug = getDebug('commonContextParser');\n\n  assert(interfaceInstance, 'interfaceInstance is required');\n\n  debug('Getting interface description');\n  const description = interfaceInstance.describe?.() || '';\n  debug('Interface description end');\n\n  debug('Uploading test info to server');\n  uploadTestInfoToServer({\n    testUrl: description,\n    serverUrl: _opt.uploadServerUrl,\n  });\n  debug('UploadTestInfoToServer end');\n\n  debug('will get size');\n  const interfaceSize = await interfaceInstance.size();\n  const { width: logicalWidth, height: logicalHeight } = interfaceSize;\n\n  if ((interfaceSize as unknown as { dpr: number }).dpr) {\n    console.warn(\n      'Warning: return value of interface.size() include a dpr property, which is not expected and ignored. ',\n    );\n  }\n\n  if (!Number.isFinite(logicalWidth) || !Number.isFinite(logicalHeight)) {\n    throw new Error(\n      `Invalid interface size: width and height must be finite numbers. Received width: ${logicalWidth}, height: ${logicalHeight}`,\n    );\n  }\n\n  if (logicalWidth <= 0 || logicalHeight <= 0) {\n    throw new Error(\n      `Invalid interface size: width and height must be positive numbers. Received width: ${logicalWidth}, height: ${logicalHeight}`,\n    );\n  }\n\n  debug(`size: ${logicalWidth}x${logicalHeight}`);\n\n  const screenshotBase64 = await interfaceInstance.screenshotBase64();\n  const screenshotCapturedAt = Date.now();\n  assert(screenshotBase64!, 'screenshotBase64 is required');\n\n  // Get physical screenshot dimensions\n  debug('will get screenshot dimensions');\n  const { width: imgWidth, height: imgHeight } =\n    await imageInfoOfBase64(screenshotBase64);\n\n  if (!Number.isFinite(imgWidth) || !Number.isFinite(imgHeight)) {\n    throw new Error(\n      `Invalid screenshot dimensions: width and height must be finite numbers. Received width: ${imgWidth}, height: ${imgHeight}`,\n    );\n  }\n  if (imgWidth <= 0 || imgHeight <= 0) {\n    throw new Error(\n      `Invalid screenshot dimensions: width and height must be positive numbers. Received width: ${imgWidth}, height: ${imgHeight}`,\n    );\n  }\n  debug('screenshot dimensions', imgWidth, 'x', imgHeight);\n\n  // Detect orientation mismatch between logical size and screenshot.\n  // Some devices (e.g. OPPO) report wrong orientation via ADB, causing\n  // size() to return portrait dimensions even when the device is landscape.\n  // We detect this by comparing aspect ratios and swap if they disagree.\n  const logicalIsPortrait = logicalWidth < logicalHeight;\n  const screenshotIsPortrait = imgWidth < imgHeight;\n  let finalLogicalWidth = logicalWidth;\n  let finalLogicalHeight = logicalHeight;\n  if (logicalIsPortrait !== screenshotIsPortrait) {\n    debug(\n      `Orientation mismatch detected: logical size ${logicalWidth}x${logicalHeight} (${logicalIsPortrait ? 'portrait' : 'landscape'}) vs screenshot ${imgWidth}x${imgHeight} (${screenshotIsPortrait ? 'portrait' : 'landscape'}). Swapping logical dimensions.`,\n    );\n    finalLogicalWidth = logicalHeight;\n    finalLogicalHeight = logicalWidth;\n  }\n\n  const userShrinkFactor = _opt.screenshotShrinkFactor ?? 1;\n\n  if (!Number.isFinite(userShrinkFactor) || userShrinkFactor < 1) {\n    throw new Error(\n      `Invalid screenshotShrinkFactor: must be a finite number >= 1. Received: ${userShrinkFactor}`,\n    );\n  }\n\n  const dpr = imgWidth / finalLogicalWidth;\n\n  debug('calculated dpr:', dpr);\n\n  const shrunkShotToLogicalRatio = dpr / userShrinkFactor;\n\n  debug('shrunkShotToLogicalRatio', shrunkShotToLogicalRatio);\n\n  if (userShrinkFactor !== 1) {\n    const targetWidth = Math.round(imgWidth / userShrinkFactor);\n    const targetHeight = Math.round(imgHeight / userShrinkFactor);\n\n    debug(\n      `Applying screenshot shrink factor: ${userShrinkFactor} (physical: ${imgWidth}x${imgHeight} -> target: ${targetWidth}x${targetHeight})`,\n    );\n\n    const resizedBase64 = await resizeImgBase64(screenshotBase64, {\n      width: targetWidth,\n      height: targetHeight,\n    });\n    return {\n      shotSize: {\n        width: targetWidth,\n        height: targetHeight,\n      },\n      deprecatedDpr: dpr,\n      screenshot: ScreenshotItem.create(resizedBase64, screenshotCapturedAt),\n      shrunkShotToLogicalRatio,\n    };\n  }\n\n  return {\n    shotSize: {\n      width: imgWidth,\n      height: imgHeight,\n    },\n    deprecatedDpr: dpr,\n    screenshot: ScreenshotItem.create(screenshotBase64, screenshotCapturedAt),\n    shrunkShotToLogicalRatio,\n  };\n}\n\nexport async function createScreenshotBoundUIContext(\n  screenshotBase64: string,\n  opt: {\n    screenshotSize?: Size;\n  },\n): Promise<UIContext> {\n  const normalizedScreenshotBase64 = normalizeBase64Image(screenshotBase64);\n  const actualScreenshotSize = await imageInfoOfBase64(\n    normalizedScreenshotBase64,\n  );\n  if (\n    opt.screenshotSize &&\n    (opt.screenshotSize.width !== actualScreenshotSize.width ||\n      opt.screenshotSize.height !== actualScreenshotSize.height)\n  ) {\n    agentDebug(\n      'describeElementAtPoint screenshotSize mismatch, use actual size',\n      {\n        provided: opt.screenshotSize,\n        actual: actualScreenshotSize,\n      },\n    );\n  }\n\n  return {\n    screenshot: ScreenshotItem.create(normalizedScreenshotBase64, Date.now()),\n    shotSize: actualScreenshotSize,\n    shrunkShotToLogicalRatio: 1,\n    _isFrozen: true,\n  };\n}\n\nexport function getReportFileName(tag = 'web') {\n  const reportTagName = globalConfigManager.getEnvConfigValue(\n    MIDSCENE_REPORT_TAG_NAME,\n  );\n  const dateTimeInFileName = dayjs().format('YYYY-MM-DD_HH-mm-ss');\n  // ensure uniqueness at the same time\n  const uniqueId = uuid().substring(0, 8);\n  return `${reportTagName || tag}-${dateTimeInFileName}-${uniqueId}`;\n}\n\nexport function printReportMsg(filepath: string) {\n  if (globalConfigManager.getEnvConfigInBoolean(MIDSCENE_REPORT_QUIET)) {\n    return;\n  }\n  logMsg(`Midscene - report file updated: ${filepath}`);\n}\n\nexport function isPixelBbox(value: unknown): value is PixelBbox {\n  return (\n    Array.isArray(value) &&\n    value.length === 4 &&\n    value.every((item) => typeof item === 'number' && Number.isFinite(item))\n  );\n}\n\ntype PlanningLocateParamWithMaybeLocatedPixelBbox = PlanningLocateParam & {\n  locatedPixelBbox?: unknown;\n};\n\nexport function ifPlanLocateParamHasLocatedPixelBbox(\n  planLocateParam: PlanningLocateParamWithMaybeLocatedPixelBbox,\n): planLocateParam is PlanningLocateParamWithLocatedPixelBbox {\n  return isPixelBbox(planLocateParam.locatedPixelBbox);\n}\n\nexport function matchElementFromPlan(\n  planLocateParam: PlanningLocateParamWithLocatedPixelBbox,\n): LocateResultElement | undefined {\n  if (!planLocateParam) {\n    return undefined;\n  }\n\n  const rect = pixelBboxToRect(planLocateParam.locatedPixelBbox);\n\n  const element = generateElementByRect(\n    rect,\n    typeof planLocateParam.prompt === 'string'\n      ? planLocateParam.prompt\n      : planLocateParam.prompt?.prompt || '',\n  );\n  return element;\n}\n\nexport async function matchElementFromCache(\n  context: {\n    taskCache?: TaskCache;\n    interfaceInstance: AbstractInterface;\n  },\n  cacheEntry: ElementCacheFeature | undefined,\n  cachePrompt: TUserPrompt,\n  cacheable: boolean | undefined,\n): Promise<LocateResultElement | undefined> {\n  if (!cacheEntry) {\n    return undefined;\n  }\n\n  if (cacheable === false) {\n    cacheDebug('cache disabled for prompt: %s', cachePrompt);\n    return undefined;\n  }\n\n  if (!context.taskCache?.isCacheResultUsed) {\n    return undefined;\n  }\n\n  if (!context.interfaceInstance.rectMatchesCacheFeature) {\n    cacheDebug(\n      'interface does not implement rectMatchesCacheFeature, skip cache',\n    );\n    return undefined;\n  }\n\n  try {\n    const rect =\n      await context.interfaceInstance.rectMatchesCacheFeature(cacheEntry);\n    const element: LocateResultElement = {\n      center: [\n        Math.round(rect.left + rect.width / 2),\n        Math.round(rect.top + rect.height / 2),\n      ],\n      rect,\n      description:\n        typeof cachePrompt === 'string'\n          ? cachePrompt\n          : cachePrompt.prompt || '',\n    };\n\n    cacheDebug('cache hit, prompt: %s', cachePrompt);\n    return element;\n  } catch (error) {\n    cacheDebug('rectMatchesCacheFeature error: %s', error);\n    return undefined;\n  }\n}\n\ndeclare const __VERSION__: string | undefined;\n\nexport const getMidsceneVersion = (): string => {\n  if (typeof __VERSION__ !== 'undefined') {\n    return __VERSION__;\n  } else if (\n    process.env.__VERSION__ &&\n    process.env.__VERSION__ !== 'undefined'\n  ) {\n    return process.env.__VERSION__;\n  }\n  throw new Error('__VERSION__ inject failed during build');\n};\n\nexport const parsePrompt = (\n  prompt: TUserPrompt,\n): {\n  textPrompt: string;\n  multimodalPrompt?: TMultimodalPrompt;\n} => {\n  if (typeof prompt === 'string') {\n    return {\n      textPrompt: prompt,\n      multimodalPrompt: undefined,\n    };\n  }\n  return {\n    textPrompt: prompt.prompt,\n    multimodalPrompt: prompt.images\n      ? {\n          images: prompt.images,\n          convertHttpImage2Base64: !!prompt.convertHttpImage2Base64,\n        }\n      : undefined,\n  };\n};\n\nexport const transformLogicalElementToScreenshot = (\n  element: LocateResultElement,\n  shrunkShotToLogicalRatio: number,\n): LocateResultElement => {\n  if (shrunkShotToLogicalRatio === 1) {\n    return element;\n  }\n\n  return {\n    ...element,\n    center: [\n      Math.round(element.center[0] * shrunkShotToLogicalRatio),\n      Math.round(element.center[1] * shrunkShotToLogicalRatio),\n    ],\n    rect: {\n      ...element.rect,\n      left: Math.round(element.rect.left * shrunkShotToLogicalRatio),\n      top: Math.round(element.rect.top * shrunkShotToLogicalRatio),\n      width: Math.round(element.rect.width * shrunkShotToLogicalRatio),\n      height: Math.round(element.rect.height * shrunkShotToLogicalRatio),\n    },\n  };\n};\n\nexport const transformLogicalRectToScreenshotRect = (\n  rect: Rect,\n  shrunkShotToLogicalRatio: number,\n): Rect => {\n  if (shrunkShotToLogicalRatio === 1) {\n    return rect;\n  }\n\n  return {\n    ...rect,\n    left: Math.round(rect.left * shrunkShotToLogicalRatio),\n    top: Math.round(rect.top * shrunkShotToLogicalRatio),\n    width: Math.round(rect.width * shrunkShotToLogicalRatio),\n    height: Math.round(rect.height * shrunkShotToLogicalRatio),\n  };\n};\n"],"names":["agentDebug","getDebug","commonContextParser","interfaceInstance","_opt","debug","assert","description","uploadTestInfoToServer","interfaceSize","logicalWidth","logicalHeight","console","Number","Error","screenshotBase64","screenshotCapturedAt","Date","imgWidth","imgHeight","imageInfoOfBase64","logicalIsPortrait","screenshotIsPortrait","finalLogicalWidth","userShrinkFactor","dpr","shrunkShotToLogicalRatio","targetWidth","Math","targetHeight","resizedBase64","resizeImgBase64","ScreenshotItem","createScreenshotBoundUIContext","opt","normalizedScreenshotBase64","normalizeBase64Image","actualScreenshotSize","getReportFileName","tag","reportTagName","globalConfigManager","MIDSCENE_REPORT_TAG_NAME","dateTimeInFileName","dayjs","uniqueId","uuid","printReportMsg","filepath","MIDSCENE_REPORT_QUIET","logMsg","isPixelBbox","value","Array","item","ifPlanLocateParamHasLocatedPixelBbox","planLocateParam","matchElementFromPlan","rect","pixelBboxToRect","element","generateElementByRect","matchElementFromCache","context","cacheEntry","cachePrompt","cacheable","cacheDebug","error","getMidsceneVersion","__VERSION__","parsePrompt","prompt","undefined","transformLogicalElementToScreenshot","transformLogicalRectToScreenshotRect"],"mappings":";;;;;;;;;;AAiCA,MAAMA,aAAaC,SAAS;AAErB,eAAeC,oBACpBC,iBAAoC,EACpCC,IAGC;IAED,MAAMC,QAAQJ,SAAS;IAEvBK,OAAOH,mBAAmB;IAE1BE,MAAM;IACN,MAAME,cAAcJ,kBAAkB,QAAQ,QAAQ;IACtDE,MAAM;IAENA,MAAM;IACNG,uBAAuB;QACrB,SAASD;QACT,WAAWH,KAAK,eAAe;IACjC;IACAC,MAAM;IAENA,MAAM;IACN,MAAMI,gBAAgB,MAAMN,kBAAkB,IAAI;IAClD,MAAM,EAAE,OAAOO,YAAY,EAAE,QAAQC,aAAa,EAAE,GAAGF;IAEvD,IAAKA,cAA6C,GAAG,EACnDG,QAAQ,IAAI,CACV;IAIJ,IAAI,CAACC,OAAO,QAAQ,CAACH,iBAAiB,CAACG,OAAO,QAAQ,CAACF,gBACrD,MAAM,IAAIG,MACR,CAAC,iFAAiF,EAAEJ,aAAa,UAAU,EAAEC,eAAe;IAIhI,IAAID,gBAAgB,KAAKC,iBAAiB,GACxC,MAAM,IAAIG,MACR,CAAC,mFAAmF,EAAEJ,aAAa,UAAU,EAAEC,eAAe;IAIlIN,MAAM,CAAC,MAAM,EAAEK,aAAa,CAAC,EAAEC,eAAe;IAE9C,MAAMI,mBAAmB,MAAMZ,kBAAkB,gBAAgB;IACjE,MAAMa,uBAAuBC,KAAK,GAAG;IACrCX,OAAOS,kBAAmB;IAG1BV,MAAM;IACN,MAAM,EAAE,OAAOa,QAAQ,EAAE,QAAQC,SAAS,EAAE,GAC1C,MAAMC,kBAAkBL;IAE1B,IAAI,CAACF,OAAO,QAAQ,CAACK,aAAa,CAACL,OAAO,QAAQ,CAACM,YACjD,MAAM,IAAIL,MACR,CAAC,wFAAwF,EAAEI,SAAS,UAAU,EAAEC,WAAW;IAG/H,IAAID,YAAY,KAAKC,aAAa,GAChC,MAAM,IAAIL,MACR,CAAC,0FAA0F,EAAEI,SAAS,UAAU,EAAEC,WAAW;IAGjId,MAAM,yBAAyBa,UAAU,KAAKC;IAM9C,MAAME,oBAAoBX,eAAeC;IACzC,MAAMW,uBAAuBJ,WAAWC;IACxC,IAAII,oBAAoBb;IAExB,IAAIW,sBAAsBC,sBAAsB;QAC9CjB,MACE,CAAC,4CAA4C,EAAEK,aAAa,CAAC,EAAEC,cAAc,EAAE,EAAEU,oBAAoB,aAAa,YAAY,gBAAgB,EAAEH,SAAS,CAAC,EAAEC,UAAU,EAAE,EAAEG,uBAAuB,aAAa,YAAY,+BAA+B,CAAC;QAE5PC,oBAAoBZ;IAEtB;IAEA,MAAMa,mBAAmBpB,KAAK,sBAAsB,IAAI;IAExD,IAAI,CAACS,OAAO,QAAQ,CAACW,qBAAqBA,mBAAmB,GAC3D,MAAM,IAAIV,MACR,CAAC,wEAAwE,EAAEU,kBAAkB;IAIjG,MAAMC,MAAMP,WAAWK;IAEvBlB,MAAM,mBAAmBoB;IAEzB,MAAMC,2BAA2BD,MAAMD;IAEvCnB,MAAM,4BAA4BqB;IAElC,IAAIF,AAAqB,MAArBA,kBAAwB;QAC1B,MAAMG,cAAcC,KAAK,KAAK,CAACV,WAAWM;QAC1C,MAAMK,eAAeD,KAAK,KAAK,CAACT,YAAYK;QAE5CnB,MACE,CAAC,mCAAmC,EAAEmB,iBAAiB,YAAY,EAAEN,SAAS,CAAC,EAAEC,UAAU,YAAY,EAAEQ,YAAY,CAAC,EAAEE,aAAa,CAAC,CAAC;QAGzI,MAAMC,gBAAgB,MAAMC,gBAAgBhB,kBAAkB;YAC5D,OAAOY;YACP,QAAQE;QACV;QACA,OAAO;YACL,UAAU;gBACR,OAAOF;gBACP,QAAQE;YACV;YACA,eAAeJ;YACf,YAAYO,eAAe,MAAM,CAACF,eAAed;YACjDU;QACF;IACF;IAEA,OAAO;QACL,UAAU;YACR,OAAOR;YACP,QAAQC;QACV;QACA,eAAeM;QACf,YAAYO,eAAe,MAAM,CAACjB,kBAAkBC;QACpDU;IACF;AACF;AAEO,eAAeO,+BACpBlB,gBAAwB,EACxBmB,GAEC;IAED,MAAMC,6BAA6BC,qBAAqBrB;IACxD,MAAMsB,uBAAuB,MAAMjB,kBACjCe;IAEF,IACED,IAAI,cAAc,IACjBA,CAAAA,IAAI,cAAc,CAAC,KAAK,KAAKG,qBAAqB,KAAK,IACtDH,IAAI,cAAc,CAAC,MAAM,KAAKG,qBAAqB,MAAK,GAE1DrC,WACE,mEACA;QACE,UAAUkC,IAAI,cAAc;QAC5B,QAAQG;IACV;IAIJ,OAAO;QACL,YAAYL,eAAe,MAAM,CAACG,4BAA4BlB,KAAK,GAAG;QACtE,UAAUoB;QACV,0BAA0B;QAC1B,WAAW;IACb;AACF;AAEO,SAASC,kBAAkBC,MAAM,KAAK;IAC3C,MAAMC,gBAAgBC,oBAAoB,iBAAiB,CACzDC;IAEF,MAAMC,qBAAqBC,QAAQ,MAAM,CAAC;IAE1C,MAAMC,WAAWC,OAAO,SAAS,CAAC,GAAG;IACrC,OAAO,GAAGN,iBAAiBD,IAAI,CAAC,EAAEI,mBAAmB,CAAC,EAAEE,UAAU;AACpE;AAEO,SAASE,eAAeC,QAAgB;IAC7C,IAAIP,oBAAoB,qBAAqB,CAACQ,wBAC5C;IAEFC,OAAO,CAAC,gCAAgC,EAAEF,UAAU;AACtD;AAEO,SAASG,YAAYC,KAAc;IACxC,OACEC,MAAM,OAAO,CAACD,UACdA,AAAiB,MAAjBA,MAAM,MAAM,IACZA,MAAM,KAAK,CAAC,CAACE,OAAS,AAAgB,YAAhB,OAAOA,QAAqBzC,OAAO,QAAQ,CAACyC;AAEtE;AAMO,SAASC,qCACdC,eAA6D;IAE7D,OAAOL,YAAYK,gBAAgB,gBAAgB;AACrD;AAEO,SAASC,qBACdD,eAAwD;IAExD,IAAI,CAACA,iBACH;IAGF,MAAME,OAAOC,gBAAgBH,gBAAgB,gBAAgB;IAE7D,MAAMI,UAAUC,sBACdH,MACA,AAAkC,YAAlC,OAAOF,gBAAgB,MAAM,GACzBA,gBAAgB,MAAM,GACtBA,gBAAgB,MAAM,EAAE,UAAU;IAExC,OAAOI;AACT;AAEO,eAAeE,sBACpBC,OAGC,EACDC,UAA2C,EAC3CC,WAAwB,EACxBC,SAA8B;IAE9B,IAAI,CAACF,YACH;IAGF,IAAIE,AAAc,UAAdA,WAAqB,YACvBC,8BAAW,iCAAiCF;IAI9C,IAAI,CAACF,QAAQ,SAAS,EAAE,mBACtB;IAGF,IAAI,CAACA,QAAQ,iBAAiB,CAAC,uBAAuB,EAAE,YACtDI,8BACE;IAKJ,IAAI;QACF,MAAMT,OACJ,MAAMK,QAAQ,iBAAiB,CAAC,uBAAuB,CAACC;QAC1D,MAAMJ,UAA+B;YACnC,QAAQ;gBACNhC,KAAK,KAAK,CAAC8B,KAAK,IAAI,GAAGA,KAAK,KAAK,GAAG;gBACpC9B,KAAK,KAAK,CAAC8B,KAAK,GAAG,GAAGA,KAAK,MAAM,GAAG;aACrC;YACDA;YACA,aACE,AAAuB,YAAvB,OAAOO,cACHA,cACAA,YAAY,MAAM,IAAI;QAC9B;QAEAE,8BAAW,yBAAyBF;QACpC,OAAOL;IACT,EAAE,OAAOQ,OAAO;QACdD,8BAAW,qCAAqCC;QAChD;IACF;AACF;AAIO,MAAMC,qBAAqB,IAEvBC;AAUJ,MAAMC,cAAc,CACzBC;IAKA,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAO;QACL,YAAYA;QACZ,kBAAkBC;IACpB;IAEF,OAAO;QACL,YAAYD,OAAO,MAAM;QACzB,kBAAkBA,OAAO,MAAM,GAC3B;YACE,QAAQA,OAAO,MAAM;YACrB,yBAAyB,CAAC,CAACA,OAAO,uBAAuB;QAC3D,IACAC;IACN;AACF;AAEO,MAAMC,sCAAsC,CACjDd,SACAlC;IAEA,IAAIA,AAA6B,MAA7BA,0BACF,OAAOkC;IAGT,OAAO;QACL,GAAGA,OAAO;QACV,QAAQ;YACNhC,KAAK,KAAK,CAACgC,QAAQ,MAAM,CAAC,EAAE,GAAGlC;YAC/BE,KAAK,KAAK,CAACgC,QAAQ,MAAM,CAAC,EAAE,GAAGlC;SAChC;QACD,MAAM;YACJ,GAAGkC,QAAQ,IAAI;YACf,MAAMhC,KAAK,KAAK,CAACgC,QAAQ,IAAI,CAAC,IAAI,GAAGlC;YACrC,KAAKE,KAAK,KAAK,CAACgC,QAAQ,IAAI,CAAC,GAAG,GAAGlC;YACnC,OAAOE,KAAK,KAAK,CAACgC,QAAQ,IAAI,CAAC,KAAK,GAAGlC;YACvC,QAAQE,KAAK,KAAK,CAACgC,QAAQ,IAAI,CAAC,MAAM,GAAGlC;QAC3C;IACF;AACF;AAEO,MAAMiD,uCAAuC,CAClDjB,MACAhC;IAEA,IAAIA,AAA6B,MAA7BA,0BACF,OAAOgC;IAGT,OAAO;QACL,GAAGA,IAAI;QACP,MAAM9B,KAAK,KAAK,CAAC8B,KAAK,IAAI,GAAGhC;QAC7B,KAAKE,KAAK,KAAK,CAAC8B,KAAK,GAAG,GAAGhC;QAC3B,OAAOE,KAAK,KAAK,CAAC8B,KAAK,KAAK,GAAGhC;QAC/B,QAAQE,KAAK,KAAK,CAAC8B,KAAK,MAAM,GAAGhC;IACnC;AACF"}
+{"version":3,"file":"agent/utils.mjs","sources":["../../../src/agent/utils.ts"],"sourcesContent":["import { pixelBboxToRect } from '@/ai-model/workflows/inspect/locate-result-rect';\nimport type { TMultimodalPrompt, TUserPrompt } from '@/common';\nimport type { AbstractInterface } from '@/device';\nimport { ScreenshotItem } from '@/screenshot-item';\nimport type {\n  ElementCacheFeature,\n  LocateResultElement,\n  PixelBbox,\n  PlanningLocateParam,\n  PlanningLocateParamWithLocatedPixelBbox,\n  Rect,\n  ScrollParam,\n  Size,\n  UIContext,\n} from '@/types';\nimport { uploadTestInfoToServer } from '@/utils';\nimport {\n  MIDSCENE_REPORT_QUIET,\n  MIDSCENE_REPORT_TAG_NAME,\n  globalConfigManager,\n} from '@midscene/shared/env';\nimport { generateElementByRect } from '@midscene/shared/extractor';\nimport {\n  imageInfoOfBase64,\n  normalizeBase64Image,\n  resizeImgBase64,\n} from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { assert, logMsg, uuid } from '@midscene/shared/utils';\nimport dayjs from 'dayjs';\nimport type { TaskCache } from './task-cache';\nimport { debug as cacheDebug } from './task-cache';\n\nconst agentDebug = getDebug('agent');\n\nconst legacyScrollTypeMap = {\n  once: 'singleAction',\n  untilBottom: 'scrollToBottom',\n  untilTop: 'scrollToTop',\n  untilRight: 'scrollToRight',\n  untilLeft: 'scrollToLeft',\n} as const;\n\nexport const normalizeScrollType = (\n  scrollType: string | undefined,\n): ScrollParam['scrollType'] | undefined => {\n  if (!scrollType) {\n    return undefined;\n  }\n\n  if (scrollType in legacyScrollTypeMap) {\n    return legacyScrollTypeMap[scrollType as keyof typeof legacyScrollTypeMap];\n  }\n\n  return scrollType as ScrollParam['scrollType'];\n};\n\nexport async function commonContextParser(\n  interfaceInstance: AbstractInterface,\n  _opt: {\n    uploadServerUrl?: string;\n    screenshotShrinkFactor?: number;\n  },\n): Promise<UIContext> {\n  const debug = getDebug('commonContextParser');\n\n  assert(interfaceInstance, 'interfaceInstance is required');\n\n  debug('Getting interface description');\n  const description = interfaceInstance.describe?.() || '';\n  debug('Interface description end');\n\n  debug('Uploading test info to server');\n  uploadTestInfoToServer({\n    testUrl: description,\n    serverUrl: _opt.uploadServerUrl,\n  });\n  debug('UploadTestInfoToServer end');\n\n  debug('will get size');\n  const interfaceSize = await interfaceInstance.size();\n  const { width: logicalWidth, height: logicalHeight } = interfaceSize;\n\n  if ((interfaceSize as unknown as { dpr: number }).dpr) {\n    console.warn(\n      'Warning: return value of interface.size() include a dpr property, which is not expected and ignored. ',\n    );\n  }\n\n  if (!Number.isFinite(logicalWidth) || !Number.isFinite(logicalHeight)) {\n    throw new Error(\n      `Invalid interface size: width and height must be finite numbers. Received width: ${logicalWidth}, height: ${logicalHeight}`,\n    );\n  }\n\n  if (logicalWidth <= 0 || logicalHeight <= 0) {\n    throw new Error(\n      `Invalid interface size: width and height must be positive numbers. Received width: ${logicalWidth}, height: ${logicalHeight}`,\n    );\n  }\n\n  debug(`size: ${logicalWidth}x${logicalHeight}`);\n\n  const screenshotBase64 = await interfaceInstance.screenshotBase64();\n  const screenshotCapturedAt = Date.now();\n  assert(screenshotBase64!, 'screenshotBase64 is required');\n\n  // Get physical screenshot dimensions\n  debug('will get screenshot dimensions');\n  const { width: imgWidth, height: imgHeight } =\n    await imageInfoOfBase64(screenshotBase64);\n\n  if (!Number.isFinite(imgWidth) || !Number.isFinite(imgHeight)) {\n    throw new Error(\n      `Invalid screenshot dimensions: width and height must be finite numbers. Received width: ${imgWidth}, height: ${imgHeight}`,\n    );\n  }\n  if (imgWidth <= 0 || imgHeight <= 0) {\n    throw new Error(\n      `Invalid screenshot dimensions: width and height must be positive numbers. Received width: ${imgWidth}, height: ${imgHeight}`,\n    );\n  }\n  debug('screenshot dimensions', imgWidth, 'x', imgHeight);\n\n  // Detect orientation mismatch between logical size and screenshot.\n  // Some devices (e.g. OPPO) report wrong orientation via ADB, causing\n  // size() to return portrait dimensions even when the device is landscape.\n  // We detect this by comparing aspect ratios and swap if they disagree.\n  const logicalIsPortrait = logicalWidth < logicalHeight;\n  const screenshotIsPortrait = imgWidth < imgHeight;\n  let finalLogicalWidth = logicalWidth;\n  let finalLogicalHeight = logicalHeight;\n  if (logicalIsPortrait !== screenshotIsPortrait) {\n    debug(\n      `Orientation mismatch detected: logical size ${logicalWidth}x${logicalHeight} (${logicalIsPortrait ? 'portrait' : 'landscape'}) vs screenshot ${imgWidth}x${imgHeight} (${screenshotIsPortrait ? 'portrait' : 'landscape'}). Swapping logical dimensions.`,\n    );\n    finalLogicalWidth = logicalHeight;\n    finalLogicalHeight = logicalWidth;\n  }\n\n  const userShrinkFactor = _opt.screenshotShrinkFactor ?? 1;\n\n  if (!Number.isFinite(userShrinkFactor) || userShrinkFactor < 1) {\n    throw new Error(\n      `Invalid screenshotShrinkFactor: must be a finite number >= 1. Received: ${userShrinkFactor}`,\n    );\n  }\n\n  const dpr = imgWidth / finalLogicalWidth;\n\n  debug('calculated dpr:', dpr);\n\n  const shrunkShotToLogicalRatio = dpr / userShrinkFactor;\n\n  debug('shrunkShotToLogicalRatio', shrunkShotToLogicalRatio);\n\n  if (userShrinkFactor !== 1) {\n    const targetWidth = Math.round(imgWidth / userShrinkFactor);\n    const targetHeight = Math.round(imgHeight / userShrinkFactor);\n\n    debug(\n      `Applying screenshot shrink factor: ${userShrinkFactor} (physical: ${imgWidth}x${imgHeight} -> target: ${targetWidth}x${targetHeight})`,\n    );\n\n    const resizedBase64 = await resizeImgBase64(screenshotBase64, {\n      width: targetWidth,\n      height: targetHeight,\n    });\n    return {\n      shotSize: {\n        width: targetWidth,\n        height: targetHeight,\n      },\n      deprecatedDpr: dpr,\n      screenshot: ScreenshotItem.create(resizedBase64, screenshotCapturedAt),\n      shrunkShotToLogicalRatio,\n    };\n  }\n\n  return {\n    shotSize: {\n      width: imgWidth,\n      height: imgHeight,\n    },\n    deprecatedDpr: dpr,\n    screenshot: ScreenshotItem.create(screenshotBase64, screenshotCapturedAt),\n    shrunkShotToLogicalRatio,\n  };\n}\n\nexport async function createScreenshotBoundUIContext(\n  screenshotBase64: string,\n  opt: {\n    screenshotSize?: Size;\n  },\n): Promise<UIContext> {\n  const normalizedScreenshotBase64 = normalizeBase64Image(screenshotBase64);\n  const actualScreenshotSize = await imageInfoOfBase64(\n    normalizedScreenshotBase64,\n  );\n  if (\n    opt.screenshotSize &&\n    (opt.screenshotSize.width !== actualScreenshotSize.width ||\n      opt.screenshotSize.height !== actualScreenshotSize.height)\n  ) {\n    agentDebug(\n      'describeElementAtPoint screenshotSize mismatch, use actual size',\n      {\n        provided: opt.screenshotSize,\n        actual: actualScreenshotSize,\n      },\n    );\n  }\n\n  return {\n    screenshot: ScreenshotItem.create(normalizedScreenshotBase64, Date.now()),\n    shotSize: actualScreenshotSize,\n    shrunkShotToLogicalRatio: 1,\n    _isFrozen: true,\n  };\n}\n\nexport function getReportFileName(tag = 'web') {\n  const reportTagName = globalConfigManager.getEnvConfigValue(\n    MIDSCENE_REPORT_TAG_NAME,\n  );\n  const dateTimeInFileName = dayjs().format('YYYY-MM-DD_HH-mm-ss');\n  // ensure uniqueness at the same time\n  const uniqueId = uuid().substring(0, 8);\n  return `${reportTagName || tag}-${dateTimeInFileName}-${uniqueId}`;\n}\n\nexport function printReportMsg(filepath: string) {\n  if (globalConfigManager.getEnvConfigInBoolean(MIDSCENE_REPORT_QUIET)) {\n    return;\n  }\n  logMsg(`Midscene - report file updated: ${filepath}`);\n}\n\nexport function isPixelBbox(value: unknown): value is PixelBbox {\n  return (\n    Array.isArray(value) &&\n    value.length === 4 &&\n    value.every((item) => typeof item === 'number' && Number.isFinite(item))\n  );\n}\n\ntype PlanningLocateParamWithMaybeLocatedPixelBbox = PlanningLocateParam & {\n  locatedPixelBbox?: unknown;\n};\n\nexport function ifPlanLocateParamHasLocatedPixelBbox(\n  planLocateParam: PlanningLocateParamWithMaybeLocatedPixelBbox,\n): planLocateParam is PlanningLocateParamWithLocatedPixelBbox {\n  return isPixelBbox(planLocateParam.locatedPixelBbox);\n}\n\nexport function matchElementFromPlan(\n  planLocateParam: PlanningLocateParamWithLocatedPixelBbox,\n): LocateResultElement | undefined {\n  if (!planLocateParam) {\n    return undefined;\n  }\n\n  const rect = pixelBboxToRect(planLocateParam.locatedPixelBbox);\n\n  const element = generateElementByRect(\n    rect,\n    typeof planLocateParam.prompt === 'string'\n      ? planLocateParam.prompt\n      : planLocateParam.prompt?.prompt || '',\n  );\n  return element;\n}\n\nexport async function matchElementFromCache(\n  context: {\n    taskCache?: TaskCache;\n    interfaceInstance: AbstractInterface;\n  },\n  cacheEntry: ElementCacheFeature | undefined,\n  cachePrompt: TUserPrompt,\n  cacheable: boolean | undefined,\n): Promise<LocateResultElement | undefined> {\n  if (!cacheEntry) {\n    return undefined;\n  }\n\n  if (cacheable === false) {\n    cacheDebug('cache disabled for prompt: %s', cachePrompt);\n    return undefined;\n  }\n\n  if (!context.taskCache?.isCacheResultUsed) {\n    return undefined;\n  }\n\n  if (!context.interfaceInstance.rectMatchesCacheFeature) {\n    cacheDebug(\n      'interface does not implement rectMatchesCacheFeature, skip cache',\n    );\n    return undefined;\n  }\n\n  try {\n    const rect =\n      await context.interfaceInstance.rectMatchesCacheFeature(cacheEntry);\n    const element: LocateResultElement = {\n      center: [\n        Math.round(rect.left + rect.width / 2),\n        Math.round(rect.top + rect.height / 2),\n      ],\n      rect,\n      description:\n        typeof cachePrompt === 'string'\n          ? cachePrompt\n          : cachePrompt.prompt || '',\n    };\n\n    cacheDebug('cache hit, prompt: %s', cachePrompt);\n    return element;\n  } catch (error) {\n    cacheDebug('rectMatchesCacheFeature error: %s', error);\n    return undefined;\n  }\n}\n\ndeclare const __VERSION__: string | undefined;\n\nexport const getMidsceneVersion = (): string => {\n  if (typeof __VERSION__ !== 'undefined') {\n    return __VERSION__;\n  } else if (\n    process.env.__VERSION__ &&\n    process.env.__VERSION__ !== 'undefined'\n  ) {\n    return process.env.__VERSION__;\n  }\n  throw new Error('__VERSION__ inject failed during build');\n};\n\nexport const parsePrompt = (\n  prompt: TUserPrompt,\n): {\n  textPrompt: string;\n  multimodalPrompt?: TMultimodalPrompt;\n} => {\n  if (typeof prompt === 'string') {\n    return {\n      textPrompt: prompt,\n      multimodalPrompt: undefined,\n    };\n  }\n  return {\n    textPrompt: prompt.prompt,\n    multimodalPrompt: prompt.images\n      ? {\n          images: prompt.images,\n          convertHttpImage2Base64: !!prompt.convertHttpImage2Base64,\n        }\n      : undefined,\n  };\n};\n\nexport const transformLogicalElementToScreenshot = (\n  element: LocateResultElement,\n  shrunkShotToLogicalRatio: number,\n): LocateResultElement => {\n  if (shrunkShotToLogicalRatio === 1) {\n    return element;\n  }\n\n  return {\n    ...element,\n    center: [\n      Math.round(element.center[0] * shrunkShotToLogicalRatio),\n      Math.round(element.center[1] * shrunkShotToLogicalRatio),\n    ],\n    rect: {\n      ...element.rect,\n      left: Math.round(element.rect.left * shrunkShotToLogicalRatio),\n      top: Math.round(element.rect.top * shrunkShotToLogicalRatio),\n      width: Math.round(element.rect.width * shrunkShotToLogicalRatio),\n      height: Math.round(element.rect.height * shrunkShotToLogicalRatio),\n    },\n  };\n};\n\nexport const transformLogicalRectToScreenshotRect = (\n  rect: Rect,\n  shrunkShotToLogicalRatio: number,\n): Rect => {\n  if (shrunkShotToLogicalRatio === 1) {\n    return rect;\n  }\n\n  return {\n    ...rect,\n    left: Math.round(rect.left * shrunkShotToLogicalRatio),\n    top: Math.round(rect.top * shrunkShotToLogicalRatio),\n    width: Math.round(rect.width * shrunkShotToLogicalRatio),\n    height: Math.round(rect.height * shrunkShotToLogicalRatio),\n  };\n};\n"],"names":["agentDebug","getDebug","legacyScrollTypeMap","normalizeScrollType","scrollType","commonContextParser","interfaceInstance","_opt","debug","assert","description","uploadTestInfoToServer","interfaceSize","logicalWidth","logicalHeight","console","Number","Error","screenshotBase64","screenshotCapturedAt","Date","imgWidth","imgHeight","imageInfoOfBase64","logicalIsPortrait","screenshotIsPortrait","finalLogicalWidth","userShrinkFactor","dpr","shrunkShotToLogicalRatio","targetWidth","Math","targetHeight","resizedBase64","resizeImgBase64","ScreenshotItem","createScreenshotBoundUIContext","opt","normalizedScreenshotBase64","normalizeBase64Image","actualScreenshotSize","getReportFileName","tag","reportTagName","globalConfigManager","MIDSCENE_REPORT_TAG_NAME","dateTimeInFileName","dayjs","uniqueId","uuid","printReportMsg","filepath","MIDSCENE_REPORT_QUIET","logMsg","isPixelBbox","value","Array","item","ifPlanLocateParamHasLocatedPixelBbox","planLocateParam","matchElementFromPlan","rect","pixelBboxToRect","element","generateElementByRect","matchElementFromCache","context","cacheEntry","cachePrompt","cacheable","cacheDebug","error","getMidsceneVersion","__VERSION__","parsePrompt","prompt","undefined","transformLogicalElementToScreenshot","transformLogicalRectToScreenshotRect"],"mappings":";;;;;;;;;;AAkCA,MAAMA,aAAaC,SAAS;AAE5B,MAAMC,sBAAsB;IAC1B,MAAM;IACN,aAAa;IACb,UAAU;IACV,YAAY;IACZ,WAAW;AACb;AAEO,MAAMC,sBAAsB,CACjCC;IAEA,IAAI,CAACA,YACH;IAGF,IAAIA,cAAcF,qBAChB,OAAOA,mBAAmB,CAACE,WAA+C;IAG5E,OAAOA;AACT;AAEO,eAAeC,oBACpBC,iBAAoC,EACpCC,IAGC;IAED,MAAMC,QAAQP,SAAS;IAEvBQ,OAAOH,mBAAmB;IAE1BE,MAAM;IACN,MAAME,cAAcJ,kBAAkB,QAAQ,QAAQ;IACtDE,MAAM;IAENA,MAAM;IACNG,uBAAuB;QACrB,SAASD;QACT,WAAWH,KAAK,eAAe;IACjC;IACAC,MAAM;IAENA,MAAM;IACN,MAAMI,gBAAgB,MAAMN,kBAAkB,IAAI;IAClD,MAAM,EAAE,OAAOO,YAAY,EAAE,QAAQC,aAAa,EAAE,GAAGF;IAEvD,IAAKA,cAA6C,GAAG,EACnDG,QAAQ,IAAI,CACV;IAIJ,IAAI,CAACC,OAAO,QAAQ,CAACH,iBAAiB,CAACG,OAAO,QAAQ,CAACF,gBACrD,MAAM,IAAIG,MACR,CAAC,iFAAiF,EAAEJ,aAAa,UAAU,EAAEC,eAAe;IAIhI,IAAID,gBAAgB,KAAKC,iBAAiB,GACxC,MAAM,IAAIG,MACR,CAAC,mFAAmF,EAAEJ,aAAa,UAAU,EAAEC,eAAe;IAIlIN,MAAM,CAAC,MAAM,EAAEK,aAAa,CAAC,EAAEC,eAAe;IAE9C,MAAMI,mBAAmB,MAAMZ,kBAAkB,gBAAgB;IACjE,MAAMa,uBAAuBC,KAAK,GAAG;IACrCX,OAAOS,kBAAmB;IAG1BV,MAAM;IACN,MAAM,EAAE,OAAOa,QAAQ,EAAE,QAAQC,SAAS,EAAE,GAC1C,MAAMC,kBAAkBL;IAE1B,IAAI,CAACF,OAAO,QAAQ,CAACK,aAAa,CAACL,OAAO,QAAQ,CAACM,YACjD,MAAM,IAAIL,MACR,CAAC,wFAAwF,EAAEI,SAAS,UAAU,EAAEC,WAAW;IAG/H,IAAID,YAAY,KAAKC,aAAa,GAChC,MAAM,IAAIL,MACR,CAAC,0FAA0F,EAAEI,SAAS,UAAU,EAAEC,WAAW;IAGjId,MAAM,yBAAyBa,UAAU,KAAKC;IAM9C,MAAME,oBAAoBX,eAAeC;IACzC,MAAMW,uBAAuBJ,WAAWC;IACxC,IAAII,oBAAoBb;IAExB,IAAIW,sBAAsBC,sBAAsB;QAC9CjB,MACE,CAAC,4CAA4C,EAAEK,aAAa,CAAC,EAAEC,cAAc,EAAE,EAAEU,oBAAoB,aAAa,YAAY,gBAAgB,EAAEH,SAAS,CAAC,EAAEC,UAAU,EAAE,EAAEG,uBAAuB,aAAa,YAAY,+BAA+B,CAAC;QAE5PC,oBAAoBZ;IAEtB;IAEA,MAAMa,mBAAmBpB,KAAK,sBAAsB,IAAI;IAExD,IAAI,CAACS,OAAO,QAAQ,CAACW,qBAAqBA,mBAAmB,GAC3D,MAAM,IAAIV,MACR,CAAC,wEAAwE,EAAEU,kBAAkB;IAIjG,MAAMC,MAAMP,WAAWK;IAEvBlB,MAAM,mBAAmBoB;IAEzB,MAAMC,2BAA2BD,MAAMD;IAEvCnB,MAAM,4BAA4BqB;IAElC,IAAIF,AAAqB,MAArBA,kBAAwB;QAC1B,MAAMG,cAAcC,KAAK,KAAK,CAACV,WAAWM;QAC1C,MAAMK,eAAeD,KAAK,KAAK,CAACT,YAAYK;QAE5CnB,MACE,CAAC,mCAAmC,EAAEmB,iBAAiB,YAAY,EAAEN,SAAS,CAAC,EAAEC,UAAU,YAAY,EAAEQ,YAAY,CAAC,EAAEE,aAAa,CAAC,CAAC;QAGzI,MAAMC,gBAAgB,MAAMC,gBAAgBhB,kBAAkB;YAC5D,OAAOY;YACP,QAAQE;QACV;QACA,OAAO;YACL,UAAU;gBACR,OAAOF;gBACP,QAAQE;YACV;YACA,eAAeJ;YACf,YAAYO,eAAe,MAAM,CAACF,eAAed;YACjDU;QACF;IACF;IAEA,OAAO;QACL,UAAU;YACR,OAAOR;YACP,QAAQC;QACV;QACA,eAAeM;QACf,YAAYO,eAAe,MAAM,CAACjB,kBAAkBC;QACpDU;IACF;AACF;AAEO,eAAeO,+BACpBlB,gBAAwB,EACxBmB,GAEC;IAED,MAAMC,6BAA6BC,qBAAqBrB;IACxD,MAAMsB,uBAAuB,MAAMjB,kBACjCe;IAEF,IACED,IAAI,cAAc,IACjBA,CAAAA,IAAI,cAAc,CAAC,KAAK,KAAKG,qBAAqB,KAAK,IACtDH,IAAI,cAAc,CAAC,MAAM,KAAKG,qBAAqB,MAAK,GAE1DxC,WACE,mEACA;QACE,UAAUqC,IAAI,cAAc;QAC5B,QAAQG;IACV;IAIJ,OAAO;QACL,YAAYL,eAAe,MAAM,CAACG,4BAA4BlB,KAAK,GAAG;QACtE,UAAUoB;QACV,0BAA0B;QAC1B,WAAW;IACb;AACF;AAEO,SAASC,kBAAkBC,MAAM,KAAK;IAC3C,MAAMC,gBAAgBC,oBAAoB,iBAAiB,CACzDC;IAEF,MAAMC,qBAAqBC,QAAQ,MAAM,CAAC;IAE1C,MAAMC,WAAWC,OAAO,SAAS,CAAC,GAAG;IACrC,OAAO,GAAGN,iBAAiBD,IAAI,CAAC,EAAEI,mBAAmB,CAAC,EAAEE,UAAU;AACpE;AAEO,SAASE,eAAeC,QAAgB;IAC7C,IAAIP,oBAAoB,qBAAqB,CAACQ,wBAC5C;IAEFC,OAAO,CAAC,gCAAgC,EAAEF,UAAU;AACtD;AAEO,SAASG,YAAYC,KAAc;IACxC,OACEC,MAAM,OAAO,CAACD,UACdA,AAAiB,MAAjBA,MAAM,MAAM,IACZA,MAAM,KAAK,CAAC,CAACE,OAAS,AAAgB,YAAhB,OAAOA,QAAqBzC,OAAO,QAAQ,CAACyC;AAEtE;AAMO,SAASC,qCACdC,eAA6D;IAE7D,OAAOL,YAAYK,gBAAgB,gBAAgB;AACrD;AAEO,SAASC,qBACdD,eAAwD;IAExD,IAAI,CAACA,iBACH;IAGF,MAAME,OAAOC,gBAAgBH,gBAAgB,gBAAgB;IAE7D,MAAMI,UAAUC,sBACdH,MACA,AAAkC,YAAlC,OAAOF,gBAAgB,MAAM,GACzBA,gBAAgB,MAAM,GACtBA,gBAAgB,MAAM,EAAE,UAAU;IAExC,OAAOI;AACT;AAEO,eAAeE,sBACpBC,OAGC,EACDC,UAA2C,EAC3CC,WAAwB,EACxBC,SAA8B;IAE9B,IAAI,CAACF,YACH;IAGF,IAAIE,AAAc,UAAdA,WAAqB,YACvBC,8BAAW,iCAAiCF;IAI9C,IAAI,CAACF,QAAQ,SAAS,EAAE,mBACtB;IAGF,IAAI,CAACA,QAAQ,iBAAiB,CAAC,uBAAuB,EAAE,YACtDI,8BACE;IAKJ,IAAI;QACF,MAAMT,OACJ,MAAMK,QAAQ,iBAAiB,CAAC,uBAAuB,CAACC;QAC1D,MAAMJ,UAA+B;YACnC,QAAQ;gBACNhC,KAAK,KAAK,CAAC8B,KAAK,IAAI,GAAGA,KAAK,KAAK,GAAG;gBACpC9B,KAAK,KAAK,CAAC8B,KAAK,GAAG,GAAGA,KAAK,MAAM,GAAG;aACrC;YACDA;YACA,aACE,AAAuB,YAAvB,OAAOO,cACHA,cACAA,YAAY,MAAM,IAAI;QAC9B;QAEAE,8BAAW,yBAAyBF;QACpC,OAAOL;IACT,EAAE,OAAOQ,OAAO;QACdD,8BAAW,qCAAqCC;QAChD;IACF;AACF;AAIO,MAAMC,qBAAqB,IAEvBC;AAUJ,MAAMC,cAAc,CACzBC;IAKA,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAO;QACL,YAAYA;QACZ,kBAAkBC;IACpB;IAEF,OAAO;QACL,YAAYD,OAAO,MAAM;QACzB,kBAAkBA,OAAO,MAAM,GAC3B;YACE,QAAQA,OAAO,MAAM;YACrB,yBAAyB,CAAC,CAACA,OAAO,uBAAuB;QAC3D,IACAC;IACN;AACF;AAEO,MAAMC,sCAAsC,CACjDd,SACAlC;IAEA,IAAIA,AAA6B,MAA7BA,0BACF,OAAOkC;IAGT,OAAO;QACL,GAAGA,OAAO;QACV,QAAQ;YACNhC,KAAK,KAAK,CAACgC,QAAQ,MAAM,CAAC,EAAE,GAAGlC;YAC/BE,KAAK,KAAK,CAACgC,QAAQ,MAAM,CAAC,EAAE,GAAGlC;SAChC;QACD,MAAM;YACJ,GAAGkC,QAAQ,IAAI;YACf,MAAMhC,KAAK,KAAK,CAACgC,QAAQ,IAAI,CAAC,IAAI,GAAGlC;YACrC,KAAKE,KAAK,KAAK,CAACgC,QAAQ,IAAI,CAAC,GAAG,GAAGlC;YACnC,OAAOE,KAAK,KAAK,CAACgC,QAAQ,IAAI,CAAC,KAAK,GAAGlC;YACvC,QAAQE,KAAK,KAAK,CAACgC,QAAQ,IAAI,CAAC,MAAM,GAAGlC;QAC3C;IACF;AACF;AAEO,MAAMiD,uCAAuC,CAClDjB,MACAhC;IAEA,IAAIA,AAA6B,MAA7BA,0BACF,OAAOgC;IAGT,OAAO;QACL,GAAGA,IAAI;QACP,MAAM9B,KAAK,KAAK,CAAC8B,KAAK,IAAI,GAAGhC;QAC7B,KAAKE,KAAK,KAAK,CAAC8B,KAAK,GAAG,GAAGhC;QAC3B,OAAOE,KAAK,KAAK,CAAC8B,KAAK,KAAK,GAAGhC;QAC/B,QAAQE,KAAK,KAAK,CAAC8B,KAAK,MAAM,GAAGhC;IACnC;AACF"}

package/dist/es/index.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"index.mjs","sources":["../../src/index.ts"],"sourcesContent":["import { z } from 'zod';\nimport Service from './service/index';\nimport { TaskRunner } from './task-runner';\nimport { getVersion } from './utils';\n\nexport {\n  plan,\n  AiLocateElement,\n  runConnectivityTest,\n  getMidsceneLocationSchema,\n  PointSchema,\n  SizeSchema,\n  RectSchema,\n  TMultimodalPromptSchema,\n  TUserPromptSchema,\n  type TMultimodalPrompt,\n  type TUserPrompt,\n  type ConnectivityCheckResultItem,\n  type ConnectivityTestConfig,\n  type ConnectivityTestResult,\n} from './ai-model/index';\n\nexport {\n  MIDSCENE_MODEL_NAME,\n  type CreateOpenAIClientFn,\n} from '@midscene/shared/env';\n\nexport type * from './types';\nexport {\n  ServiceError,\n  ExecutionDump,\n  ReportActionDump,\n  GroupedActionDump,\n  type IExecutionDump,\n  type IReportActionDump,\n  type IGroupedActionDump,\n  type ReportMeta,\n  type GroupMeta,\n} from './types';\n\nexport { z };\n\nexport default Service;\nexport { TaskRunner, Service, getVersion };\n\nexport type {\n  MidsceneYamlScript,\n  MidsceneYamlTask,\n  MidsceneYamlFlowItem,\n  MidsceneYamlConfigResult,\n  MidsceneYamlConfig,\n  MidsceneYamlScriptWebEnv,\n  MidsceneYamlScriptAndroidEnv,\n  MidsceneYamlScriptIOSEnv,\n  MidsceneYamlScriptEnv,\n  LocateOption,\n  DetailedLocateParam,\n} from './yaml';\n\nexport { Agent, type AgentOpt, type AiActOptions, createAgent } from './agent';\n\n// Dump utilities\nexport {\n  restoreImageReferences,\n  escapeContent,\n  unescapeContent,\n  parseImageScripts,\n  parseDumpScript,\n  parseDumpScriptAttributes,\n  generateImageScriptTag,\n  generateDumpScriptTag,\n} from './dump';\nexport {\n  getTaskSearchArea,\n  getTaskServiceDump,\n} from './dump/task-service-dump';\n\n// Report generator\nexport type { IReportGenerator } from './report-generator';\nexport { ReportGenerator, nullReportGenerator } from './report-generator';\nexport {\n  collectDedupedExecutions,\n  ReportMergingTool,\n  dedupeExecutionsKeepLatest,\n  splitReportHtmlByExecution,\n} from './report';\nexport {\n  createReportCliCommands,\n  reportFileToMarkdown,\n  splitReportFile,\n  mergeReportFiles,\n  type ConsumeReportFileAction,\n  type ReportFileToMarkdownOptions,\n  type ReportCliCommandDefinition,\n  type ReportCliCommandEntry,\n  type SplitReportFileOptions,\n  type MergeReportFilesOptions,\n  type MergeReportFilesResult,\n} from './report-cli';\n\n// ScreenshotItem\nexport { ScreenshotItem } from './screenshot-item';\nexport { ScreenshotStore, type ScreenshotRef } from './dump/screenshot-store';\n\nexport {\n  executionToMarkdown,\n  reportToMarkdown,\n  type ExecutionMarkdownOptions,\n  type ExecutionMarkdownResult,\n  type ReportMarkdownResult,\n  type MarkdownAttachment,\n} from './report-markdown';\n"],"names":["Service"],"mappings":";;;;;;;;;;;;;;;;AA0CA,YAAeA"}
+{"version":3,"file":"index.mjs","sources":["../../src/index.ts"],"sourcesContent":["import { z } from 'zod';\nimport Service from './service/index';\nimport { TaskRunner } from './task-runner';\nimport { getVersion } from './utils';\n\nexport {\n  plan,\n  AiLocateElement,\n  runConnectivityTest,\n  getMidsceneLocationSchema,\n  PointSchema,\n  SizeSchema,\n  RectSchema,\n  TMultimodalPromptSchema,\n  TUserPromptSchema,\n  type TMultimodalPrompt,\n  type TUserPrompt,\n  type ConnectivityCheckResultItem,\n  type ConnectivityTestConfig,\n  type ConnectivityTestResult,\n} from './ai-model/index';\n\nexport {\n  MIDSCENE_MODEL_NAME,\n  type CreateOpenAIClientFn,\n} from '@midscene/shared/env';\n\nexport type * from './types';\nexport {\n  ServiceError,\n  ExecutionDump,\n  ReportActionDump,\n  GroupedActionDump,\n  type IExecutionDump,\n  type IReportActionDump,\n  type IGroupedActionDump,\n  type ReportMeta,\n  type GroupMeta,\n} from './types';\n\nexport { z };\n\nexport default Service;\nexport { TaskRunner, Service, getVersion };\n\nexport type {\n  MidsceneYamlScript,\n  MidsceneYamlTask,\n  MidsceneYamlFlowItem,\n  MidsceneYamlConfigResult,\n  MidsceneYamlConfig,\n  MidsceneYamlScriptWebEnv,\n  MidsceneYamlScriptAndroidEnv,\n  MidsceneYamlScriptIOSEnv,\n  MidsceneYamlScriptEnv,\n  LocateOption,\n  DetailedLocateParam,\n} from './yaml';\n\nexport {\n  Agent,\n  type AgentOpt,\n  type AiActOptions,\n  createAgent,\n} from './agent';\n\n// Dump utilities\nexport {\n  restoreImageReferences,\n  escapeContent,\n  unescapeContent,\n  parseImageScripts,\n  parseDumpScript,\n  parseDumpScriptAttributes,\n  generateImageScriptTag,\n  generateDumpScriptTag,\n} from './dump';\nexport {\n  getTaskSearchArea,\n  getTaskServiceDump,\n} from './dump/task-service-dump';\n\n// Report generator\nexport type { IReportGenerator } from './report-generator';\nexport { ReportGenerator, nullReportGenerator } from './report-generator';\nexport {\n  collectDedupedExecutions,\n  ReportMergingTool,\n  dedupeExecutionsKeepLatest,\n  splitReportHtmlByExecution,\n} from './report';\nexport {\n  createReportCliCommands,\n  reportFileToMarkdown,\n  splitReportFile,\n  mergeReportFiles,\n  type ConsumeReportFileAction,\n  type ReportFileToMarkdownOptions,\n  type ReportCliCommandDefinition,\n  type ReportCliCommandEntry,\n  type SplitReportFileOptions,\n  type MergeReportFilesOptions,\n  type MergeReportFilesResult,\n} from './report-cli';\n\n// ScreenshotItem\nexport { ScreenshotItem } from './screenshot-item';\nexport { ScreenshotStore, type ScreenshotRef } from './dump/screenshot-store';\n\nexport {\n  executionToMarkdown,\n  reportToMarkdown,\n  type ExecutionMarkdownOptions,\n  type ExecutionMarkdownResult,\n  type ReportMarkdownResult,\n  type MarkdownAttachment,\n} from './report-markdown';\n"],"names":["Service"],"mappings":";;;;;;;;;;;;;;;;AA0CA,YAAeA"}

package/dist/es/report-markdown.mjs

@@ -124,7 +124,8 @@ function recorderMarkdownSection(recorder, screenshotBaseDir, executionIndex, ta
     ];
     const attachments = [];
     recorder.forEach((item, recorderIndex)=>{
-        lines.push(`- #${recorderIndex + 1} type=${item.type}, ts=${formatTime(item.ts)}, timing=${item.timing || 'N/A'}`);
+        const descriptionText = item.description ? `, description=${item.description}` : '';
+        lines.push(`- #${recorderIndex + 1} type=${item.type}, ts=${formatTime(item.ts)}, timing=${item.timing || 'N/A'}${descriptionText}`);
         if (!item.screenshot) return;
         const imageResult = screenshotAttachment(item.screenshot, screenshotBaseDir, executionIndex, taskIndex);
         lines.push(imageResult.markdown);

package/dist/es/report-markdown.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"report-markdown.mjs","sources":["../../src/report-markdown.ts"],"sourcesContent":["import { basename } from 'node:path';\nimport { extractInsightParam, paramStr, typeStr } from '@/agent/ui-utils';\nimport { ScreenshotItem } from '@/screenshot-item';\nimport type {\n  ExecutionDump,\n  ExecutionRecorderItem,\n  ExecutionTask,\n  IExecutionDump,\n  IReportActionDump,\n  ReportActionDump,\n} from '@/types';\nimport type { ScreenshotRef } from './dump/screenshot-store';\nimport { normalizeScreenshotRef } from './dump/screenshot-store';\n\nexport interface MarkdownAttachment {\n  id: string;\n  /**\n   * Stable, prefixed file name for the exported copy. The markdown image link\n   * always points at `${screenshotBaseDir}/${suggestedFileName}`, so consumers\n   * write the screenshot under this name to keep links in sync. See #2392.\n   */\n  suggestedFileName: string;\n  mimeType?: string;\n  /**\n   * Reference to the screenshot in the source report, used to locate the\n   * original bytes when copying them to the exported name. Absent for in-memory\n   * screenshots, which carry their data in `base64Data` instead.\n   */\n  sourceRef?: ScreenshotRef;\n  executionIndex: number;\n  taskIndex: number;\n  /** Populated when screenshot data is available in memory (e.g. browser context). */\n  base64Data?: string;\n}\n\nexport interface ExecutionMarkdownOptions {\n  screenshotBaseDir?: string;\n}\n\nexport interface ExecutionMarkdownResult {\n  markdown: string;\n  attachments: MarkdownAttachment[];\n}\n\nexport interface ReportMarkdownResult {\n  markdown: string;\n  attachments: MarkdownAttachment[];\n}\n\nfunction toExecutionDump(\n  execution: ExecutionDump | IExecutionDump,\n): IExecutionDump {\n  if (!execution || typeof execution !== 'object') {\n    throw new Error('executionToMarkdown: execution is required');\n  }\n\n  if (!Array.isArray(execution.tasks)) {\n    throw new Error('executionToMarkdown: execution.tasks must be an array');\n  }\n\n  if (!execution.name) {\n    throw new Error('executionToMarkdown: execution.name is required');\n  }\n\n  return execution;\n}\n\nfunction toReportDump(\n  report: ReportActionDump | IReportActionDump,\n): IReportActionDump {\n  if (!report || typeof report !== 'object') {\n    throw new Error('reportToMarkdown: report is required');\n  }\n\n  if (!Array.isArray(report.executions)) {\n    throw new Error('reportToMarkdown: report.executions must be an array');\n  }\n\n  return report;\n}\n\nfunction formatTime(ts?: number): string {\n  if (typeof ts !== 'number' || Number.isNaN(ts)) {\n    return 'N/A';\n  }\n  return new Date(ts).toISOString();\n}\n\nfunction resolveTaskTiming(task: ExecutionTask): {\n  start?: number;\n  end?: number;\n  cost?: number;\n} {\n  const timing = task.timing;\n  if (!timing) {\n    return {};\n  }\n\n  const start = timing.start ?? timing.callAiStart ?? timing.callActionStart;\n  const end =\n    timing.end ??\n    timing.callAiEnd ??\n    timing.callActionEnd ??\n    timing.captureAfterCallingSnapshotEnd;\n  const cost =\n    timing.cost ??\n    (typeof start === 'number' && typeof end === 'number'\n      ? end - start\n      : undefined);\n\n  return { start, end, cost };\n}\n\nfunction safeTaskParam(task: ExecutionTask): string {\n  const readable = paramStr(task);\n  if (readable) {\n    return readable;\n  }\n\n  if (task.type === 'Insight') {\n    return extractInsightParam((task as any).param).content;\n  }\n\n  return '';\n}\n\nfunction formatSize(\n  size?: { width?: number; height?: number } | null,\n): string | undefined {\n  if (\n    !size ||\n    typeof size.width !== 'number' ||\n    typeof size.height !== 'number' ||\n    Number.isNaN(size.width) ||\n    Number.isNaN(size.height)\n  ) {\n    return undefined;\n  }\n\n  return `${size.width} x ${size.height}`;\n}\n\nfunction extractLocateCenter(\n  task: ExecutionTask,\n): [number, number] | undefined {\n  const outputCenter = (task.output as { element?: { center?: unknown } })\n    ?.element?.center;\n  if (\n    Array.isArray(outputCenter) &&\n    outputCenter.length >= 2 &&\n    typeof outputCenter[0] === 'number' &&\n    typeof outputCenter[1] === 'number'\n  ) {\n    return [outputCenter[0], outputCenter[1]];\n  }\n\n  const paramLocateCenter = (task.param as { locate?: { center?: unknown } })\n    ?.locate?.center;\n  if (\n    Array.isArray(paramLocateCenter) &&\n    paramLocateCenter.length >= 2 &&\n    typeof paramLocateCenter[0] === 'number' &&\n    typeof paramLocateCenter[1] === 'number'\n  ) {\n    return [paramLocateCenter[0], paramLocateCenter[1]];\n  }\n\n  const paramCenter = (task.param as { center?: unknown })?.center;\n  if (\n    Array.isArray(paramCenter) &&\n    paramCenter.length >= 2 &&\n    typeof paramCenter[0] === 'number' &&\n    typeof paramCenter[1] === 'number'\n  ) {\n    return [paramCenter[0], paramCenter[1]];\n  }\n\n  return undefined;\n}\n\nfunction tryExtractBase64(screenshot: unknown): string | undefined {\n  if (!screenshot || typeof screenshot !== 'object') return undefined;\n  const s = screenshot as Record<string, unknown>;\n  if (typeof s.base64 === 'string' && s.base64.length > 0) {\n    return s.base64;\n  }\n  return undefined;\n}\n\nfunction screenshotAttachment(\n  screenshot: unknown,\n  screenshotBaseDir: string,\n  executionIndex: number,\n  taskIndex: number,\n): { markdown: string; attachment: MarkdownAttachment } {\n  if (screenshot instanceof ScreenshotItem) {\n    const ext = screenshot.extension;\n    const suggestedFileName = `execution-${executionIndex + 1}-task-${taskIndex + 1}-${screenshot.id}.${ext}`;\n    return {\n      markdown: `\\n![task-${taskIndex + 1}](${screenshotBaseDir}/${suggestedFileName})`,\n      attachment: {\n        id: screenshot.id,\n        suggestedFileName,\n        mimeType: `image/${ext === 'jpeg' ? 'jpeg' : 'png'}`,\n        executionIndex,\n        taskIndex,\n        base64Data: tryExtractBase64(screenshot),\n      },\n    };\n  }\n\n  const ref = normalizeScreenshotRef(screenshot);\n  if (ref) {\n    const ext = ref.mimeType === 'image/jpeg' ? 'jpeg' : 'png';\n    const suggestedFileName = `execution-${executionIndex + 1}-task-${taskIndex + 1}-${ref.id}.${ext}`;\n    return {\n      markdown: `\\n![task-${taskIndex + 1}](${screenshotBaseDir}/${suggestedFileName})`,\n      attachment: {\n        id: ref.id,\n        suggestedFileName,\n        sourceRef: ref,\n        mimeType: ref.mimeType,\n        executionIndex,\n        taskIndex,\n        base64Data: tryExtractBase64(screenshot),\n      },\n    };\n  }\n\n  const base64 = tryExtractBase64(screenshot);\n  if (base64) {\n    const ext = base64.startsWith('data:image/jpeg') ? 'jpeg' : 'png';\n    const id = `restored-${executionIndex + 1}-${taskIndex + 1}`;\n    const suggestedFileName = `execution-${executionIndex + 1}-task-${taskIndex + 1}-${id}.${ext}`;\n    return {\n      markdown: `\\n![task-${taskIndex + 1}](${screenshotBaseDir}/${suggestedFileName})`,\n      attachment: {\n        id,\n        suggestedFileName,\n        mimeType: `image/${ext}`,\n        executionIndex,\n        taskIndex,\n        base64Data: base64,\n      },\n    };\n  }\n\n  throw new Error(\n    `executionToMarkdown: missing screenshot for execution #${executionIndex + 1} task #${taskIndex + 1}`,\n  );\n}\n\nfunction recorderMarkdownSection(\n  recorder: ExecutionRecorderItem[] | undefined,\n  screenshotBaseDir: string,\n  executionIndex: number,\n  taskIndex: number,\n): { lines: string[]; attachments: MarkdownAttachment[] } {\n  if (!recorder?.length) {\n    return { lines: [], attachments: [] };\n  }\n\n  const lines: string[] = ['', '### Recorder'];\n  const attachments: MarkdownAttachment[] = [];\n\n  recorder.forEach((item, recorderIndex) => {\n    lines.push(\n      `- #${recorderIndex + 1} type=${item.type}, ts=${formatTime(item.ts)}, timing=${item.timing || 'N/A'}`,\n    );\n\n    if (!item.screenshot) {\n      return;\n    }\n\n    const imageResult = screenshotAttachment(\n      item.screenshot,\n      screenshotBaseDir,\n      executionIndex,\n      taskIndex,\n    );\n\n    lines.push(imageResult.markdown);\n    attachments.push(imageResult.attachment);\n  });\n\n  return { lines, attachments };\n}\n\nfunction renderExecution(\n  executionRaw: ExecutionDump | IExecutionDump,\n  executionIndex: number,\n  options?: ExecutionMarkdownOptions,\n): ExecutionMarkdownResult {\n  const execution = toExecutionDump(executionRaw);\n  const screenshotBaseDir = options?.screenshotBaseDir ?? './screenshots';\n\n  const lines: string[] = [];\n  const attachments: MarkdownAttachment[] = [];\n\n  lines.push(`# ${execution.name}`);\n  if (execution.description) {\n    lines.push('', execution.description);\n  }\n\n  lines.push('', `- Execution start: ${formatTime(execution.logTime)}`);\n  lines.push(`- Task count: ${execution.tasks.length}`);\n\n  execution.tasks.forEach((task, taskIndex) => {\n    const title = typeStr(task);\n    const detail = safeTaskParam(task);\n    const time = resolveTaskTiming(task);\n\n    lines.push(\n      '',\n      `## ${taskIndex + 1}. ${title}${detail ? ` - ${detail}` : ''}`,\n    );\n    lines.push(`- Status: ${task.status || 'unknown'}`);\n    lines.push(`- Start: ${formatTime(time.start)}`);\n    lines.push(`- End: ${formatTime(time.end)}`);\n    lines.push(\n      `- Cost(ms): ${typeof time.cost === 'number' ? time.cost : 'N/A'}`,\n    );\n    lines.push(\n      `- Screen size: ${formatSize(task.uiContext?.shotSize) || 'N/A'}`,\n    );\n\n    if (task.subType === 'Locate') {\n      const locateCenter = extractLocateCenter(task);\n      if (locateCenter) {\n        lines.push(`- Locate center: (${locateCenter[0]}, ${locateCenter[1]})`);\n      }\n    }\n\n    if (task.errorMessage) {\n      lines.push(`- Error: ${task.errorMessage}`);\n    }\n\n    if (task.uiContext?.screenshot) {\n      const imageResult = screenshotAttachment(\n        task.uiContext.screenshot,\n        screenshotBaseDir,\n        executionIndex,\n        taskIndex,\n      );\n\n      lines.push(imageResult.markdown);\n      attachments.push(imageResult.attachment);\n    }\n\n    const recorderSection = recorderMarkdownSection(\n      task.recorder,\n      screenshotBaseDir,\n      executionIndex,\n      taskIndex,\n    );\n    if (recorderSection.lines.length) {\n      lines.push(...recorderSection.lines);\n      attachments.push(...recorderSection.attachments);\n    }\n  });\n\n  return {\n    markdown: lines.join('\\n'),\n    attachments,\n  };\n}\n\nfunction reportFileName(\n  execution: IExecutionDump,\n  executionIndex: number,\n): string {\n  const safeName =\n    execution.name\n      .trim()\n      .replace(/\\s+/g, '-')\n      .replace(/[^a-zA-Z0-9-_]/g, '') || `execution-${executionIndex + 1}`;\n  return `${executionIndex + 1}-${basename(safeName)}.md`;\n}\n\nexport function executionToMarkdown(\n  execution: ExecutionDump | IExecutionDump,\n  options?: ExecutionMarkdownOptions,\n): ExecutionMarkdownResult {\n  return renderExecution(execution, 0, options);\n}\n\nexport function reportToMarkdown(\n  report: ReportActionDump | IReportActionDump,\n): ReportMarkdownResult {\n  const reportDump = toReportDump(report);\n\n  const executionResults = reportDump.executions.map((execution, index) => {\n    const rendered = renderExecution(execution, index);\n    return {\n      executionIndex: index,\n      executionName: execution.name,\n      markdown: rendered.markdown,\n      attachments: rendered.attachments,\n      suggestedFileName: reportFileName(execution, index),\n    };\n  });\n\n  const attachments = executionResults.flatMap((item) => item.attachments);\n\n  const header = [\n    `# ${reportDump.groupName}`,\n    reportDump.groupDescription ? `\\n${reportDump.groupDescription}` : '',\n    `\\n- SDK Version: ${reportDump.sdkVersion}`,\n    `- Execution count: ${reportDump.executions.length}`,\n    '\\n## Suggested execution markdown files',\n    ...executionResults.map(\n      (item) => `- ${item.suggestedFileName} (${item.executionName})`,\n    ),\n  ]\n    .filter(Boolean)\n    .join('\\n');\n\n  return {\n    markdown: `${header}\\n\\n${executionResults.map((item) => item.markdown).join('\\n\\n---\\n\\n')}`,\n    attachments,\n  };\n}\n"],"names":["toExecutionDump","execution","Error","Array","toReportDump","report","formatTime","ts","Number","Date","resolveTaskTiming","task","timing","start","end","cost","undefined","safeTaskParam","readable","paramStr","extractInsightParam","formatSize","size","extractLocateCenter","outputCenter","paramLocateCenter","paramCenter","tryExtractBase64","screenshot","s","screenshotAttachment","screenshotBaseDir","executionIndex","taskIndex","ScreenshotItem","ext","suggestedFileName","ref","normalizeScreenshotRef","base64","id","recorderMarkdownSection","recorder","lines","attachments","item","recorderIndex","imageResult","renderExecution","executionRaw","options","title","typeStr","detail","time","locateCenter","recorderSection","reportFileName","safeName","basename","executionToMarkdown","reportToMarkdown","reportDump","executionResults","index","rendered","header","Boolean"],"mappings":";;;;AAiDA,SAASA,gBACPC,SAAyC;IAEzC,IAAI,CAACA,aAAa,AAAqB,YAArB,OAAOA,WACvB,MAAM,IAAIC,MAAM;IAGlB,IAAI,CAACC,MAAM,OAAO,CAACF,UAAU,KAAK,GAChC,MAAM,IAAIC,MAAM;IAGlB,IAAI,CAACD,UAAU,IAAI,EACjB,MAAM,IAAIC,MAAM;IAGlB,OAAOD;AACT;AAEA,SAASG,aACPC,MAA4C;IAE5C,IAAI,CAACA,UAAU,AAAkB,YAAlB,OAAOA,QACpB,MAAM,IAAIH,MAAM;IAGlB,IAAI,CAACC,MAAM,OAAO,CAACE,OAAO,UAAU,GAClC,MAAM,IAAIH,MAAM;IAGlB,OAAOG;AACT;AAEA,SAASC,WAAWC,EAAW;IAC7B,IAAI,AAAc,YAAd,OAAOA,MAAmBC,OAAO,KAAK,CAACD,KACzC,OAAO;IAET,OAAO,IAAIE,KAAKF,IAAI,WAAW;AACjC;AAEA,SAASG,kBAAkBC,IAAmB;IAK5C,MAAMC,SAASD,KAAK,MAAM;IAC1B,IAAI,CAACC,QACH,OAAO,CAAC;IAGV,MAAMC,QAAQD,OAAO,KAAK,IAAIA,OAAO,WAAW,IAAIA,OAAO,eAAe;IAC1E,MAAME,MACJF,OAAO,GAAG,IACVA,OAAO,SAAS,IAChBA,OAAO,aAAa,IACpBA,OAAO,8BAA8B;IACvC,MAAMG,OACJH,OAAO,IAAI,IACV,CAAiB,YAAjB,OAAOC,SAAsB,AAAe,YAAf,OAAOC,MACjCA,MAAMD,QACNG,MAAQ;IAEd,OAAO;QAAEH;QAAOC;QAAKC;IAAK;AAC5B;AAEA,SAASE,cAAcN,IAAmB;IACxC,MAAMO,WAAWC,SAASR;IAC1B,IAAIO,UACF,OAAOA;IAGT,IAAIP,AAAc,cAAdA,KAAK,IAAI,EACX,OAAOS,oBAAqBT,KAAa,KAAK,EAAE,OAAO;IAGzD,OAAO;AACT;AAEA,SAASU,WACPC,IAAiD;IAEjD,IACE,CAACA,QACD,AAAsB,YAAtB,OAAOA,KAAK,KAAK,IACjB,AAAuB,YAAvB,OAAOA,KAAK,MAAM,IAClBd,OAAO,KAAK,CAACc,KAAK,KAAK,KACvBd,OAAO,KAAK,CAACc,KAAK,MAAM,GAExB;IAGF,OAAO,GAAGA,KAAK,KAAK,CAAC,GAAG,EAAEA,KAAK,MAAM,EAAE;AACzC;AAEA,SAASC,oBACPZ,IAAmB;IAEnB,MAAMa,eAAgBb,KAAK,MAAM,EAC7B,SAAS;IACb,IACER,MAAM,OAAO,CAACqB,iBACdA,aAAa,MAAM,IAAI,KACvB,AAA2B,YAA3B,OAAOA,YAAY,CAAC,EAAE,IACtB,AAA2B,YAA3B,OAAOA,YAAY,CAAC,EAAE,EAEtB,OAAO;QAACA,YAAY,CAAC,EAAE;QAAEA,YAAY,CAAC,EAAE;KAAC;IAG3C,MAAMC,oBAAqBd,KAAK,KAAK,EACjC,QAAQ;IACZ,IACER,MAAM,OAAO,CAACsB,sBACdA,kBAAkB,MAAM,IAAI,KAC5B,AAAgC,YAAhC,OAAOA,iBAAiB,CAAC,EAAE,IAC3B,AAAgC,YAAhC,OAAOA,iBAAiB,CAAC,EAAE,EAE3B,OAAO;QAACA,iBAAiB,CAAC,EAAE;QAAEA,iBAAiB,CAAC,EAAE;KAAC;IAGrD,MAAMC,cAAef,KAAK,KAAK,EAA2B;IAC1D,IACER,MAAM,OAAO,CAACuB,gBACdA,YAAY,MAAM,IAAI,KACtB,AAA0B,YAA1B,OAAOA,WAAW,CAAC,EAAE,IACrB,AAA0B,YAA1B,OAAOA,WAAW,CAAC,EAAE,EAErB,OAAO;QAACA,WAAW,CAAC,EAAE;QAAEA,WAAW,CAAC,EAAE;KAAC;AAI3C;AAEA,SAASC,iBAAiBC,UAAmB;IAC3C,IAAI,CAACA,cAAc,AAAsB,YAAtB,OAAOA,YAAyB;IACnD,MAAMC,IAAID;IACV,IAAI,AAAoB,YAApB,OAAOC,EAAE,MAAM,IAAiBA,EAAE,MAAM,CAAC,MAAM,GAAG,GACpD,OAAOA,EAAE,MAAM;AAGnB;AAEA,SAASC,qBACPF,UAAmB,EACnBG,iBAAyB,EACzBC,cAAsB,EACtBC,SAAiB;IAEjB,IAAIL,sBAAsBM,gBAAgB;QACxC,MAAMC,MAAMP,WAAW,SAAS;QAChC,MAAMQ,oBAAoB,CAAC,UAAU,EAAEJ,iBAAiB,EAAE,MAAM,EAAEC,YAAY,EAAE,CAAC,EAAEL,WAAW,EAAE,CAAC,CAAC,EAAEO,KAAK;QACzG,OAAO;YACL,UAAU,CAAC,SAAS,EAAEF,YAAY,EAAE,EAAE,EAAEF,kBAAkB,CAAC,EAAEK,kBAAkB,CAAC,CAAC;YACjF,YAAY;gBACV,IAAIR,WAAW,EAAE;gBACjBQ;gBACA,UAAU,CAAC,MAAM,EAAED,AAAQ,WAARA,MAAiB,SAAS,OAAO;gBACpDH;gBACAC;gBACA,YAAYN,iBAAiBC;YAC/B;QACF;IACF;IAEA,MAAMS,MAAMC,uBAAuBV;IACnC,IAAIS,KAAK;QACP,MAAMF,MAAME,AAAiB,iBAAjBA,IAAI,QAAQ,GAAoB,SAAS;QACrD,MAAMD,oBAAoB,CAAC,UAAU,EAAEJ,iBAAiB,EAAE,MAAM,EAAEC,YAAY,EAAE,CAAC,EAAEI,IAAI,EAAE,CAAC,CAAC,EAAEF,KAAK;QAClG,OAAO;YACL,UAAU,CAAC,SAAS,EAAEF,YAAY,EAAE,EAAE,EAAEF,kBAAkB,CAAC,EAAEK,kBAAkB,CAAC,CAAC;YACjF,YAAY;gBACV,IAAIC,IAAI,EAAE;gBACVD;gBACA,WAAWC;gBACX,UAAUA,IAAI,QAAQ;gBACtBL;gBACAC;gBACA,YAAYN,iBAAiBC;YAC/B;QACF;IACF;IAEA,MAAMW,SAASZ,iBAAiBC;IAChC,IAAIW,QAAQ;QACV,MAAMJ,MAAMI,OAAO,UAAU,CAAC,qBAAqB,SAAS;QAC5D,MAAMC,KAAK,CAAC,SAAS,EAAER,iBAAiB,EAAE,CAAC,EAAEC,YAAY,GAAG;QAC5D,MAAMG,oBAAoB,CAAC,UAAU,EAAEJ,iBAAiB,EAAE,MAAM,EAAEC,YAAY,EAAE,CAAC,EAAEO,GAAG,CAAC,EAAEL,KAAK;QAC9F,OAAO;YACL,UAAU,CAAC,SAAS,EAAEF,YAAY,EAAE,EAAE,EAAEF,kBAAkB,CAAC,EAAEK,kBAAkB,CAAC,CAAC;YACjF,YAAY;gBACVI;gBACAJ;gBACA,UAAU,CAAC,MAAM,EAAED,KAAK;gBACxBH;gBACAC;gBACA,YAAYM;YACd;QACF;IACF;IAEA,MAAM,IAAIrC,MACR,CAAC,uDAAuD,EAAE8B,iBAAiB,EAAE,OAAO,EAAEC,YAAY,GAAG;AAEzG;AAEA,SAASQ,wBACPC,QAA6C,EAC7CX,iBAAyB,EACzBC,cAAsB,EACtBC,SAAiB;IAEjB,IAAI,CAACS,UAAU,QACb,OAAO;QAAE,OAAO,EAAE;QAAE,aAAa,EAAE;IAAC;IAGtC,MAAMC,QAAkB;QAAC;QAAI;KAAe;IAC5C,MAAMC,cAAoC,EAAE;IAE5CF,SAAS,OAAO,CAAC,CAACG,MAAMC;QACtBH,MAAM,IAAI,CACR,CAAC,GAAG,EAAEG,gBAAgB,EAAE,MAAM,EAAED,KAAK,IAAI,CAAC,KAAK,EAAEvC,WAAWuC,KAAK,EAAE,EAAE,SAAS,EAAEA,KAAK,MAAM,IAAI,OAAO;QAGxG,IAAI,CAACA,KAAK,UAAU,EAClB;QAGF,MAAME,cAAcjB,qBAClBe,KAAK,UAAU,EACfd,mBACAC,gBACAC;QAGFU,MAAM,IAAI,CAACI,YAAY,QAAQ;QAC/BH,YAAY,IAAI,CAACG,YAAY,UAAU;IACzC;IAEA,OAAO;QAAEJ;QAAOC;IAAY;AAC9B;AAEA,SAASI,gBACPC,YAA4C,EAC5CjB,cAAsB,EACtBkB,OAAkC;IAElC,MAAMjD,YAAYD,gBAAgBiD;IAClC,MAAMlB,oBAAoBmB,SAAS,qBAAqB;IAExD,MAAMP,QAAkB,EAAE;IAC1B,MAAMC,cAAoC,EAAE;IAE5CD,MAAM,IAAI,CAAC,CAAC,EAAE,EAAE1C,UAAU,IAAI,EAAE;IAChC,IAAIA,UAAU,WAAW,EACvB0C,MAAM,IAAI,CAAC,IAAI1C,UAAU,WAAW;IAGtC0C,MAAM,IAAI,CAAC,IAAI,CAAC,mBAAmB,EAAErC,WAAWL,UAAU,OAAO,GAAG;IACpE0C,MAAM,IAAI,CAAC,CAAC,cAAc,EAAE1C,UAAU,KAAK,CAAC,MAAM,EAAE;IAEpDA,UAAU,KAAK,CAAC,OAAO,CAAC,CAACU,MAAMsB;QAC7B,MAAMkB,QAAQC,QAAQzC;QACtB,MAAM0C,SAASpC,cAAcN;QAC7B,MAAM2C,OAAO5C,kBAAkBC;QAE/BgC,MAAM,IAAI,CACR,IACA,CAAC,GAAG,EAAEV,YAAY,EAAE,EAAE,EAAEkB,QAAQE,SAAS,CAAC,GAAG,EAAEA,QAAQ,GAAG,IAAI;QAEhEV,MAAM,IAAI,CAAC,CAAC,UAAU,EAAEhC,KAAK,MAAM,IAAI,WAAW;QAClDgC,MAAM,IAAI,CAAC,CAAC,SAAS,EAAErC,WAAWgD,KAAK,KAAK,GAAG;QAC/CX,MAAM,IAAI,CAAC,CAAC,OAAO,EAAErC,WAAWgD,KAAK,GAAG,GAAG;QAC3CX,MAAM,IAAI,CACR,CAAC,YAAY,EAAE,AAAqB,YAArB,OAAOW,KAAK,IAAI,GAAgBA,KAAK,IAAI,GAAG,OAAO;QAEpEX,MAAM,IAAI,CACR,CAAC,eAAe,EAAEtB,WAAWV,KAAK,SAAS,EAAE,aAAa,OAAO;QAGnE,IAAIA,AAAiB,aAAjBA,KAAK,OAAO,EAAe;YAC7B,MAAM4C,eAAehC,oBAAoBZ;YACzC,IAAI4C,cACFZ,MAAM,IAAI,CAAC,CAAC,kBAAkB,EAAEY,YAAY,CAAC,EAAE,CAAC,EAAE,EAAEA,YAAY,CAAC,EAAE,CAAC,CAAC,CAAC;QAE1E;QAEA,IAAI5C,KAAK,YAAY,EACnBgC,MAAM,IAAI,CAAC,CAAC,SAAS,EAAEhC,KAAK,YAAY,EAAE;QAG5C,IAAIA,KAAK,SAAS,EAAE,YAAY;YAC9B,MAAMoC,cAAcjB,qBAClBnB,KAAK,SAAS,CAAC,UAAU,EACzBoB,mBACAC,gBACAC;YAGFU,MAAM,IAAI,CAACI,YAAY,QAAQ;YAC/BH,YAAY,IAAI,CAACG,YAAY,UAAU;QACzC;QAEA,MAAMS,kBAAkBf,wBACtB9B,KAAK,QAAQ,EACboB,mBACAC,gBACAC;QAEF,IAAIuB,gBAAgB,KAAK,CAAC,MAAM,EAAE;YAChCb,MAAM,IAAI,IAAIa,gBAAgB,KAAK;YACnCZ,YAAY,IAAI,IAAIY,gBAAgB,WAAW;QACjD;IACF;IAEA,OAAO;QACL,UAAUb,MAAM,IAAI,CAAC;QACrBC;IACF;AACF;AAEA,SAASa,eACPxD,SAAyB,EACzB+B,cAAsB;IAEtB,MAAM0B,WACJzD,UAAU,IAAI,CACX,IAAI,GACJ,OAAO,CAAC,QAAQ,KAChB,OAAO,CAAC,mBAAmB,OAAO,CAAC,UAAU,EAAE+B,iBAAiB,GAAG;IACxE,OAAO,GAAGA,iBAAiB,EAAE,CAAC,EAAE2B,SAASD,UAAU,GAAG,CAAC;AACzD;AAEO,SAASE,oBACd3D,SAAyC,EACzCiD,OAAkC;IAElC,OAAOF,gBAAgB/C,WAAW,GAAGiD;AACvC;AAEO,SAASW,iBACdxD,MAA4C;IAE5C,MAAMyD,aAAa1D,aAAaC;IAEhC,MAAM0D,mBAAmBD,WAAW,UAAU,CAAC,GAAG,CAAC,CAAC7D,WAAW+D;QAC7D,MAAMC,WAAWjB,gBAAgB/C,WAAW+D;QAC5C,OAAO;YACL,gBAAgBA;YAChB,eAAe/D,UAAU,IAAI;YAC7B,UAAUgE,SAAS,QAAQ;YAC3B,aAAaA,SAAS,WAAW;YACjC,mBAAmBR,eAAexD,WAAW+D;QAC/C;IACF;IAEA,MAAMpB,cAAcmB,iBAAiB,OAAO,CAAC,CAAClB,OAASA,KAAK,WAAW;IAEvE,MAAMqB,SAAS;QACb,CAAC,EAAE,EAAEJ,WAAW,SAAS,EAAE;QAC3BA,WAAW,gBAAgB,GAAG,CAAC,EAAE,EAAEA,WAAW,gBAAgB,EAAE,GAAG;QACnE,CAAC,iBAAiB,EAAEA,WAAW,UAAU,EAAE;QAC3C,CAAC,mBAAmB,EAAEA,WAAW,UAAU,CAAC,MAAM,EAAE;QACpD;WACGC,iBAAiB,GAAG,CACrB,CAAClB,OAAS,CAAC,EAAE,EAAEA,KAAK,iBAAiB,CAAC,EAAE,EAAEA,KAAK,aAAa,CAAC,CAAC,CAAC;KAElE,CACE,MAAM,CAACsB,SACP,IAAI,CAAC;IAER,OAAO;QACL,UAAU,GAAGD,OAAO,IAAI,EAAEH,iBAAiB,GAAG,CAAC,CAAClB,OAASA,KAAK,QAAQ,EAAE,IAAI,CAAC,gBAAgB;QAC7FD;IACF;AACF"}
+{"version":3,"file":"report-markdown.mjs","sources":["../../src/report-markdown.ts"],"sourcesContent":["import { basename } from 'node:path';\nimport { extractInsightParam, paramStr, typeStr } from '@/agent/ui-utils';\nimport { ScreenshotItem } from '@/screenshot-item';\nimport type {\n  ExecutionDump,\n  ExecutionRecorderItem,\n  ExecutionTask,\n  IExecutionDump,\n  IReportActionDump,\n  ReportActionDump,\n} from '@/types';\nimport type { ScreenshotRef } from './dump/screenshot-store';\nimport { normalizeScreenshotRef } from './dump/screenshot-store';\n\nexport interface MarkdownAttachment {\n  id: string;\n  /**\n   * Stable, prefixed file name for the exported copy. The markdown image link\n   * always points at `${screenshotBaseDir}/${suggestedFileName}`, so consumers\n   * write the screenshot under this name to keep links in sync. See #2392.\n   */\n  suggestedFileName: string;\n  mimeType?: string;\n  /**\n   * Reference to the screenshot in the source report, used to locate the\n   * original bytes when copying them to the exported name. Absent for in-memory\n   * screenshots, which carry their data in `base64Data` instead.\n   */\n  sourceRef?: ScreenshotRef;\n  executionIndex: number;\n  taskIndex: number;\n  /** Populated when screenshot data is available in memory (e.g. browser context). */\n  base64Data?: string;\n}\n\nexport interface ExecutionMarkdownOptions {\n  screenshotBaseDir?: string;\n}\n\nexport interface ExecutionMarkdownResult {\n  markdown: string;\n  attachments: MarkdownAttachment[];\n}\n\nexport interface ReportMarkdownResult {\n  markdown: string;\n  attachments: MarkdownAttachment[];\n}\n\nfunction toExecutionDump(\n  execution: ExecutionDump | IExecutionDump,\n): IExecutionDump {\n  if (!execution || typeof execution !== 'object') {\n    throw new Error('executionToMarkdown: execution is required');\n  }\n\n  if (!Array.isArray(execution.tasks)) {\n    throw new Error('executionToMarkdown: execution.tasks must be an array');\n  }\n\n  if (!execution.name) {\n    throw new Error('executionToMarkdown: execution.name is required');\n  }\n\n  return execution;\n}\n\nfunction toReportDump(\n  report: ReportActionDump | IReportActionDump,\n): IReportActionDump {\n  if (!report || typeof report !== 'object') {\n    throw new Error('reportToMarkdown: report is required');\n  }\n\n  if (!Array.isArray(report.executions)) {\n    throw new Error('reportToMarkdown: report.executions must be an array');\n  }\n\n  return report;\n}\n\nfunction formatTime(ts?: number): string {\n  if (typeof ts !== 'number' || Number.isNaN(ts)) {\n    return 'N/A';\n  }\n  return new Date(ts).toISOString();\n}\n\nfunction resolveTaskTiming(task: ExecutionTask): {\n  start?: number;\n  end?: number;\n  cost?: number;\n} {\n  const timing = task.timing;\n  if (!timing) {\n    return {};\n  }\n\n  const start = timing.start ?? timing.callAiStart ?? timing.callActionStart;\n  const end =\n    timing.end ??\n    timing.callAiEnd ??\n    timing.callActionEnd ??\n    timing.captureAfterCallingSnapshotEnd;\n  const cost =\n    timing.cost ??\n    (typeof start === 'number' && typeof end === 'number'\n      ? end - start\n      : undefined);\n\n  return { start, end, cost };\n}\n\nfunction safeTaskParam(task: ExecutionTask): string {\n  const readable = paramStr(task);\n  if (readable) {\n    return readable;\n  }\n\n  if (task.type === 'Insight') {\n    return extractInsightParam((task as any).param).content;\n  }\n\n  return '';\n}\n\nfunction formatSize(\n  size?: { width?: number; height?: number } | null,\n): string | undefined {\n  if (\n    !size ||\n    typeof size.width !== 'number' ||\n    typeof size.height !== 'number' ||\n    Number.isNaN(size.width) ||\n    Number.isNaN(size.height)\n  ) {\n    return undefined;\n  }\n\n  return `${size.width} x ${size.height}`;\n}\n\nfunction extractLocateCenter(\n  task: ExecutionTask,\n): [number, number] | undefined {\n  const outputCenter = (task.output as { element?: { center?: unknown } })\n    ?.element?.center;\n  if (\n    Array.isArray(outputCenter) &&\n    outputCenter.length >= 2 &&\n    typeof outputCenter[0] === 'number' &&\n    typeof outputCenter[1] === 'number'\n  ) {\n    return [outputCenter[0], outputCenter[1]];\n  }\n\n  const paramLocateCenter = (task.param as { locate?: { center?: unknown } })\n    ?.locate?.center;\n  if (\n    Array.isArray(paramLocateCenter) &&\n    paramLocateCenter.length >= 2 &&\n    typeof paramLocateCenter[0] === 'number' &&\n    typeof paramLocateCenter[1] === 'number'\n  ) {\n    return [paramLocateCenter[0], paramLocateCenter[1]];\n  }\n\n  const paramCenter = (task.param as { center?: unknown })?.center;\n  if (\n    Array.isArray(paramCenter) &&\n    paramCenter.length >= 2 &&\n    typeof paramCenter[0] === 'number' &&\n    typeof paramCenter[1] === 'number'\n  ) {\n    return [paramCenter[0], paramCenter[1]];\n  }\n\n  return undefined;\n}\n\nfunction tryExtractBase64(screenshot: unknown): string | undefined {\n  if (!screenshot || typeof screenshot !== 'object') return undefined;\n  const s = screenshot as Record<string, unknown>;\n  if (typeof s.base64 === 'string' && s.base64.length > 0) {\n    return s.base64;\n  }\n  return undefined;\n}\n\nfunction screenshotAttachment(\n  screenshot: unknown,\n  screenshotBaseDir: string,\n  executionIndex: number,\n  taskIndex: number,\n): { markdown: string; attachment: MarkdownAttachment } {\n  if (screenshot instanceof ScreenshotItem) {\n    const ext = screenshot.extension;\n    const suggestedFileName = `execution-${executionIndex + 1}-task-${taskIndex + 1}-${screenshot.id}.${ext}`;\n    return {\n      markdown: `\\n![task-${taskIndex + 1}](${screenshotBaseDir}/${suggestedFileName})`,\n      attachment: {\n        id: screenshot.id,\n        suggestedFileName,\n        mimeType: `image/${ext === 'jpeg' ? 'jpeg' : 'png'}`,\n        executionIndex,\n        taskIndex,\n        base64Data: tryExtractBase64(screenshot),\n      },\n    };\n  }\n\n  const ref = normalizeScreenshotRef(screenshot);\n  if (ref) {\n    const ext = ref.mimeType === 'image/jpeg' ? 'jpeg' : 'png';\n    const suggestedFileName = `execution-${executionIndex + 1}-task-${taskIndex + 1}-${ref.id}.${ext}`;\n    return {\n      markdown: `\\n![task-${taskIndex + 1}](${screenshotBaseDir}/${suggestedFileName})`,\n      attachment: {\n        id: ref.id,\n        suggestedFileName,\n        sourceRef: ref,\n        mimeType: ref.mimeType,\n        executionIndex,\n        taskIndex,\n        base64Data: tryExtractBase64(screenshot),\n      },\n    };\n  }\n\n  const base64 = tryExtractBase64(screenshot);\n  if (base64) {\n    const ext = base64.startsWith('data:image/jpeg') ? 'jpeg' : 'png';\n    const id = `restored-${executionIndex + 1}-${taskIndex + 1}`;\n    const suggestedFileName = `execution-${executionIndex + 1}-task-${taskIndex + 1}-${id}.${ext}`;\n    return {\n      markdown: `\\n![task-${taskIndex + 1}](${screenshotBaseDir}/${suggestedFileName})`,\n      attachment: {\n        id,\n        suggestedFileName,\n        mimeType: `image/${ext}`,\n        executionIndex,\n        taskIndex,\n        base64Data: base64,\n      },\n    };\n  }\n\n  throw new Error(\n    `executionToMarkdown: missing screenshot for execution #${executionIndex + 1} task #${taskIndex + 1}`,\n  );\n}\n\nfunction recorderMarkdownSection(\n  recorder: ExecutionRecorderItem[] | undefined,\n  screenshotBaseDir: string,\n  executionIndex: number,\n  taskIndex: number,\n): { lines: string[]; attachments: MarkdownAttachment[] } {\n  if (!recorder?.length) {\n    return { lines: [], attachments: [] };\n  }\n\n  const lines: string[] = ['', '### Recorder'];\n  const attachments: MarkdownAttachment[] = [];\n\n  recorder.forEach((item, recorderIndex) => {\n    const descriptionText = item.description\n      ? `, description=${item.description}`\n      : '';\n    lines.push(\n      `- #${recorderIndex + 1} type=${item.type}, ts=${formatTime(item.ts)}, timing=${item.timing || 'N/A'}${descriptionText}`,\n    );\n\n    if (!item.screenshot) {\n      return;\n    }\n\n    const imageResult = screenshotAttachment(\n      item.screenshot,\n      screenshotBaseDir,\n      executionIndex,\n      taskIndex,\n    );\n\n    lines.push(imageResult.markdown);\n    attachments.push(imageResult.attachment);\n  });\n\n  return { lines, attachments };\n}\n\nfunction renderExecution(\n  executionRaw: ExecutionDump | IExecutionDump,\n  executionIndex: number,\n  options?: ExecutionMarkdownOptions,\n): ExecutionMarkdownResult {\n  const execution = toExecutionDump(executionRaw);\n  const screenshotBaseDir = options?.screenshotBaseDir ?? './screenshots';\n\n  const lines: string[] = [];\n  const attachments: MarkdownAttachment[] = [];\n\n  lines.push(`# ${execution.name}`);\n  if (execution.description) {\n    lines.push('', execution.description);\n  }\n\n  lines.push('', `- Execution start: ${formatTime(execution.logTime)}`);\n  lines.push(`- Task count: ${execution.tasks.length}`);\n\n  execution.tasks.forEach((task, taskIndex) => {\n    const title = typeStr(task);\n    const detail = safeTaskParam(task);\n    const time = resolveTaskTiming(task);\n\n    lines.push(\n      '',\n      `## ${taskIndex + 1}. ${title}${detail ? ` - ${detail}` : ''}`,\n    );\n    lines.push(`- Status: ${task.status || 'unknown'}`);\n    lines.push(`- Start: ${formatTime(time.start)}`);\n    lines.push(`- End: ${formatTime(time.end)}`);\n    lines.push(\n      `- Cost(ms): ${typeof time.cost === 'number' ? time.cost : 'N/A'}`,\n    );\n    lines.push(\n      `- Screen size: ${formatSize(task.uiContext?.shotSize) || 'N/A'}`,\n    );\n\n    if (task.subType === 'Locate') {\n      const locateCenter = extractLocateCenter(task);\n      if (locateCenter) {\n        lines.push(`- Locate center: (${locateCenter[0]}, ${locateCenter[1]})`);\n      }\n    }\n\n    if (task.errorMessage) {\n      lines.push(`- Error: ${task.errorMessage}`);\n    }\n\n    if (task.uiContext?.screenshot) {\n      const imageResult = screenshotAttachment(\n        task.uiContext.screenshot,\n        screenshotBaseDir,\n        executionIndex,\n        taskIndex,\n      );\n\n      lines.push(imageResult.markdown);\n      attachments.push(imageResult.attachment);\n    }\n\n    const recorderSection = recorderMarkdownSection(\n      task.recorder,\n      screenshotBaseDir,\n      executionIndex,\n      taskIndex,\n    );\n    if (recorderSection.lines.length) {\n      lines.push(...recorderSection.lines);\n      attachments.push(...recorderSection.attachments);\n    }\n  });\n\n  return {\n    markdown: lines.join('\\n'),\n    attachments,\n  };\n}\n\nfunction reportFileName(\n  execution: IExecutionDump,\n  executionIndex: number,\n): string {\n  const safeName =\n    execution.name\n      .trim()\n      .replace(/\\s+/g, '-')\n      .replace(/[^a-zA-Z0-9-_]/g, '') || `execution-${executionIndex + 1}`;\n  return `${executionIndex + 1}-${basename(safeName)}.md`;\n}\n\nexport function executionToMarkdown(\n  execution: ExecutionDump | IExecutionDump,\n  options?: ExecutionMarkdownOptions,\n): ExecutionMarkdownResult {\n  return renderExecution(execution, 0, options);\n}\n\nexport function reportToMarkdown(\n  report: ReportActionDump | IReportActionDump,\n): ReportMarkdownResult {\n  const reportDump = toReportDump(report);\n\n  const executionResults = reportDump.executions.map((execution, index) => {\n    const rendered = renderExecution(execution, index);\n    return {\n      executionIndex: index,\n      executionName: execution.name,\n      markdown: rendered.markdown,\n      attachments: rendered.attachments,\n      suggestedFileName: reportFileName(execution, index),\n    };\n  });\n\n  const attachments = executionResults.flatMap((item) => item.attachments);\n\n  const header = [\n    `# ${reportDump.groupName}`,\n    reportDump.groupDescription ? `\\n${reportDump.groupDescription}` : '',\n    `\\n- SDK Version: ${reportDump.sdkVersion}`,\n    `- Execution count: ${reportDump.executions.length}`,\n    '\\n## Suggested execution markdown files',\n    ...executionResults.map(\n      (item) => `- ${item.suggestedFileName} (${item.executionName})`,\n    ),\n  ]\n    .filter(Boolean)\n    .join('\\n');\n\n  return {\n    markdown: `${header}\\n\\n${executionResults.map((item) => item.markdown).join('\\n\\n---\\n\\n')}`,\n    attachments,\n  };\n}\n"],"names":["toExecutionDump","execution","Error","Array","toReportDump","report","formatTime","ts","Number","Date","resolveTaskTiming","task","timing","start","end","cost","undefined","safeTaskParam","readable","paramStr","extractInsightParam","formatSize","size","extractLocateCenter","outputCenter","paramLocateCenter","paramCenter","tryExtractBase64","screenshot","s","screenshotAttachment","screenshotBaseDir","executionIndex","taskIndex","ScreenshotItem","ext","suggestedFileName","ref","normalizeScreenshotRef","base64","id","recorderMarkdownSection","recorder","lines","attachments","item","recorderIndex","descriptionText","imageResult","renderExecution","executionRaw","options","title","typeStr","detail","time","locateCenter","recorderSection","reportFileName","safeName","basename","executionToMarkdown","reportToMarkdown","reportDump","executionResults","index","rendered","header","Boolean"],"mappings":";;;;AAiDA,SAASA,gBACPC,SAAyC;IAEzC,IAAI,CAACA,aAAa,AAAqB,YAArB,OAAOA,WACvB,MAAM,IAAIC,MAAM;IAGlB,IAAI,CAACC,MAAM,OAAO,CAACF,UAAU,KAAK,GAChC,MAAM,IAAIC,MAAM;IAGlB,IAAI,CAACD,UAAU,IAAI,EACjB,MAAM,IAAIC,MAAM;IAGlB,OAAOD;AACT;AAEA,SAASG,aACPC,MAA4C;IAE5C,IAAI,CAACA,UAAU,AAAkB,YAAlB,OAAOA,QACpB,MAAM,IAAIH,MAAM;IAGlB,IAAI,CAACC,MAAM,OAAO,CAACE,OAAO,UAAU,GAClC,MAAM,IAAIH,MAAM;IAGlB,OAAOG;AACT;AAEA,SAASC,WAAWC,EAAW;IAC7B,IAAI,AAAc,YAAd,OAAOA,MAAmBC,OAAO,KAAK,CAACD,KACzC,OAAO;IAET,OAAO,IAAIE,KAAKF,IAAI,WAAW;AACjC;AAEA,SAASG,kBAAkBC,IAAmB;IAK5C,MAAMC,SAASD,KAAK,MAAM;IAC1B,IAAI,CAACC,QACH,OAAO,CAAC;IAGV,MAAMC,QAAQD,OAAO,KAAK,IAAIA,OAAO,WAAW,IAAIA,OAAO,eAAe;IAC1E,MAAME,MACJF,OAAO,GAAG,IACVA,OAAO,SAAS,IAChBA,OAAO,aAAa,IACpBA,OAAO,8BAA8B;IACvC,MAAMG,OACJH,OAAO,IAAI,IACV,CAAiB,YAAjB,OAAOC,SAAsB,AAAe,YAAf,OAAOC,MACjCA,MAAMD,QACNG,MAAQ;IAEd,OAAO;QAAEH;QAAOC;QAAKC;IAAK;AAC5B;AAEA,SAASE,cAAcN,IAAmB;IACxC,MAAMO,WAAWC,SAASR;IAC1B,IAAIO,UACF,OAAOA;IAGT,IAAIP,AAAc,cAAdA,KAAK,IAAI,EACX,OAAOS,oBAAqBT,KAAa,KAAK,EAAE,OAAO;IAGzD,OAAO;AACT;AAEA,SAASU,WACPC,IAAiD;IAEjD,IACE,CAACA,QACD,AAAsB,YAAtB,OAAOA,KAAK,KAAK,IACjB,AAAuB,YAAvB,OAAOA,KAAK,MAAM,IAClBd,OAAO,KAAK,CAACc,KAAK,KAAK,KACvBd,OAAO,KAAK,CAACc,KAAK,MAAM,GAExB;IAGF,OAAO,GAAGA,KAAK,KAAK,CAAC,GAAG,EAAEA,KAAK,MAAM,EAAE;AACzC;AAEA,SAASC,oBACPZ,IAAmB;IAEnB,MAAMa,eAAgBb,KAAK,MAAM,EAC7B,SAAS;IACb,IACER,MAAM,OAAO,CAACqB,iBACdA,aAAa,MAAM,IAAI,KACvB,AAA2B,YAA3B,OAAOA,YAAY,CAAC,EAAE,IACtB,AAA2B,YAA3B,OAAOA,YAAY,CAAC,EAAE,EAEtB,OAAO;QAACA,YAAY,CAAC,EAAE;QAAEA,YAAY,CAAC,EAAE;KAAC;IAG3C,MAAMC,oBAAqBd,KAAK,KAAK,EACjC,QAAQ;IACZ,IACER,MAAM,OAAO,CAACsB,sBACdA,kBAAkB,MAAM,IAAI,KAC5B,AAAgC,YAAhC,OAAOA,iBAAiB,CAAC,EAAE,IAC3B,AAAgC,YAAhC,OAAOA,iBAAiB,CAAC,EAAE,EAE3B,OAAO;QAACA,iBAAiB,CAAC,EAAE;QAAEA,iBAAiB,CAAC,EAAE;KAAC;IAGrD,MAAMC,cAAef,KAAK,KAAK,EAA2B;IAC1D,IACER,MAAM,OAAO,CAACuB,gBACdA,YAAY,MAAM,IAAI,KACtB,AAA0B,YAA1B,OAAOA,WAAW,CAAC,EAAE,IACrB,AAA0B,YAA1B,OAAOA,WAAW,CAAC,EAAE,EAErB,OAAO;QAACA,WAAW,CAAC,EAAE;QAAEA,WAAW,CAAC,EAAE;KAAC;AAI3C;AAEA,SAASC,iBAAiBC,UAAmB;IAC3C,IAAI,CAACA,cAAc,AAAsB,YAAtB,OAAOA,YAAyB;IACnD,MAAMC,IAAID;IACV,IAAI,AAAoB,YAApB,OAAOC,EAAE,MAAM,IAAiBA,EAAE,MAAM,CAAC,MAAM,GAAG,GACpD,OAAOA,EAAE,MAAM;AAGnB;AAEA,SAASC,qBACPF,UAAmB,EACnBG,iBAAyB,EACzBC,cAAsB,EACtBC,SAAiB;IAEjB,IAAIL,sBAAsBM,gBAAgB;QACxC,MAAMC,MAAMP,WAAW,SAAS;QAChC,MAAMQ,oBAAoB,CAAC,UAAU,EAAEJ,iBAAiB,EAAE,MAAM,EAAEC,YAAY,EAAE,CAAC,EAAEL,WAAW,EAAE,CAAC,CAAC,EAAEO,KAAK;QACzG,OAAO;YACL,UAAU,CAAC,SAAS,EAAEF,YAAY,EAAE,EAAE,EAAEF,kBAAkB,CAAC,EAAEK,kBAAkB,CAAC,CAAC;YACjF,YAAY;gBACV,IAAIR,WAAW,EAAE;gBACjBQ;gBACA,UAAU,CAAC,MAAM,EAAED,AAAQ,WAARA,MAAiB,SAAS,OAAO;gBACpDH;gBACAC;gBACA,YAAYN,iBAAiBC;YAC/B;QACF;IACF;IAEA,MAAMS,MAAMC,uBAAuBV;IACnC,IAAIS,KAAK;QACP,MAAMF,MAAME,AAAiB,iBAAjBA,IAAI,QAAQ,GAAoB,SAAS;QACrD,MAAMD,oBAAoB,CAAC,UAAU,EAAEJ,iBAAiB,EAAE,MAAM,EAAEC,YAAY,EAAE,CAAC,EAAEI,IAAI,EAAE,CAAC,CAAC,EAAEF,KAAK;QAClG,OAAO;YACL,UAAU,CAAC,SAAS,EAAEF,YAAY,EAAE,EAAE,EAAEF,kBAAkB,CAAC,EAAEK,kBAAkB,CAAC,CAAC;YACjF,YAAY;gBACV,IAAIC,IAAI,EAAE;gBACVD;gBACA,WAAWC;gBACX,UAAUA,IAAI,QAAQ;gBACtBL;gBACAC;gBACA,YAAYN,iBAAiBC;YAC/B;QACF;IACF;IAEA,MAAMW,SAASZ,iBAAiBC;IAChC,IAAIW,QAAQ;QACV,MAAMJ,MAAMI,OAAO,UAAU,CAAC,qBAAqB,SAAS;QAC5D,MAAMC,KAAK,CAAC,SAAS,EAAER,iBAAiB,EAAE,CAAC,EAAEC,YAAY,GAAG;QAC5D,MAAMG,oBAAoB,CAAC,UAAU,EAAEJ,iBAAiB,EAAE,MAAM,EAAEC,YAAY,EAAE,CAAC,EAAEO,GAAG,CAAC,EAAEL,KAAK;QAC9F,OAAO;YACL,UAAU,CAAC,SAAS,EAAEF,YAAY,EAAE,EAAE,EAAEF,kBAAkB,CAAC,EAAEK,kBAAkB,CAAC,CAAC;YACjF,YAAY;gBACVI;gBACAJ;gBACA,UAAU,CAAC,MAAM,EAAED,KAAK;gBACxBH;gBACAC;gBACA,YAAYM;YACd;QACF;IACF;IAEA,MAAM,IAAIrC,MACR,CAAC,uDAAuD,EAAE8B,iBAAiB,EAAE,OAAO,EAAEC,YAAY,GAAG;AAEzG;AAEA,SAASQ,wBACPC,QAA6C,EAC7CX,iBAAyB,EACzBC,cAAsB,EACtBC,SAAiB;IAEjB,IAAI,CAACS,UAAU,QACb,OAAO;QAAE,OAAO,EAAE;QAAE,aAAa,EAAE;IAAC;IAGtC,MAAMC,QAAkB;QAAC;QAAI;KAAe;IAC5C,MAAMC,cAAoC,EAAE;IAE5CF,SAAS,OAAO,CAAC,CAACG,MAAMC;QACtB,MAAMC,kBAAkBF,KAAK,WAAW,GACpC,CAAC,cAAc,EAAEA,KAAK,WAAW,EAAE,GACnC;QACJF,MAAM,IAAI,CACR,CAAC,GAAG,EAAEG,gBAAgB,EAAE,MAAM,EAAED,KAAK,IAAI,CAAC,KAAK,EAAEvC,WAAWuC,KAAK,EAAE,EAAE,SAAS,EAAEA,KAAK,MAAM,IAAI,QAAQE,iBAAiB;QAG1H,IAAI,CAACF,KAAK,UAAU,EAClB;QAGF,MAAMG,cAAclB,qBAClBe,KAAK,UAAU,EACfd,mBACAC,gBACAC;QAGFU,MAAM,IAAI,CAACK,YAAY,QAAQ;QAC/BJ,YAAY,IAAI,CAACI,YAAY,UAAU;IACzC;IAEA,OAAO;QAAEL;QAAOC;IAAY;AAC9B;AAEA,SAASK,gBACPC,YAA4C,EAC5ClB,cAAsB,EACtBmB,OAAkC;IAElC,MAAMlD,YAAYD,gBAAgBkD;IAClC,MAAMnB,oBAAoBoB,SAAS,qBAAqB;IAExD,MAAMR,QAAkB,EAAE;IAC1B,MAAMC,cAAoC,EAAE;IAE5CD,MAAM,IAAI,CAAC,CAAC,EAAE,EAAE1C,UAAU,IAAI,EAAE;IAChC,IAAIA,UAAU,WAAW,EACvB0C,MAAM,IAAI,CAAC,IAAI1C,UAAU,WAAW;IAGtC0C,MAAM,IAAI,CAAC,IAAI,CAAC,mBAAmB,EAAErC,WAAWL,UAAU,OAAO,GAAG;IACpE0C,MAAM,IAAI,CAAC,CAAC,cAAc,EAAE1C,UAAU,KAAK,CAAC,MAAM,EAAE;IAEpDA,UAAU,KAAK,CAAC,OAAO,CAAC,CAACU,MAAMsB;QAC7B,MAAMmB,QAAQC,QAAQ1C;QACtB,MAAM2C,SAASrC,cAAcN;QAC7B,MAAM4C,OAAO7C,kBAAkBC;QAE/BgC,MAAM,IAAI,CACR,IACA,CAAC,GAAG,EAAEV,YAAY,EAAE,EAAE,EAAEmB,QAAQE,SAAS,CAAC,GAAG,EAAEA,QAAQ,GAAG,IAAI;QAEhEX,MAAM,IAAI,CAAC,CAAC,UAAU,EAAEhC,KAAK,MAAM,IAAI,WAAW;QAClDgC,MAAM,IAAI,CAAC,CAAC,SAAS,EAAErC,WAAWiD,KAAK,KAAK,GAAG;QAC/CZ,MAAM,IAAI,CAAC,CAAC,OAAO,EAAErC,WAAWiD,KAAK,GAAG,GAAG;QAC3CZ,MAAM,IAAI,CACR,CAAC,YAAY,EAAE,AAAqB,YAArB,OAAOY,KAAK,IAAI,GAAgBA,KAAK,IAAI,GAAG,OAAO;QAEpEZ,MAAM,IAAI,CACR,CAAC,eAAe,EAAEtB,WAAWV,KAAK,SAAS,EAAE,aAAa,OAAO;QAGnE,IAAIA,AAAiB,aAAjBA,KAAK,OAAO,EAAe;YAC7B,MAAM6C,eAAejC,oBAAoBZ;YACzC,IAAI6C,cACFb,MAAM,IAAI,CAAC,CAAC,kBAAkB,EAAEa,YAAY,CAAC,EAAE,CAAC,EAAE,EAAEA,YAAY,CAAC,EAAE,CAAC,CAAC,CAAC;QAE1E;QAEA,IAAI7C,KAAK,YAAY,EACnBgC,MAAM,IAAI,CAAC,CAAC,SAAS,EAAEhC,KAAK,YAAY,EAAE;QAG5C,IAAIA,KAAK,SAAS,EAAE,YAAY;YAC9B,MAAMqC,cAAclB,qBAClBnB,KAAK,SAAS,CAAC,UAAU,EACzBoB,mBACAC,gBACAC;YAGFU,MAAM,IAAI,CAACK,YAAY,QAAQ;YAC/BJ,YAAY,IAAI,CAACI,YAAY,UAAU;QACzC;QAEA,MAAMS,kBAAkBhB,wBACtB9B,KAAK,QAAQ,EACboB,mBACAC,gBACAC;QAEF,IAAIwB,gBAAgB,KAAK,CAAC,MAAM,EAAE;YAChCd,MAAM,IAAI,IAAIc,gBAAgB,KAAK;YACnCb,YAAY,IAAI,IAAIa,gBAAgB,WAAW;QACjD;IACF;IAEA,OAAO;QACL,UAAUd,MAAM,IAAI,CAAC;QACrBC;IACF;AACF;AAEA,SAASc,eACPzD,SAAyB,EACzB+B,cAAsB;IAEtB,MAAM2B,WACJ1D,UAAU,IAAI,CACX,IAAI,GACJ,OAAO,CAAC,QAAQ,KAChB,OAAO,CAAC,mBAAmB,OAAO,CAAC,UAAU,EAAE+B,iBAAiB,GAAG;IACxE,OAAO,GAAGA,iBAAiB,EAAE,CAAC,EAAE4B,SAASD,UAAU,GAAG,CAAC;AACzD;AAEO,SAASE,oBACd5D,SAAyC,EACzCkD,OAAkC;IAElC,OAAOF,gBAAgBhD,WAAW,GAAGkD;AACvC;AAEO,SAASW,iBACdzD,MAA4C;IAE5C,MAAM0D,aAAa3D,aAAaC;IAEhC,MAAM2D,mBAAmBD,WAAW,UAAU,CAAC,GAAG,CAAC,CAAC9D,WAAWgE;QAC7D,MAAMC,WAAWjB,gBAAgBhD,WAAWgE;QAC5C,OAAO;YACL,gBAAgBA;YAChB,eAAehE,UAAU,IAAI;YAC7B,UAAUiE,SAAS,QAAQ;YAC3B,aAAaA,SAAS,WAAW;YACjC,mBAAmBR,eAAezD,WAAWgE;QAC/C;IACF;IAEA,MAAMrB,cAAcoB,iBAAiB,OAAO,CAAC,CAACnB,OAASA,KAAK,WAAW;IAEvE,MAAMsB,SAAS;QACb,CAAC,EAAE,EAAEJ,WAAW,SAAS,EAAE;QAC3BA,WAAW,gBAAgB,GAAG,CAAC,EAAE,EAAEA,WAAW,gBAAgB,EAAE,GAAG;QACnE,CAAC,iBAAiB,EAAEA,WAAW,UAAU,EAAE;QAC3C,CAAC,mBAAmB,EAAEA,WAAW,UAAU,CAAC,MAAM,EAAE;QACpD;WACGC,iBAAiB,GAAG,CACrB,CAACnB,OAAS,CAAC,EAAE,EAAEA,KAAK,iBAAiB,CAAC,EAAE,EAAEA,KAAK,aAAa,CAAC,CAAC,CAAC;KAElE,CACE,MAAM,CAACuB,SACP,IAAI,CAAC;IAER,OAAO;QACL,UAAU,GAAGD,OAAO,IAAI,EAAEH,iBAAiB,GAAG,CAAC,CAACnB,OAASA,KAAK,QAAQ,EAAE,IAAI,CAAC,gBAAgB;QAC7FD;IACF;AACF"}

package/dist/es/skill/index.mjs

@@ -1,5 +1,5 @@
+import { BaseMidsceneTools } from "@midscene/shared/agent-tools/base-tools";
 import { reportCLIError, runToolsCLI } from "@midscene/shared/cli";
-import { BaseMidsceneTools } from "@midscene/shared/mcp/base-tools";
 import { Agent } from "../agent/agent.mjs";
 function _define_property(obj, key, value) {
     if (key in obj) Object.defineProperty(obj, key, {

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

@@ -1 +1 @@
-{"version":3,"file":"skill/index.mjs","sources":["../../../src/skill/index.ts"],"sourcesContent":["import { reportCLIError, runToolsCLI } from '@midscene/shared/cli';\nimport { BaseMidsceneTools } from '@midscene/shared/mcp/base-tools';\nimport type { BaseAgent, BaseDevice } from '@midscene/shared/mcp/types';\nimport { Agent } from '../agent/agent';\nimport type { AbstractInterface } from '../device';\n\ntype DeviceClass = new (...args: any[]) => AbstractInterface;\n\n/**\n * Skill tools manager that lazily creates Agent from a Device class.\n * Used by runSkillCLI for CLI / Agent Skills scenarios where no agent exists at startup.\n */\nclass SkillMidsceneTools extends BaseMidsceneTools<BaseAgent> {\n  constructor(private DeviceClass: DeviceClass) {\n    super();\n  }\n\n  protected createTemporaryDevice(): BaseDevice {\n    return new this.DeviceClass() as unknown as BaseDevice;\n  }\n\n  protected async ensureAgent(): Promise<BaseAgent> {\n    if (!this.agent) {\n      const device = new this.DeviceClass();\n      this.agent = new Agent(device) as unknown as BaseAgent;\n    }\n    return this.agent;\n  }\n}\n\nexport interface SkillCLIOptions {\n  scriptName: string;\n  DeviceClass: DeviceClass;\n}\n\n/**\n * Launch a Skill CLI for a custom interface Device class.\n * This enables AI coding assistants (Claude Code, Cline, etc.) to control\n * your custom interface through CLI commands.\n *\n * @example\n * ```typescript\n * #!/usr/bin/env node\n * import { runSkillCLI } from '@midscene/core/skill';\n * import { SampleDevice } from './sample-device';\n *\n * runSkillCLI({\n *   DeviceClass: SampleDevice,\n *   scriptName: 'my-device',\n * });\n * ```\n */\nexport function runSkillCLI(options: SkillCLIOptions): Promise<void> {\n  const tools = new SkillMidsceneTools(options.DeviceClass);\n  return runToolsCLI(tools, options.scriptName).catch((e) => {\n    process.exit(reportCLIError(e));\n  });\n}\n"],"names":["SkillMidsceneTools","BaseMidsceneTools","device","Agent","DeviceClass","runSkillCLI","options","tools","runToolsCLI","e","process","reportCLIError"],"mappings":";;;;;;;;;;;;;AAYA,MAAMA,2BAA2BC;IAKrB,wBAAoC;QAC5C,OAAO,IAAI,IAAI,CAAC,WAAW;IAC7B;IAEA,MAAgB,cAAkC;QAChD,IAAI,CAAC,IAAI,CAAC,KAAK,EAAE;YACf,MAAMC,SAAS,IAAI,IAAI,CAAC,WAAW;YACnC,IAAI,CAAC,KAAK,GAAG,IAAIC,MAAMD;QACzB;QACA,OAAO,IAAI,CAAC,KAAK;IACnB;IAdA,YAAoBE,WAAwB,CAAE;QAC5C,KAAK,wDADaA,WAAW,GAAXA;IAEpB;AAaF;AAwBO,SAASC,YAAYC,OAAwB;IAClD,MAAMC,QAAQ,IAAIP,mBAAmBM,QAAQ,WAAW;IACxD,OAAOE,YAAYD,OAAOD,QAAQ,UAAU,EAAE,KAAK,CAAC,CAACG;QACnDC,QAAQ,IAAI,CAACC,eAAeF;IAC9B;AACF"}
+{"version":3,"file":"skill/index.mjs","sources":["../../../src/skill/index.ts"],"sourcesContent":["import { BaseMidsceneTools } from '@midscene/shared/agent-tools/base-tools';\nimport type { BaseAgent, BaseDevice } from '@midscene/shared/agent-tools/types';\nimport { reportCLIError, runToolsCLI } from '@midscene/shared/cli';\nimport { Agent } from '../agent/agent';\nimport type { AbstractInterface } from '../device';\n\ntype DeviceClass = new (...args: any[]) => AbstractInterface;\n\n/**\n * Skill tools manager that lazily creates Agent from a Device class.\n * Used by runSkillCLI for CLI / Agent Skills scenarios where no agent exists at startup.\n */\nclass SkillMidsceneTools extends BaseMidsceneTools<BaseAgent> {\n  constructor(private DeviceClass: DeviceClass) {\n    super();\n  }\n\n  protected createTemporaryDevice(): BaseDevice {\n    return new this.DeviceClass() as unknown as BaseDevice;\n  }\n\n  protected async ensureAgent(): Promise<BaseAgent> {\n    if (!this.agent) {\n      const device = new this.DeviceClass();\n      this.agent = new Agent(device) as unknown as BaseAgent;\n    }\n    return this.agent;\n  }\n}\n\nexport interface SkillCLIOptions {\n  scriptName: string;\n  DeviceClass: DeviceClass;\n}\n\n/**\n * Launch a Skill CLI for a custom interface Device class.\n * This enables AI coding assistants (Claude Code, Cline, etc.) to control\n * your custom interface through CLI commands.\n *\n * @example\n * ```typescript\n * #!/usr/bin/env node\n * import { runSkillCLI } from '@midscene/core/skill';\n * import { SampleDevice } from './sample-device';\n *\n * runSkillCLI({\n *   DeviceClass: SampleDevice,\n *   scriptName: 'my-device',\n * });\n * ```\n */\nexport function runSkillCLI(options: SkillCLIOptions): Promise<void> {\n  const tools = new SkillMidsceneTools(options.DeviceClass);\n  return runToolsCLI(tools, options.scriptName).catch((e) => {\n    process.exit(reportCLIError(e));\n  });\n}\n"],"names":["SkillMidsceneTools","BaseMidsceneTools","device","Agent","DeviceClass","runSkillCLI","options","tools","runToolsCLI","e","process","reportCLIError"],"mappings":";;;;;;;;;;;;;AAYA,MAAMA,2BAA2BC;IAKrB,wBAAoC;QAC5C,OAAO,IAAI,IAAI,CAAC,WAAW;IAC7B;IAEA,MAAgB,cAAkC;QAChD,IAAI,CAAC,IAAI,CAAC,KAAK,EAAE;YACf,MAAMC,SAAS,IAAI,IAAI,CAAC,WAAW;YACnC,IAAI,CAAC,KAAK,GAAG,IAAIC,MAAMD;QACzB;QACA,OAAO,IAAI,CAAC,KAAK;IACnB;IAdA,YAAoBE,WAAwB,CAAE;QAC5C,KAAK,wDADaA,WAAW,GAAXA;IAEpB;AAaF;AAwBO,SAASC,YAAYC,OAAwB;IAClD,MAAMC,QAAQ,IAAIP,mBAAmBM,QAAQ,WAAW;IACxD,OAAOE,YAAYD,OAAOD,QAAQ,UAAU,EAAE,KAAK,CAAC,CAACG;QACnDC,QAAQ,IAAI,CAACC,eAAeF;IAC9B;AACF"}

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 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 type { 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 { ServiceError } from './errors';\nexport {\n  ExecutionDump,\n  ReportActionDump,\n  GroupedActionDump,\n} from './dump/report-action-dump';\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  /**\n   * Raw top-level `.model` value returned by the model service response.\n   */\n  response_model_name: string | undefined;\n  /**\n   * Semantic intent of the model call, such as default, planning, or insight.\n   */\n  intent: string | undefined;\n  /**\n   * Config slot where the model config was resolved from. For example, a\n   * planning call may use the default slot when no planning model is configured.\n   */\n  slot: 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 type LocateResultPoint = [number, number];\nexport type Bbox = [number, number, number, number];\nexport type LocateResultBbox = Bbox;\nexport type PixelBbox = Bbox;\n\nexport interface AIElementLocateResponse {\n  bbox?: LocateResultBbox;\n  point?: LocateResultPoint;\n  errors?: string[];\n}\n\nexport interface AIDataExtractionResponse<DataDemand> {\n  data: DataDemand;\n  errors?: string[];\n  thought?: string;\n}\n\nexport interface AISectionLocatorResponse {\n  bbox?: LocateResultBbox;\n  point?: LocateResultPoint;\n  references_bbox?: LocateResultBbox[];\n  references_point?: LocateResultPoint[];\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  includedInRect?: boolean;\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  /**\n   * Adapter-extracted content used by Midscene for parsing. This is not the\n   * full provider response or choices[0].message.\n   */\n  rawResponse?: string;\n  rawChoiceMessage?: unknown;\n  usage?: AIUsageInfo;\n  searchArea?: Rect;\n  /**\n   * Adapter-extracted content from the search-area model call. This is not the\n   * full provider response or choices[0].message.\n   */\n  searchAreaRawResponse?: string;\n  searchAreaRawChoiceMessage?: unknown;\n  searchAreaUsage?: AIUsageInfo;\n  reasoning_content?: string;\n}\n\nexport interface DumpMeta {\n  logTime: number;\n}\n\nexport type ReportAttributes = Record<\n  string,\n  string | number | boolean | null | undefined\n>;\n\nexport interface ReportDumpWithAttributes {\n  dumpString: string;\n  attributes?: ReportAttributes;\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\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?: LocateResultBbox;\n  point?: LocateResultPoint;\n}\n\nexport type PlanningLocateParamWithLocatedPixelBbox = PlanningLocateParam & {\n  /** Pixel bbox of the located element in screenshot coordinates. */\n  locatedPixelBbox: PixelBbox;\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  /**\n   * Adapter-extracted content used by Midscene for parsing. This is not the\n   * full provider response or choices[0].message.\n   */\n  rawResponse?: string;\n  rawChoiceMessage?: unknown;\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    /**\n     * Optional feedback produced by a task for the next planning round.\n     * This is execution metadata, not part of the action return value.\n     */\n    planningFeedback?: string;\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    /**\n     * Pixel rect of the deepLocate first-stage search area in screenshot\n     * coordinates. Used by reports to explain the crop/zoom area that the\n     * final locate ran against.\n     */\n    searchArea?: Rect;\n    searchAreaUsage?: AIUsageInfo;\n    reasoning_content?: string;\n  };\n\nexport interface IExecutionDump extends DumpMeta {\n  /** Stable unique identifier for this execution run */\n  id?: string;\n  name: string;\n  description?: string;\n  tasks: ExecutionTask[];\n  aiActContext?: string;\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  domIncluded?: boolean | 'visible-only';\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 interface ExecutionTaskPlanningParam {\n  userInstruction: TUserPrompt;\n  userInstructionDisplay?: string;\n  aiActContext?: string;\n  imagesIncludeCount?: number;\n  deepThink?: DeepThinkOption;\n  subGoalStatus?: string;\n  memoriesStatus?: string;\n}\n\nexport type ExecutionTaskPlanningApply = ExecutionTaskApply<\n  'Planning',\n  ExecutionTaskPlanningParam,\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/*\nReport metadata - extracted from ReportActionDump for per-execution writes\n*/\nexport interface ReportMeta {\n  groupName: string;\n  groupDescription?: string;\n  sdkVersion: string;\n  modelBriefs: ModelBrief[];\n  deviceType?: string;\n}\n\n// Backward-compatible aliases for existing external consumers.\nexport type GroupMeta = ReportMeta;\n\n/*\nReport dump\n*/\nexport interface IReportActionDump {\n  sdkVersion: string;\n  groupName: string;\n  groupDescription?: string;\n  modelBriefs: ModelBrief[];\n  executions: IExecutionDump[];\n  deviceType?: string;\n}\n\n// Backward-compatible aliases for existing external consumers.\nexport type IGroupedActionDump = IReportActionDump;\n\nexport interface ModelBrief {\n  /**\n   * The intent/category of the model call, for example \"planning\" or \"insight\".\n   */\n  intent?: string;\n\n  /**\n   * The model name returned by usage metadata, for example \"gpt-4o\".\n   */\n  name?: string;\n\n  /**\n   * Optional human-readable model description, for example \"qwen2.5-vl mode\".\n   */\n  modelDescription?: string;\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: (\n    param: TParam,\n    context?: ExecutorContext,\n  ) => Promise<TReturn> | TReturn;\n  delayBeforeRunner?: number;\n  delayAfterRunner?: number;\n  /**\n   * An example param object for this action.\n   * Locate fields with { prompt } may be resolved to internal pixel bboxes 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  /**\n   * Optional cache directory path.\n   * When set, cache files are written to this directory instead of\n   * `<MIDSCENE_RUN_DIR>/cache`.\n   */\n  cacheDir?: string;\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  // @deprecated Use `reportFileName` and `cache.id` instead.\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 persist per-execution dump files next to the report, default false */\n  persistExecutionDump?: 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  reportAttributes?: ReportAttributes;\n  modelConfig?: TModelConfig;\n  cache?: Cache;\n  /**\n   * Maximum number of replanning cycles for aiAct.\n   * Defaults are resolved by the active model adapter: 20 for standard planning,\n   * 40 for UI-TARS, and 100 for Auto-GLM.\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 formatted local\n   * time instead of the runtime system time. The target interface must implement\n   * getDeviceLocalTimeString to provide device-local wall-clock time.\n   * Default: false\n   */\n  useDeviceTime?: 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 ReportFileAttributes {\n  testDuration: number;\n  testStatus: TestStatus;\n  testTitle: string;\n  testId: string;\n  testDescription: string;\n}\n\nexport type ReportFileWithAttributes =\n  | {\n      reportFilePath: string;\n      reportAttributes: ReportFileAttributes;\n    }\n  | {\n      reportFilePath?: string;\n      reportAttributes: ReportFileAttributes & { testStatus: 'skipped' };\n    };\n"],"names":["UIContext"],"mappings":";;;AAgIO,MAAeA;AA4BtB"}
+{"version":3,"file":"types.mjs","sources":["../../src/types.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n\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 type { 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 { ServiceError } from './errors';\nexport {\n  ExecutionDump,\n  ReportActionDump,\n  GroupedActionDump,\n} from './dump/report-action-dump';\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  /**\n   * Raw top-level `.model` value returned by the model service response.\n   */\n  response_model_name: string | undefined;\n  /**\n   * Semantic intent of the model call, such as default, planning, or insight.\n   */\n  intent: string | undefined;\n  /**\n   * Config slot where the model config was resolved from. For example, a\n   * planning call may use the default slot when no planning model is configured.\n   */\n  slot: 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 type LocateResultPoint = [number, number];\nexport type Bbox = [number, number, number, number];\nexport type LocateResultBbox = Bbox;\nexport type PixelBbox = Bbox;\n\nexport interface AIElementLocateResponse {\n  bbox?: LocateResultBbox;\n  point?: LocateResultPoint;\n  errors?: string[];\n}\n\nexport interface AIDataExtractionResponse<DataDemand> {\n  data: DataDemand;\n  errors?: string[];\n  thought?: string;\n}\n\nexport interface AISectionLocatorResponse {\n  bbox?: LocateResultBbox;\n  point?: LocateResultPoint;\n  references_bbox?: LocateResultBbox[];\n  references_point?: LocateResultPoint[];\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  includedInRect?: boolean;\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  /**\n   * Adapter-extracted content used by Midscene for parsing. This is not the\n   * full provider response or choices[0].message.\n   */\n  rawResponse?: string;\n  rawChoiceMessage?: unknown;\n  usage?: AIUsageInfo;\n  searchArea?: Rect;\n  /**\n   * Adapter-extracted content from the search-area model call. This is not the\n   * full provider response or choices[0].message.\n   */\n  searchAreaRawResponse?: string;\n  searchAreaRawChoiceMessage?: unknown;\n  searchAreaUsage?: AIUsageInfo;\n  reasoning_content?: string;\n}\n\nexport interface DumpMeta {\n  logTime: number;\n}\n\nexport type ReportAttributes = Record<\n  string,\n  string | number | boolean | null | undefined\n>;\n\nexport interface ReportDumpWithAttributes {\n  dumpString: string;\n  attributes?: ReportAttributes;\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\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?: LocateResultBbox;\n  point?: LocateResultPoint;\n}\n\nexport type PlanningLocateParamWithLocatedPixelBbox = PlanningLocateParam & {\n  /** Pixel bbox of the located element in screenshot coordinates. */\n  locatedPixelBbox: PixelBbox;\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  /**\n   * Adapter-extracted content used by Midscene for parsing. This is not the\n   * full provider response or choices[0].message.\n   */\n  rawResponse?: string;\n  rawChoiceMessage?: unknown;\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  description?: string;\n  timing?: string;\n}\n\nexport interface RecordToReportScreenshot {\n  /**\n   * PNG/JPEG data URI, or raw PNG base64 body.\n   */\n  base64: string;\n  description?: string;\n}\n\nexport interface RecordToReportOptions {\n  content?: string;\n  /**\n   * @deprecated Use `screenshots: [{ base64 }]` instead.\n   */\n  screenshotBase64?: string;\n  /**\n   * Custom screenshots to display under a single report entry.\n   */\n  screenshots?: RecordToReportScreenshot[];\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    /**\n     * Optional feedback produced by a task for the next planning round.\n     * This is execution metadata, not part of the action return value.\n     */\n    planningFeedback?: string;\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    /**\n     * Pixel rect of the deepLocate first-stage search area in screenshot\n     * coordinates. Used by reports to explain the crop/zoom area that the\n     * final locate ran against.\n     */\n    searchArea?: Rect;\n    searchAreaUsage?: AIUsageInfo;\n    reasoning_content?: string;\n  };\n\nexport interface IExecutionDump extends DumpMeta {\n  /** Stable unique identifier for this execution run */\n  id?: string;\n  name: string;\n  description?: string;\n  tasks: ExecutionTask[];\n  aiActContext?: string;\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  domIncluded?: boolean | 'visible-only';\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 interface ExecutionTaskPlanningParam {\n  userInstruction: TUserPrompt;\n  userInstructionDisplay?: string;\n  aiActContext?: string;\n  imagesIncludeCount?: number;\n  deepThink?: DeepThinkOption;\n  subGoalStatus?: string;\n  memoriesStatus?: string;\n}\n\nexport type ExecutionTaskPlanningApply = ExecutionTaskApply<\n  'Planning',\n  ExecutionTaskPlanningParam,\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/*\nReport metadata - extracted from ReportActionDump for per-execution writes\n*/\nexport interface ReportMeta {\n  groupName: string;\n  groupDescription?: string;\n  sdkVersion: string;\n  modelBriefs: ModelBrief[];\n  deviceType?: string;\n}\n\n// Backward-compatible aliases for existing external consumers.\nexport type GroupMeta = ReportMeta;\n\n/*\nReport dump\n*/\nexport interface IReportActionDump {\n  sdkVersion: string;\n  groupName: string;\n  groupDescription?: string;\n  modelBriefs: ModelBrief[];\n  executions: IExecutionDump[];\n  deviceType?: string;\n}\n\n// Backward-compatible aliases for existing external consumers.\nexport type IGroupedActionDump = IReportActionDump;\n\nexport interface ModelBrief {\n  /**\n   * The intent/category of the model call, for example \"planning\" or \"insight\".\n   */\n  intent?: string;\n\n  /**\n   * The model name returned by usage metadata, for example \"gpt-4o\".\n   */\n  name?: string;\n\n  /**\n   * Optional human-readable model description, for example \"qwen2.5-vl mode\".\n   */\n  modelDescription?: string;\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: (\n    param: TParam,\n    context?: ExecutorContext,\n  ) => Promise<TReturn> | TReturn;\n  delayBeforeRunner?: number;\n  delayAfterRunner?: number;\n  /**\n   * An example param object for this action.\n   * Locate fields with { prompt } may be resolved to internal pixel bboxes 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  /**\n   * Optional cache directory path.\n   * When set, cache files are written to this directory instead of\n   * `<MIDSCENE_RUN_DIR>/cache`.\n   */\n  cacheDir?: string;\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  // @deprecated Use `reportFileName` and `cache.id` instead.\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 persist per-execution dump files next to the report, default false */\n  persistExecutionDump?: 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  reportAttributes?: ReportAttributes;\n  modelConfig?: TModelConfig;\n  cache?: Cache;\n  /**\n   * Maximum number of replanning cycles for aiAct.\n   * Defaults are resolved by the active model adapter: 20 for standard planning,\n   * 40 for UI-TARS, and 100 for Auto-GLM.\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 formatted local\n   * time instead of the runtime system time. The target interface must implement\n   * getDeviceLocalTimeString to provide device-local wall-clock time.\n   * Default: false\n   */\n  useDeviceTime?: 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 ReportFileAttributes {\n  testDuration: number;\n  testStatus: TestStatus;\n  testTitle: string;\n  testId: string;\n  testDescription: string;\n}\n\nexport type ReportFileWithAttributes =\n  | {\n      reportFilePath: string;\n      reportAttributes: ReportFileAttributes;\n    }\n  | {\n      reportFilePath?: string;\n      reportAttributes: ReportFileAttributes & { testStatus: 'skipped' };\n    };\n"],"names":["UIContext"],"mappings":";;;AAgIO,MAAeA;AA4BtB"}