@midscene/core 1.2.2-beta-20260119111553.0 → 1.2.2-beta-20260119114334.0

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

@@ -1 +1 @@
-{"version":3,"file":"agent/agent.mjs","sources":["../../../src/agent/agent.ts"],"sourcesContent":["import {\n  type ActionParam,\n  type ActionReturn,\n  type AgentAssertOpt,\n  type AgentDescribeElementAtPointResult,\n  type AgentOpt,\n  type AgentWaitForOpt,\n  type CacheConfig,\n  type DeepThinkOption,\n  type DetailedLocateParam,\n  type DeviceAction,\n  ExecutionDump,\n  type ExecutionRecorderItem,\n  type ExecutionTask,\n  type ExecutionTaskLog,\n  type ExecutionTaskPlanning,\n  GroupedActionDump,\n  type LocateOption,\n  type LocateResultElement,\n  type LocateValidatorResult,\n  type LocatorValidatorOption,\n  type MidsceneYamlScript,\n  type OnTaskStartTip,\n  type PlanningAction,\n  type Rect,\n  ScreenshotItem,\n  type ScrollParam,\n  Service,\n  type ServiceAction,\n  type ServiceExtractOption,\n  type ServiceExtractParam,\n  type TUserPrompt,\n  type UIContext,\n} from '../index';\nexport type TestStatus =\n  | 'passed'\n  | 'failed'\n  | 'timedOut'\n  | 'skipped'\n  | 'interrupted';\nimport { isAutoGLM, isUITars } from '@/ai-model/auto-glm/util';\nimport yaml from 'js-yaml';\n\nimport {\n  getVersion,\n  groupedActionDumpFileExt,\n  processCacheConfig,\n  reportHTMLContent,\n  writeLogFile,\n} from '@/utils';\nimport {\n  ScriptPlayer,\n  buildDetailedLocateParam,\n  parseYamlScript,\n} from '../yaml/index';\n\nimport { existsSync } from 'node:fs';\nimport { resolve } from 'node:path';\nimport type { AbstractInterface } from '@/device';\nimport type { TaskRunner } from '@/task-runner';\nimport {\n  type IModelConfig,\n  MIDSCENE_REPLANNING_CYCLE_LIMIT,\n  ModelConfigManager,\n  globalConfigManager,\n  globalModelConfigManager,\n} from '@midscene/shared/env';\nimport { imageInfoOfBase64, resizeImgBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { defineActionAssert, defineActionFinalize } from '../device';\nimport { TaskCache } from './task-cache';\nimport {\n  TaskExecutionError,\n  TaskExecutor,\n  locatePlanForLocate,\n  withFileChooser,\n} from './tasks';\nimport { locateParamStr, paramStr, taskTitleStr, typeStr } from './ui-utils';\nimport {\n  commonContextParser,\n  getReportFileName,\n  parsePrompt,\n  printReportMsg,\n} from './utils';\n\nconst debug = getDebug('agent');\n\nconst distanceOfTwoPoints = (p1: [number, number], p2: [number, number]) => {\n  const [x1, y1] = p1;\n  const [x2, y2] = p2;\n  return Math.round(Math.sqrt((x1 - x2) ** 2 + (y1 - y2) ** 2));\n};\n\nconst includedInRect = (point: [number, number], rect: Rect) => {\n  const [x, y] = point;\n  const { left, top, width, height } = rect;\n  return x >= left && x <= left + width && y >= top && y <= top + height;\n};\n\nconst defaultServiceExtractOption: ServiceExtractOption = {\n  domIncluded: false,\n  screenshotIncluded: true,\n};\n\ntype CacheStrategy = NonNullable<CacheConfig['strategy']>;\n\nconst CACHE_STRATEGIES: readonly CacheStrategy[] = [\n  'read-only',\n  'read-write',\n  'write-only',\n];\n\nconst isValidCacheStrategy = (strategy: string): strategy is CacheStrategy =>\n  CACHE_STRATEGIES.some((value) => value === strategy);\n\nconst CACHE_STRATEGY_VALUES = CACHE_STRATEGIES.map(\n  (value) => `\"${value}\"`,\n).join(', ');\n\nconst legacyScrollTypeMap = {\n  once: 'singleAction',\n  untilBottom: 'scrollToBottom',\n  untilTop: 'scrollToTop',\n  untilRight: 'scrollToRight',\n  untilLeft: 'scrollToLeft',\n} as const;\n\ntype LegacyScrollType = keyof typeof legacyScrollTypeMap;\n\nconst normalizeScrollType = (\n  scrollType: ScrollParam['scrollType'] | LegacyScrollType | undefined,\n): ScrollParam['scrollType'] | undefined => {\n  if (!scrollType) {\n    return scrollType;\n  }\n\n  if (scrollType in legacyScrollTypeMap) {\n    return legacyScrollTypeMap[scrollType as LegacyScrollType];\n  }\n\n  return scrollType as ScrollParam['scrollType'];\n};\n\nconst defaultReplanningCycleLimit = 20;\nconst defaultVlmUiTarsReplanningCycleLimit = 40;\nconst defaultAutoGlmReplanningCycleLimit = 100;\n\nexport type AiActOptions = {\n  cacheable?: boolean;\n  fileChooserAccept?: string | string[];\n  deepThink?: DeepThinkOption;\n};\n\nexport class Agent<\n  InterfaceType extends AbstractInterface = AbstractInterface,\n> {\n  interface: InterfaceType;\n\n  service: Service;\n\n  dump: GroupedActionDump;\n\n  reportFile?: string | null;\n\n  reportFileName?: string;\n\n  taskExecutor: TaskExecutor;\n\n  opts: AgentOpt;\n\n  /**\n   * If true, the agent will not perform any actions\n   */\n  dryMode = false;\n\n  onTaskStartTip?: OnTaskStartTip;\n\n  taskCache?: TaskCache;\n\n  private dumpUpdateListeners: Array<\n    (dump: string, executionDump?: ExecutionDump) => void\n  > = [];\n\n  get onDumpUpdate():\n    | ((dump: string, executionDump?: ExecutionDump) => void)\n    | undefined {\n    return this.dumpUpdateListeners[0];\n  }\n\n  set onDumpUpdate(callback:\n    | ((dump: string, executionDump?: ExecutionDump) => void)\n    | undefined) {\n    // Clear existing listeners\n    this.dumpUpdateListeners = [];\n    // Add callback to array if provided\n    if (callback) {\n      this.dumpUpdateListeners.push(callback);\n    }\n  }\n\n  destroyed = false;\n\n  modelConfigManager: ModelConfigManager;\n\n  /**\n   * Frozen page context for consistent AI operations\n   */\n  private frozenUIContext?: UIContext;\n\n  private get aiActContext(): string | undefined {\n    return this.opts.aiActContext ?? this.opts.aiActionContext;\n  }\n\n  /**\n   * Flag to track if VL model warning has been shown\n   */\n  private hasWarnedNonVLModel = false;\n\n  /**\n   * Screenshot scale factor derived from actual screenshot dimensions\n   */\n  private screenshotScale?: number;\n\n  /**\n   * Internal promise to deduplicate screenshot scale computation\n   */\n  private screenshotScalePromise?: Promise<number>;\n\n  private executionDumpIndexByRunner = new WeakMap<TaskRunner, number>();\n\n  private fullActionSpace: DeviceAction[];\n\n  // @deprecated use .interface instead\n  get page() {\n    return this.interface;\n  }\n\n  /**\n   * Ensures VL model warning is shown once when needed\n   */\n  private ensureVLModelWarning() {\n    if (\n      !this.hasWarnedNonVLModel &&\n      this.interface.interfaceType !== 'puppeteer' &&\n      this.interface.interfaceType !== 'playwright' &&\n      this.interface.interfaceType !== 'static' &&\n      this.interface.interfaceType !== 'chrome-extension-proxy' &&\n      this.interface.interfaceType !== 'page-over-chrome-extension-bridge'\n    ) {\n      this.modelConfigManager.throwErrorIfNonVLModel();\n      this.hasWarnedNonVLModel = true;\n    }\n  }\n\n  /**\n   * Lazily compute the ratio between the physical screenshot width and the logical page width\n   */\n  private async getScreenshotScale(context: UIContext): Promise<number> {\n    if (this.screenshotScale !== undefined) {\n      return this.screenshotScale;\n    }\n\n    if (!this.screenshotScalePromise) {\n      this.screenshotScalePromise = (async () => {\n        const pageWidth = context.size?.width;\n        assert(\n          pageWidth && pageWidth > 0,\n          `Invalid page width when computing screenshot scale: ${pageWidth}`,\n        );\n\n        debug('will get image info of base64');\n        const screenshotBase64 = context.screenshot.base64;\n        const { width: screenshotWidth } =\n          await imageInfoOfBase64(screenshotBase64);\n        debug('image info of base64 done');\n\n        assert(\n          Number.isFinite(screenshotWidth) && screenshotWidth > 0,\n          `Invalid screenshot width when computing screenshot scale: ${screenshotWidth}`,\n        );\n\n        const computedScale = screenshotWidth / pageWidth;\n        assert(\n          Number.isFinite(computedScale) && computedScale > 0,\n          `Invalid computed screenshot scale: ${computedScale}`,\n        );\n\n        debug(\n          `Computed screenshot scale ${computedScale} from screenshot width ${screenshotWidth} and page width ${pageWidth}`,\n        );\n        return computedScale;\n      })();\n    }\n\n    try {\n      this.screenshotScale = await this.screenshotScalePromise;\n      return this.screenshotScale;\n    } finally {\n      this.screenshotScalePromise = undefined;\n    }\n  }\n\n  private resolveReplanningCycleLimit(\n    modelConfigForPlanning: IModelConfig,\n  ): number {\n    if (this.opts.replanningCycleLimit !== undefined) {\n      return this.opts.replanningCycleLimit;\n    }\n\n    return isUITars(modelConfigForPlanning.modelFamily)\n      ? defaultVlmUiTarsReplanningCycleLimit\n      : isAutoGLM(modelConfigForPlanning.modelFamily)\n        ? defaultAutoGlmReplanningCycleLimit\n        : defaultReplanningCycleLimit;\n  }\n\n  constructor(interfaceInstance: InterfaceType, opts?: AgentOpt) {\n    this.interface = interfaceInstance;\n\n    const envReplanningCycleLimit =\n      globalConfigManager.getEnvConfigValueAsNumber(\n        MIDSCENE_REPLANNING_CYCLE_LIMIT,\n      );\n\n    this.opts = Object.assign(\n      {\n        generateReport: true,\n        autoPrintReportMsg: true,\n        groupName: 'Midscene Report',\n        groupDescription: '',\n      },\n      opts || {},\n      opts?.replanningCycleLimit === undefined &&\n        envReplanningCycleLimit !== undefined &&\n        !Number.isNaN(envReplanningCycleLimit)\n        ? { replanningCycleLimit: envReplanningCycleLimit }\n        : {},\n    );\n\n    const resolvedAiActContext =\n      this.opts.aiActContext ?? this.opts.aiActionContext;\n    if (resolvedAiActContext !== undefined) {\n      this.opts.aiActContext = resolvedAiActContext;\n      this.opts.aiActionContext ??= resolvedAiActContext;\n    }\n\n    if (\n      opts?.modelConfig &&\n      (typeof opts?.modelConfig !== 'object' || Array.isArray(opts.modelConfig))\n    ) {\n      throw new Error(\n        `opts.modelConfig must be a plain object map of env keys to values, but got ${typeof opts?.modelConfig}`,\n      );\n    }\n    // Create ModelConfigManager if modelConfig or createOpenAIClient is provided\n    // Otherwise, use the global config manager\n    const hasCustomConfig = opts?.modelConfig || opts?.createOpenAIClient;\n    this.modelConfigManager = hasCustomConfig\n      ? new ModelConfigManager(opts?.modelConfig, opts?.createOpenAIClient)\n      : globalModelConfigManager;\n\n    this.onTaskStartTip = this.opts.onTaskStartTip;\n\n    this.service = new Service(async () => {\n      return this.getUIContext();\n    });\n\n    // Process cache configuration\n    const cacheConfigObj = this.processCacheConfig(opts || {});\n    if (cacheConfigObj) {\n      this.taskCache = new TaskCache(\n        cacheConfigObj.id,\n        cacheConfigObj.enabled,\n        undefined, // cacheFilePath\n        {\n          readOnly: cacheConfigObj.readOnly,\n          writeOnly: cacheConfigObj.writeOnly,\n        },\n      );\n    }\n\n    const baseActionSpace = this.interface.actionSpace();\n    this.fullActionSpace = [\n      ...baseActionSpace,\n      defineActionAssert(),\n      defineActionFinalize(),\n    ];\n\n    this.taskExecutor = new TaskExecutor(this.interface, this.service, {\n      taskCache: this.taskCache,\n      onTaskStart: this.callbackOnTaskStartTip.bind(this),\n      replanningCycleLimit: this.opts.replanningCycleLimit,\n      actionSpace: this.fullActionSpace,\n      hooks: {\n        onTaskUpdate: (runner) => {\n          const executionDump = runner.dump();\n          this.appendExecutionDump(executionDump, runner);\n\n          // Call all registered dump update listeners\n          const dumpString = this.dumpDataString();\n          for (const listener of this.dumpUpdateListeners) {\n            try {\n              listener(dumpString, executionDump);\n            } catch (error) {\n              console.error('Error in onDumpUpdate listener', error);\n            }\n          }\n\n          this.writeOutActionDumps();\n        },\n      },\n    });\n    this.dump = this.resetDump();\n    this.reportFileName =\n      opts?.reportFileName ||\n      getReportFileName(opts?.testId || this.interface.interfaceType || 'web');\n  }\n\n  async getActionSpace(): Promise<DeviceAction[]> {\n    return this.fullActionSpace;\n  }\n\n  async getUIContext(action?: ServiceAction): Promise<UIContext> {\n    // Check VL model configuration when UI context is first needed\n    this.ensureVLModelWarning();\n\n    // If page context is frozen, return the frozen context for all actions\n    if (this.frozenUIContext) {\n      debug('Using frozen page context for action:', action);\n      return this.frozenUIContext;\n    }\n\n    // Get original context\n    let context: UIContext;\n    if (this.interface.getContext) {\n      debug('Using page.getContext for action:', action);\n      context = await this.interface.getContext();\n    } else {\n      debug('Using commonContextParser');\n      context = await commonContextParser(this.interface, {\n        uploadServerUrl: this.modelConfigManager.getUploadTestServerUrl(),\n      });\n    }\n\n    debug('will get screenshot scale');\n    const computedScreenshotScale = await this.getScreenshotScale(context);\n    debug('computedScreenshotScale', computedScreenshotScale);\n\n    if (computedScreenshotScale !== 1) {\n      const scaleForLog = Number.parseFloat(computedScreenshotScale.toFixed(4));\n      debug(\n        `Applying computed screenshot scale: ${scaleForLog} (resize to logical size)`,\n      );\n      const targetWidth = Math.round(context.size.width);\n      const targetHeight = Math.round(context.size.height);\n      debug(`Resizing screenshot to ${targetWidth}x${targetHeight}`);\n      const currentScreenshotBase64 = context.screenshot.base64;\n      const resizedBase64 = await resizeImgBase64(currentScreenshotBase64, {\n        width: targetWidth,\n        height: targetHeight,\n      });\n      context.screenshot = ScreenshotItem.create(resizedBase64);\n    } else {\n      debug(`screenshot scale=${computedScreenshotScale}`);\n    }\n\n    return context;\n  }\n\n  async _snapshotContext(): Promise<UIContext> {\n    return await this.getUIContext('locate');\n  }\n\n  /**\n   * @deprecated Use {@link setAIActContext} instead.\n   */\n  async setAIActionContext(prompt: string) {\n    await this.setAIActContext(prompt);\n  }\n\n  async setAIActContext(prompt: string) {\n    if (this.aiActContext) {\n      console.warn(\n        'aiActContext is already set, and it is called again, will override the previous setting',\n      );\n    }\n    this.opts.aiActContext = prompt;\n    this.opts.aiActionContext = prompt;\n  }\n\n  resetDump() {\n    this.dump = new GroupedActionDump({\n      sdkVersion: getVersion(),\n      groupName: this.opts.groupName!,\n      groupDescription: this.opts.groupDescription,\n      executions: [],\n      modelBriefs: [],\n    });\n    this.executionDumpIndexByRunner = new WeakMap<TaskRunner, number>();\n\n    return this.dump;\n  }\n\n  appendExecutionDump(execution: ExecutionDump, runner?: TaskRunner) {\n    const currentDump = this.dump;\n    if (runner) {\n      const existingIndex = this.executionDumpIndexByRunner.get(runner);\n      if (existingIndex !== undefined) {\n        currentDump.executions[existingIndex] = execution;\n        return;\n      }\n      currentDump.executions.push(execution);\n      this.executionDumpIndexByRunner.set(\n        runner,\n        currentDump.executions.length - 1,\n      );\n      return;\n    }\n    currentDump.executions.push(execution);\n  }\n\n  dumpDataString() {\n    // update dump info\n    this.dump.groupName = this.opts.groupName!;\n    this.dump.groupDescription = this.opts.groupDescription;\n    return this.dump.serialize();\n  }\n\n  reportHTMLString() {\n    return reportHTMLContent(this.dumpDataString());\n  }\n\n  writeOutActionDumps() {\n    if (this.destroyed) {\n      throw new Error(\n        'PageAgent has been destroyed. Cannot update report file.',\n      );\n    }\n    const { generateReport, autoPrintReportMsg } = this.opts;\n    this.reportFile = writeLogFile({\n      fileName: this.reportFileName!,\n      fileExt: groupedActionDumpFileExt,\n      fileContent: this.dumpDataString(),\n      type: 'dump',\n      generateReport,\n    });\n    debug('writeOutActionDumps', this.reportFile);\n    if (generateReport && autoPrintReportMsg && this.reportFile) {\n      printReportMsg(this.reportFile);\n    }\n  }\n\n  private async callbackOnTaskStartTip(task: ExecutionTask) {\n    const param = paramStr(task);\n    const tip = param ? `${typeStr(task)} - ${param}` : typeStr(task);\n\n    if (this.onTaskStartTip) {\n      await this.onTaskStartTip(tip);\n    }\n  }\n\n  wrapActionInActionSpace<T extends DeviceAction>(\n    name: string,\n  ): (param: ActionParam<T>) => Promise<ActionReturn<T>> {\n    return async (param: ActionParam<T>) => {\n      return await this.callActionInActionSpace<ActionReturn<T>>(name, param);\n    };\n  }\n\n  async callActionInActionSpace<T = any>(\n    type: string,\n    opt?: T, // and all other action params\n  ) {\n    debug('callActionInActionSpace', type, ',', opt);\n\n    const actionPlan: PlanningAction<T> = {\n      type: type as any,\n      param: (opt as any) || {},\n      thought: '',\n    };\n    debug('actionPlan', actionPlan); // , ', in which the locateParam is', locateParam);\n\n    const plans: PlanningAction[] = [actionPlan].filter(\n      Boolean,\n    ) as PlanningAction[];\n\n    const title = taskTitleStr(\n      type as any,\n      locateParamStr((opt as any)?.locate || {}),\n    );\n\n    // assume all operation in action space is related to locating\n    const defaultIntentModelConfig =\n      this.modelConfigManager.getModelConfig('default');\n    const modelConfigForPlanning =\n      this.modelConfigManager.getModelConfig('planning');\n\n    const { output } = await this.taskExecutor.runPlans(\n      title,\n      plans,\n      modelConfigForPlanning,\n      defaultIntentModelConfig,\n    );\n    return output;\n  }\n\n  async aiTap(\n    locatePrompt: TUserPrompt,\n    opt?: LocateOption & { fileChooserAccept?: string | string[] },\n  ) {\n    assert(locatePrompt, 'missing locate prompt for tap');\n\n    const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n    const fileChooserAccept = opt?.fileChooserAccept\n      ? this.normalizeFileInput(opt.fileChooserAccept)\n      : undefined;\n\n    return withFileChooser(this.interface, fileChooserAccept, async () => {\n      return this.callActionInActionSpace('Tap', {\n        locate: detailedLocateParam,\n      });\n    });\n  }\n\n  async aiRightClick(locatePrompt: TUserPrompt, opt?: LocateOption) {\n    assert(locatePrompt, 'missing locate prompt for right click');\n\n    const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n    return this.callActionInActionSpace('RightClick', {\n      locate: detailedLocateParam,\n    });\n  }\n\n  async aiDoubleClick(locatePrompt: TUserPrompt, opt?: LocateOption) {\n    assert(locatePrompt, 'missing locate prompt for double click');\n\n    const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n    return this.callActionInActionSpace('DoubleClick', {\n      locate: detailedLocateParam,\n    });\n  }\n\n  async aiHover(locatePrompt: TUserPrompt, opt?: LocateOption) {\n    assert(locatePrompt, 'missing locate prompt for hover');\n\n    const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n    return this.callActionInActionSpace('Hover', {\n      locate: detailedLocateParam,\n    });\n  }\n\n  // New signature, always use locatePrompt as the first param\n  async aiInput(\n    locatePrompt: TUserPrompt,\n    opt: LocateOption & { value: string | number } & {\n      autoDismissKeyboard?: boolean;\n    } & { mode?: 'replace' | 'clear' | 'typeOnly' | 'append' },\n  ): Promise<any>;\n\n  // Legacy signature - deprecated\n  /**\n   * @deprecated Use aiInput(locatePrompt, opt) instead where opt contains the value\n   */\n  async aiInput(\n    value: string | number,\n    locatePrompt: TUserPrompt,\n    opt?: LocateOption & { autoDismissKeyboard?: boolean } & {\n      mode?: 'replace' | 'clear' | 'typeOnly' | 'append';\n    }, // AndroidDeviceInputOpt &\n  ): Promise<any>;\n\n  // Implementation\n  async aiInput(\n    locatePromptOrValue: TUserPrompt | string | number,\n    locatePromptOrOpt:\n      | TUserPrompt\n      | (LocateOption & { value: string | number } & {\n          autoDismissKeyboard?: boolean;\n        } & { mode?: 'replace' | 'clear' | 'typeOnly' | 'append' }) // AndroidDeviceInputOpt &\n      | undefined,\n    optOrUndefined?: LocateOption, // AndroidDeviceInputOpt &\n  ) {\n    let value: string | number;\n    let locatePrompt: TUserPrompt;\n    let opt:\n      | (LocateOption & { value: string | number } & {\n          autoDismissKeyboard?: boolean;\n        } & { mode?: 'replace' | 'clear' | 'typeOnly' | 'append' }) // AndroidDeviceInputOpt &\n      | undefined;\n\n    // Check if using new signature (first param is locatePrompt, second has value)\n    if (\n      typeof locatePromptOrOpt === 'object' &&\n      locatePromptOrOpt !== null &&\n      'value' in locatePromptOrOpt\n    ) {\n      // New signature: aiInput(locatePrompt, opt)\n      locatePrompt = locatePromptOrValue as TUserPrompt;\n      const optWithValue = locatePromptOrOpt as LocateOption & {\n        // AndroidDeviceInputOpt &\n        value: string | number;\n        autoDismissKeyboard?: boolean;\n      };\n      value = optWithValue.value;\n      opt = optWithValue;\n    } else {\n      // Legacy signature: aiInput(value, locatePrompt, opt)\n      value = locatePromptOrValue as string | number;\n      locatePrompt = locatePromptOrOpt as TUserPrompt;\n      opt = {\n        ...optOrUndefined,\n        value,\n      };\n    }\n\n    assert(\n      typeof value === 'string' || typeof value === 'number',\n      'input value must be a string or number, use empty string if you want to clear the input',\n    );\n    assert(locatePrompt, 'missing locate prompt for input');\n\n    const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n    // Convert value to string to ensure consistency\n    const stringValue = typeof value === 'number' ? String(value) : value;\n\n    return this.callActionInActionSpace('Input', {\n      ...(opt || {}),\n      value: stringValue,\n      locate: detailedLocateParam,\n    });\n  }\n\n  // New signature\n  async aiKeyboardPress(\n    locatePrompt: TUserPrompt,\n    opt: LocateOption & { keyName: string },\n  ): Promise<any>;\n\n  // Legacy signature - deprecated\n  /**\n   * @deprecated Use aiKeyboardPress(locatePrompt, opt) instead where opt contains the keyName\n   */\n  async aiKeyboardPress(\n    keyName: string,\n    locatePrompt?: TUserPrompt,\n    opt?: LocateOption,\n  ): Promise<any>;\n\n  // Implementation\n  async aiKeyboardPress(\n    locatePromptOrKeyName: TUserPrompt | string,\n    locatePromptOrOpt:\n      | TUserPrompt\n      | (LocateOption & { keyName: string })\n      | undefined,\n    optOrUndefined?: LocateOption,\n  ) {\n    let keyName: string;\n    let locatePrompt: TUserPrompt | undefined;\n    let opt: (LocateOption & { keyName: string }) | undefined;\n\n    // Check if using new signature (first param is locatePrompt, second has keyName)\n    if (\n      typeof locatePromptOrOpt === 'object' &&\n      locatePromptOrOpt !== null &&\n      'keyName' in locatePromptOrOpt\n    ) {\n      // New signature: aiKeyboardPress(locatePrompt, opt)\n      locatePrompt = locatePromptOrKeyName as TUserPrompt;\n      opt = locatePromptOrOpt as LocateOption & {\n        keyName: string;\n      };\n    } else {\n      // Legacy signature: aiKeyboardPress(keyName, locatePrompt, opt)\n      keyName = locatePromptOrKeyName as string;\n      locatePrompt = locatePromptOrOpt as TUserPrompt | undefined;\n      opt = {\n        ...(optOrUndefined || {}),\n        keyName,\n      };\n    }\n\n    assert(opt?.keyName, 'missing keyName for keyboard press');\n\n    const detailedLocateParam = locatePrompt\n      ? buildDetailedLocateParam(locatePrompt, opt)\n      : undefined;\n\n    return this.callActionInActionSpace('KeyboardPress', {\n      ...(opt || {}),\n      locate: detailedLocateParam,\n    });\n  }\n\n  // New signature\n  async aiScroll(\n    locatePrompt: TUserPrompt | undefined,\n    opt: LocateOption & ScrollParam,\n  ): Promise<any>;\n\n  // Legacy signature - deprecated\n  /**\n   * @deprecated Use aiScroll(locatePrompt, opt) instead where opt contains the scroll parameters\n   */\n  async aiScroll(\n    scrollParam: ScrollParam,\n    locatePrompt?: TUserPrompt,\n    opt?: LocateOption,\n  ): Promise<any>;\n\n  // Implementation\n  async aiScroll(\n    locatePromptOrScrollParam: TUserPrompt | ScrollParam | undefined,\n    locatePromptOrOpt: TUserPrompt | (LocateOption & ScrollParam) | undefined,\n    optOrUndefined?: LocateOption,\n  ) {\n    let scrollParam: ScrollParam;\n    let locatePrompt: TUserPrompt | undefined;\n    let opt: LocateOption | undefined;\n\n    // Check if using new signature (first param is locatePrompt, second has scroll params)\n    if (\n      typeof locatePromptOrOpt === 'object' &&\n      ('direction' in locatePromptOrOpt ||\n        'scrollType' in locatePromptOrOpt ||\n        'distance' in locatePromptOrOpt)\n    ) {\n      // New signature: aiScroll(locatePrompt, opt)\n      locatePrompt = locatePromptOrScrollParam as TUserPrompt;\n      opt = locatePromptOrOpt as LocateOption & ScrollParam;\n    } else {\n      // Legacy signature: aiScroll(scrollParam, locatePrompt, opt)\n      scrollParam = locatePromptOrScrollParam as ScrollParam;\n      locatePrompt = locatePromptOrOpt as TUserPrompt | undefined;\n      opt = {\n        ...(optOrUndefined || {}),\n        ...(scrollParam || {}),\n      };\n    }\n\n    if (opt) {\n      const normalizedScrollType = normalizeScrollType(\n        (opt as ScrollParam).scrollType as\n          | ScrollParam['scrollType']\n          | LegacyScrollType\n          | undefined,\n      );\n\n      if (normalizedScrollType !== (opt as ScrollParam).scrollType) {\n        (opt as ScrollParam) = {\n          ...(opt || {}),\n          scrollType: normalizedScrollType as ScrollParam['scrollType'],\n        };\n      }\n    }\n\n    const detailedLocateParam = buildDetailedLocateParam(\n      locatePrompt || '',\n      opt,\n    );\n\n    return this.callActionInActionSpace('Scroll', {\n      ...(opt || {}),\n      locate: detailedLocateParam,\n    });\n  }\n\n  async aiAct(\n    taskPrompt: string,\n    opt?: AiActOptions,\n  ): Promise<string | undefined> {\n    const fileChooserAccept = opt?.fileChooserAccept\n      ? this.normalizeFileInput(opt.fileChooserAccept)\n      : undefined;\n\n    const runAiAct = async () => {\n      const modelConfigForPlanning =\n        this.modelConfigManager.getModelConfig('planning');\n      const defaultIntentModelConfig =\n        this.modelConfigManager.getModelConfig('default');\n\n      const includeBboxInPlanning =\n        modelConfigForPlanning.modelName ===\n          defaultIntentModelConfig.modelName &&\n        modelConfigForPlanning.openaiBaseURL ===\n          defaultIntentModelConfig.openaiBaseURL;\n      debug('setting includeBboxInPlanning to', includeBboxInPlanning);\n\n      const cacheable = opt?.cacheable;\n      const deepThink = opt?.deepThink === 'unset' ? undefined : opt?.deepThink;\n      const replanningCycleLimit = this.resolveReplanningCycleLimit(\n        modelConfigForPlanning,\n      );\n      // if vlm-ui-tars or auto-glm, plan cache is not used\n      const isVlmUiTars = isUITars(modelConfigForPlanning.modelFamily);\n      const isAutoGlm = isAutoGLM(modelConfigForPlanning.modelFamily);\n      const matchedCache =\n        isVlmUiTars || isAutoGlm || cacheable === false\n          ? undefined\n          : this.taskCache?.matchPlanCache(taskPrompt);\n      if (\n        matchedCache &&\n        this.taskCache?.isCacheResultUsed &&\n        matchedCache.cacheContent?.yamlWorkflow?.trim()\n      ) {\n        // log into report file\n        await this.taskExecutor.loadYamlFlowAsPlanning(\n          taskPrompt,\n          matchedCache.cacheContent.yamlWorkflow,\n        );\n\n        debug('matched cache, will call .runYaml to run the action');\n        const yaml = matchedCache.cacheContent.yamlWorkflow;\n        await this.runYaml(yaml);\n        return;\n      }\n\n      // If cache matched but yamlWorkflow is empty, fall through to normal execution\n\n      const useDeepThink = (this.opts as any)?._deepThink;\n      if (useDeepThink) {\n        debug('using deep think planning settings');\n      }\n      const imagesIncludeCount: number | undefined = useDeepThink\n        ? undefined\n        : 2;\n      const { output: actionOutput } = await this.taskExecutor.action(\n        taskPrompt,\n        modelConfigForPlanning,\n        defaultIntentModelConfig,\n        includeBboxInPlanning,\n        this.aiActContext,\n        cacheable,\n        replanningCycleLimit,\n        imagesIncludeCount,\n        deepThink,\n        fileChooserAccept,\n      );\n\n      // update cache\n      if (this.taskCache && actionOutput?.yamlFlow && cacheable !== false) {\n        const yamlContent: MidsceneYamlScript = {\n          tasks: [\n            {\n              name: taskPrompt,\n              flow: actionOutput.yamlFlow,\n            },\n          ],\n        };\n        const yamlFlowStr = yaml.dump(yamlContent);\n        this.taskCache.updateOrAppendCacheRecord(\n          {\n            type: 'plan',\n            prompt: taskPrompt,\n            yamlWorkflow: yamlFlowStr,\n          },\n          matchedCache,\n        );\n      }\n\n      return actionOutput?.output;\n    };\n\n    return await runAiAct();\n  }\n\n  /**\n   * @deprecated Use {@link Agent.aiAct} instead.\n   */\n  async aiAction(taskPrompt: string, opt?: AiActOptions) {\n    return this.aiAct(taskPrompt, opt);\n  }\n\n  async aiQuery<ReturnType = any>(\n    demand: ServiceExtractParam,\n    opt: ServiceExtractOption = defaultServiceExtractOption,\n  ): Promise<ReturnType> {\n    const modelConfig = this.modelConfigManager.getModelConfig('insight');\n    const { output } = await this.taskExecutor.createTypeQueryExecution(\n      'Query',\n      demand,\n      modelConfig,\n      opt,\n    );\n    return output as ReturnType;\n  }\n\n  async aiBoolean(\n    prompt: TUserPrompt,\n    opt: ServiceExtractOption = defaultServiceExtractOption,\n  ): Promise<boolean> {\n    const modelConfig = this.modelConfigManager.getModelConfig('insight');\n\n    const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n    const { output } = await this.taskExecutor.createTypeQueryExecution(\n      'Boolean',\n      textPrompt,\n      modelConfig,\n      opt,\n      multimodalPrompt,\n    );\n    return output as boolean;\n  }\n\n  async aiNumber(\n    prompt: TUserPrompt,\n    opt: ServiceExtractOption = defaultServiceExtractOption,\n  ): Promise<number> {\n    const modelConfig = this.modelConfigManager.getModelConfig('insight');\n\n    const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n    const { output } = await this.taskExecutor.createTypeQueryExecution(\n      'Number',\n      textPrompt,\n      modelConfig,\n      opt,\n      multimodalPrompt,\n    );\n    return output as number;\n  }\n\n  async aiString(\n    prompt: TUserPrompt,\n    opt: ServiceExtractOption = defaultServiceExtractOption,\n  ): Promise<string> {\n    const modelConfig = this.modelConfigManager.getModelConfig('insight');\n\n    const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n    const { output } = await this.taskExecutor.createTypeQueryExecution(\n      'String',\n      textPrompt,\n      modelConfig,\n      opt,\n      multimodalPrompt,\n    );\n    return output as string;\n  }\n\n  async aiAsk(\n    prompt: TUserPrompt,\n    opt: ServiceExtractOption = defaultServiceExtractOption,\n  ): Promise<string> {\n    return this.aiString(prompt, opt);\n  }\n\n  async describeElementAtPoint(\n    center: [number, number],\n    opt?: {\n      verifyPrompt?: boolean;\n      retryLimit?: number;\n      deepThink?: boolean;\n    } & LocatorValidatorOption,\n  ): Promise<AgentDescribeElementAtPointResult> {\n    const { verifyPrompt = true, retryLimit = 3 } = opt || {};\n\n    let success = false;\n    let retryCount = 0;\n    let resultPrompt = '';\n    let deepThink = opt?.deepThink || false;\n    let verifyResult: LocateValidatorResult | undefined;\n\n    while (!success && retryCount < retryLimit) {\n      if (retryCount >= 2) {\n        deepThink = true;\n      }\n      debug(\n        'aiDescribe',\n        center,\n        'verifyPrompt',\n        verifyPrompt,\n        'retryCount',\n        retryCount,\n        'deepThink',\n        deepThink,\n      );\n      // use same intent as aiLocate\n      const modelConfig = this.modelConfigManager.getModelConfig('insight');\n\n      const text = await this.service.describe(center, modelConfig, {\n        deepThink,\n      });\n      debug('aiDescribe text', text);\n      assert(text.description, `failed to describe element at [${center}]`);\n      resultPrompt = text.description;\n\n      verifyResult = await this.verifyLocator(\n        resultPrompt,\n        deepThink ? { deepThink: true } : undefined,\n        center,\n        opt,\n      );\n      if (verifyResult.pass) {\n        success = true;\n      } else {\n        retryCount++;\n      }\n    }\n\n    return {\n      prompt: resultPrompt,\n      deepThink,\n      verifyResult,\n    };\n  }\n\n  async verifyLocator(\n    prompt: string,\n    locateOpt: LocateOption | undefined,\n    expectCenter: [number, number],\n    verifyLocateOption?: LocatorValidatorOption,\n  ): Promise<LocateValidatorResult> {\n    debug('verifyLocator', prompt, locateOpt, expectCenter, verifyLocateOption);\n\n    const { center: verifyCenter, rect: verifyRect } = await this.aiLocate(\n      prompt,\n      locateOpt,\n    );\n    const distance = distanceOfTwoPoints(expectCenter, verifyCenter);\n    const included = includedInRect(expectCenter, verifyRect);\n    const pass =\n      distance <= (verifyLocateOption?.centerDistanceThreshold || 20) ||\n      included;\n    const verifyResult = {\n      pass,\n      rect: verifyRect,\n      center: verifyCenter,\n      centerDistance: distance,\n    };\n    debug('aiDescribe verifyResult', verifyResult);\n    return verifyResult;\n  }\n\n  async aiLocate(prompt: TUserPrompt, opt?: LocateOption) {\n    const locateParam = buildDetailedLocateParam(prompt, opt);\n    assert(locateParam, 'cannot get locate param for aiLocate');\n    const locatePlan = locatePlanForLocate(locateParam);\n    const plans = [locatePlan];\n    const defaultIntentModelConfig =\n      this.modelConfigManager.getModelConfig('default');\n    const modelConfigForPlanning =\n      this.modelConfigManager.getModelConfig('planning');\n\n    const { output } = await this.taskExecutor.runPlans(\n      taskTitleStr('Locate', locateParamStr(locateParam)),\n      plans,\n      modelConfigForPlanning,\n      defaultIntentModelConfig,\n    );\n\n    const { element } = output;\n\n    const dprValue = await (this.interface.size() as any).dpr;\n    const dprEntry = dprValue\n      ? {\n          dpr: dprValue,\n        }\n      : {};\n    return {\n      rect: element?.rect,\n      center: element?.center,\n      ...dprEntry,\n    } as Pick<LocateResultElement, 'rect' | 'center'> & {\n      dpr?: number; // this field is deprecated\n    };\n  }\n\n  async aiAssert(\n    assertion: TUserPrompt,\n    msg?: string,\n    opt?: AgentAssertOpt & ServiceExtractOption,\n  ) {\n    const modelConfig = this.modelConfigManager.getModelConfig('insight');\n\n    const serviceOpt: ServiceExtractOption = {\n      domIncluded: opt?.domIncluded ?? defaultServiceExtractOption.domIncluded,\n      screenshotIncluded:\n        opt?.screenshotIncluded ??\n        defaultServiceExtractOption.screenshotIncluded,\n    };\n\n    const { textPrompt, multimodalPrompt } = parsePrompt(assertion);\n    const assertionText =\n      typeof assertion === 'string' ? assertion : assertion.prompt;\n\n    try {\n      const { output, thought } =\n        await this.taskExecutor.createTypeQueryExecution<boolean>(\n          'Assert',\n          textPrompt,\n          modelConfig,\n          serviceOpt,\n          multimodalPrompt,\n        );\n\n      const pass = Boolean(output);\n      const message = pass\n        ? undefined\n        : `Assertion failed: ${msg || assertionText}\\nReason: ${thought || '(no_reason)'}`;\n\n      if (opt?.keepRawResponse) {\n        return {\n          pass,\n          thought,\n          message,\n        };\n      }\n\n      if (!pass) {\n        throw new Error(message);\n      }\n    } catch (error) {\n      if (error instanceof TaskExecutionError) {\n        const errorTask = error.errorTask;\n        const thought = errorTask?.thought;\n        const rawError = errorTask?.error;\n        const rawMessage =\n          errorTask?.errorMessage ||\n          (rawError instanceof Error\n            ? rawError.message\n            : rawError\n              ? String(rawError)\n              : undefined);\n        const reason = thought || rawMessage || '(no_reason)';\n        const message = `Assertion failed: ${msg || assertionText}\\nReason: ${reason}`;\n\n        if (opt?.keepRawResponse) {\n          return {\n            pass: false,\n            thought,\n            message,\n          };\n        }\n\n        throw new Error(message, {\n          cause: rawError ?? error,\n        });\n      }\n\n      throw error;\n    }\n  }\n\n  async aiWaitFor(assertion: TUserPrompt, opt?: AgentWaitForOpt) {\n    const modelConfig = this.modelConfigManager.getModelConfig('insight');\n    await this.taskExecutor.waitFor(\n      assertion,\n      {\n        ...opt,\n        timeoutMs: opt?.timeoutMs || 15 * 1000,\n        checkIntervalMs: opt?.checkIntervalMs || 3 * 1000,\n      },\n      modelConfig,\n    );\n  }\n\n  async ai(...args: Parameters<typeof this.aiAct>) {\n    return this.aiAct(...args);\n  }\n\n  async runYaml(yamlScriptContent: string): Promise<{\n    result: Record<string, any>;\n  }> {\n    const script = parseYamlScript(yamlScriptContent, 'yaml');\n    const player = new ScriptPlayer(script, async () => {\n      return { agent: this, freeFn: [] };\n    });\n    await player.run();\n\n    if (player.status === 'error') {\n      const errors = player.taskStatusList\n        .filter((task) => task.status === 'error')\n        .map((task) => {\n          return `task - ${task.name}: ${task.error?.message}`;\n        })\n        .join('\\n');\n      throw new Error(`Error(s) occurred in running yaml script:\\n${errors}`);\n    }\n\n    return {\n      result: player.result,\n    };\n  }\n\n  async evaluateJavaScript(script: string) {\n    assert(\n      this.interface.evaluateJavaScript,\n      'evaluateJavaScript is not supported in current agent',\n    );\n    return this.interface.evaluateJavaScript(script);\n  }\n\n  /**\n   * Add a dump update listener\n   * @param listener Listener function\n   * @returns A remove function that can be called to remove this listener\n   */\n  addDumpUpdateListener(\n    listener: (dump: string, executionDump?: ExecutionDump) => void,\n  ): () => void {\n    this.dumpUpdateListeners.push(listener);\n\n    // Return remove function\n    return () => {\n      this.removeDumpUpdateListener(listener);\n    };\n  }\n\n  /**\n   * Remove a dump update listener\n   * @param listener The listener function to remove\n   */\n  removeDumpUpdateListener(\n    listener: (dump: string, executionDump?: ExecutionDump) => void,\n  ): void {\n    const index = this.dumpUpdateListeners.indexOf(listener);\n    if (index > -1) {\n      this.dumpUpdateListeners.splice(index, 1);\n    }\n  }\n\n  /**\n   * Clear all dump update listeners\n   */\n  clearDumpUpdateListeners(): void {\n    this.dumpUpdateListeners = [];\n  }\n\n  async destroy() {\n    // Early return if already destroyed\n    if (this.destroyed) {\n      return;\n    }\n\n    await this.interface.destroy?.();\n    this.resetDump(); // reset dump to release memory\n    this.destroyed = true;\n  }\n\n  async recordToReport(\n    title?: string,\n    opt?: {\n      content: string;\n    },\n  ) {\n    // 1. screenshot\n    const base64 = await this.interface.screenshotBase64();\n    const screenshot = ScreenshotItem.create(base64);\n    const now = Date.now();\n    // 2. build recorder\n    const recorder: ExecutionRecorderItem[] = [\n      {\n        type: 'screenshot',\n        ts: now,\n        screenshot,\n      },\n    ];\n    // 3. build ExecutionTaskLog\n    const task: ExecutionTaskLog = {\n      type: 'Log',\n      subType: 'Screenshot',\n      status: 'finished',\n      recorder,\n      timing: {\n        start: now,\n        end: now,\n        cost: 0,\n      },\n      param: {\n        content: opt?.content || '',\n      },\n      executor: async () => {},\n    };\n    // 4. build ExecutionDump\n    const executionDump = new ExecutionDump({\n      logTime: now,\n      name: `Log - ${title || 'untitled'}`,\n      description: opt?.content || '',\n      tasks: [task],\n    });\n    // 5. append to execution dump\n    this.appendExecutionDump(executionDump);\n\n    // Call all registered dump update listeners\n    const dumpString = this.dumpDataString();\n    for (const listener of this.dumpUpdateListeners) {\n      try {\n        listener(dumpString);\n      } catch (error) {\n        console.error('Error in onDumpUpdate listener', error);\n      }\n    }\n\n    this.writeOutActionDumps();\n  }\n\n  /**\n   * @deprecated Use {@link Agent.recordToReport} instead.\n   */\n  async logScreenshot(\n    title?: string,\n    opt?: {\n      content: string;\n    },\n  ) {\n    await this.recordToReport(title, opt);\n  }\n\n  _unstableLogContent() {\n    const { groupName, groupDescription, executions } = this.dump;\n    return {\n      groupName,\n      groupDescription,\n      executions: executions || [],\n    };\n  }\n\n  /**\n   * Freezes the current page context to be reused in subsequent AI operations\n   * This avoids recalculating page context for each operation\n   */\n  async freezePageContext(): Promise<void> {\n    debug('Freezing page context');\n    const context = await this._snapshotContext();\n    // Mark the context as frozen\n    context._isFrozen = true;\n    this.frozenUIContext = context;\n    debug('Page context frozen successfully');\n  }\n\n  /**\n   * Unfreezes the page context, allowing AI operations to calculate context dynamically\n   */\n  async unfreezePageContext(): Promise<void> {\n    debug('Unfreezing page context');\n    this.frozenUIContext = undefined;\n    debug('Page context unfrozen successfully');\n  }\n\n  /**\n   * Process cache configuration and return normalized cache settings\n   */\n  private processCacheConfig(opts: AgentOpt): {\n    id: string;\n    enabled: boolean;\n    readOnly: boolean;\n    writeOnly: boolean;\n  } | null {\n    // Validate original cache config before processing\n    // Agent requires explicit IDs - don't allow auto-generation\n    if (opts.cache === true) {\n      throw new Error(\n        'cache: true requires an explicit cache ID. Please provide:\\n' +\n          'Example: cache: { id: \"my-cache-id\" }',\n      );\n    }\n\n    // Check if cache config object is missing ID\n    if (\n      opts.cache &&\n      typeof opts.cache === 'object' &&\n      opts.cache !== null &&\n      !opts.cache.id\n    ) {\n      throw new Error(\n        'cache configuration requires an explicit id.\\n' +\n          'Example: cache: { id: \"my-cache-id\" }',\n      );\n    }\n\n    // Use the unified utils function to process cache configuration\n    const cacheConfig = processCacheConfig(\n      opts.cache,\n      opts.cacheId || opts.testId || 'default',\n    );\n\n    if (!cacheConfig) {\n      return null;\n    }\n\n    // Handle cache configuration object\n    if (typeof cacheConfig === 'object' && cacheConfig !== null) {\n      const id = cacheConfig.id;\n      const rawStrategy = cacheConfig.strategy as unknown;\n      let strategyValue: string;\n\n      if (rawStrategy === undefined) {\n        strategyValue = 'read-write';\n      } else if (typeof rawStrategy === 'string') {\n        strategyValue = rawStrategy;\n      } else {\n        throw new Error(\n          `cache.strategy must be a string when provided, but received type ${typeof rawStrategy}`,\n        );\n      }\n\n      if (!isValidCacheStrategy(strategyValue)) {\n        throw new Error(\n          `cache.strategy must be one of ${CACHE_STRATEGY_VALUES}, but received \"${strategyValue}\"`,\n        );\n      }\n\n      const isReadOnly = strategyValue === 'read-only';\n      const isWriteOnly = strategyValue === 'write-only';\n\n      return {\n        id,\n        enabled: !isWriteOnly,\n        readOnly: isReadOnly,\n        writeOnly: isWriteOnly,\n      };\n    }\n\n    return null;\n  }\n\n  private normalizeFilePaths(files: string[]): string[] {\n    return files.map((file) => {\n      const absolutePath = resolve(file);\n      if (!existsSync(absolutePath)) {\n        throw new Error(`File not found: ${file}`);\n      }\n      return absolutePath;\n    });\n  }\n\n  private normalizeFileInput(files: string | string[]): string[] {\n    const filesArray = Array.isArray(files) ? files : [files];\n    return this.normalizeFilePaths(filesArray);\n  }\n\n  /**\n   * Manually flush cache to file\n   * @param options - Optional configuration\n   * @param options.cleanUnused - If true, removes unused cache records before flushing\n   */\n  async flushCache(options?: { cleanUnused?: boolean }): Promise<void> {\n    if (!this.taskCache) {\n      throw new Error('Cache is not configured');\n    }\n\n    this.taskCache.flushCacheToFile(options);\n  }\n}\n\nexport const createAgent = (\n  interfaceInstance: AbstractInterface,\n  opts?: AgentOpt,\n) => {\n  return new Agent(interfaceInstance, opts);\n};\n"],"names":["debug","getDebug","distanceOfTwoPoints","p1","p2","x1","y1","x2","y2","Math","includedInRect","point","rect","x","y","left","top","width","height","defaultServiceExtractOption","CACHE_STRATEGIES","isValidCacheStrategy","strategy","value","CACHE_STRATEGY_VALUES","legacyScrollTypeMap","normalizeScrollType","scrollType","defaultReplanningCycleLimit","defaultVlmUiTarsReplanningCycleLimit","defaultAutoGlmReplanningCycleLimit","Agent","callback","context","undefined","pageWidth","assert","screenshotBase64","screenshotWidth","imageInfoOfBase64","Number","computedScale","modelConfigForPlanning","isUITars","isAutoGLM","action","commonContextParser","computedScreenshotScale","scaleForLog","targetWidth","targetHeight","currentScreenshotBase64","resizedBase64","resizeImgBase64","ScreenshotItem","prompt","console","GroupedActionDump","getVersion","WeakMap","execution","runner","currentDump","existingIndex","reportHTMLContent","Error","generateReport","autoPrintReportMsg","writeLogFile","groupedActionDumpFileExt","printReportMsg","task","param","paramStr","tip","typeStr","name","type","opt","actionPlan","plans","Boolean","title","taskTitleStr","locateParamStr","defaultIntentModelConfig","output","locatePrompt","detailedLocateParam","buildDetailedLocateParam","fileChooserAccept","withFileChooser","locatePromptOrValue","locatePromptOrOpt","optOrUndefined","optWithValue","stringValue","String","locatePromptOrKeyName","keyName","locatePromptOrScrollParam","scrollParam","normalizedScrollType","taskPrompt","runAiAct","includeBboxInPlanning","cacheable","deepThink","replanningCycleLimit","isVlmUiTars","isAutoGlm","matchedCache","yaml","useDeepThink","imagesIncludeCount","actionOutput","yamlContent","yamlFlowStr","demand","modelConfig","textPrompt","multimodalPrompt","parsePrompt","center","verifyPrompt","retryLimit","success","retryCount","resultPrompt","verifyResult","text","locateOpt","expectCenter","verifyLocateOption","verifyCenter","verifyRect","distance","included","pass","locateParam","locatePlan","locatePlanForLocate","element","dprValue","dprEntry","assertion","msg","serviceOpt","assertionText","thought","message","error","TaskExecutionError","errorTask","rawError","rawMessage","reason","args","yamlScriptContent","script","parseYamlScript","player","ScriptPlayer","errors","listener","index","base64","screenshot","now","Date","recorder","executionDump","ExecutionDump","dumpString","groupName","groupDescription","executions","opts","cacheConfig","processCacheConfig","id","rawStrategy","strategyValue","isReadOnly","isWriteOnly","files","file","absolutePath","resolve","existsSync","filesArray","Array","options","interfaceInstance","envReplanningCycleLimit","globalConfigManager","MIDSCENE_REPLANNING_CYCLE_LIMIT","Object","resolvedAiActContext","hasCustomConfig","ModelConfigManager","globalModelConfigManager","Service","cacheConfigObj","TaskCache","baseActionSpace","defineActionAssert","defineActionFinalize","TaskExecutor","getReportFileName","createAgent"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;AAsFA,MAAMA,QAAQC,SAAS;AAEvB,MAAMC,sBAAsB,CAACC,IAAsBC;IACjD,MAAM,CAACC,IAAIC,GAAG,GAAGH;IACjB,MAAM,CAACI,IAAIC,GAAG,GAAGJ;IACjB,OAAOK,KAAK,KAAK,CAACA,KAAK,IAAI,CAAEJ,AAAAA,CAAAA,KAAKE,EAAC,KAAM,IAAKD,AAAAA,CAAAA,KAAKE,EAAC,KAAM;AAC5D;AAEA,MAAME,iBAAiB,CAACC,OAAyBC;IAC/C,MAAM,CAACC,GAAGC,EAAE,GAAGH;IACf,MAAM,EAAEI,IAAI,EAAEC,GAAG,EAAEC,KAAK,EAAEC,MAAM,EAAE,GAAGN;IACrC,OAAOC,KAAKE,QAAQF,KAAKE,OAAOE,SAASH,KAAKE,OAAOF,KAAKE,MAAME;AAClE;AAEA,MAAMC,8BAAoD;IACxD,aAAa;IACb,oBAAoB;AACtB;AAIA,MAAMC,mBAA6C;IACjD;IACA;IACA;CACD;AAED,MAAMC,uBAAuB,CAACC,WAC5BF,iBAAiB,IAAI,CAAC,CAACG,QAAUA,UAAUD;AAE7C,MAAME,wBAAwBJ,iBAAiB,GAAG,CAChD,CAACG,QAAU,CAAC,CAAC,EAAEA,MAAM,CAAC,CAAC,EACvB,IAAI,CAAC;AAEP,MAAME,sBAAsB;IAC1B,MAAM;IACN,aAAa;IACb,UAAU;IACV,YAAY;IACZ,WAAW;AACb;AAIA,MAAMC,sBAAsB,CAC1BC;IAEA,IAAI,CAACA,YACH,OAAOA;IAGT,IAAIA,cAAcF,qBAChB,OAAOA,mBAAmB,CAACE,WAA+B;IAG5D,OAAOA;AACT;AAEA,MAAMC,8BAA8B;AACpC,MAAMC,uCAAuC;AAC7C,MAAMC,qCAAqC;AAQpC,MAAMC;IA8BX,IAAI,eAEU;QACZ,OAAO,IAAI,CAAC,mBAAmB,CAAC,EAAE;IACpC;IAEA,IAAI,aAAaC,QAEJ,EAAE;QAEb,IAAI,CAAC,mBAAmB,GAAG,EAAE;QAE7B,IAAIA,UACF,IAAI,CAAC,mBAAmB,CAAC,IAAI,CAACA;IAElC;IAWA,IAAY,eAAmC;QAC7C,OAAO,IAAI,CAAC,IAAI,CAAC,YAAY,IAAI,IAAI,CAAC,IAAI,CAAC,eAAe;IAC5D;IAsBA,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IAKQ,uBAAuB;QAC7B,IACE,CAAC,IAAI,CAAC,mBAAmB,IACzB,AAAiC,gBAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,IAC5B,AAAiC,iBAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,IAC5B,AAAiC,aAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,IAC5B,AAAiC,6BAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,IAC5B,AAAiC,wCAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,EAC5B;YACA,IAAI,CAAC,kBAAkB,CAAC,sBAAsB;YAC9C,IAAI,CAAC,mBAAmB,GAAG;QAC7B;IACF;IAKA,MAAc,mBAAmBC,OAAkB,EAAmB;QACpE,IAAI,AAAyBC,WAAzB,IAAI,CAAC,eAAe,EACtB,OAAO,IAAI,CAAC,eAAe;QAG7B,IAAI,CAAC,IAAI,CAAC,sBAAsB,EAC9B,IAAI,CAAC,sBAAsB,GAAI;YAC7B,MAAMC,YAAYF,QAAQ,IAAI,EAAE;YAChCG,OACED,aAAaA,YAAY,GACzB,CAAC,oDAAoD,EAAEA,WAAW;YAGpEnC,MAAM;YACN,MAAMqC,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;YAClD,MAAM,EAAE,OAAOK,eAAe,EAAE,GAC9B,MAAMC,kBAAkBF;YAC1BrC,MAAM;YAENoC,OACEI,OAAO,QAAQ,CAACF,oBAAoBA,kBAAkB,GACtD,CAAC,0DAA0D,EAAEA,iBAAiB;YAGhF,MAAMG,gBAAgBH,kBAAkBH;YACxCC,OACEI,OAAO,QAAQ,CAACC,kBAAkBA,gBAAgB,GAClD,CAAC,mCAAmC,EAAEA,eAAe;YAGvDzC,MACE,CAAC,0BAA0B,EAAEyC,cAAc,uBAAuB,EAAEH,gBAAgB,gBAAgB,EAAEH,WAAW;YAEnH,OAAOM;QACT;QAGF,IAAI;YACF,IAAI,CAAC,eAAe,GAAG,MAAM,IAAI,CAAC,sBAAsB;YACxD,OAAO,IAAI,CAAC,eAAe;QAC7B,SAAU;YACR,IAAI,CAAC,sBAAsB,GAAGP;QAChC;IACF;IAEQ,4BACNQ,sBAAoC,EAC5B;QACR,IAAI,AAAmCR,WAAnC,IAAI,CAAC,IAAI,CAAC,oBAAoB,EAChC,OAAO,IAAI,CAAC,IAAI,CAAC,oBAAoB;QAGvC,OAAOS,SAASD,uBAAuB,WAAW,IAC9Cb,uCACAe,UAAUF,uBAAuB,WAAW,IAC1CZ,qCACAF;IACR;IAwGA,MAAM,iBAA0C;QAC9C,OAAO,IAAI,CAAC,eAAe;IAC7B;IAEA,MAAM,aAAaiB,MAAsB,EAAsB;QAE7D,IAAI,CAAC,oBAAoB;QAGzB,IAAI,IAAI,CAAC,eAAe,EAAE;YACxB7C,MAAM,yCAAyC6C;YAC/C,OAAO,IAAI,CAAC,eAAe;QAC7B;QAGA,IAAIZ;QACJ,IAAI,IAAI,CAAC,SAAS,CAAC,UAAU,EAAE;YAC7BjC,MAAM,qCAAqC6C;YAC3CZ,UAAU,MAAM,IAAI,CAAC,SAAS,CAAC,UAAU;QAC3C,OAAO;YACLjC,MAAM;YACNiC,UAAU,MAAMa,oBAAoB,IAAI,CAAC,SAAS,EAAE;gBAClD,iBAAiB,IAAI,CAAC,kBAAkB,CAAC,sBAAsB;YACjE;QACF;QAEA9C,MAAM;QACN,MAAM+C,0BAA0B,MAAM,IAAI,CAAC,kBAAkB,CAACd;QAC9DjC,MAAM,2BAA2B+C;QAEjC,IAAIA,AAA4B,MAA5BA,yBAA+B;YACjC,MAAMC,cAAcR,OAAO,UAAU,CAACO,wBAAwB,OAAO,CAAC;YACtE/C,MACE,CAAC,oCAAoC,EAAEgD,YAAY,yBAAyB,CAAC;YAE/E,MAAMC,cAAcxC,KAAK,KAAK,CAACwB,QAAQ,IAAI,CAAC,KAAK;YACjD,MAAMiB,eAAezC,KAAK,KAAK,CAACwB,QAAQ,IAAI,CAAC,MAAM;YACnDjC,MAAM,CAAC,uBAAuB,EAAEiD,YAAY,CAAC,EAAEC,cAAc;YAC7D,MAAMC,0BAA0BlB,QAAQ,UAAU,CAAC,MAAM;YACzD,MAAMmB,gBAAgB,MAAMC,gBAAgBF,yBAAyB;gBACnE,OAAOF;gBACP,QAAQC;YACV;YACAjB,QAAQ,UAAU,GAAGqB,eAAe,MAAM,CAACF;QAC7C,OACEpD,MAAM,CAAC,iBAAiB,EAAE+C,yBAAyB;QAGrD,OAAOd;IACT;IAEA,MAAM,mBAAuC;QAC3C,OAAO,MAAM,IAAI,CAAC,YAAY,CAAC;IACjC;IAKA,MAAM,mBAAmBsB,MAAc,EAAE;QACvC,MAAM,IAAI,CAAC,eAAe,CAACA;IAC7B;IAEA,MAAM,gBAAgBA,MAAc,EAAE;QACpC,IAAI,IAAI,CAAC,YAAY,EACnBC,QAAQ,IAAI,CACV;QAGJ,IAAI,CAAC,IAAI,CAAC,YAAY,GAAGD;QACzB,IAAI,CAAC,IAAI,CAAC,eAAe,GAAGA;IAC9B;IAEA,YAAY;QACV,IAAI,CAAC,IAAI,GAAG,IAAIE,kBAAkB;YAChC,YAAYC;YACZ,WAAW,IAAI,CAAC,IAAI,CAAC,SAAS;YAC9B,kBAAkB,IAAI,CAAC,IAAI,CAAC,gBAAgB;YAC5C,YAAY,EAAE;YACd,aAAa,EAAE;QACjB;QACA,IAAI,CAAC,0BAA0B,GAAG,IAAIC;QAEtC,OAAO,IAAI,CAAC,IAAI;IAClB;IAEA,oBAAoBC,SAAwB,EAAEC,MAAmB,EAAE;QACjE,MAAMC,cAAc,IAAI,CAAC,IAAI;QAC7B,IAAID,QAAQ;YACV,MAAME,gBAAgB,IAAI,CAAC,0BAA0B,CAAC,GAAG,CAACF;YAC1D,IAAIE,AAAkB7B,WAAlB6B,eAA6B;gBAC/BD,YAAY,UAAU,CAACC,cAAc,GAAGH;gBACxC;YACF;YACAE,YAAY,UAAU,CAAC,IAAI,CAACF;YAC5B,IAAI,CAAC,0BAA0B,CAAC,GAAG,CACjCC,QACAC,YAAY,UAAU,CAAC,MAAM,GAAG;YAElC;QACF;QACAA,YAAY,UAAU,CAAC,IAAI,CAACF;IAC9B;IAEA,iBAAiB;QAEf,IAAI,CAAC,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,IAAI,CAAC,SAAS;QACzC,IAAI,CAAC,IAAI,CAAC,gBAAgB,GAAG,IAAI,CAAC,IAAI,CAAC,gBAAgB;QACvD,OAAO,IAAI,CAAC,IAAI,CAAC,SAAS;IAC5B;IAEA,mBAAmB;QACjB,OAAOI,kBAAkB,IAAI,CAAC,cAAc;IAC9C;IAEA,sBAAsB;QACpB,IAAI,IAAI,CAAC,SAAS,EAChB,MAAM,IAAIC,MACR;QAGJ,MAAM,EAAEC,cAAc,EAAEC,kBAAkB,EAAE,GAAG,IAAI,CAAC,IAAI;QACxD,IAAI,CAAC,UAAU,GAAGC,aAAa;YAC7B,UAAU,IAAI,CAAC,cAAc;YAC7B,SAASC;YACT,aAAa,IAAI,CAAC,cAAc;YAChC,MAAM;YACNH;QACF;QACAlE,MAAM,uBAAuB,IAAI,CAAC,UAAU;QAC5C,IAAIkE,kBAAkBC,sBAAsB,IAAI,CAAC,UAAU,EACzDG,eAAe,IAAI,CAAC,UAAU;IAElC;IAEA,MAAc,uBAAuBC,IAAmB,EAAE;QACxD,MAAMC,QAAQC,SAASF;QACvB,MAAMG,MAAMF,QAAQ,GAAGG,QAAQJ,MAAM,GAAG,EAAEC,OAAO,GAAGG,QAAQJ;QAE5D,IAAI,IAAI,CAAC,cAAc,EACrB,MAAM,IAAI,CAAC,cAAc,CAACG;IAE9B;IAEA,wBACEE,IAAY,EACyC;QACrD,OAAO,OAAOJ,QACL,MAAM,IAAI,CAAC,uBAAuB,CAAkBI,MAAMJ;IAErE;IAEA,MAAM,wBACJK,IAAY,EACZC,GAAO,EACP;QACA9E,MAAM,2BAA2B6E,MAAM,KAAKC;QAE5C,MAAMC,aAAgC;YACpC,MAAMF;YACN,OAAQC,OAAe,CAAC;YACxB,SAAS;QACX;QACA9E,MAAM,cAAc+E;QAEpB,MAAMC,QAA0B;YAACD;SAAW,CAAC,MAAM,CACjDE;QAGF,MAAMC,QAAQC,aACZN,MACAO,eAAgBN,KAAa,UAAU,CAAC;QAI1C,MAAMO,2BACJ,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QACzC,MAAM3C,yBACJ,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAEzC,MAAM,EAAE4C,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,CACjDJ,OACAF,OACAtC,wBACA2C;QAEF,OAAOC;IACT;IAEA,MAAM,MACJC,YAAyB,EACzBT,GAA8D,EAC9D;QACA1C,OAAOmD,cAAc;QAErB,MAAMC,sBAAsBC,yBAAyBF,cAAcT;QAEnE,MAAMY,oBAAoBZ,KAAK,oBAC3B,IAAI,CAAC,kBAAkB,CAACA,IAAI,iBAAiB,IAC7C5C;QAEJ,OAAOyD,gBAAgB,IAAI,CAAC,SAAS,EAAED,mBAAmB,UACjD,IAAI,CAAC,uBAAuB,CAAC,OAAO;gBACzC,QAAQF;YACV;IAEJ;IAEA,MAAM,aAAaD,YAAyB,EAAET,GAAkB,EAAE;QAChE1C,OAAOmD,cAAc;QAErB,MAAMC,sBAAsBC,yBAAyBF,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,cAAc;YAChD,QAAQU;QACV;IACF;IAEA,MAAM,cAAcD,YAAyB,EAAET,GAAkB,EAAE;QACjE1C,OAAOmD,cAAc;QAErB,MAAMC,sBAAsBC,yBAAyBF,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,eAAe;YACjD,QAAQU;QACV;IACF;IAEA,MAAM,QAAQD,YAAyB,EAAET,GAAkB,EAAE;QAC3D1C,OAAOmD,cAAc;QAErB,MAAMC,sBAAsBC,yBAAyBF,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,SAAS;YAC3C,QAAQU;QACV;IACF;IAuBA,MAAM,QACJI,mBAAkD,EAClDC,iBAKa,EACbC,cAA6B,EAC7B;QACA,IAAIvE;QACJ,IAAIgE;QACJ,IAAIT;QAOJ,IACE,AAA6B,YAA7B,OAAOe,qBACPA,AAAsB,SAAtBA,qBACA,WAAWA,mBACX;YAEAN,eAAeK;YACf,MAAMG,eAAeF;YAKrBtE,QAAQwE,aAAa,KAAK;YAC1BjB,MAAMiB;QACR,OAAO;YAELxE,QAAQqE;YACRL,eAAeM;YACff,MAAM;gBACJ,GAAGgB,cAAc;gBACjBvE;YACF;QACF;QAEAa,OACE,AAAiB,YAAjB,OAAOb,SAAsB,AAAiB,YAAjB,OAAOA,OACpC;QAEFa,OAAOmD,cAAc;QAErB,MAAMC,sBAAsBC,yBAAyBF,cAAcT;QAGnE,MAAMkB,cAAc,AAAiB,YAAjB,OAAOzE,QAAqB0E,OAAO1E,SAASA;QAEhE,OAAO,IAAI,CAAC,uBAAuB,CAAC,SAAS;YAC3C,GAAIuD,OAAO,CAAC,CAAC;YACb,OAAOkB;YACP,QAAQR;QACV;IACF;IAmBA,MAAM,gBACJU,qBAA2C,EAC3CL,iBAGa,EACbC,cAA6B,EAC7B;QACA,IAAIK;QACJ,IAAIZ;QACJ,IAAIT;QAGJ,IACE,AAA6B,YAA7B,OAAOe,qBACPA,AAAsB,SAAtBA,qBACA,aAAaA,mBACb;YAEAN,eAAeW;YACfpB,MAAMe;QAGR,OAAO;YAELM,UAAUD;YACVX,eAAeM;YACff,MAAM;gBACJ,GAAIgB,kBAAkB,CAAC,CAAC;gBACxBK;YACF;QACF;QAEA/D,OAAO0C,KAAK,SAAS;QAErB,MAAMU,sBAAsBD,eACxBE,yBAAyBF,cAAcT,OACvC5C;QAEJ,OAAO,IAAI,CAAC,uBAAuB,CAAC,iBAAiB;YACnD,GAAI4C,OAAO,CAAC,CAAC;YACb,QAAQU;QACV;IACF;IAmBA,MAAM,SACJY,yBAAgE,EAChEP,iBAAyE,EACzEC,cAA6B,EAC7B;QACA,IAAIO;QACJ,IAAId;QACJ,IAAIT;QAGJ,IACE,AAA6B,YAA7B,OAAOe,qBACN,gBAAeA,qBACd,gBAAgBA,qBAChB,cAAcA,iBAAgB,GAChC;YAEAN,eAAea;YACftB,MAAMe;QACR,OAAO;YAELQ,cAAcD;YACdb,eAAeM;YACff,MAAM;gBACJ,GAAIgB,kBAAkB,CAAC,CAAC;gBACxB,GAAIO,eAAe,CAAC,CAAC;YACvB;QACF;QAEA,IAAIvB,KAAK;YACP,MAAMwB,uBAAuB5E,oBAC1BoD,IAAoB,UAAU;YAMjC,IAAIwB,yBAA0BxB,IAAoB,UAAU,EACzDA,MAAsB;gBACrB,GAAIA,OAAO,CAAC,CAAC;gBACb,YAAYwB;YACd;QAEJ;QAEA,MAAMd,sBAAsBC,yBAC1BF,gBAAgB,IAChBT;QAGF,OAAO,IAAI,CAAC,uBAAuB,CAAC,UAAU;YAC5C,GAAIA,OAAO,CAAC,CAAC;YACb,QAAQU;QACV;IACF;IAEA,MAAM,MACJe,UAAkB,EAClBzB,GAAkB,EACW;QAC7B,MAAMY,oBAAoBZ,KAAK,oBAC3B,IAAI,CAAC,kBAAkB,CAACA,IAAI,iBAAiB,IAC7C5C;QAEJ,MAAMsE,WAAW;YACf,MAAM9D,yBACJ,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;YACzC,MAAM2C,2BACJ,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;YAEzC,MAAMoB,wBACJ/D,uBAAuB,SAAS,KAC9B2C,yBAAyB,SAAS,IACpC3C,uBAAuB,aAAa,KAClC2C,yBAAyB,aAAa;YAC1CrF,MAAM,oCAAoCyG;YAE1C,MAAMC,YAAY5B,KAAK;YACvB,MAAM6B,YAAY7B,KAAK,cAAc,UAAU5C,SAAY4C,KAAK;YAChE,MAAM8B,uBAAuB,IAAI,CAAC,2BAA2B,CAC3DlE;YAGF,MAAMmE,cAAclE,SAASD,uBAAuB,WAAW;YAC/D,MAAMoE,YAAYlE,UAAUF,uBAAuB,WAAW;YAC9D,MAAMqE,eACJF,eAAeC,aAAaJ,AAAc,UAAdA,YACxBxE,SACA,IAAI,CAAC,SAAS,EAAE,eAAeqE;YACrC,IACEQ,gBACA,IAAI,CAAC,SAAS,EAAE,qBAChBA,aAAa,YAAY,EAAE,cAAc,QACzC;gBAEA,MAAM,IAAI,CAAC,YAAY,CAAC,sBAAsB,CAC5CR,YACAQ,aAAa,YAAY,CAAC,YAAY;gBAGxC/G,MAAM;gBACN,MAAMgH,OAAOD,aAAa,YAAY,CAAC,YAAY;gBACnD,MAAM,IAAI,CAAC,OAAO,CAACC;gBACnB;YACF;YAIA,MAAMC,eAAgB,IAAI,CAAC,IAAI,EAAU;YACzC,IAAIA,cACFjH,MAAM;YAER,MAAMkH,qBAAyCD,eAC3C/E,SACA;YACJ,MAAM,EAAE,QAAQiF,YAAY,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,MAAM,CAC7DZ,YACA7D,wBACA2C,0BACAoB,uBACA,IAAI,CAAC,YAAY,EACjBC,WACAE,sBACAM,oBACAP,WACAjB;YAIF,IAAI,IAAI,CAAC,SAAS,IAAIyB,cAAc,YAAYT,AAAc,UAAdA,WAAqB;gBACnE,MAAMU,cAAkC;oBACtC,OAAO;wBACL;4BACE,MAAMb;4BACN,MAAMY,aAAa,QAAQ;wBAC7B;qBACD;gBACH;gBACA,MAAME,cAAcL,QAAAA,IAAS,CAACI;gBAC9B,IAAI,CAAC,SAAS,CAAC,yBAAyB,CACtC;oBACE,MAAM;oBACN,QAAQb;oBACR,cAAcc;gBAChB,GACAN;YAEJ;YAEA,OAAOI,cAAc;QACvB;QAEA,OAAO,MAAMX;IACf;IAKA,MAAM,SAASD,UAAkB,EAAEzB,GAAkB,EAAE;QACrD,OAAO,IAAI,CAAC,KAAK,CAACyB,YAAYzB;IAChC;IAEA,MAAM,QACJwC,MAA2B,EAC3BxC,MAA4B3D,2BAA2B,EAClC;QACrB,MAAMoG,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAC3D,MAAM,EAAEjC,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CACjE,SACAgC,QACAC,aACAzC;QAEF,OAAOQ;IACT;IAEA,MAAM,UACJ/B,MAAmB,EACnBuB,MAA4B3D,2BAA2B,EACrC;QAClB,MAAMoG,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAEC,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAYnE;QACrD,MAAM,EAAE+B,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CACjE,WACAkC,YACAD,aACAzC,KACA2C;QAEF,OAAOnC;IACT;IAEA,MAAM,SACJ/B,MAAmB,EACnBuB,MAA4B3D,2BAA2B,EACtC;QACjB,MAAMoG,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAEC,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAYnE;QACrD,MAAM,EAAE+B,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CACjE,UACAkC,YACAD,aACAzC,KACA2C;QAEF,OAAOnC;IACT;IAEA,MAAM,SACJ/B,MAAmB,EACnBuB,MAA4B3D,2BAA2B,EACtC;QACjB,MAAMoG,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAEC,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAYnE;QACrD,MAAM,EAAE+B,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CACjE,UACAkC,YACAD,aACAzC,KACA2C;QAEF,OAAOnC;IACT;IAEA,MAAM,MACJ/B,MAAmB,EACnBuB,MAA4B3D,2BAA2B,EACtC;QACjB,OAAO,IAAI,CAAC,QAAQ,CAACoC,QAAQuB;IAC/B;IAEA,MAAM,uBACJ6C,MAAwB,EACxB7C,GAI0B,EACkB;QAC5C,MAAM,EAAE8C,eAAe,IAAI,EAAEC,aAAa,CAAC,EAAE,GAAG/C,OAAO,CAAC;QAExD,IAAIgD,UAAU;QACd,IAAIC,aAAa;QACjB,IAAIC,eAAe;QACnB,IAAIrB,YAAY7B,KAAK,aAAa;QAClC,IAAImD;QAEJ,MAAO,CAACH,WAAWC,aAAaF,WAAY;YAC1C,IAAIE,cAAc,GAChBpB,YAAY;YAEd3G,MACE,cACA2H,QACA,gBACAC,cACA,cACAG,YACA,aACApB;YAGF,MAAMY,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;YAE3D,MAAMW,OAAO,MAAM,IAAI,CAAC,OAAO,CAAC,QAAQ,CAACP,QAAQJ,aAAa;gBAC5DZ;YACF;YACA3G,MAAM,mBAAmBkI;YACzB9F,OAAO8F,KAAK,WAAW,EAAE,CAAC,+BAA+B,EAAEP,OAAO,CAAC,CAAC;YACpEK,eAAeE,KAAK,WAAW;YAE/BD,eAAe,MAAM,IAAI,CAAC,aAAa,CACrCD,cACArB,YAAY;gBAAE,WAAW;YAAK,IAAIzE,QAClCyF,QACA7C;YAEF,IAAImD,aAAa,IAAI,EACnBH,UAAU;iBAEVC;QAEJ;QAEA,OAAO;YACL,QAAQC;YACRrB;YACAsB;QACF;IACF;IAEA,MAAM,cACJ1E,MAAc,EACd4E,SAAmC,EACnCC,YAA8B,EAC9BC,kBAA2C,EACX;QAChCrI,MAAM,iBAAiBuD,QAAQ4E,WAAWC,cAAcC;QAExD,MAAM,EAAE,QAAQC,YAAY,EAAE,MAAMC,UAAU,EAAE,GAAG,MAAM,IAAI,CAAC,QAAQ,CACpEhF,QACA4E;QAEF,MAAMK,WAAWtI,oBAAoBkI,cAAcE;QACnD,MAAMG,WAAW/H,eAAe0H,cAAcG;QAC9C,MAAMG,OACJF,YAAaH,CAAAA,oBAAoB,2BAA2B,EAAC,KAC7DI;QACF,MAAMR,eAAe;YACnBS;YACA,MAAMH;YACN,QAAQD;YACR,gBAAgBE;QAClB;QACAxI,MAAM,2BAA2BiI;QACjC,OAAOA;IACT;IAEA,MAAM,SAAS1E,MAAmB,EAAEuB,GAAkB,EAAE;QACtD,MAAM6D,cAAclD,yBAAyBlC,QAAQuB;QACrD1C,OAAOuG,aAAa;QACpB,MAAMC,aAAaC,oBAAoBF;QACvC,MAAM3D,QAAQ;YAAC4D;SAAW;QAC1B,MAAMvD,2BACJ,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QACzC,MAAM3C,yBACJ,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAEzC,MAAM,EAAE4C,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,CACjDH,aAAa,UAAUC,eAAeuD,eACtC3D,OACAtC,wBACA2C;QAGF,MAAM,EAAEyD,OAAO,EAAE,GAAGxD;QAEpB,MAAMyD,WAAW,MAAO,IAAI,CAAC,SAAS,CAAC,IAAI,GAAW,GAAG;QACzD,MAAMC,WAAWD,WACb;YACE,KAAKA;QACP,IACA,CAAC;QACL,OAAO;YACL,MAAMD,SAAS;YACf,QAAQA,SAAS;YACjB,GAAGE,QAAQ;QACb;IAGF;IAEA,MAAM,SACJC,SAAsB,EACtBC,GAAY,EACZpE,GAA2C,EAC3C;QACA,MAAMyC,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM4B,aAAmC;YACvC,aAAarE,KAAK,eAAe3D,4BAA4B,WAAW;YACxE,oBACE2D,KAAK,sBACL3D,4BAA4B,kBAAkB;QAClD;QAEA,MAAM,EAAEqG,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAYuB;QACrD,MAAMG,gBACJ,AAAqB,YAArB,OAAOH,YAAyBA,YAAYA,UAAU,MAAM;QAE9D,IAAI;YACF,MAAM,EAAE3D,MAAM,EAAE+D,OAAO,EAAE,GACvB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,UACA7B,YACAD,aACA4B,YACA1B;YAGJ,MAAMiB,OAAOzD,QAAQK;YACrB,MAAMgE,UAAUZ,OACZxG,SACA,CAAC,kBAAkB,EAAEgH,OAAOE,cAAc,UAAU,EAAEC,WAAW,eAAe;YAEpF,IAAIvE,KAAK,iBACP,OAAO;gBACL4D;gBACAW;gBACAC;YACF;YAGF,IAAI,CAACZ,MACH,MAAM,IAAIzE,MAAMqF;QAEpB,EAAE,OAAOC,OAAO;YACd,IAAIA,iBAAiBC,oBAAoB;gBACvC,MAAMC,YAAYF,MAAM,SAAS;gBACjC,MAAMF,UAAUI,WAAW;gBAC3B,MAAMC,WAAWD,WAAW;gBAC5B,MAAME,aACJF,WAAW,gBACVC,CAAAA,oBAAoBzF,QACjByF,SAAS,OAAO,GAChBA,WACEzD,OAAOyD,YACPxH,MAAQ;gBAChB,MAAM0H,SAASP,WAAWM,cAAc;gBACxC,MAAML,UAAU,CAAC,kBAAkB,EAAEJ,OAAOE,cAAc,UAAU,EAAEQ,QAAQ;gBAE9E,IAAI9E,KAAK,iBACP,OAAO;oBACL,MAAM;oBACNuE;oBACAC;gBACF;gBAGF,MAAM,IAAIrF,MAAMqF,SAAS;oBACvB,OAAOI,YAAYH;gBACrB;YACF;YAEA,MAAMA;QACR;IACF;IAEA,MAAM,UAAUN,SAAsB,EAAEnE,GAAqB,EAAE;QAC7D,MAAMyC,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAC3D,MAAM,IAAI,CAAC,YAAY,CAAC,OAAO,CAC7B0B,WACA;YACE,GAAGnE,GAAG;YACN,WAAWA,KAAK,aAAa;YAC7B,iBAAiBA,KAAK,mBAAmB;QAC3C,GACAyC;IAEJ;IAEA,MAAM,GAAG,GAAGsC,IAAmC,EAAE;QAC/C,OAAO,IAAI,CAAC,KAAK,IAAIA;IACvB;IAEA,MAAM,QAAQC,iBAAyB,EAEpC;QACD,MAAMC,SAASC,gBAAgBF,mBAAmB;QAClD,MAAMG,SAAS,IAAIC,aAAaH,QAAQ,UAC/B;gBAAE,OAAO,IAAI;gBAAE,QAAQ,EAAE;YAAC;QAEnC,MAAME,OAAO,GAAG;QAEhB,IAAIA,AAAkB,YAAlBA,OAAO,MAAM,EAAc;YAC7B,MAAME,SAASF,OAAO,cAAc,CACjC,MAAM,CAAC,CAAC1F,OAASA,AAAgB,YAAhBA,KAAK,MAAM,EAC5B,GAAG,CAAC,CAACA,OACG,CAAC,OAAO,EAAEA,KAAK,IAAI,CAAC,EAAE,EAAEA,KAAK,KAAK,EAAE,SAAS,EAErD,IAAI,CAAC;YACR,MAAM,IAAIN,MAAM,CAAC,2CAA2C,EAAEkG,QAAQ;QACxE;QAEA,OAAO;YACL,QAAQF,OAAO,MAAM;QACvB;IACF;IAEA,MAAM,mBAAmBF,MAAc,EAAE;QACvC3H,OACE,IAAI,CAAC,SAAS,CAAC,kBAAkB,EACjC;QAEF,OAAO,IAAI,CAAC,SAAS,CAAC,kBAAkB,CAAC2H;IAC3C;IAOA,sBACEK,QAA+D,EACnD;QACZ,IAAI,CAAC,mBAAmB,CAAC,IAAI,CAACA;QAG9B,OAAO;YACL,IAAI,CAAC,wBAAwB,CAACA;QAChC;IACF;IAMA,yBACEA,QAA+D,EACzD;QACN,MAAMC,QAAQ,IAAI,CAAC,mBAAmB,CAAC,OAAO,CAACD;QAC/C,IAAIC,QAAQ,IACV,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAACA,OAAO;IAE3C;IAKA,2BAAiC;QAC/B,IAAI,CAAC,mBAAmB,GAAG,EAAE;IAC/B;IAEA,MAAM,UAAU;QAEd,IAAI,IAAI,CAAC,SAAS,EAChB;QAGF,MAAM,IAAI,CAAC,SAAS,CAAC,OAAO;QAC5B,IAAI,CAAC,SAAS;QACd,IAAI,CAAC,SAAS,GAAG;IACnB;IAEA,MAAM,eACJnF,KAAc,EACdJ,GAEC,EACD;QAEA,MAAMwF,SAAS,MAAM,IAAI,CAAC,SAAS,CAAC,gBAAgB;QACpD,MAAMC,aAAajH,eAAe,MAAM,CAACgH;QACzC,MAAME,MAAMC,KAAK,GAAG;QAEpB,MAAMC,WAAoC;YACxC;gBACE,MAAM;gBACN,IAAIF;gBACJD;YACF;SACD;QAED,MAAMhG,OAAyB;YAC7B,MAAM;YACN,SAAS;YACT,QAAQ;YACRmG;YACA,QAAQ;gBACN,OAAOF;gBACP,KAAKA;gBACL,MAAM;YACR;YACA,OAAO;gBACL,SAAS1F,KAAK,WAAW;YAC3B;YACA,UAAU,WAAa;QACzB;QAEA,MAAM6F,gBAAgB,IAAIC,cAAc;YACtC,SAASJ;YACT,MAAM,CAAC,MAAM,EAAEtF,SAAS,YAAY;YACpC,aAAaJ,KAAK,WAAW;YAC7B,OAAO;gBAACP;aAAK;QACf;QAEA,IAAI,CAAC,mBAAmB,CAACoG;QAGzB,MAAME,aAAa,IAAI,CAAC,cAAc;QACtC,KAAK,MAAMT,YAAY,IAAI,CAAC,mBAAmB,CAC7C,IAAI;YACFA,SAASS;QACX,EAAE,OAAOtB,OAAO;YACd/F,QAAQ,KAAK,CAAC,kCAAkC+F;QAClD;QAGF,IAAI,CAAC,mBAAmB;IAC1B;IAKA,MAAM,cACJrE,KAAc,EACdJ,GAEC,EACD;QACA,MAAM,IAAI,CAAC,cAAc,CAACI,OAAOJ;IACnC;IAEA,sBAAsB;QACpB,MAAM,EAAEgG,SAAS,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAG,IAAI,CAAC,IAAI;QAC7D,OAAO;YACLF;YACAC;YACA,YAAYC,cAAc,EAAE;QAC9B;IACF;IAMA,MAAM,oBAAmC;QACvChL,MAAM;QACN,MAAMiC,UAAU,MAAM,IAAI,CAAC,gBAAgB;QAE3CA,QAAQ,SAAS,GAAG;QACpB,IAAI,CAAC,eAAe,GAAGA;QACvBjC,MAAM;IACR;IAKA,MAAM,sBAAqC;QACzCA,MAAM;QACN,IAAI,CAAC,eAAe,GAAGkC;QACvBlC,MAAM;IACR;IAKQ,mBAAmBiL,IAAc,EAKhC;QAGP,IAAIA,AAAe,SAAfA,KAAK,KAAK,EACZ,MAAM,IAAIhH,MACR;QAMJ,IACEgH,KAAK,KAAK,IACV,AAAsB,YAAtB,OAAOA,KAAK,KAAK,IACjBA,AAAe,SAAfA,KAAK,KAAK,IACV,CAACA,KAAK,KAAK,CAAC,EAAE,EAEd,MAAM,IAAIhH,MACR;QAMJ,MAAMiH,cAAcC,mBAClBF,KAAK,KAAK,EACVA,KAAK,OAAO,IAAIA,KAAK,MAAM,IAAI;QAGjC,IAAI,CAACC,aACH,OAAO;QAIT,IAAI,AAAuB,YAAvB,OAAOA,eAA4BA,AAAgB,SAAhBA,aAAsB;YAC3D,MAAME,KAAKF,YAAY,EAAE;YACzB,MAAMG,cAAcH,YAAY,QAAQ;YACxC,IAAII;YAEJ,IAAID,AAAgBnJ,WAAhBmJ,aACFC,gBAAgB;iBACX,IAAI,AAAuB,YAAvB,OAAOD,aAChBC,gBAAgBD;iBAEhB,MAAM,IAAIpH,MACR,CAAC,iEAAiE,EAAE,OAAOoH,aAAa;YAI5F,IAAI,CAAChK,qBAAqBiK,gBACxB,MAAM,IAAIrH,MACR,CAAC,8BAA8B,EAAEzC,sBAAsB,gBAAgB,EAAE8J,cAAc,CAAC,CAAC;YAI7F,MAAMC,aAAaD,AAAkB,gBAAlBA;YACnB,MAAME,cAAcF,AAAkB,iBAAlBA;YAEpB,OAAO;gBACLF;gBACA,SAAS,CAACI;gBACV,UAAUD;gBACV,WAAWC;YACb;QACF;QAEA,OAAO;IACT;IAEQ,mBAAmBC,KAAe,EAAY;QACpD,OAAOA,MAAM,GAAG,CAAC,CAACC;YAChB,MAAMC,eAAeC,QAAQF;YAC7B,IAAI,CAACG,WAAWF,eACd,MAAM,IAAI1H,MAAM,CAAC,gBAAgB,EAAEyH,MAAM;YAE3C,OAAOC;QACT;IACF;IAEQ,mBAAmBF,KAAwB,EAAY;QAC7D,MAAMK,aAAaC,MAAM,OAAO,CAACN,SAASA,QAAQ;YAACA;SAAM;QACzD,OAAO,IAAI,CAAC,kBAAkB,CAACK;IACjC;IAOA,MAAM,WAAWE,OAAmC,EAAiB;QACnE,IAAI,CAAC,IAAI,CAAC,SAAS,EACjB,MAAM,IAAI/H,MAAM;QAGlB,IAAI,CAAC,SAAS,CAAC,gBAAgB,CAAC+H;IAClC;IA9sCA,YAAYC,iBAAgC,EAAEhB,IAAe,CAAE;QAhK/D;QAEA;QAEA;QAEA;QAEA;QAEA;QAEA;QAKA,kCAAU;QAEV;QAEA;QAEA,uBAAQ,uBAEJ,EAAE;QAmBN,oCAAY;QAEZ;QAKA,uBAAQ,mBAAR;QASA,uBAAQ,uBAAsB;QAK9B,uBAAQ,mBAAR;QAKA,uBAAQ,0BAAR;QAEA,uBAAQ,8BAA6B,IAAItH;QAEzC,uBAAQ,mBAAR;QAuFE,IAAI,CAAC,SAAS,GAAGsI;QAEjB,MAAMC,0BACJC,oBAAoB,yBAAyB,CAC3CC;QAGJ,IAAI,CAAC,IAAI,GAAGC,OAAO,MAAM,CACvB;YACE,gBAAgB;YAChB,oBAAoB;YACpB,WAAW;YACX,kBAAkB;QACpB,GACApB,QAAQ,CAAC,GACTA,MAAM,yBAAyB/I,UAC7BgK,AAA4BhK,WAA5BgK,2BACC1J,OAAO,KAAK,CAAC0J,2BAEZ,CAAC,IADD;YAAE,sBAAsBA;QAAwB;QAItD,MAAMI,uBACJ,IAAI,CAAC,IAAI,CAAC,YAAY,IAAI,IAAI,CAAC,IAAI,CAAC,eAAe;QACrD,IAAIA,AAAyBpK,WAAzBoK,sBAAoC;YACtC,IAAI,CAAC,IAAI,CAAC,YAAY,GAAGA;YACzB,IAAI,CAAC,IAAI,CAAC,eAAe,KAAKA;QAChC;QAEA,IACErB,MAAM,eACL,CAA6B,YAA7B,OAAOA,MAAM,eAA4Bc,MAAM,OAAO,CAACd,KAAK,WAAW,IAExE,MAAM,IAAIhH,MACR,CAAC,2EAA2E,EAAE,OAAOgH,MAAM,aAAa;QAK5G,MAAMsB,kBAAkBtB,MAAM,eAAeA,MAAM;QACnD,IAAI,CAAC,kBAAkB,GAAGsB,kBACtB,IAAIC,mBAAmBvB,MAAM,aAAaA,MAAM,sBAChDwB;QAEJ,IAAI,CAAC,cAAc,GAAG,IAAI,CAAC,IAAI,CAAC,cAAc;QAE9C,IAAI,CAAC,OAAO,GAAG,IAAIC,QAAQ,UAClB,IAAI,CAAC,YAAY;QAI1B,MAAMC,iBAAiB,IAAI,CAAC,kBAAkB,CAAC1B,QAAQ,CAAC;QACxD,IAAI0B,gBACF,IAAI,CAAC,SAAS,GAAG,IAAIC,UACnBD,eAAe,EAAE,EACjBA,eAAe,OAAO,EACtBzK,QACA;YACE,UAAUyK,eAAe,QAAQ;YACjC,WAAWA,eAAe,SAAS;QACrC;QAIJ,MAAME,kBAAkB,IAAI,CAAC,SAAS,CAAC,WAAW;QAClD,IAAI,CAAC,eAAe,GAAG;eAClBA;YACHC;YACAC;SACD;QAED,IAAI,CAAC,YAAY,GAAG,IAAIC,aAAa,IAAI,CAAC,SAAS,EAAE,IAAI,CAAC,OAAO,EAAE;YACjE,WAAW,IAAI,CAAC,SAAS;YACzB,aAAa,IAAI,CAAC,sBAAsB,CAAC,IAAI,CAAC,IAAI;YAClD,sBAAsB,IAAI,CAAC,IAAI,CAAC,oBAAoB;YACpD,aAAa,IAAI,CAAC,eAAe;YACjC,OAAO;gBACL,cAAc,CAACnJ;oBACb,MAAM8G,gBAAgB9G,OAAO,IAAI;oBACjC,IAAI,CAAC,mBAAmB,CAAC8G,eAAe9G;oBAGxC,MAAMgH,aAAa,IAAI,CAAC,cAAc;oBACtC,KAAK,MAAMT,YAAY,IAAI,CAAC,mBAAmB,CAC7C,IAAI;wBACFA,SAASS,YAAYF;oBACvB,EAAE,OAAOpB,OAAO;wBACd/F,QAAQ,KAAK,CAAC,kCAAkC+F;oBAClD;oBAGF,IAAI,CAAC,mBAAmB;gBAC1B;YACF;QACF;QACA,IAAI,CAAC,IAAI,GAAG,IAAI,CAAC,SAAS;QAC1B,IAAI,CAAC,cAAc,GACjB0B,MAAM,kBACNgC,kBAAkBhC,MAAM,UAAU,IAAI,CAAC,SAAS,CAAC,aAAa,IAAI;IACtE;AA2mCF;AAEO,MAAMiC,cAAc,CACzBjB,mBACAhB,OAEO,IAAIlJ,MAAMkK,mBAAmBhB"}
+{"version":3,"file":"agent/agent.mjs","sources":["../../../src/agent/agent.ts"],"sourcesContent":["import {\n  type ActionParam,\n  type ActionReturn,\n  type AgentAssertOpt,\n  type AgentDescribeElementAtPointResult,\n  type AgentOpt,\n  type AgentWaitForOpt,\n  type CacheConfig,\n  type DeepThinkOption,\n  type DetailedLocateParam,\n  type DeviceAction,\n  ExecutionDump,\n  type ExecutionRecorderItem,\n  type ExecutionTask,\n  type ExecutionTaskLog,\n  type ExecutionTaskPlanning,\n  GroupedActionDump,\n  type LocateOption,\n  type LocateResultElement,\n  type LocateValidatorResult,\n  type LocatorValidatorOption,\n  type MidsceneYamlScript,\n  type OnTaskStartTip,\n  type PlanningAction,\n  type Rect,\n  ScreenshotItem,\n  type ScrollParam,\n  Service,\n  type ServiceAction,\n  type ServiceExtractOption,\n  type ServiceExtractParam,\n  type TUserPrompt,\n  type UIContext,\n} from '../index';\nexport type TestStatus =\n  | 'passed'\n  | 'failed'\n  | 'timedOut'\n  | 'skipped'\n  | 'interrupted';\nimport { isAutoGLM, isUITars } from '@/ai-model/auto-glm/util';\nimport yaml from 'js-yaml';\n\nimport {\n  getVersion,\n  groupedActionDumpFileExt,\n  processCacheConfig,\n  reportHTMLContent,\n  writeLogFile,\n} from '@/utils';\nimport {\n  ScriptPlayer,\n  buildDetailedLocateParam,\n  parseYamlScript,\n} from '../yaml/index';\n\nimport { existsSync } from 'node:fs';\nimport { resolve } from 'node:path';\nimport type { AbstractInterface } from '@/device';\nimport type { TaskRunner } from '@/task-runner';\nimport {\n  type IModelConfig,\n  MIDSCENE_REPLANNING_CYCLE_LIMIT,\n  ModelConfigManager,\n  globalConfigManager,\n  globalModelConfigManager,\n} from '@midscene/shared/env';\nimport { imageInfoOfBase64, resizeImgBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { defineActionAssert, defineActionFinalize } from '../device';\nimport { TaskCache } from './task-cache';\nimport {\n  TaskExecutionError,\n  TaskExecutor,\n  locatePlanForLocate,\n  withFileChooser,\n} from './tasks';\nimport { locateParamStr, paramStr, taskTitleStr, typeStr } from './ui-utils';\nimport {\n  commonContextParser,\n  getReportFileName,\n  parsePrompt,\n  printReportMsg,\n} from './utils';\n\nconst debug = getDebug('agent');\n\nconst distanceOfTwoPoints = (p1: [number, number], p2: [number, number]) => {\n  const [x1, y1] = p1;\n  const [x2, y2] = p2;\n  return Math.round(Math.sqrt((x1 - x2) ** 2 + (y1 - y2) ** 2));\n};\n\nconst includedInRect = (point: [number, number], rect: Rect) => {\n  const [x, y] = point;\n  const { left, top, width, height } = rect;\n  return x >= left && x <= left + width && y >= top && y <= top + height;\n};\n\nconst defaultServiceExtractOption: ServiceExtractOption = {\n  domIncluded: false,\n  screenshotIncluded: true,\n};\n\ntype CacheStrategy = NonNullable<CacheConfig['strategy']>;\n\nconst CACHE_STRATEGIES: readonly CacheStrategy[] = [\n  'read-only',\n  'read-write',\n  'write-only',\n];\n\nconst isValidCacheStrategy = (strategy: string): strategy is CacheStrategy =>\n  CACHE_STRATEGIES.some((value) => value === strategy);\n\nconst CACHE_STRATEGY_VALUES = CACHE_STRATEGIES.map(\n  (value) => `\"${value}\"`,\n).join(', ');\n\nconst legacyScrollTypeMap = {\n  once: 'singleAction',\n  untilBottom: 'scrollToBottom',\n  untilTop: 'scrollToTop',\n  untilRight: 'scrollToRight',\n  untilLeft: 'scrollToLeft',\n} as const;\n\ntype LegacyScrollType = keyof typeof legacyScrollTypeMap;\n\nconst normalizeScrollType = (\n  scrollType: ScrollParam['scrollType'] | LegacyScrollType | undefined,\n): ScrollParam['scrollType'] | undefined => {\n  if (!scrollType) {\n    return scrollType;\n  }\n\n  if (scrollType in legacyScrollTypeMap) {\n    return legacyScrollTypeMap[scrollType as LegacyScrollType];\n  }\n\n  return scrollType as ScrollParam['scrollType'];\n};\n\nconst defaultReplanningCycleLimit = 20;\nconst defaultVlmUiTarsReplanningCycleLimit = 40;\nconst defaultAutoGlmReplanningCycleLimit = 100;\n\nexport type AiActOptions = {\n  cacheable?: boolean;\n  fileChooserAccept?: string | string[];\n  deepThink?: DeepThinkOption;\n};\n\nexport class Agent<\n  InterfaceType extends AbstractInterface = AbstractInterface,\n> {\n  interface: InterfaceType;\n\n  service: Service;\n\n  dump: GroupedActionDump;\n\n  reportFile?: string | null;\n\n  reportFileName?: string;\n\n  taskExecutor: TaskExecutor;\n\n  opts: AgentOpt;\n\n  /**\n   * If true, the agent will not perform any actions\n   */\n  dryMode = false;\n\n  onTaskStartTip?: OnTaskStartTip;\n\n  taskCache?: TaskCache;\n\n  private dumpUpdateListeners: Array<\n    (dump: string, executionDump?: ExecutionDump) => void\n  > = [];\n\n  get onDumpUpdate():\n    | ((dump: string, executionDump?: ExecutionDump) => void)\n    | undefined {\n    return this.dumpUpdateListeners[0];\n  }\n\n  set onDumpUpdate(callback:\n    | ((dump: string, executionDump?: ExecutionDump) => void)\n    | undefined) {\n    // Clear existing listeners\n    this.dumpUpdateListeners = [];\n    // Add callback to array if provided\n    if (callback) {\n      this.dumpUpdateListeners.push(callback);\n    }\n  }\n\n  destroyed = false;\n\n  modelConfigManager: ModelConfigManager;\n\n  /**\n   * Frozen page context for consistent AI operations\n   */\n  private frozenUIContext?: UIContext;\n\n  private get aiActContext(): string | undefined {\n    return this.opts.aiActContext ?? this.opts.aiActionContext;\n  }\n\n  /**\n   * Flag to track if VL model warning has been shown\n   */\n  private hasWarnedNonVLModel = false;\n\n  /**\n   * Screenshot scale factor derived from actual screenshot dimensions\n   */\n  private screenshotScale?: number;\n\n  /**\n   * Internal promise to deduplicate screenshot scale computation\n   */\n  private screenshotScalePromise?: Promise<number>;\n\n  private executionDumpIndexByRunner = new WeakMap<TaskRunner, number>();\n\n  private fullActionSpace: DeviceAction[];\n\n  // @deprecated use .interface instead\n  get page() {\n    return this.interface;\n  }\n\n  /**\n   * Ensures VL model warning is shown once when needed\n   */\n  private ensureVLModelWarning() {\n    if (\n      !this.hasWarnedNonVLModel &&\n      this.interface.interfaceType !== 'puppeteer' &&\n      this.interface.interfaceType !== 'playwright' &&\n      this.interface.interfaceType !== 'static' &&\n      this.interface.interfaceType !== 'chrome-extension-proxy' &&\n      this.interface.interfaceType !== 'page-over-chrome-extension-bridge'\n    ) {\n      this.modelConfigManager.throwErrorIfNonVLModel();\n      this.hasWarnedNonVLModel = true;\n    }\n  }\n\n  /**\n   * Lazily compute the ratio between the physical screenshot width and the logical page width\n   */\n  private async getScreenshotScale(context: UIContext): Promise<number> {\n    if (this.screenshotScale !== undefined) {\n      return this.screenshotScale;\n    }\n\n    if (!this.screenshotScalePromise) {\n      this.screenshotScalePromise = (async () => {\n        const pageWidth = context.size?.width;\n        assert(\n          pageWidth && pageWidth > 0,\n          `Invalid page width when computing screenshot scale: ${pageWidth}`,\n        );\n\n        debug('will get image info of base64');\n        const screenshotBase64 = context.screenshot.base64;\n        const { width: screenshotWidth } =\n          await imageInfoOfBase64(screenshotBase64);\n        debug('image info of base64 done');\n\n        assert(\n          Number.isFinite(screenshotWidth) && screenshotWidth > 0,\n          `Invalid screenshot width when computing screenshot scale: ${screenshotWidth}`,\n        );\n\n        const computedScale = screenshotWidth / pageWidth;\n        assert(\n          Number.isFinite(computedScale) && computedScale > 0,\n          `Invalid computed screenshot scale: ${computedScale}`,\n        );\n\n        debug(\n          `Computed screenshot scale ${computedScale} from screenshot width ${screenshotWidth} and page width ${pageWidth}`,\n        );\n        return computedScale;\n      })();\n    }\n\n    try {\n      this.screenshotScale = await this.screenshotScalePromise;\n      return this.screenshotScale;\n    } finally {\n      this.screenshotScalePromise = undefined;\n    }\n  }\n\n  private resolveReplanningCycleLimit(\n    modelConfigForPlanning: IModelConfig,\n  ): number {\n    if (this.opts.replanningCycleLimit !== undefined) {\n      return this.opts.replanningCycleLimit;\n    }\n\n    return isUITars(modelConfigForPlanning.modelFamily)\n      ? defaultVlmUiTarsReplanningCycleLimit\n      : isAutoGLM(modelConfigForPlanning.modelFamily)\n        ? defaultAutoGlmReplanningCycleLimit\n        : defaultReplanningCycleLimit;\n  }\n\n  constructor(interfaceInstance: InterfaceType, opts?: AgentOpt) {\n    this.interface = interfaceInstance;\n\n    const envReplanningCycleLimit =\n      globalConfigManager.getEnvConfigValueAsNumber(\n        MIDSCENE_REPLANNING_CYCLE_LIMIT,\n      );\n\n    this.opts = Object.assign(\n      {\n        generateReport: true,\n        autoPrintReportMsg: true,\n        groupName: 'Midscene Report',\n        groupDescription: '',\n      },\n      opts || {},\n      opts?.replanningCycleLimit === undefined &&\n        envReplanningCycleLimit !== undefined &&\n        !Number.isNaN(envReplanningCycleLimit)\n        ? { replanningCycleLimit: envReplanningCycleLimit }\n        : {},\n    );\n\n    const resolvedAiActContext =\n      this.opts.aiActContext ?? this.opts.aiActionContext;\n    if (resolvedAiActContext !== undefined) {\n      this.opts.aiActContext = resolvedAiActContext;\n      this.opts.aiActionContext ??= resolvedAiActContext;\n    }\n\n    if (\n      opts?.modelConfig &&\n      (typeof opts?.modelConfig !== 'object' || Array.isArray(opts.modelConfig))\n    ) {\n      throw new Error(\n        `opts.modelConfig must be a plain object map of env keys to values, but got ${typeof opts?.modelConfig}`,\n      );\n    }\n    // Create ModelConfigManager if modelConfig or createOpenAIClient is provided\n    // Otherwise, use the global config manager\n    const hasCustomConfig = opts?.modelConfig || opts?.createOpenAIClient;\n    this.modelConfigManager = hasCustomConfig\n      ? new ModelConfigManager(opts?.modelConfig, opts?.createOpenAIClient)\n      : globalModelConfigManager;\n\n    this.onTaskStartTip = this.opts.onTaskStartTip;\n\n    this.service = new Service(async () => {\n      return this.getUIContext();\n    });\n\n    // Process cache configuration\n    const cacheConfigObj = this.processCacheConfig(opts || {});\n    if (cacheConfigObj) {\n      this.taskCache = new TaskCache(\n        cacheConfigObj.id,\n        cacheConfigObj.enabled,\n        undefined, // cacheFilePath\n        {\n          readOnly: cacheConfigObj.readOnly,\n          writeOnly: cacheConfigObj.writeOnly,\n        },\n      );\n    }\n\n    const baseActionSpace = this.interface.actionSpace();\n    this.fullActionSpace = [\n      ...baseActionSpace,\n      defineActionAssert(),\n      defineActionFinalize(),\n    ];\n\n    this.taskExecutor = new TaskExecutor(this.interface, this.service, {\n      taskCache: this.taskCache,\n      onTaskStart: this.callbackOnTaskStartTip.bind(this),\n      replanningCycleLimit: this.opts.replanningCycleLimit,\n      actionSpace: this.fullActionSpace,\n      hooks: {\n        onTaskUpdate: (runner) => {\n          const executionDump = runner.dump();\n          this.appendExecutionDump(executionDump, runner);\n\n          // Call all registered dump update listeners\n          const dumpString = this.dumpDataString();\n          for (const listener of this.dumpUpdateListeners) {\n            try {\n              listener(dumpString, executionDump);\n            } catch (error) {\n              console.error('Error in onDumpUpdate listener', error);\n            }\n          }\n\n          this.writeOutActionDumps();\n        },\n      },\n    });\n    this.dump = this.resetDump();\n    this.reportFileName =\n      opts?.reportFileName ||\n      getReportFileName(opts?.testId || this.interface.interfaceType || 'web');\n  }\n\n  async getActionSpace(): Promise<DeviceAction[]> {\n    return this.fullActionSpace;\n  }\n\n  async getUIContext(action?: ServiceAction): Promise<UIContext> {\n    // Check VL model configuration when UI context is first needed\n    this.ensureVLModelWarning();\n\n    // If page context is frozen, return the frozen context for all actions\n    if (this.frozenUIContext) {\n      debug('Using frozen page context for action:', action);\n      return this.frozenUIContext;\n    }\n\n    // Get original context\n    let context: UIContext;\n    if (this.interface.getContext) {\n      debug('Using page.getContext for action:', action);\n      context = await this.interface.getContext();\n    } else {\n      debug('Using commonContextParser');\n      context = await commonContextParser(this.interface, {\n        uploadServerUrl: this.modelConfigManager.getUploadTestServerUrl(),\n      });\n    }\n\n    debug('will get screenshot scale');\n    const computedScreenshotScale = await this.getScreenshotScale(context);\n    debug('computedScreenshotScale', computedScreenshotScale);\n\n    if (computedScreenshotScale !== 1) {\n      const scaleForLog = Number.parseFloat(computedScreenshotScale.toFixed(4));\n      debug(\n        `Applying computed screenshot scale: ${scaleForLog} (resize to logical size)`,\n      );\n      const targetWidth = Math.round(context.size.width);\n      const targetHeight = Math.round(context.size.height);\n      debug(`Resizing screenshot to ${targetWidth}x${targetHeight}`);\n      const currentScreenshotBase64 = context.screenshot.base64;\n      const resizedBase64 = await resizeImgBase64(currentScreenshotBase64, {\n        width: targetWidth,\n        height: targetHeight,\n      });\n      context.screenshot = ScreenshotItem.create(resizedBase64);\n    } else {\n      debug(`screenshot scale=${computedScreenshotScale}`);\n    }\n\n    return context;\n  }\n\n  async _snapshotContext(): Promise<UIContext> {\n    return await this.getUIContext('locate');\n  }\n\n  /**\n   * @deprecated Use {@link setAIActContext} instead.\n   */\n  async setAIActionContext(prompt: string) {\n    await this.setAIActContext(prompt);\n  }\n\n  async setAIActContext(prompt: string) {\n    if (this.aiActContext) {\n      console.warn(\n        'aiActContext is already set, and it is called again, will override the previous setting',\n      );\n    }\n    this.opts.aiActContext = prompt;\n    this.opts.aiActionContext = prompt;\n  }\n\n  resetDump() {\n    this.dump = new GroupedActionDump({\n      sdkVersion: getVersion(),\n      groupName: this.opts.groupName!,\n      groupDescription: this.opts.groupDescription,\n      executions: [],\n      modelBriefs: [],\n    });\n    this.executionDumpIndexByRunner = new WeakMap<TaskRunner, number>();\n\n    return this.dump;\n  }\n\n  appendExecutionDump(execution: ExecutionDump, runner?: TaskRunner) {\n    const currentDump = this.dump;\n    if (runner) {\n      const existingIndex = this.executionDumpIndexByRunner.get(runner);\n      if (existingIndex !== undefined) {\n        currentDump.executions[existingIndex] = execution;\n        return;\n      }\n      currentDump.executions.push(execution);\n      this.executionDumpIndexByRunner.set(\n        runner,\n        currentDump.executions.length - 1,\n      );\n      return;\n    }\n    currentDump.executions.push(execution);\n  }\n\n  dumpDataString() {\n    // update dump info\n    this.dump.groupName = this.opts.groupName!;\n    this.dump.groupDescription = this.opts.groupDescription;\n    return this.dump.serialize();\n  }\n\n  reportHTMLString() {\n    return reportHTMLContent(this.dumpDataString());\n  }\n\n  writeOutActionDumps() {\n    if (this.destroyed) {\n      throw new Error(\n        'PageAgent has been destroyed. Cannot update report file.',\n      );\n    }\n    const { generateReport, autoPrintReportMsg } = this.opts;\n    this.reportFile = writeLogFile({\n      fileName: this.reportFileName!,\n      fileExt: groupedActionDumpFileExt,\n      fileContent: this.dumpDataString(),\n      type: 'dump',\n      generateReport,\n    });\n    debug('writeOutActionDumps', this.reportFile);\n    if (generateReport && autoPrintReportMsg && this.reportFile) {\n      printReportMsg(this.reportFile);\n    }\n  }\n\n  private async callbackOnTaskStartTip(task: ExecutionTask) {\n    const param = paramStr(task);\n    const tip = param ? `${typeStr(task)} - ${param}` : typeStr(task);\n\n    if (this.onTaskStartTip) {\n      await this.onTaskStartTip(tip);\n    }\n  }\n\n  wrapActionInActionSpace<T extends DeviceAction>(\n    name: string,\n  ): (param: ActionParam<T>) => Promise<ActionReturn<T>> {\n    return async (param: ActionParam<T>) => {\n      return await this.callActionInActionSpace<ActionReturn<T>>(name, param);\n    };\n  }\n\n  async callActionInActionSpace<T = any>(\n    type: string,\n    opt?: T, // and all other action params\n  ) {\n    debug('callActionInActionSpace', type, ',', opt);\n\n    const actionPlan: PlanningAction<T> = {\n      type: type as any,\n      param: (opt as any) || {},\n      thought: '',\n    };\n    debug('actionPlan', actionPlan); // , ', in which the locateParam is', locateParam);\n\n    const plans: PlanningAction[] = [actionPlan].filter(\n      Boolean,\n    ) as PlanningAction[];\n\n    const title = taskTitleStr(\n      type as any,\n      locateParamStr((opt as any)?.locate || {}),\n    );\n\n    // assume all operation in action space is related to locating\n    const defaultIntentModelConfig =\n      this.modelConfigManager.getModelConfig('default');\n    const modelConfigForPlanning =\n      this.modelConfigManager.getModelConfig('planning');\n\n    const { output } = await this.taskExecutor.runPlans(\n      title,\n      plans,\n      modelConfigForPlanning,\n      defaultIntentModelConfig,\n    );\n    return output;\n  }\n\n  async aiTap(\n    locatePrompt: TUserPrompt,\n    opt?: LocateOption & { fileChooserAccept?: string | string[] },\n  ) {\n    assert(locatePrompt, 'missing locate prompt for tap');\n\n    const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n    const fileChooserAccept = opt?.fileChooserAccept\n      ? this.normalizeFileInput(opt.fileChooserAccept)\n      : undefined;\n\n    return withFileChooser(this.interface, fileChooserAccept, async () => {\n      return this.callActionInActionSpace('Tap', {\n        locate: detailedLocateParam,\n      });\n    });\n  }\n\n  async aiRightClick(locatePrompt: TUserPrompt, opt?: LocateOption) {\n    assert(locatePrompt, 'missing locate prompt for right click');\n\n    const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n    return this.callActionInActionSpace('RightClick', {\n      locate: detailedLocateParam,\n    });\n  }\n\n  async aiDoubleClick(locatePrompt: TUserPrompt, opt?: LocateOption) {\n    assert(locatePrompt, 'missing locate prompt for double click');\n\n    const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n    return this.callActionInActionSpace('DoubleClick', {\n      locate: detailedLocateParam,\n    });\n  }\n\n  async aiHover(locatePrompt: TUserPrompt, opt?: LocateOption) {\n    assert(locatePrompt, 'missing locate prompt for hover');\n\n    const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n    return this.callActionInActionSpace('Hover', {\n      locate: detailedLocateParam,\n    });\n  }\n\n  // New signature, always use locatePrompt as the first param\n  async aiInput(\n    locatePrompt: TUserPrompt,\n    opt: LocateOption & { value: string | number } & {\n      autoDismissKeyboard?: boolean;\n    } & { mode?: 'replace' | 'clear' | 'append' },\n  ): Promise<any>;\n\n  // Legacy signature - deprecated\n  /**\n   * @deprecated Use aiInput(locatePrompt, opt) instead where opt contains the value\n   */\n  async aiInput(\n    value: string | number,\n    locatePrompt: TUserPrompt,\n    opt?: LocateOption & { autoDismissKeyboard?: boolean } & {\n      mode?: 'replace' | 'clear' | 'append';\n    }, // AndroidDeviceInputOpt &\n  ): Promise<any>;\n\n  // Implementation\n  async aiInput(\n    locatePromptOrValue: TUserPrompt | string | number,\n    locatePromptOrOpt:\n      | TUserPrompt\n      | (LocateOption & { value: string | number } & {\n          autoDismissKeyboard?: boolean;\n        } & { mode?: 'replace' | 'clear' | 'append' }) // AndroidDeviceInputOpt &\n      | undefined,\n    optOrUndefined?: LocateOption, // AndroidDeviceInputOpt &\n  ) {\n    let value: string | number;\n    let locatePrompt: TUserPrompt;\n    let opt:\n      | (LocateOption & { value: string | number } & {\n          autoDismissKeyboard?: boolean;\n        } & { mode?: 'replace' | 'clear' | 'append' }) // AndroidDeviceInputOpt &\n      | undefined;\n\n    // Check if using new signature (first param is locatePrompt, second has value)\n    if (\n      typeof locatePromptOrOpt === 'object' &&\n      locatePromptOrOpt !== null &&\n      'value' in locatePromptOrOpt\n    ) {\n      // New signature: aiInput(locatePrompt, opt)\n      locatePrompt = locatePromptOrValue as TUserPrompt;\n      const optWithValue = locatePromptOrOpt as LocateOption & {\n        // AndroidDeviceInputOpt &\n        value: string | number;\n        autoDismissKeyboard?: boolean;\n      };\n      value = optWithValue.value;\n      opt = optWithValue;\n    } else {\n      // Legacy signature: aiInput(value, locatePrompt, opt)\n      value = locatePromptOrValue as string | number;\n      locatePrompt = locatePromptOrOpt as TUserPrompt;\n      opt = {\n        ...optOrUndefined,\n        value,\n      };\n    }\n\n    assert(\n      typeof value === 'string' || typeof value === 'number',\n      'input value must be a string or number, use empty string if you want to clear the input',\n    );\n    assert(locatePrompt, 'missing locate prompt for input');\n\n    const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n    // Convert value to string to ensure consistency\n    const stringValue = typeof value === 'number' ? String(value) : value;\n\n    return this.callActionInActionSpace('Input', {\n      ...(opt || {}),\n      value: stringValue,\n      locate: detailedLocateParam,\n    });\n  }\n\n  // New signature\n  async aiKeyboardPress(\n    locatePrompt: TUserPrompt,\n    opt: LocateOption & { keyName: string },\n  ): Promise<any>;\n\n  // Legacy signature - deprecated\n  /**\n   * @deprecated Use aiKeyboardPress(locatePrompt, opt) instead where opt contains the keyName\n   */\n  async aiKeyboardPress(\n    keyName: string,\n    locatePrompt?: TUserPrompt,\n    opt?: LocateOption,\n  ): Promise<any>;\n\n  // Implementation\n  async aiKeyboardPress(\n    locatePromptOrKeyName: TUserPrompt | string,\n    locatePromptOrOpt:\n      | TUserPrompt\n      | (LocateOption & { keyName: string })\n      | undefined,\n    optOrUndefined?: LocateOption,\n  ) {\n    let keyName: string;\n    let locatePrompt: TUserPrompt | undefined;\n    let opt: (LocateOption & { keyName: string }) | undefined;\n\n    // Check if using new signature (first param is locatePrompt, second has keyName)\n    if (\n      typeof locatePromptOrOpt === 'object' &&\n      locatePromptOrOpt !== null &&\n      'keyName' in locatePromptOrOpt\n    ) {\n      // New signature: aiKeyboardPress(locatePrompt, opt)\n      locatePrompt = locatePromptOrKeyName as TUserPrompt;\n      opt = locatePromptOrOpt as LocateOption & {\n        keyName: string;\n      };\n    } else {\n      // Legacy signature: aiKeyboardPress(keyName, locatePrompt, opt)\n      keyName = locatePromptOrKeyName as string;\n      locatePrompt = locatePromptOrOpt as TUserPrompt | undefined;\n      opt = {\n        ...(optOrUndefined || {}),\n        keyName,\n      };\n    }\n\n    assert(opt?.keyName, 'missing keyName for keyboard press');\n\n    const detailedLocateParam = locatePrompt\n      ? buildDetailedLocateParam(locatePrompt, opt)\n      : undefined;\n\n    return this.callActionInActionSpace('KeyboardPress', {\n      ...(opt || {}),\n      locate: detailedLocateParam,\n    });\n  }\n\n  // New signature\n  async aiScroll(\n    locatePrompt: TUserPrompt | undefined,\n    opt: LocateOption & ScrollParam,\n  ): Promise<any>;\n\n  // Legacy signature - deprecated\n  /**\n   * @deprecated Use aiScroll(locatePrompt, opt) instead where opt contains the scroll parameters\n   */\n  async aiScroll(\n    scrollParam: ScrollParam,\n    locatePrompt?: TUserPrompt,\n    opt?: LocateOption,\n  ): Promise<any>;\n\n  // Implementation\n  async aiScroll(\n    locatePromptOrScrollParam: TUserPrompt | ScrollParam | undefined,\n    locatePromptOrOpt: TUserPrompt | (LocateOption & ScrollParam) | undefined,\n    optOrUndefined?: LocateOption,\n  ) {\n    let scrollParam: ScrollParam;\n    let locatePrompt: TUserPrompt | undefined;\n    let opt: LocateOption | undefined;\n\n    // Check if using new signature (first param is locatePrompt, second has scroll params)\n    if (\n      typeof locatePromptOrOpt === 'object' &&\n      ('direction' in locatePromptOrOpt ||\n        'scrollType' in locatePromptOrOpt ||\n        'distance' in locatePromptOrOpt)\n    ) {\n      // New signature: aiScroll(locatePrompt, opt)\n      locatePrompt = locatePromptOrScrollParam as TUserPrompt;\n      opt = locatePromptOrOpt as LocateOption & ScrollParam;\n    } else {\n      // Legacy signature: aiScroll(scrollParam, locatePrompt, opt)\n      scrollParam = locatePromptOrScrollParam as ScrollParam;\n      locatePrompt = locatePromptOrOpt as TUserPrompt | undefined;\n      opt = {\n        ...(optOrUndefined || {}),\n        ...(scrollParam || {}),\n      };\n    }\n\n    if (opt) {\n      const normalizedScrollType = normalizeScrollType(\n        (opt as ScrollParam).scrollType as\n          | ScrollParam['scrollType']\n          | LegacyScrollType\n          | undefined,\n      );\n\n      if (normalizedScrollType !== (opt as ScrollParam).scrollType) {\n        (opt as ScrollParam) = {\n          ...(opt || {}),\n          scrollType: normalizedScrollType as ScrollParam['scrollType'],\n        };\n      }\n    }\n\n    const detailedLocateParam = buildDetailedLocateParam(\n      locatePrompt || '',\n      opt,\n    );\n\n    return this.callActionInActionSpace('Scroll', {\n      ...(opt || {}),\n      locate: detailedLocateParam,\n    });\n  }\n\n  async aiAct(\n    taskPrompt: string,\n    opt?: AiActOptions,\n  ): Promise<string | undefined> {\n    const fileChooserAccept = opt?.fileChooserAccept\n      ? this.normalizeFileInput(opt.fileChooserAccept)\n      : undefined;\n\n    const runAiAct = async () => {\n      const modelConfigForPlanning =\n        this.modelConfigManager.getModelConfig('planning');\n      const defaultIntentModelConfig =\n        this.modelConfigManager.getModelConfig('default');\n\n      const includeBboxInPlanning =\n        modelConfigForPlanning.modelName ===\n          defaultIntentModelConfig.modelName &&\n        modelConfigForPlanning.openaiBaseURL ===\n          defaultIntentModelConfig.openaiBaseURL;\n      debug('setting includeBboxInPlanning to', includeBboxInPlanning);\n\n      const cacheable = opt?.cacheable;\n      const deepThink = opt?.deepThink === 'unset' ? undefined : opt?.deepThink;\n      const replanningCycleLimit = this.resolveReplanningCycleLimit(\n        modelConfigForPlanning,\n      );\n      // if vlm-ui-tars or auto-glm, plan cache is not used\n      const isVlmUiTars = isUITars(modelConfigForPlanning.modelFamily);\n      const isAutoGlm = isAutoGLM(modelConfigForPlanning.modelFamily);\n      const matchedCache =\n        isVlmUiTars || isAutoGlm || cacheable === false\n          ? undefined\n          : this.taskCache?.matchPlanCache(taskPrompt);\n      if (\n        matchedCache &&\n        this.taskCache?.isCacheResultUsed &&\n        matchedCache.cacheContent?.yamlWorkflow?.trim()\n      ) {\n        // log into report file\n        await this.taskExecutor.loadYamlFlowAsPlanning(\n          taskPrompt,\n          matchedCache.cacheContent.yamlWorkflow,\n        );\n\n        debug('matched cache, will call .runYaml to run the action');\n        const yaml = matchedCache.cacheContent.yamlWorkflow;\n        await this.runYaml(yaml);\n        return;\n      }\n\n      // If cache matched but yamlWorkflow is empty, fall through to normal execution\n\n      const useDeepThink = (this.opts as any)?._deepThink;\n      if (useDeepThink) {\n        debug('using deep think planning settings');\n      }\n      const imagesIncludeCount: number | undefined = useDeepThink\n        ? undefined\n        : 2;\n      const { output: actionOutput } = await this.taskExecutor.action(\n        taskPrompt,\n        modelConfigForPlanning,\n        defaultIntentModelConfig,\n        includeBboxInPlanning,\n        this.aiActContext,\n        cacheable,\n        replanningCycleLimit,\n        imagesIncludeCount,\n        deepThink,\n        fileChooserAccept,\n      );\n\n      // update cache\n      if (this.taskCache && actionOutput?.yamlFlow && cacheable !== false) {\n        const yamlContent: MidsceneYamlScript = {\n          tasks: [\n            {\n              name: taskPrompt,\n              flow: actionOutput.yamlFlow,\n            },\n          ],\n        };\n        const yamlFlowStr = yaml.dump(yamlContent);\n        this.taskCache.updateOrAppendCacheRecord(\n          {\n            type: 'plan',\n            prompt: taskPrompt,\n            yamlWorkflow: yamlFlowStr,\n          },\n          matchedCache,\n        );\n      }\n\n      return actionOutput?.output;\n    };\n\n    return await runAiAct();\n  }\n\n  /**\n   * @deprecated Use {@link Agent.aiAct} instead.\n   */\n  async aiAction(taskPrompt: string, opt?: AiActOptions) {\n    return this.aiAct(taskPrompt, opt);\n  }\n\n  async aiQuery<ReturnType = any>(\n    demand: ServiceExtractParam,\n    opt: ServiceExtractOption = defaultServiceExtractOption,\n  ): Promise<ReturnType> {\n    const modelConfig = this.modelConfigManager.getModelConfig('insight');\n    const { output } = await this.taskExecutor.createTypeQueryExecution(\n      'Query',\n      demand,\n      modelConfig,\n      opt,\n    );\n    return output as ReturnType;\n  }\n\n  async aiBoolean(\n    prompt: TUserPrompt,\n    opt: ServiceExtractOption = defaultServiceExtractOption,\n  ): Promise<boolean> {\n    const modelConfig = this.modelConfigManager.getModelConfig('insight');\n\n    const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n    const { output } = await this.taskExecutor.createTypeQueryExecution(\n      'Boolean',\n      textPrompt,\n      modelConfig,\n      opt,\n      multimodalPrompt,\n    );\n    return output as boolean;\n  }\n\n  async aiNumber(\n    prompt: TUserPrompt,\n    opt: ServiceExtractOption = defaultServiceExtractOption,\n  ): Promise<number> {\n    const modelConfig = this.modelConfigManager.getModelConfig('insight');\n\n    const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n    const { output } = await this.taskExecutor.createTypeQueryExecution(\n      'Number',\n      textPrompt,\n      modelConfig,\n      opt,\n      multimodalPrompt,\n    );\n    return output as number;\n  }\n\n  async aiString(\n    prompt: TUserPrompt,\n    opt: ServiceExtractOption = defaultServiceExtractOption,\n  ): Promise<string> {\n    const modelConfig = this.modelConfigManager.getModelConfig('insight');\n\n    const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n    const { output } = await this.taskExecutor.createTypeQueryExecution(\n      'String',\n      textPrompt,\n      modelConfig,\n      opt,\n      multimodalPrompt,\n    );\n    return output as string;\n  }\n\n  async aiAsk(\n    prompt: TUserPrompt,\n    opt: ServiceExtractOption = defaultServiceExtractOption,\n  ): Promise<string> {\n    return this.aiString(prompt, opt);\n  }\n\n  async describeElementAtPoint(\n    center: [number, number],\n    opt?: {\n      verifyPrompt?: boolean;\n      retryLimit?: number;\n      deepThink?: boolean;\n    } & LocatorValidatorOption,\n  ): Promise<AgentDescribeElementAtPointResult> {\n    const { verifyPrompt = true, retryLimit = 3 } = opt || {};\n\n    let success = false;\n    let retryCount = 0;\n    let resultPrompt = '';\n    let deepThink = opt?.deepThink || false;\n    let verifyResult: LocateValidatorResult | undefined;\n\n    while (!success && retryCount < retryLimit) {\n      if (retryCount >= 2) {\n        deepThink = true;\n      }\n      debug(\n        'aiDescribe',\n        center,\n        'verifyPrompt',\n        verifyPrompt,\n        'retryCount',\n        retryCount,\n        'deepThink',\n        deepThink,\n      );\n      // use same intent as aiLocate\n      const modelConfig = this.modelConfigManager.getModelConfig('insight');\n\n      const text = await this.service.describe(center, modelConfig, {\n        deepThink,\n      });\n      debug('aiDescribe text', text);\n      assert(text.description, `failed to describe element at [${center}]`);\n      resultPrompt = text.description;\n\n      verifyResult = await this.verifyLocator(\n        resultPrompt,\n        deepThink ? { deepThink: true } : undefined,\n        center,\n        opt,\n      );\n      if (verifyResult.pass) {\n        success = true;\n      } else {\n        retryCount++;\n      }\n    }\n\n    return {\n      prompt: resultPrompt,\n      deepThink,\n      verifyResult,\n    };\n  }\n\n  async verifyLocator(\n    prompt: string,\n    locateOpt: LocateOption | undefined,\n    expectCenter: [number, number],\n    verifyLocateOption?: LocatorValidatorOption,\n  ): Promise<LocateValidatorResult> {\n    debug('verifyLocator', prompt, locateOpt, expectCenter, verifyLocateOption);\n\n    const { center: verifyCenter, rect: verifyRect } = await this.aiLocate(\n      prompt,\n      locateOpt,\n    );\n    const distance = distanceOfTwoPoints(expectCenter, verifyCenter);\n    const included = includedInRect(expectCenter, verifyRect);\n    const pass =\n      distance <= (verifyLocateOption?.centerDistanceThreshold || 20) ||\n      included;\n    const verifyResult = {\n      pass,\n      rect: verifyRect,\n      center: verifyCenter,\n      centerDistance: distance,\n    };\n    debug('aiDescribe verifyResult', verifyResult);\n    return verifyResult;\n  }\n\n  async aiLocate(prompt: TUserPrompt, opt?: LocateOption) {\n    const locateParam = buildDetailedLocateParam(prompt, opt);\n    assert(locateParam, 'cannot get locate param for aiLocate');\n    const locatePlan = locatePlanForLocate(locateParam);\n    const plans = [locatePlan];\n    const defaultIntentModelConfig =\n      this.modelConfigManager.getModelConfig('default');\n    const modelConfigForPlanning =\n      this.modelConfigManager.getModelConfig('planning');\n\n    const { output } = await this.taskExecutor.runPlans(\n      taskTitleStr('Locate', locateParamStr(locateParam)),\n      plans,\n      modelConfigForPlanning,\n      defaultIntentModelConfig,\n    );\n\n    const { element } = output;\n\n    const dprValue = await (this.interface.size() as any).dpr;\n    const dprEntry = dprValue\n      ? {\n          dpr: dprValue,\n        }\n      : {};\n    return {\n      rect: element?.rect,\n      center: element?.center,\n      ...dprEntry,\n    } as Pick<LocateResultElement, 'rect' | 'center'> & {\n      dpr?: number; // this field is deprecated\n    };\n  }\n\n  async aiAssert(\n    assertion: TUserPrompt,\n    msg?: string,\n    opt?: AgentAssertOpt & ServiceExtractOption,\n  ) {\n    const modelConfig = this.modelConfigManager.getModelConfig('insight');\n\n    const serviceOpt: ServiceExtractOption = {\n      domIncluded: opt?.domIncluded ?? defaultServiceExtractOption.domIncluded,\n      screenshotIncluded:\n        opt?.screenshotIncluded ??\n        defaultServiceExtractOption.screenshotIncluded,\n    };\n\n    const { textPrompt, multimodalPrompt } = parsePrompt(assertion);\n    const assertionText =\n      typeof assertion === 'string' ? assertion : assertion.prompt;\n\n    try {\n      const { output, thought } =\n        await this.taskExecutor.createTypeQueryExecution<boolean>(\n          'Assert',\n          textPrompt,\n          modelConfig,\n          serviceOpt,\n          multimodalPrompt,\n        );\n\n      const pass = Boolean(output);\n      const message = pass\n        ? undefined\n        : `Assertion failed: ${msg || assertionText}\\nReason: ${thought || '(no_reason)'}`;\n\n      if (opt?.keepRawResponse) {\n        return {\n          pass,\n          thought,\n          message,\n        };\n      }\n\n      if (!pass) {\n        throw new Error(message);\n      }\n    } catch (error) {\n      if (error instanceof TaskExecutionError) {\n        const errorTask = error.errorTask;\n        const thought = errorTask?.thought;\n        const rawError = errorTask?.error;\n        const rawMessage =\n          errorTask?.errorMessage ||\n          (rawError instanceof Error\n            ? rawError.message\n            : rawError\n              ? String(rawError)\n              : undefined);\n        const reason = thought || rawMessage || '(no_reason)';\n        const message = `Assertion failed: ${msg || assertionText}\\nReason: ${reason}`;\n\n        if (opt?.keepRawResponse) {\n          return {\n            pass: false,\n            thought,\n            message,\n          };\n        }\n\n        throw new Error(message, {\n          cause: rawError ?? error,\n        });\n      }\n\n      throw error;\n    }\n  }\n\n  async aiWaitFor(assertion: TUserPrompt, opt?: AgentWaitForOpt) {\n    const modelConfig = this.modelConfigManager.getModelConfig('insight');\n    await this.taskExecutor.waitFor(\n      assertion,\n      {\n        ...opt,\n        timeoutMs: opt?.timeoutMs || 15 * 1000,\n        checkIntervalMs: opt?.checkIntervalMs || 3 * 1000,\n      },\n      modelConfig,\n    );\n  }\n\n  async ai(...args: Parameters<typeof this.aiAct>) {\n    return this.aiAct(...args);\n  }\n\n  async runYaml(yamlScriptContent: string): Promise<{\n    result: Record<string, any>;\n  }> {\n    const script = parseYamlScript(yamlScriptContent, 'yaml');\n    const player = new ScriptPlayer(script, async () => {\n      return { agent: this, freeFn: [] };\n    });\n    await player.run();\n\n    if (player.status === 'error') {\n      const errors = player.taskStatusList\n        .filter((task) => task.status === 'error')\n        .map((task) => {\n          return `task - ${task.name}: ${task.error?.message}`;\n        })\n        .join('\\n');\n      throw new Error(`Error(s) occurred in running yaml script:\\n${errors}`);\n    }\n\n    return {\n      result: player.result,\n    };\n  }\n\n  async evaluateJavaScript(script: string) {\n    assert(\n      this.interface.evaluateJavaScript,\n      'evaluateJavaScript is not supported in current agent',\n    );\n    return this.interface.evaluateJavaScript(script);\n  }\n\n  /**\n   * Add a dump update listener\n   * @param listener Listener function\n   * @returns A remove function that can be called to remove this listener\n   */\n  addDumpUpdateListener(\n    listener: (dump: string, executionDump?: ExecutionDump) => void,\n  ): () => void {\n    this.dumpUpdateListeners.push(listener);\n\n    // Return remove function\n    return () => {\n      this.removeDumpUpdateListener(listener);\n    };\n  }\n\n  /**\n   * Remove a dump update listener\n   * @param listener The listener function to remove\n   */\n  removeDumpUpdateListener(\n    listener: (dump: string, executionDump?: ExecutionDump) => void,\n  ): void {\n    const index = this.dumpUpdateListeners.indexOf(listener);\n    if (index > -1) {\n      this.dumpUpdateListeners.splice(index, 1);\n    }\n  }\n\n  /**\n   * Clear all dump update listeners\n   */\n  clearDumpUpdateListeners(): void {\n    this.dumpUpdateListeners = [];\n  }\n\n  async destroy() {\n    // Early return if already destroyed\n    if (this.destroyed) {\n      return;\n    }\n\n    await this.interface.destroy?.();\n    this.resetDump(); // reset dump to release memory\n    this.destroyed = true;\n  }\n\n  async recordToReport(\n    title?: string,\n    opt?: {\n      content: string;\n    },\n  ) {\n    // 1. screenshot\n    const base64 = await this.interface.screenshotBase64();\n    const screenshot = ScreenshotItem.create(base64);\n    const now = Date.now();\n    // 2. build recorder\n    const recorder: ExecutionRecorderItem[] = [\n      {\n        type: 'screenshot',\n        ts: now,\n        screenshot,\n      },\n    ];\n    // 3. build ExecutionTaskLog\n    const task: ExecutionTaskLog = {\n      type: 'Log',\n      subType: 'Screenshot',\n      status: 'finished',\n      recorder,\n      timing: {\n        start: now,\n        end: now,\n        cost: 0,\n      },\n      param: {\n        content: opt?.content || '',\n      },\n      executor: async () => {},\n    };\n    // 4. build ExecutionDump\n    const executionDump = new ExecutionDump({\n      logTime: now,\n      name: `Log - ${title || 'untitled'}`,\n      description: opt?.content || '',\n      tasks: [task],\n    });\n    // 5. append to execution dump\n    this.appendExecutionDump(executionDump);\n\n    // Call all registered dump update listeners\n    const dumpString = this.dumpDataString();\n    for (const listener of this.dumpUpdateListeners) {\n      try {\n        listener(dumpString);\n      } catch (error) {\n        console.error('Error in onDumpUpdate listener', error);\n      }\n    }\n\n    this.writeOutActionDumps();\n  }\n\n  /**\n   * @deprecated Use {@link Agent.recordToReport} instead.\n   */\n  async logScreenshot(\n    title?: string,\n    opt?: {\n      content: string;\n    },\n  ) {\n    await this.recordToReport(title, opt);\n  }\n\n  _unstableLogContent() {\n    const { groupName, groupDescription, executions } = this.dump;\n    return {\n      groupName,\n      groupDescription,\n      executions: executions || [],\n    };\n  }\n\n  /**\n   * Freezes the current page context to be reused in subsequent AI operations\n   * This avoids recalculating page context for each operation\n   */\n  async freezePageContext(): Promise<void> {\n    debug('Freezing page context');\n    const context = await this._snapshotContext();\n    // Mark the context as frozen\n    context._isFrozen = true;\n    this.frozenUIContext = context;\n    debug('Page context frozen successfully');\n  }\n\n  /**\n   * Unfreezes the page context, allowing AI operations to calculate context dynamically\n   */\n  async unfreezePageContext(): Promise<void> {\n    debug('Unfreezing page context');\n    this.frozenUIContext = undefined;\n    debug('Page context unfrozen successfully');\n  }\n\n  /**\n   * Process cache configuration and return normalized cache settings\n   */\n  private processCacheConfig(opts: AgentOpt): {\n    id: string;\n    enabled: boolean;\n    readOnly: boolean;\n    writeOnly: boolean;\n  } | null {\n    // Validate original cache config before processing\n    // Agent requires explicit IDs - don't allow auto-generation\n    if (opts.cache === true) {\n      throw new Error(\n        'cache: true requires an explicit cache ID. Please provide:\\n' +\n          'Example: cache: { id: \"my-cache-id\" }',\n      );\n    }\n\n    // Check if cache config object is missing ID\n    if (\n      opts.cache &&\n      typeof opts.cache === 'object' &&\n      opts.cache !== null &&\n      !opts.cache.id\n    ) {\n      throw new Error(\n        'cache configuration requires an explicit id.\\n' +\n          'Example: cache: { id: \"my-cache-id\" }',\n      );\n    }\n\n    // Use the unified utils function to process cache configuration\n    const cacheConfig = processCacheConfig(\n      opts.cache,\n      opts.cacheId || opts.testId || 'default',\n    );\n\n    if (!cacheConfig) {\n      return null;\n    }\n\n    // Handle cache configuration object\n    if (typeof cacheConfig === 'object' && cacheConfig !== null) {\n      const id = cacheConfig.id;\n      const rawStrategy = cacheConfig.strategy as unknown;\n      let strategyValue: string;\n\n      if (rawStrategy === undefined) {\n        strategyValue = 'read-write';\n      } else if (typeof rawStrategy === 'string') {\n        strategyValue = rawStrategy;\n      } else {\n        throw new Error(\n          `cache.strategy must be a string when provided, but received type ${typeof rawStrategy}`,\n        );\n      }\n\n      if (!isValidCacheStrategy(strategyValue)) {\n        throw new Error(\n          `cache.strategy must be one of ${CACHE_STRATEGY_VALUES}, but received \"${strategyValue}\"`,\n        );\n      }\n\n      const isReadOnly = strategyValue === 'read-only';\n      const isWriteOnly = strategyValue === 'write-only';\n\n      return {\n        id,\n        enabled: !isWriteOnly,\n        readOnly: isReadOnly,\n        writeOnly: isWriteOnly,\n      };\n    }\n\n    return null;\n  }\n\n  private normalizeFilePaths(files: string[]): string[] {\n    return files.map((file) => {\n      const absolutePath = resolve(file);\n      if (!existsSync(absolutePath)) {\n        throw new Error(`File not found: ${file}`);\n      }\n      return absolutePath;\n    });\n  }\n\n  private normalizeFileInput(files: string | string[]): string[] {\n    const filesArray = Array.isArray(files) ? files : [files];\n    return this.normalizeFilePaths(filesArray);\n  }\n\n  /**\n   * Manually flush cache to file\n   * @param options - Optional configuration\n   * @param options.cleanUnused - If true, removes unused cache records before flushing\n   */\n  async flushCache(options?: { cleanUnused?: boolean }): Promise<void> {\n    if (!this.taskCache) {\n      throw new Error('Cache is not configured');\n    }\n\n    this.taskCache.flushCacheToFile(options);\n  }\n}\n\nexport const createAgent = (\n  interfaceInstance: AbstractInterface,\n  opts?: AgentOpt,\n) => {\n  return new Agent(interfaceInstance, opts);\n};\n"],"names":["debug","getDebug","distanceOfTwoPoints","p1","p2","x1","y1","x2","y2","Math","includedInRect","point","rect","x","y","left","top","width","height","defaultServiceExtractOption","CACHE_STRATEGIES","isValidCacheStrategy","strategy","value","CACHE_STRATEGY_VALUES","legacyScrollTypeMap","normalizeScrollType","scrollType","defaultReplanningCycleLimit","defaultVlmUiTarsReplanningCycleLimit","defaultAutoGlmReplanningCycleLimit","Agent","callback","context","undefined","pageWidth","assert","screenshotBase64","screenshotWidth","imageInfoOfBase64","Number","computedScale","modelConfigForPlanning","isUITars","isAutoGLM","action","commonContextParser","computedScreenshotScale","scaleForLog","targetWidth","targetHeight","currentScreenshotBase64","resizedBase64","resizeImgBase64","ScreenshotItem","prompt","console","GroupedActionDump","getVersion","WeakMap","execution","runner","currentDump","existingIndex","reportHTMLContent","Error","generateReport","autoPrintReportMsg","writeLogFile","groupedActionDumpFileExt","printReportMsg","task","param","paramStr","tip","typeStr","name","type","opt","actionPlan","plans","Boolean","title","taskTitleStr","locateParamStr","defaultIntentModelConfig","output","locatePrompt","detailedLocateParam","buildDetailedLocateParam","fileChooserAccept","withFileChooser","locatePromptOrValue","locatePromptOrOpt","optOrUndefined","optWithValue","stringValue","String","locatePromptOrKeyName","keyName","locatePromptOrScrollParam","scrollParam","normalizedScrollType","taskPrompt","runAiAct","includeBboxInPlanning","cacheable","deepThink","replanningCycleLimit","isVlmUiTars","isAutoGlm","matchedCache","yaml","useDeepThink","imagesIncludeCount","actionOutput","yamlContent","yamlFlowStr","demand","modelConfig","textPrompt","multimodalPrompt","parsePrompt","center","verifyPrompt","retryLimit","success","retryCount","resultPrompt","verifyResult","text","locateOpt","expectCenter","verifyLocateOption","verifyCenter","verifyRect","distance","included","pass","locateParam","locatePlan","locatePlanForLocate","element","dprValue","dprEntry","assertion","msg","serviceOpt","assertionText","thought","message","error","TaskExecutionError","errorTask","rawError","rawMessage","reason","args","yamlScriptContent","script","parseYamlScript","player","ScriptPlayer","errors","listener","index","base64","screenshot","now","Date","recorder","executionDump","ExecutionDump","dumpString","groupName","groupDescription","executions","opts","cacheConfig","processCacheConfig","id","rawStrategy","strategyValue","isReadOnly","isWriteOnly","files","file","absolutePath","resolve","existsSync","filesArray","Array","options","interfaceInstance","envReplanningCycleLimit","globalConfigManager","MIDSCENE_REPLANNING_CYCLE_LIMIT","Object","resolvedAiActContext","hasCustomConfig","ModelConfigManager","globalModelConfigManager","Service","cacheConfigObj","TaskCache","baseActionSpace","defineActionAssert","defineActionFinalize","TaskExecutor","getReportFileName","createAgent"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;AAsFA,MAAMA,QAAQC,SAAS;AAEvB,MAAMC,sBAAsB,CAACC,IAAsBC;IACjD,MAAM,CAACC,IAAIC,GAAG,GAAGH;IACjB,MAAM,CAACI,IAAIC,GAAG,GAAGJ;IACjB,OAAOK,KAAK,KAAK,CAACA,KAAK,IAAI,CAAEJ,AAAAA,CAAAA,KAAKE,EAAC,KAAM,IAAKD,AAAAA,CAAAA,KAAKE,EAAC,KAAM;AAC5D;AAEA,MAAME,iBAAiB,CAACC,OAAyBC;IAC/C,MAAM,CAACC,GAAGC,EAAE,GAAGH;IACf,MAAM,EAAEI,IAAI,EAAEC,GAAG,EAAEC,KAAK,EAAEC,MAAM,EAAE,GAAGN;IACrC,OAAOC,KAAKE,QAAQF,KAAKE,OAAOE,SAASH,KAAKE,OAAOF,KAAKE,MAAME;AAClE;AAEA,MAAMC,8BAAoD;IACxD,aAAa;IACb,oBAAoB;AACtB;AAIA,MAAMC,mBAA6C;IACjD;IACA;IACA;CACD;AAED,MAAMC,uBAAuB,CAACC,WAC5BF,iBAAiB,IAAI,CAAC,CAACG,QAAUA,UAAUD;AAE7C,MAAME,wBAAwBJ,iBAAiB,GAAG,CAChD,CAACG,QAAU,CAAC,CAAC,EAAEA,MAAM,CAAC,CAAC,EACvB,IAAI,CAAC;AAEP,MAAME,sBAAsB;IAC1B,MAAM;IACN,aAAa;IACb,UAAU;IACV,YAAY;IACZ,WAAW;AACb;AAIA,MAAMC,sBAAsB,CAC1BC;IAEA,IAAI,CAACA,YACH,OAAOA;IAGT,IAAIA,cAAcF,qBAChB,OAAOA,mBAAmB,CAACE,WAA+B;IAG5D,OAAOA;AACT;AAEA,MAAMC,8BAA8B;AACpC,MAAMC,uCAAuC;AAC7C,MAAMC,qCAAqC;AAQpC,MAAMC;IA8BX,IAAI,eAEU;QACZ,OAAO,IAAI,CAAC,mBAAmB,CAAC,EAAE;IACpC;IAEA,IAAI,aAAaC,QAEJ,EAAE;QAEb,IAAI,CAAC,mBAAmB,GAAG,EAAE;QAE7B,IAAIA,UACF,IAAI,CAAC,mBAAmB,CAAC,IAAI,CAACA;IAElC;IAWA,IAAY,eAAmC;QAC7C,OAAO,IAAI,CAAC,IAAI,CAAC,YAAY,IAAI,IAAI,CAAC,IAAI,CAAC,eAAe;IAC5D;IAsBA,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IAKQ,uBAAuB;QAC7B,IACE,CAAC,IAAI,CAAC,mBAAmB,IACzB,AAAiC,gBAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,IAC5B,AAAiC,iBAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,IAC5B,AAAiC,aAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,IAC5B,AAAiC,6BAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,IAC5B,AAAiC,wCAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,EAC5B;YACA,IAAI,CAAC,kBAAkB,CAAC,sBAAsB;YAC9C,IAAI,CAAC,mBAAmB,GAAG;QAC7B;IACF;IAKA,MAAc,mBAAmBC,OAAkB,EAAmB;QACpE,IAAI,AAAyBC,WAAzB,IAAI,CAAC,eAAe,EACtB,OAAO,IAAI,CAAC,eAAe;QAG7B,IAAI,CAAC,IAAI,CAAC,sBAAsB,EAC9B,IAAI,CAAC,sBAAsB,GAAI;YAC7B,MAAMC,YAAYF,QAAQ,IAAI,EAAE;YAChCG,OACED,aAAaA,YAAY,GACzB,CAAC,oDAAoD,EAAEA,WAAW;YAGpEnC,MAAM;YACN,MAAMqC,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;YAClD,MAAM,EAAE,OAAOK,eAAe,EAAE,GAC9B,MAAMC,kBAAkBF;YAC1BrC,MAAM;YAENoC,OACEI,OAAO,QAAQ,CAACF,oBAAoBA,kBAAkB,GACtD,CAAC,0DAA0D,EAAEA,iBAAiB;YAGhF,MAAMG,gBAAgBH,kBAAkBH;YACxCC,OACEI,OAAO,QAAQ,CAACC,kBAAkBA,gBAAgB,GAClD,CAAC,mCAAmC,EAAEA,eAAe;YAGvDzC,MACE,CAAC,0BAA0B,EAAEyC,cAAc,uBAAuB,EAAEH,gBAAgB,gBAAgB,EAAEH,WAAW;YAEnH,OAAOM;QACT;QAGF,IAAI;YACF,IAAI,CAAC,eAAe,GAAG,MAAM,IAAI,CAAC,sBAAsB;YACxD,OAAO,IAAI,CAAC,eAAe;QAC7B,SAAU;YACR,IAAI,CAAC,sBAAsB,GAAGP;QAChC;IACF;IAEQ,4BACNQ,sBAAoC,EAC5B;QACR,IAAI,AAAmCR,WAAnC,IAAI,CAAC,IAAI,CAAC,oBAAoB,EAChC,OAAO,IAAI,CAAC,IAAI,CAAC,oBAAoB;QAGvC,OAAOS,SAASD,uBAAuB,WAAW,IAC9Cb,uCACAe,UAAUF,uBAAuB,WAAW,IAC1CZ,qCACAF;IACR;IAwGA,MAAM,iBAA0C;QAC9C,OAAO,IAAI,CAAC,eAAe;IAC7B;IAEA,MAAM,aAAaiB,MAAsB,EAAsB;QAE7D,IAAI,CAAC,oBAAoB;QAGzB,IAAI,IAAI,CAAC,eAAe,EAAE;YACxB7C,MAAM,yCAAyC6C;YAC/C,OAAO,IAAI,CAAC,eAAe;QAC7B;QAGA,IAAIZ;QACJ,IAAI,IAAI,CAAC,SAAS,CAAC,UAAU,EAAE;YAC7BjC,MAAM,qCAAqC6C;YAC3CZ,UAAU,MAAM,IAAI,CAAC,SAAS,CAAC,UAAU;QAC3C,OAAO;YACLjC,MAAM;YACNiC,UAAU,MAAMa,oBAAoB,IAAI,CAAC,SAAS,EAAE;gBAClD,iBAAiB,IAAI,CAAC,kBAAkB,CAAC,sBAAsB;YACjE;QACF;QAEA9C,MAAM;QACN,MAAM+C,0BAA0B,MAAM,IAAI,CAAC,kBAAkB,CAACd;QAC9DjC,MAAM,2BAA2B+C;QAEjC,IAAIA,AAA4B,MAA5BA,yBAA+B;YACjC,MAAMC,cAAcR,OAAO,UAAU,CAACO,wBAAwB,OAAO,CAAC;YACtE/C,MACE,CAAC,oCAAoC,EAAEgD,YAAY,yBAAyB,CAAC;YAE/E,MAAMC,cAAcxC,KAAK,KAAK,CAACwB,QAAQ,IAAI,CAAC,KAAK;YACjD,MAAMiB,eAAezC,KAAK,KAAK,CAACwB,QAAQ,IAAI,CAAC,MAAM;YACnDjC,MAAM,CAAC,uBAAuB,EAAEiD,YAAY,CAAC,EAAEC,cAAc;YAC7D,MAAMC,0BAA0BlB,QAAQ,UAAU,CAAC,MAAM;YACzD,MAAMmB,gBAAgB,MAAMC,gBAAgBF,yBAAyB;gBACnE,OAAOF;gBACP,QAAQC;YACV;YACAjB,QAAQ,UAAU,GAAGqB,eAAe,MAAM,CAACF;QAC7C,OACEpD,MAAM,CAAC,iBAAiB,EAAE+C,yBAAyB;QAGrD,OAAOd;IACT;IAEA,MAAM,mBAAuC;QAC3C,OAAO,MAAM,IAAI,CAAC,YAAY,CAAC;IACjC;IAKA,MAAM,mBAAmBsB,MAAc,EAAE;QACvC,MAAM,IAAI,CAAC,eAAe,CAACA;IAC7B;IAEA,MAAM,gBAAgBA,MAAc,EAAE;QACpC,IAAI,IAAI,CAAC,YAAY,EACnBC,QAAQ,IAAI,CACV;QAGJ,IAAI,CAAC,IAAI,CAAC,YAAY,GAAGD;QACzB,IAAI,CAAC,IAAI,CAAC,eAAe,GAAGA;IAC9B;IAEA,YAAY;QACV,IAAI,CAAC,IAAI,GAAG,IAAIE,kBAAkB;YAChC,YAAYC;YACZ,WAAW,IAAI,CAAC,IAAI,CAAC,SAAS;YAC9B,kBAAkB,IAAI,CAAC,IAAI,CAAC,gBAAgB;YAC5C,YAAY,EAAE;YACd,aAAa,EAAE;QACjB;QACA,IAAI,CAAC,0BAA0B,GAAG,IAAIC;QAEtC,OAAO,IAAI,CAAC,IAAI;IAClB;IAEA,oBAAoBC,SAAwB,EAAEC,MAAmB,EAAE;QACjE,MAAMC,cAAc,IAAI,CAAC,IAAI;QAC7B,IAAID,QAAQ;YACV,MAAME,gBAAgB,IAAI,CAAC,0BAA0B,CAAC,GAAG,CAACF;YAC1D,IAAIE,AAAkB7B,WAAlB6B,eAA6B;gBAC/BD,YAAY,UAAU,CAACC,cAAc,GAAGH;gBACxC;YACF;YACAE,YAAY,UAAU,CAAC,IAAI,CAACF;YAC5B,IAAI,CAAC,0BAA0B,CAAC,GAAG,CACjCC,QACAC,YAAY,UAAU,CAAC,MAAM,GAAG;YAElC;QACF;QACAA,YAAY,UAAU,CAAC,IAAI,CAACF;IAC9B;IAEA,iBAAiB;QAEf,IAAI,CAAC,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,IAAI,CAAC,SAAS;QACzC,IAAI,CAAC,IAAI,CAAC,gBAAgB,GAAG,IAAI,CAAC,IAAI,CAAC,gBAAgB;QACvD,OAAO,IAAI,CAAC,IAAI,CAAC,SAAS;IAC5B;IAEA,mBAAmB;QACjB,OAAOI,kBAAkB,IAAI,CAAC,cAAc;IAC9C;IAEA,sBAAsB;QACpB,IAAI,IAAI,CAAC,SAAS,EAChB,MAAM,IAAIC,MACR;QAGJ,MAAM,EAAEC,cAAc,EAAEC,kBAAkB,EAAE,GAAG,IAAI,CAAC,IAAI;QACxD,IAAI,CAAC,UAAU,GAAGC,aAAa;YAC7B,UAAU,IAAI,CAAC,cAAc;YAC7B,SAASC;YACT,aAAa,IAAI,CAAC,cAAc;YAChC,MAAM;YACNH;QACF;QACAlE,MAAM,uBAAuB,IAAI,CAAC,UAAU;QAC5C,IAAIkE,kBAAkBC,sBAAsB,IAAI,CAAC,UAAU,EACzDG,eAAe,IAAI,CAAC,UAAU;IAElC;IAEA,MAAc,uBAAuBC,IAAmB,EAAE;QACxD,MAAMC,QAAQC,SAASF;QACvB,MAAMG,MAAMF,QAAQ,GAAGG,QAAQJ,MAAM,GAAG,EAAEC,OAAO,GAAGG,QAAQJ;QAE5D,IAAI,IAAI,CAAC,cAAc,EACrB,MAAM,IAAI,CAAC,cAAc,CAACG;IAE9B;IAEA,wBACEE,IAAY,EACyC;QACrD,OAAO,OAAOJ,QACL,MAAM,IAAI,CAAC,uBAAuB,CAAkBI,MAAMJ;IAErE;IAEA,MAAM,wBACJK,IAAY,EACZC,GAAO,EACP;QACA9E,MAAM,2BAA2B6E,MAAM,KAAKC;QAE5C,MAAMC,aAAgC;YACpC,MAAMF;YACN,OAAQC,OAAe,CAAC;YACxB,SAAS;QACX;QACA9E,MAAM,cAAc+E;QAEpB,MAAMC,QAA0B;YAACD;SAAW,CAAC,MAAM,CACjDE;QAGF,MAAMC,QAAQC,aACZN,MACAO,eAAgBN,KAAa,UAAU,CAAC;QAI1C,MAAMO,2BACJ,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QACzC,MAAM3C,yBACJ,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAEzC,MAAM,EAAE4C,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,CACjDJ,OACAF,OACAtC,wBACA2C;QAEF,OAAOC;IACT;IAEA,MAAM,MACJC,YAAyB,EACzBT,GAA8D,EAC9D;QACA1C,OAAOmD,cAAc;QAErB,MAAMC,sBAAsBC,yBAAyBF,cAAcT;QAEnE,MAAMY,oBAAoBZ,KAAK,oBAC3B,IAAI,CAAC,kBAAkB,CAACA,IAAI,iBAAiB,IAC7C5C;QAEJ,OAAOyD,gBAAgB,IAAI,CAAC,SAAS,EAAED,mBAAmB,UACjD,IAAI,CAAC,uBAAuB,CAAC,OAAO;gBACzC,QAAQF;YACV;IAEJ;IAEA,MAAM,aAAaD,YAAyB,EAAET,GAAkB,EAAE;QAChE1C,OAAOmD,cAAc;QAErB,MAAMC,sBAAsBC,yBAAyBF,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,cAAc;YAChD,QAAQU;QACV;IACF;IAEA,MAAM,cAAcD,YAAyB,EAAET,GAAkB,EAAE;QACjE1C,OAAOmD,cAAc;QAErB,MAAMC,sBAAsBC,yBAAyBF,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,eAAe;YACjD,QAAQU;QACV;IACF;IAEA,MAAM,QAAQD,YAAyB,EAAET,GAAkB,EAAE;QAC3D1C,OAAOmD,cAAc;QAErB,MAAMC,sBAAsBC,yBAAyBF,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,SAAS;YAC3C,QAAQU;QACV;IACF;IAuBA,MAAM,QACJI,mBAAkD,EAClDC,iBAKa,EACbC,cAA6B,EAC7B;QACA,IAAIvE;QACJ,IAAIgE;QACJ,IAAIT;QAOJ,IACE,AAA6B,YAA7B,OAAOe,qBACPA,AAAsB,SAAtBA,qBACA,WAAWA,mBACX;YAEAN,eAAeK;YACf,MAAMG,eAAeF;YAKrBtE,QAAQwE,aAAa,KAAK;YAC1BjB,MAAMiB;QACR,OAAO;YAELxE,QAAQqE;YACRL,eAAeM;YACff,MAAM;gBACJ,GAAGgB,cAAc;gBACjBvE;YACF;QACF;QAEAa,OACE,AAAiB,YAAjB,OAAOb,SAAsB,AAAiB,YAAjB,OAAOA,OACpC;QAEFa,OAAOmD,cAAc;QAErB,MAAMC,sBAAsBC,yBAAyBF,cAAcT;QAGnE,MAAMkB,cAAc,AAAiB,YAAjB,OAAOzE,QAAqB0E,OAAO1E,SAASA;QAEhE,OAAO,IAAI,CAAC,uBAAuB,CAAC,SAAS;YAC3C,GAAIuD,OAAO,CAAC,CAAC;YACb,OAAOkB;YACP,QAAQR;QACV;IACF;IAmBA,MAAM,gBACJU,qBAA2C,EAC3CL,iBAGa,EACbC,cAA6B,EAC7B;QACA,IAAIK;QACJ,IAAIZ;QACJ,IAAIT;QAGJ,IACE,AAA6B,YAA7B,OAAOe,qBACPA,AAAsB,SAAtBA,qBACA,aAAaA,mBACb;YAEAN,eAAeW;YACfpB,MAAMe;QAGR,OAAO;YAELM,UAAUD;YACVX,eAAeM;YACff,MAAM;gBACJ,GAAIgB,kBAAkB,CAAC,CAAC;gBACxBK;YACF;QACF;QAEA/D,OAAO0C,KAAK,SAAS;QAErB,MAAMU,sBAAsBD,eACxBE,yBAAyBF,cAAcT,OACvC5C;QAEJ,OAAO,IAAI,CAAC,uBAAuB,CAAC,iBAAiB;YACnD,GAAI4C,OAAO,CAAC,CAAC;YACb,QAAQU;QACV;IACF;IAmBA,MAAM,SACJY,yBAAgE,EAChEP,iBAAyE,EACzEC,cAA6B,EAC7B;QACA,IAAIO;QACJ,IAAId;QACJ,IAAIT;QAGJ,IACE,AAA6B,YAA7B,OAAOe,qBACN,gBAAeA,qBACd,gBAAgBA,qBAChB,cAAcA,iBAAgB,GAChC;YAEAN,eAAea;YACftB,MAAMe;QACR,OAAO;YAELQ,cAAcD;YACdb,eAAeM;YACff,MAAM;gBACJ,GAAIgB,kBAAkB,CAAC,CAAC;gBACxB,GAAIO,eAAe,CAAC,CAAC;YACvB;QACF;QAEA,IAAIvB,KAAK;YACP,MAAMwB,uBAAuB5E,oBAC1BoD,IAAoB,UAAU;YAMjC,IAAIwB,yBAA0BxB,IAAoB,UAAU,EACzDA,MAAsB;gBACrB,GAAIA,OAAO,CAAC,CAAC;gBACb,YAAYwB;YACd;QAEJ;QAEA,MAAMd,sBAAsBC,yBAC1BF,gBAAgB,IAChBT;QAGF,OAAO,IAAI,CAAC,uBAAuB,CAAC,UAAU;YAC5C,GAAIA,OAAO,CAAC,CAAC;YACb,QAAQU;QACV;IACF;IAEA,MAAM,MACJe,UAAkB,EAClBzB,GAAkB,EACW;QAC7B,MAAMY,oBAAoBZ,KAAK,oBAC3B,IAAI,CAAC,kBAAkB,CAACA,IAAI,iBAAiB,IAC7C5C;QAEJ,MAAMsE,WAAW;YACf,MAAM9D,yBACJ,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;YACzC,MAAM2C,2BACJ,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;YAEzC,MAAMoB,wBACJ/D,uBAAuB,SAAS,KAC9B2C,yBAAyB,SAAS,IACpC3C,uBAAuB,aAAa,KAClC2C,yBAAyB,aAAa;YAC1CrF,MAAM,oCAAoCyG;YAE1C,MAAMC,YAAY5B,KAAK;YACvB,MAAM6B,YAAY7B,KAAK,cAAc,UAAU5C,SAAY4C,KAAK;YAChE,MAAM8B,uBAAuB,IAAI,CAAC,2BAA2B,CAC3DlE;YAGF,MAAMmE,cAAclE,SAASD,uBAAuB,WAAW;YAC/D,MAAMoE,YAAYlE,UAAUF,uBAAuB,WAAW;YAC9D,MAAMqE,eACJF,eAAeC,aAAaJ,AAAc,UAAdA,YACxBxE,SACA,IAAI,CAAC,SAAS,EAAE,eAAeqE;YACrC,IACEQ,gBACA,IAAI,CAAC,SAAS,EAAE,qBAChBA,aAAa,YAAY,EAAE,cAAc,QACzC;gBAEA,MAAM,IAAI,CAAC,YAAY,CAAC,sBAAsB,CAC5CR,YACAQ,aAAa,YAAY,CAAC,YAAY;gBAGxC/G,MAAM;gBACN,MAAMgH,OAAOD,aAAa,YAAY,CAAC,YAAY;gBACnD,MAAM,IAAI,CAAC,OAAO,CAACC;gBACnB;YACF;YAIA,MAAMC,eAAgB,IAAI,CAAC,IAAI,EAAU;YACzC,IAAIA,cACFjH,MAAM;YAER,MAAMkH,qBAAyCD,eAC3C/E,SACA;YACJ,MAAM,EAAE,QAAQiF,YAAY,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,MAAM,CAC7DZ,YACA7D,wBACA2C,0BACAoB,uBACA,IAAI,CAAC,YAAY,EACjBC,WACAE,sBACAM,oBACAP,WACAjB;YAIF,IAAI,IAAI,CAAC,SAAS,IAAIyB,cAAc,YAAYT,AAAc,UAAdA,WAAqB;gBACnE,MAAMU,cAAkC;oBACtC,OAAO;wBACL;4BACE,MAAMb;4BACN,MAAMY,aAAa,QAAQ;wBAC7B;qBACD;gBACH;gBACA,MAAME,cAAcL,QAAAA,IAAS,CAACI;gBAC9B,IAAI,CAAC,SAAS,CAAC,yBAAyB,CACtC;oBACE,MAAM;oBACN,QAAQb;oBACR,cAAcc;gBAChB,GACAN;YAEJ;YAEA,OAAOI,cAAc;QACvB;QAEA,OAAO,MAAMX;IACf;IAKA,MAAM,SAASD,UAAkB,EAAEzB,GAAkB,EAAE;QACrD,OAAO,IAAI,CAAC,KAAK,CAACyB,YAAYzB;IAChC;IAEA,MAAM,QACJwC,MAA2B,EAC3BxC,MAA4B3D,2BAA2B,EAClC;QACrB,MAAMoG,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAC3D,MAAM,EAAEjC,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CACjE,SACAgC,QACAC,aACAzC;QAEF,OAAOQ;IACT;IAEA,MAAM,UACJ/B,MAAmB,EACnBuB,MAA4B3D,2BAA2B,EACrC;QAClB,MAAMoG,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAEC,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAYnE;QACrD,MAAM,EAAE+B,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CACjE,WACAkC,YACAD,aACAzC,KACA2C;QAEF,OAAOnC;IACT;IAEA,MAAM,SACJ/B,MAAmB,EACnBuB,MAA4B3D,2BAA2B,EACtC;QACjB,MAAMoG,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAEC,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAYnE;QACrD,MAAM,EAAE+B,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CACjE,UACAkC,YACAD,aACAzC,KACA2C;QAEF,OAAOnC;IACT;IAEA,MAAM,SACJ/B,MAAmB,EACnBuB,MAA4B3D,2BAA2B,EACtC;QACjB,MAAMoG,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAEC,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAYnE;QACrD,MAAM,EAAE+B,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CACjE,UACAkC,YACAD,aACAzC,KACA2C;QAEF,OAAOnC;IACT;IAEA,MAAM,MACJ/B,MAAmB,EACnBuB,MAA4B3D,2BAA2B,EACtC;QACjB,OAAO,IAAI,CAAC,QAAQ,CAACoC,QAAQuB;IAC/B;IAEA,MAAM,uBACJ6C,MAAwB,EACxB7C,GAI0B,EACkB;QAC5C,MAAM,EAAE8C,eAAe,IAAI,EAAEC,aAAa,CAAC,EAAE,GAAG/C,OAAO,CAAC;QAExD,IAAIgD,UAAU;QACd,IAAIC,aAAa;QACjB,IAAIC,eAAe;QACnB,IAAIrB,YAAY7B,KAAK,aAAa;QAClC,IAAImD;QAEJ,MAAO,CAACH,WAAWC,aAAaF,WAAY;YAC1C,IAAIE,cAAc,GAChBpB,YAAY;YAEd3G,MACE,cACA2H,QACA,gBACAC,cACA,cACAG,YACA,aACApB;YAGF,MAAMY,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;YAE3D,MAAMW,OAAO,MAAM,IAAI,CAAC,OAAO,CAAC,QAAQ,CAACP,QAAQJ,aAAa;gBAC5DZ;YACF;YACA3G,MAAM,mBAAmBkI;YACzB9F,OAAO8F,KAAK,WAAW,EAAE,CAAC,+BAA+B,EAAEP,OAAO,CAAC,CAAC;YACpEK,eAAeE,KAAK,WAAW;YAE/BD,eAAe,MAAM,IAAI,CAAC,aAAa,CACrCD,cACArB,YAAY;gBAAE,WAAW;YAAK,IAAIzE,QAClCyF,QACA7C;YAEF,IAAImD,aAAa,IAAI,EACnBH,UAAU;iBAEVC;QAEJ;QAEA,OAAO;YACL,QAAQC;YACRrB;YACAsB;QACF;IACF;IAEA,MAAM,cACJ1E,MAAc,EACd4E,SAAmC,EACnCC,YAA8B,EAC9BC,kBAA2C,EACX;QAChCrI,MAAM,iBAAiBuD,QAAQ4E,WAAWC,cAAcC;QAExD,MAAM,EAAE,QAAQC,YAAY,EAAE,MAAMC,UAAU,EAAE,GAAG,MAAM,IAAI,CAAC,QAAQ,CACpEhF,QACA4E;QAEF,MAAMK,WAAWtI,oBAAoBkI,cAAcE;QACnD,MAAMG,WAAW/H,eAAe0H,cAAcG;QAC9C,MAAMG,OACJF,YAAaH,CAAAA,oBAAoB,2BAA2B,EAAC,KAC7DI;QACF,MAAMR,eAAe;YACnBS;YACA,MAAMH;YACN,QAAQD;YACR,gBAAgBE;QAClB;QACAxI,MAAM,2BAA2BiI;QACjC,OAAOA;IACT;IAEA,MAAM,SAAS1E,MAAmB,EAAEuB,GAAkB,EAAE;QACtD,MAAM6D,cAAclD,yBAAyBlC,QAAQuB;QACrD1C,OAAOuG,aAAa;QACpB,MAAMC,aAAaC,oBAAoBF;QACvC,MAAM3D,QAAQ;YAAC4D;SAAW;QAC1B,MAAMvD,2BACJ,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QACzC,MAAM3C,yBACJ,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAEzC,MAAM,EAAE4C,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,CACjDH,aAAa,UAAUC,eAAeuD,eACtC3D,OACAtC,wBACA2C;QAGF,MAAM,EAAEyD,OAAO,EAAE,GAAGxD;QAEpB,MAAMyD,WAAW,MAAO,IAAI,CAAC,SAAS,CAAC,IAAI,GAAW,GAAG;QACzD,MAAMC,WAAWD,WACb;YACE,KAAKA;QACP,IACA,CAAC;QACL,OAAO;YACL,MAAMD,SAAS;YACf,QAAQA,SAAS;YACjB,GAAGE,QAAQ;QACb;IAGF;IAEA,MAAM,SACJC,SAAsB,EACtBC,GAAY,EACZpE,GAA2C,EAC3C;QACA,MAAMyC,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM4B,aAAmC;YACvC,aAAarE,KAAK,eAAe3D,4BAA4B,WAAW;YACxE,oBACE2D,KAAK,sBACL3D,4BAA4B,kBAAkB;QAClD;QAEA,MAAM,EAAEqG,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAYuB;QACrD,MAAMG,gBACJ,AAAqB,YAArB,OAAOH,YAAyBA,YAAYA,UAAU,MAAM;QAE9D,IAAI;YACF,MAAM,EAAE3D,MAAM,EAAE+D,OAAO,EAAE,GACvB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,UACA7B,YACAD,aACA4B,YACA1B;YAGJ,MAAMiB,OAAOzD,QAAQK;YACrB,MAAMgE,UAAUZ,OACZxG,SACA,CAAC,kBAAkB,EAAEgH,OAAOE,cAAc,UAAU,EAAEC,WAAW,eAAe;YAEpF,IAAIvE,KAAK,iBACP,OAAO;gBACL4D;gBACAW;gBACAC;YACF;YAGF,IAAI,CAACZ,MACH,MAAM,IAAIzE,MAAMqF;QAEpB,EAAE,OAAOC,OAAO;YACd,IAAIA,iBAAiBC,oBAAoB;gBACvC,MAAMC,YAAYF,MAAM,SAAS;gBACjC,MAAMF,UAAUI,WAAW;gBAC3B,MAAMC,WAAWD,WAAW;gBAC5B,MAAME,aACJF,WAAW,gBACVC,CAAAA,oBAAoBzF,QACjByF,SAAS,OAAO,GAChBA,WACEzD,OAAOyD,YACPxH,MAAQ;gBAChB,MAAM0H,SAASP,WAAWM,cAAc;gBACxC,MAAML,UAAU,CAAC,kBAAkB,EAAEJ,OAAOE,cAAc,UAAU,EAAEQ,QAAQ;gBAE9E,IAAI9E,KAAK,iBACP,OAAO;oBACL,MAAM;oBACNuE;oBACAC;gBACF;gBAGF,MAAM,IAAIrF,MAAMqF,SAAS;oBACvB,OAAOI,YAAYH;gBACrB;YACF;YAEA,MAAMA;QACR;IACF;IAEA,MAAM,UAAUN,SAAsB,EAAEnE,GAAqB,EAAE;QAC7D,MAAMyC,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAC3D,MAAM,IAAI,CAAC,YAAY,CAAC,OAAO,CAC7B0B,WACA;YACE,GAAGnE,GAAG;YACN,WAAWA,KAAK,aAAa;YAC7B,iBAAiBA,KAAK,mBAAmB;QAC3C,GACAyC;IAEJ;IAEA,MAAM,GAAG,GAAGsC,IAAmC,EAAE;QAC/C,OAAO,IAAI,CAAC,KAAK,IAAIA;IACvB;IAEA,MAAM,QAAQC,iBAAyB,EAEpC;QACD,MAAMC,SAASC,gBAAgBF,mBAAmB;QAClD,MAAMG,SAAS,IAAIC,aAAaH,QAAQ,UAC/B;gBAAE,OAAO,IAAI;gBAAE,QAAQ,EAAE;YAAC;QAEnC,MAAME,OAAO,GAAG;QAEhB,IAAIA,AAAkB,YAAlBA,OAAO,MAAM,EAAc;YAC7B,MAAME,SAASF,OAAO,cAAc,CACjC,MAAM,CAAC,CAAC1F,OAASA,AAAgB,YAAhBA,KAAK,MAAM,EAC5B,GAAG,CAAC,CAACA,OACG,CAAC,OAAO,EAAEA,KAAK,IAAI,CAAC,EAAE,EAAEA,KAAK,KAAK,EAAE,SAAS,EAErD,IAAI,CAAC;YACR,MAAM,IAAIN,MAAM,CAAC,2CAA2C,EAAEkG,QAAQ;QACxE;QAEA,OAAO;YACL,QAAQF,OAAO,MAAM;QACvB;IACF;IAEA,MAAM,mBAAmBF,MAAc,EAAE;QACvC3H,OACE,IAAI,CAAC,SAAS,CAAC,kBAAkB,EACjC;QAEF,OAAO,IAAI,CAAC,SAAS,CAAC,kBAAkB,CAAC2H;IAC3C;IAOA,sBACEK,QAA+D,EACnD;QACZ,IAAI,CAAC,mBAAmB,CAAC,IAAI,CAACA;QAG9B,OAAO;YACL,IAAI,CAAC,wBAAwB,CAACA;QAChC;IACF;IAMA,yBACEA,QAA+D,EACzD;QACN,MAAMC,QAAQ,IAAI,CAAC,mBAAmB,CAAC,OAAO,CAACD;QAC/C,IAAIC,QAAQ,IACV,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAACA,OAAO;IAE3C;IAKA,2BAAiC;QAC/B,IAAI,CAAC,mBAAmB,GAAG,EAAE;IAC/B;IAEA,MAAM,UAAU;QAEd,IAAI,IAAI,CAAC,SAAS,EAChB;QAGF,MAAM,IAAI,CAAC,SAAS,CAAC,OAAO;QAC5B,IAAI,CAAC,SAAS;QACd,IAAI,CAAC,SAAS,GAAG;IACnB;IAEA,MAAM,eACJnF,KAAc,EACdJ,GAEC,EACD;QAEA,MAAMwF,SAAS,MAAM,IAAI,CAAC,SAAS,CAAC,gBAAgB;QACpD,MAAMC,aAAajH,eAAe,MAAM,CAACgH;QACzC,MAAME,MAAMC,KAAK,GAAG;QAEpB,MAAMC,WAAoC;YACxC;gBACE,MAAM;gBACN,IAAIF;gBACJD;YACF;SACD;QAED,MAAMhG,OAAyB;YAC7B,MAAM;YACN,SAAS;YACT,QAAQ;YACRmG;YACA,QAAQ;gBACN,OAAOF;gBACP,KAAKA;gBACL,MAAM;YACR;YACA,OAAO;gBACL,SAAS1F,KAAK,WAAW;YAC3B;YACA,UAAU,WAAa;QACzB;QAEA,MAAM6F,gBAAgB,IAAIC,cAAc;YACtC,SAASJ;YACT,MAAM,CAAC,MAAM,EAAEtF,SAAS,YAAY;YACpC,aAAaJ,KAAK,WAAW;YAC7B,OAAO;gBAACP;aAAK;QACf;QAEA,IAAI,CAAC,mBAAmB,CAACoG;QAGzB,MAAME,aAAa,IAAI,CAAC,cAAc;QACtC,KAAK,MAAMT,YAAY,IAAI,CAAC,mBAAmB,CAC7C,IAAI;YACFA,SAASS;QACX,EAAE,OAAOtB,OAAO;YACd/F,QAAQ,KAAK,CAAC,kCAAkC+F;QAClD;QAGF,IAAI,CAAC,mBAAmB;IAC1B;IAKA,MAAM,cACJrE,KAAc,EACdJ,GAEC,EACD;QACA,MAAM,IAAI,CAAC,cAAc,CAACI,OAAOJ;IACnC;IAEA,sBAAsB;QACpB,MAAM,EAAEgG,SAAS,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAG,IAAI,CAAC,IAAI;QAC7D,OAAO;YACLF;YACAC;YACA,YAAYC,cAAc,EAAE;QAC9B;IACF;IAMA,MAAM,oBAAmC;QACvChL,MAAM;QACN,MAAMiC,UAAU,MAAM,IAAI,CAAC,gBAAgB;QAE3CA,QAAQ,SAAS,GAAG;QACpB,IAAI,CAAC,eAAe,GAAGA;QACvBjC,MAAM;IACR;IAKA,MAAM,sBAAqC;QACzCA,MAAM;QACN,IAAI,CAAC,eAAe,GAAGkC;QACvBlC,MAAM;IACR;IAKQ,mBAAmBiL,IAAc,EAKhC;QAGP,IAAIA,AAAe,SAAfA,KAAK,KAAK,EACZ,MAAM,IAAIhH,MACR;QAMJ,IACEgH,KAAK,KAAK,IACV,AAAsB,YAAtB,OAAOA,KAAK,KAAK,IACjBA,AAAe,SAAfA,KAAK,KAAK,IACV,CAACA,KAAK,KAAK,CAAC,EAAE,EAEd,MAAM,IAAIhH,MACR;QAMJ,MAAMiH,cAAcC,mBAClBF,KAAK,KAAK,EACVA,KAAK,OAAO,IAAIA,KAAK,MAAM,IAAI;QAGjC,IAAI,CAACC,aACH,OAAO;QAIT,IAAI,AAAuB,YAAvB,OAAOA,eAA4BA,AAAgB,SAAhBA,aAAsB;YAC3D,MAAME,KAAKF,YAAY,EAAE;YACzB,MAAMG,cAAcH,YAAY,QAAQ;YACxC,IAAII;YAEJ,IAAID,AAAgBnJ,WAAhBmJ,aACFC,gBAAgB;iBACX,IAAI,AAAuB,YAAvB,OAAOD,aAChBC,gBAAgBD;iBAEhB,MAAM,IAAIpH,MACR,CAAC,iEAAiE,EAAE,OAAOoH,aAAa;YAI5F,IAAI,CAAChK,qBAAqBiK,gBACxB,MAAM,IAAIrH,MACR,CAAC,8BAA8B,EAAEzC,sBAAsB,gBAAgB,EAAE8J,cAAc,CAAC,CAAC;YAI7F,MAAMC,aAAaD,AAAkB,gBAAlBA;YACnB,MAAME,cAAcF,AAAkB,iBAAlBA;YAEpB,OAAO;gBACLF;gBACA,SAAS,CAACI;gBACV,UAAUD;gBACV,WAAWC;YACb;QACF;QAEA,OAAO;IACT;IAEQ,mBAAmBC,KAAe,EAAY;QACpD,OAAOA,MAAM,GAAG,CAAC,CAACC;YAChB,MAAMC,eAAeC,QAAQF;YAC7B,IAAI,CAACG,WAAWF,eACd,MAAM,IAAI1H,MAAM,CAAC,gBAAgB,EAAEyH,MAAM;YAE3C,OAAOC;QACT;IACF;IAEQ,mBAAmBF,KAAwB,EAAY;QAC7D,MAAMK,aAAaC,MAAM,OAAO,CAACN,SAASA,QAAQ;YAACA;SAAM;QACzD,OAAO,IAAI,CAAC,kBAAkB,CAACK;IACjC;IAOA,MAAM,WAAWE,OAAmC,EAAiB;QACnE,IAAI,CAAC,IAAI,CAAC,SAAS,EACjB,MAAM,IAAI/H,MAAM;QAGlB,IAAI,CAAC,SAAS,CAAC,gBAAgB,CAAC+H;IAClC;IA9sCA,YAAYC,iBAAgC,EAAEhB,IAAe,CAAE;QAhK/D;QAEA;QAEA;QAEA;QAEA;QAEA;QAEA;QAKA,kCAAU;QAEV;QAEA;QAEA,uBAAQ,uBAEJ,EAAE;QAmBN,oCAAY;QAEZ;QAKA,uBAAQ,mBAAR;QASA,uBAAQ,uBAAsB;QAK9B,uBAAQ,mBAAR;QAKA,uBAAQ,0BAAR;QAEA,uBAAQ,8BAA6B,IAAItH;QAEzC,uBAAQ,mBAAR;QAuFE,IAAI,CAAC,SAAS,GAAGsI;QAEjB,MAAMC,0BACJC,oBAAoB,yBAAyB,CAC3CC;QAGJ,IAAI,CAAC,IAAI,GAAGC,OAAO,MAAM,CACvB;YACE,gBAAgB;YAChB,oBAAoB;YACpB,WAAW;YACX,kBAAkB;QACpB,GACApB,QAAQ,CAAC,GACTA,MAAM,yBAAyB/I,UAC7BgK,AAA4BhK,WAA5BgK,2BACC1J,OAAO,KAAK,CAAC0J,2BAEZ,CAAC,IADD;YAAE,sBAAsBA;QAAwB;QAItD,MAAMI,uBACJ,IAAI,CAAC,IAAI,CAAC,YAAY,IAAI,IAAI,CAAC,IAAI,CAAC,eAAe;QACrD,IAAIA,AAAyBpK,WAAzBoK,sBAAoC;YACtC,IAAI,CAAC,IAAI,CAAC,YAAY,GAAGA;YACzB,IAAI,CAAC,IAAI,CAAC,eAAe,KAAKA;QAChC;QAEA,IACErB,MAAM,eACL,CAA6B,YAA7B,OAAOA,MAAM,eAA4Bc,MAAM,OAAO,CAACd,KAAK,WAAW,IAExE,MAAM,IAAIhH,MACR,CAAC,2EAA2E,EAAE,OAAOgH,MAAM,aAAa;QAK5G,MAAMsB,kBAAkBtB,MAAM,eAAeA,MAAM;QACnD,IAAI,CAAC,kBAAkB,GAAGsB,kBACtB,IAAIC,mBAAmBvB,MAAM,aAAaA,MAAM,sBAChDwB;QAEJ,IAAI,CAAC,cAAc,GAAG,IAAI,CAAC,IAAI,CAAC,cAAc;QAE9C,IAAI,CAAC,OAAO,GAAG,IAAIC,QAAQ,UAClB,IAAI,CAAC,YAAY;QAI1B,MAAMC,iBAAiB,IAAI,CAAC,kBAAkB,CAAC1B,QAAQ,CAAC;QACxD,IAAI0B,gBACF,IAAI,CAAC,SAAS,GAAG,IAAIC,UACnBD,eAAe,EAAE,EACjBA,eAAe,OAAO,EACtBzK,QACA;YACE,UAAUyK,eAAe,QAAQ;YACjC,WAAWA,eAAe,SAAS;QACrC;QAIJ,MAAME,kBAAkB,IAAI,CAAC,SAAS,CAAC,WAAW;QAClD,IAAI,CAAC,eAAe,GAAG;eAClBA;YACHC;YACAC;SACD;QAED,IAAI,CAAC,YAAY,GAAG,IAAIC,aAAa,IAAI,CAAC,SAAS,EAAE,IAAI,CAAC,OAAO,EAAE;YACjE,WAAW,IAAI,CAAC,SAAS;YACzB,aAAa,IAAI,CAAC,sBAAsB,CAAC,IAAI,CAAC,IAAI;YAClD,sBAAsB,IAAI,CAAC,IAAI,CAAC,oBAAoB;YACpD,aAAa,IAAI,CAAC,eAAe;YACjC,OAAO;gBACL,cAAc,CAACnJ;oBACb,MAAM8G,gBAAgB9G,OAAO,IAAI;oBACjC,IAAI,CAAC,mBAAmB,CAAC8G,eAAe9G;oBAGxC,MAAMgH,aAAa,IAAI,CAAC,cAAc;oBACtC,KAAK,MAAMT,YAAY,IAAI,CAAC,mBAAmB,CAC7C,IAAI;wBACFA,SAASS,YAAYF;oBACvB,EAAE,OAAOpB,OAAO;wBACd/F,QAAQ,KAAK,CAAC,kCAAkC+F;oBAClD;oBAGF,IAAI,CAAC,mBAAmB;gBAC1B;YACF;QACF;QACA,IAAI,CAAC,IAAI,GAAG,IAAI,CAAC,SAAS;QAC1B,IAAI,CAAC,cAAc,GACjB0B,MAAM,kBACNgC,kBAAkBhC,MAAM,UAAU,IAAI,CAAC,SAAS,CAAC,aAAa,IAAI;IACtE;AA2mCF;AAEO,MAAMiC,cAAc,CACzBjB,mBACAhB,OAEO,IAAIlJ,MAAMkK,mBAAmBhB"}

package/dist/es/agent/utils.mjs

@@ -103,7 +103,7 @@ async function matchElementFromCache(context, cacheEntry, cachePrompt, cacheable
         return;
     }
 }
-const getMidsceneVersion = ()=>"1.2.2-beta-20260119111553.0";
+const getMidsceneVersion = ()=>"1.2.2-beta-20260119114334.0";
 const parsePrompt = (prompt)=>{
     if ('string' == typeof prompt) return {
         textPrompt: prompt,

package/dist/es/ai-model/llm-planning.mjs

@@ -3,8 +3,51 @@ import { getDebug } from "@midscene/shared/logger";
 import { assert } from "@midscene/shared/utils";
 import { buildYamlFlowFromPlans, fillBboxParam, finalizeActionName, findAllMidsceneLocatorField } from "../common.mjs";
 import { systemPromptToTaskPlanning } from "./prompt/llm-planning.mjs";
-import { callAIWithObjectResponse } from "./service-caller/index.mjs";
+import { callAI, safeParseJson } from "./service-caller/index.mjs";
 const debug = getDebug('planning');
+function parseXMLPlanningResponse(xmlString) {
+    const extractTag = (tagName)=>{
+        const regex = new RegExp(`<${tagName}>([\\s\\S]*?)</${tagName}>`, 'i');
+        const match = xmlString.match(regex);
+        return match ? match[1].trim() : void 0;
+    };
+    const thought = extractTag('thought');
+    const note = extractTag('note');
+    const log = extractTag('log');
+    const error = extractTag('error');
+    const actionType = extractTag('action-type');
+    const actionParamStr = extractTag('action-param-json');
+    if (!log) throw new Error('Missing required field: log');
+    let action = null;
+    if (actionType && 'null' !== actionType.toLowerCase()) {
+        const type = actionType.trim();
+        let param;
+        if (actionParamStr) try {
+            param = safeParseJson(actionParamStr, void 0);
+        } catch (e) {
+            throw new Error(`Failed to parse action-param-json: ${e}`);
+        }
+        action = {
+            type,
+            ...void 0 !== param ? {
+                param
+            } : {}
+        };
+    }
+    return {
+        ...thought ? {
+            thought
+        } : {},
+        ...note ? {
+            note
+        } : {},
+        log,
+        ...error ? {
+            error
+        } : {},
+        action
+    };
+}
 async function plan(userInstruction, opts) {
     const { context, modelConfig, conversationHistory } = opts;
     const { size } = context;
@@ -84,9 +127,10 @@ async function plan(userInstruction, opts) {
         ...instruction,
         ...historyLog
     ];
-    const { content: planFromAI, contentString: rawResponse, usage, reasoning_content } = await callAIWithObjectResponse(msgs, modelConfig, {
+    const { content: rawResponse, usage, reasoning_content } = await callAI(msgs, modelConfig, {
         deepThink: 'unset' === opts.deepThink ? void 0 : opts.deepThink
     });
+    const planFromAI = parseXMLPlanningResponse(rawResponse);
     const actions = planFromAI.action ? [
         planFromAI.action
     ] : [];
@@ -127,6 +171,6 @@ async function plan(userInstruction, opts) {
     });
     return returnValue;
 }
-export { plan };
+export { parseXMLPlanningResponse, plan };
 
 //# sourceMappingURL=llm-planning.mjs.map

package/dist/es/ai-model/llm-planning.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/llm-planning.mjs","sources":["../../../src/ai-model/llm-planning.ts"],"sourcesContent":["import type {\n  DeepThinkOption,\n  DeviceAction,\n  InterfaceType,\n  PlanningAIResponse,\n  RawResponsePlanningAIResponse,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { paddingToMatchBlockByBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport {\n  buildYamlFlowFromPlans,\n  fillBboxParam,\n  finalizeActionName,\n  findAllMidsceneLocatorField,\n} from '../common';\nimport type { ConversationHistory } from './conversation-history';\nimport { systemPromptToTaskPlanning } from './prompt/llm-planning';\nimport { callAIWithObjectResponse } from './service-caller/index';\n\nconst debug = getDebug('planning');\n\nexport async function plan(\n  userInstruction: string,\n  opts: {\n    context: UIContext;\n    interfaceType: InterfaceType;\n    actionSpace: DeviceAction<any>[];\n    actionContext?: string;\n    modelConfig: IModelConfig;\n    conversationHistory: ConversationHistory;\n    includeBbox: boolean;\n    imagesIncludeCount?: number;\n    deepThink?: DeepThinkOption;\n  },\n): Promise<PlanningAIResponse> {\n  const { context, modelConfig, conversationHistory } = opts;\n  const { size } = context;\n  const screenshotBase64 = context.screenshot.base64;\n\n  const { modelFamily } = modelConfig;\n\n  const systemPrompt = await systemPromptToTaskPlanning({\n    actionSpace: opts.actionSpace,\n    modelFamily,\n    includeBbox: opts.includeBbox,\n    includeThought: opts.deepThink !== true,\n  });\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = size.width;\n  let imageHeight = size.height;\n  const rightLimit = imageWidth;\n  const bottomLimit = imageHeight;\n\n  // Process image based on VL mode requirements\n  if (modelFamily === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  }\n\n  const actionContext = opts.actionContext\n    ? `<high_priority_knowledge>${opts.actionContext}</high_priority_knowledge>\\n`\n    : '';\n\n  const instruction: ChatCompletionMessageParam[] = [\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${actionContext}<user_instruction>${userInstruction}</user_instruction>`,\n        },\n      ],\n    },\n  ];\n\n  let latestFeedbackMessage: ChatCompletionMessageParam;\n\n  if (conversationHistory.pendingFeedbackMessage) {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${conversationHistory.pendingFeedbackMessage}. The last screenshot is attached. Please going on according to the instruction.`,\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n\n    conversationHistory.resetPendingFeedbackMessageIfExists();\n  } else {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: 'this is the latest screenshot',\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n  }\n  conversationHistory.append(latestFeedbackMessage);\n  const historyLog = conversationHistory.snapshot(opts.imagesIncludeCount);\n\n  const msgs: ChatCompletionMessageParam[] = [\n    { role: 'system', content: systemPrompt },\n    ...instruction,\n    ...historyLog,\n  ];\n\n  const {\n    content: planFromAI,\n    contentString: rawResponse,\n    usage,\n    reasoning_content,\n  } = await callAIWithObjectResponse<RawResponsePlanningAIResponse>(\n    msgs,\n    modelConfig,\n    {\n      deepThink: opts.deepThink === 'unset' ? undefined : opts.deepThink,\n    },\n  );\n\n  const actions = planFromAI.action ? [planFromAI.action] : [];\n  let shouldContinuePlanning = true;\n  if (actions[0]?.type === finalizeActionName) {\n    debug('finalize action planned, stop planning');\n    shouldContinuePlanning = false;\n  }\n  const returnValue: PlanningAIResponse = {\n    ...planFromAI,\n    actions,\n    rawResponse,\n    usage,\n    reasoning_content,\n    yamlFlow: buildYamlFlowFromPlans(actions, opts.actionSpace),\n    shouldContinuePlanning,\n  };\n\n  assert(planFromAI, \"can't get plans from AI\");\n\n  actions.forEach((action) => {\n    const type = action.type;\n    const actionInActionSpace = opts.actionSpace.find(\n      (action) => action.name === type,\n    );\n\n    debug('actionInActionSpace matched', actionInActionSpace);\n    const locateFields = actionInActionSpace\n      ? findAllMidsceneLocatorField(actionInActionSpace.paramSchema)\n      : [];\n\n    debug('locateFields', locateFields);\n\n    locateFields.forEach((field) => {\n      const locateResult = action.param[field];\n      if (locateResult && modelFamily !== undefined) {\n        // Always use model family to fill bbox parameters\n        action.param[field] = fillBboxParam(\n          locateResult,\n          imageWidth,\n          imageHeight,\n          rightLimit,\n          bottomLimit,\n          modelFamily,\n        );\n      }\n    });\n  });\n\n  conversationHistory.append({\n    role: 'assistant',\n    content: [\n      {\n        type: 'text',\n        text: rawResponse,\n      },\n    ],\n  });\n\n  return returnValue;\n}\n"],"names":["debug","getDebug","plan","userInstruction","opts","context","modelConfig","conversationHistory","size","screenshotBase64","modelFamily","systemPrompt","systemPromptToTaskPlanning","imagePayload","imageWidth","imageHeight","rightLimit","bottomLimit","paddedResult","paddingToMatchBlockByBase64","actionContext","instruction","latestFeedbackMessage","historyLog","msgs","planFromAI","rawResponse","usage","reasoning_content","callAIWithObjectResponse","undefined","actions","shouldContinuePlanning","finalizeActionName","returnValue","buildYamlFlowFromPlans","assert","action","type","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam"],"mappings":";;;;;;AAuBA,MAAMA,QAAQC,SAAS;AAEhB,eAAeC,KACpBC,eAAuB,EACvBC,IAUC;IAED,MAAM,EAAEC,OAAO,EAAEC,WAAW,EAAEC,mBAAmB,EAAE,GAAGH;IACtD,MAAM,EAAEI,IAAI,EAAE,GAAGH;IACjB,MAAMI,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM,EAAEK,WAAW,EAAE,GAAGJ;IAExB,MAAMK,eAAe,MAAMC,2BAA2B;QACpD,aAAaR,KAAK,WAAW;QAC7BM;QACA,aAAaN,KAAK,WAAW;QAC7B,gBAAgBA,AAAmB,SAAnBA,KAAK,SAAS;IAChC;IAEA,IAAIS,eAAeJ;IACnB,IAAIK,aAAaN,KAAK,KAAK;IAC3B,IAAIO,cAAcP,KAAK,MAAM;IAC7B,MAAMQ,aAAaF;IACnB,MAAMG,cAAcF;IAGpB,IAAIL,AAAgB,iBAAhBA,aAA8B;QAChC,MAAMQ,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAME,gBAAgBhB,KAAK,aAAa,GACpC,CAAC,yBAAyB,EAAEA,KAAK,aAAa,CAAC,4BAA4B,CAAC,GAC5E;IAEJ,MAAMiB,cAA4C;QAChD;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGD,cAAc,kBAAkB,EAAEjB,gBAAgB,mBAAmB,CAAC;gBACjF;aACD;QACH;KACD;IAED,IAAImB;IAEJ,IAAIf,oBAAoB,sBAAsB,EAAE;QAC9Ce,wBAAwB;YACtB,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGf,oBAAoB,sBAAsB,CAAC,gFAAgF,CAAC;gBACvI;gBACA;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKM;wBACL,QAAQ;oBACV;gBACF;aACD;QACH;QAEAN,oBAAoB,mCAAmC;IACzD,OACEe,wBAAwB;QACtB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAM;YACR;YACA;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKT;oBACL,QAAQ;gBACV;YACF;SACD;IACH;IAEFN,oBAAoB,MAAM,CAACe;IAC3B,MAAMC,aAAahB,oBAAoB,QAAQ,CAACH,KAAK,kBAAkB;IAEvE,MAAMoB,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASb;QAAa;WACrCU;WACAE;KACJ;IAED,MAAM,EACJ,SAASE,UAAU,EACnB,eAAeC,WAAW,EAC1BC,KAAK,EACLC,iBAAiB,EAClB,GAAG,MAAMC,yBACRL,MACAlB,aACA;QACE,WAAWF,AAAmB,YAAnBA,KAAK,SAAS,GAAe0B,SAAY1B,KAAK,SAAS;IACpE;IAGF,MAAM2B,UAAUN,WAAW,MAAM,GAAG;QAACA,WAAW,MAAM;KAAC,GAAG,EAAE;IAC5D,IAAIO,yBAAyB;IAC7B,IAAID,OAAO,CAAC,EAAE,EAAE,SAASE,oBAAoB;QAC3CjC,MAAM;QACNgC,yBAAyB;IAC3B;IACA,MAAME,cAAkC;QACtC,GAAGT,UAAU;QACbM;QACAL;QACAC;QACAC;QACA,UAAUO,uBAAuBJ,SAAS3B,KAAK,WAAW;QAC1D4B;IACF;IAEAI,OAAOX,YAAY;IAEnBM,QAAQ,OAAO,CAAC,CAACM;QACf,MAAMC,OAAOD,OAAO,IAAI;QACxB,MAAME,sBAAsBnC,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACiC,SAAWA,OAAO,IAAI,KAAKC;QAG9BtC,MAAM,+BAA+BuC;QACrC,MAAMC,eAAeD,sBACjBE,4BAA4BF,oBAAoB,WAAW,IAC3D,EAAE;QAENvC,MAAM,gBAAgBwC;QAEtBA,aAAa,OAAO,CAAC,CAACE;YACpB,MAAMC,eAAeN,OAAO,KAAK,CAACK,MAAM;YACxC,IAAIC,gBAAgBjC,AAAgBoB,WAAhBpB,aAElB2B,OAAO,KAAK,CAACK,MAAM,GAAGE,cACpBD,cACA7B,YACAC,aACAC,YACAC,aACAP;QAGN;IACF;IAEAH,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAMmB;YACR;SACD;IACH;IAEA,OAAOQ;AACT"}
+{"version":3,"file":"ai-model/llm-planning.mjs","sources":["../../../src/ai-model/llm-planning.ts"],"sourcesContent":["import type {\n  DeepThinkOption,\n  DeviceAction,\n  InterfaceType,\n  PlanningAIResponse,\n  RawResponsePlanningAIResponse,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { paddingToMatchBlockByBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport {\n  buildYamlFlowFromPlans,\n  fillBboxParam,\n  finalizeActionName,\n  findAllMidsceneLocatorField,\n} from '../common';\nimport type { ConversationHistory } from './conversation-history';\nimport { systemPromptToTaskPlanning } from './prompt/llm-planning';\nimport { callAI } from './service-caller/index';\nimport { safeParseJson } from './service-caller/index';\n\nconst debug = getDebug('planning');\n\n/**\n * Parse XML response from LLM and convert to RawResponsePlanningAIResponse\n */\nexport function parseXMLPlanningResponse(xmlString: string): RawResponsePlanningAIResponse {\n  // Extract content from XML tags\n  const extractTag = (tagName: string): string | undefined => {\n    const regex = new RegExp(`<${tagName}>([\\\\s\\\\S]*?)</${tagName}>`, 'i');\n    const match = xmlString.match(regex);\n    return match ? match[1].trim() : undefined;\n  };\n\n  const thought = extractTag('thought');\n  const note = extractTag('note');\n  const log = extractTag('log');\n  const error = extractTag('error');\n  const actionType = extractTag('action-type');\n  const actionParamStr = extractTag('action-param-json');\n\n  // Validate required fields\n  if (!log) {\n    throw new Error('Missing required field: log');\n  }\n\n  // Parse action\n  let action: any = null;\n  if (actionType && actionType.toLowerCase() !== 'null') {\n    const type = actionType.trim();\n    let param: any = undefined;\n\n    if (actionParamStr) {\n      try {\n        // Parse the JSON string in action-param-json\n        param = safeParseJson(actionParamStr, undefined);\n      } catch (e) {\n        throw new Error(`Failed to parse action-param-json: ${e}`);\n      }\n    }\n\n    action = {\n      type,\n      ...(param !== undefined ? { param } : {}),\n    };\n  }\n\n  return {\n    ...(thought ? { thought } : {}),\n    ...(note ? { note } : {}),\n    log,\n    ...(error ? { error } : {}),\n    action,\n  };\n}\n\nexport async function plan(\n  userInstruction: string,\n  opts: {\n    context: UIContext;\n    interfaceType: InterfaceType;\n    actionSpace: DeviceAction<any>[];\n    actionContext?: string;\n    modelConfig: IModelConfig;\n    conversationHistory: ConversationHistory;\n    includeBbox: boolean;\n    imagesIncludeCount?: number;\n    deepThink?: DeepThinkOption;\n  },\n): Promise<PlanningAIResponse> {\n  const { context, modelConfig, conversationHistory } = opts;\n  const { size } = context;\n  const screenshotBase64 = context.screenshot.base64;\n\n  const { modelFamily } = modelConfig;\n\n  const systemPrompt = await systemPromptToTaskPlanning({\n    actionSpace: opts.actionSpace,\n    modelFamily,\n    includeBbox: opts.includeBbox,\n    includeThought: opts.deepThink !== true,\n  });\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = size.width;\n  let imageHeight = size.height;\n  const rightLimit = imageWidth;\n  const bottomLimit = imageHeight;\n\n  // Process image based on VL mode requirements\n  if (modelFamily === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  }\n\n  const actionContext = opts.actionContext\n    ? `<high_priority_knowledge>${opts.actionContext}</high_priority_knowledge>\\n`\n    : '';\n\n  const instruction: ChatCompletionMessageParam[] = [\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${actionContext}<user_instruction>${userInstruction}</user_instruction>`,\n        },\n      ],\n    },\n  ];\n\n  let latestFeedbackMessage: ChatCompletionMessageParam;\n\n  if (conversationHistory.pendingFeedbackMessage) {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${conversationHistory.pendingFeedbackMessage}. The last screenshot is attached. Please going on according to the instruction.`,\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n\n    conversationHistory.resetPendingFeedbackMessageIfExists();\n  } else {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: 'this is the latest screenshot',\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n  }\n  conversationHistory.append(latestFeedbackMessage);\n  const historyLog = conversationHistory.snapshot(opts.imagesIncludeCount);\n\n  const msgs: ChatCompletionMessageParam[] = [\n    { role: 'system', content: systemPrompt },\n    ...instruction,\n    ...historyLog,\n  ];\n\n  const {\n    content: rawResponse,\n    usage,\n    reasoning_content,\n  } = await callAI(msgs, modelConfig, {\n    deepThink: opts.deepThink === 'unset' ? undefined : opts.deepThink,\n  });\n\n  // Parse XML response to JSON object\n  const planFromAI = parseXMLPlanningResponse(rawResponse);\n\n  const actions = planFromAI.action ? [planFromAI.action] : [];\n  let shouldContinuePlanning = true;\n  if (actions[0]?.type === finalizeActionName) {\n    debug('finalize action planned, stop planning');\n    shouldContinuePlanning = false;\n  }\n  const returnValue: PlanningAIResponse = {\n    ...planFromAI,\n    actions,\n    rawResponse,\n    usage,\n    reasoning_content,\n    yamlFlow: buildYamlFlowFromPlans(actions, opts.actionSpace),\n    shouldContinuePlanning,\n  };\n\n  assert(planFromAI, \"can't get plans from AI\");\n\n  actions.forEach((action) => {\n    const type = action.type;\n    const actionInActionSpace = opts.actionSpace.find(\n      (action) => action.name === type,\n    );\n\n    debug('actionInActionSpace matched', actionInActionSpace);\n    const locateFields = actionInActionSpace\n      ? findAllMidsceneLocatorField(actionInActionSpace.paramSchema)\n      : [];\n\n    debug('locateFields', locateFields);\n\n    locateFields.forEach((field) => {\n      const locateResult = action.param[field];\n      if (locateResult && modelFamily !== undefined) {\n        // Always use model family to fill bbox parameters\n        action.param[field] = fillBboxParam(\n          locateResult,\n          imageWidth,\n          imageHeight,\n          rightLimit,\n          bottomLimit,\n          modelFamily,\n        );\n      }\n    });\n  });\n\n  conversationHistory.append({\n    role: 'assistant',\n    content: [\n      {\n        type: 'text',\n        text: rawResponse,\n      },\n    ],\n  });\n\n  return returnValue;\n}\n"],"names":["debug","getDebug","parseXMLPlanningResponse","xmlString","extractTag","tagName","regex","RegExp","match","undefined","thought","note","log","error","actionType","actionParamStr","Error","action","type","param","safeParseJson","e","plan","userInstruction","opts","context","modelConfig","conversationHistory","size","screenshotBase64","modelFamily","systemPrompt","systemPromptToTaskPlanning","imagePayload","imageWidth","imageHeight","rightLimit","bottomLimit","paddedResult","paddingToMatchBlockByBase64","actionContext","instruction","latestFeedbackMessage","historyLog","msgs","rawResponse","usage","reasoning_content","callAI","planFromAI","actions","shouldContinuePlanning","finalizeActionName","returnValue","buildYamlFlowFromPlans","assert","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam"],"mappings":";;;;;;AAwBA,MAAMA,QAAQC,SAAS;AAKhB,SAASC,yBAAyBC,SAAiB;IAExD,MAAMC,aAAa,CAACC;QAClB,MAAMC,QAAQ,IAAIC,OAAO,CAAC,CAAC,EAAEF,QAAQ,eAAe,EAAEA,QAAQ,CAAC,CAAC,EAAE;QAClE,MAAMG,QAAQL,UAAU,KAAK,CAACG;QAC9B,OAAOE,QAAQA,KAAK,CAAC,EAAE,CAAC,IAAI,KAAKC;IACnC;IAEA,MAAMC,UAAUN,WAAW;IAC3B,MAAMO,OAAOP,WAAW;IACxB,MAAMQ,MAAMR,WAAW;IACvB,MAAMS,QAAQT,WAAW;IACzB,MAAMU,aAAaV,WAAW;IAC9B,MAAMW,iBAAiBX,WAAW;IAGlC,IAAI,CAACQ,KACH,MAAM,IAAII,MAAM;IAIlB,IAAIC,SAAc;IAClB,IAAIH,cAAcA,AAA6B,WAA7BA,WAAW,WAAW,IAAe;QACrD,MAAMI,OAAOJ,WAAW,IAAI;QAC5B,IAAIK;QAEJ,IAAIJ,gBACF,IAAI;YAEFI,QAAQC,cAAcL,gBAAgBN;QACxC,EAAE,OAAOY,GAAG;YACV,MAAM,IAAIL,MAAM,CAAC,mCAAmC,EAAEK,GAAG;QAC3D;QAGFJ,SAAS;YACPC;YACA,GAAIC,AAAUV,WAAVU,QAAsB;gBAAEA;YAAM,IAAI,CAAC,CAAC;QAC1C;IACF;IAEA,OAAO;QACL,GAAIT,UAAU;YAAEA;QAAQ,IAAI,CAAC,CAAC;QAC9B,GAAIC,OAAO;YAAEA;QAAK,IAAI,CAAC,CAAC;QACxBC;QACA,GAAIC,QAAQ;YAAEA;QAAM,IAAI,CAAC,CAAC;QAC1BI;IACF;AACF;AAEO,eAAeK,KACpBC,eAAuB,EACvBC,IAUC;IAED,MAAM,EAAEC,OAAO,EAAEC,WAAW,EAAEC,mBAAmB,EAAE,GAAGH;IACtD,MAAM,EAAEI,IAAI,EAAE,GAAGH;IACjB,MAAMI,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM,EAAEK,WAAW,EAAE,GAAGJ;IAExB,MAAMK,eAAe,MAAMC,2BAA2B;QACpD,aAAaR,KAAK,WAAW;QAC7BM;QACA,aAAaN,KAAK,WAAW;QAC7B,gBAAgBA,AAAmB,SAAnBA,KAAK,SAAS;IAChC;IAEA,IAAIS,eAAeJ;IACnB,IAAIK,aAAaN,KAAK,KAAK;IAC3B,IAAIO,cAAcP,KAAK,MAAM;IAC7B,MAAMQ,aAAaF;IACnB,MAAMG,cAAcF;IAGpB,IAAIL,AAAgB,iBAAhBA,aAA8B;QAChC,MAAMQ,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAME,gBAAgBhB,KAAK,aAAa,GACpC,CAAC,yBAAyB,EAAEA,KAAK,aAAa,CAAC,4BAA4B,CAAC,GAC5E;IAEJ,MAAMiB,cAA4C;QAChD;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGD,cAAc,kBAAkB,EAAEjB,gBAAgB,mBAAmB,CAAC;gBACjF;aACD;QACH;KACD;IAED,IAAImB;IAEJ,IAAIf,oBAAoB,sBAAsB,EAAE;QAC9Ce,wBAAwB;YACtB,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGf,oBAAoB,sBAAsB,CAAC,gFAAgF,CAAC;gBACvI;gBACA;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKM;wBACL,QAAQ;oBACV;gBACF;aACD;QACH;QAEAN,oBAAoB,mCAAmC;IACzD,OACEe,wBAAwB;QACtB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAM;YACR;YACA;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKT;oBACL,QAAQ;gBACV;YACF;SACD;IACH;IAEFN,oBAAoB,MAAM,CAACe;IAC3B,MAAMC,aAAahB,oBAAoB,QAAQ,CAACH,KAAK,kBAAkB;IAEvE,MAAMoB,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASb;QAAa;WACrCU;WACAE;KACJ;IAED,MAAM,EACJ,SAASE,WAAW,EACpBC,KAAK,EACLC,iBAAiB,EAClB,GAAG,MAAMC,OAAOJ,MAAMlB,aAAa;QAClC,WAAWF,AAAmB,YAAnBA,KAAK,SAAS,GAAef,SAAYe,KAAK,SAAS;IACpE;IAGA,MAAMyB,aAAa/C,yBAAyB2C;IAE5C,MAAMK,UAAUD,WAAW,MAAM,GAAG;QAACA,WAAW,MAAM;KAAC,GAAG,EAAE;IAC5D,IAAIE,yBAAyB;IAC7B,IAAID,OAAO,CAAC,EAAE,EAAE,SAASE,oBAAoB;QAC3CpD,MAAM;QACNmD,yBAAyB;IAC3B;IACA,MAAME,cAAkC;QACtC,GAAGJ,UAAU;QACbC;QACAL;QACAC;QACAC;QACA,UAAUO,uBAAuBJ,SAAS1B,KAAK,WAAW;QAC1D2B;IACF;IAEAI,OAAON,YAAY;IAEnBC,QAAQ,OAAO,CAAC,CAACjC;QACf,MAAMC,OAAOD,OAAO,IAAI;QACxB,MAAMuC,sBAAsBhC,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACP,SAAWA,OAAO,IAAI,KAAKC;QAG9BlB,MAAM,+BAA+BwD;QACrC,MAAMC,eAAeD,sBACjBE,4BAA4BF,oBAAoB,WAAW,IAC3D,EAAE;QAENxD,MAAM,gBAAgByD;QAEtBA,aAAa,OAAO,CAAC,CAACE;YACpB,MAAMC,eAAe3C,OAAO,KAAK,CAAC0C,MAAM;YACxC,IAAIC,gBAAgB9B,AAAgBrB,WAAhBqB,aAElBb,OAAO,KAAK,CAAC0C,MAAM,GAAGE,cACpBD,cACA1B,YACAC,aACAC,YACAC,aACAP;QAGN;IACF;IAEAH,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAMkB;YACR;SACD;IACH;IAEA,OAAOQ;AACT"}

package/dist/es/ai-model/prompt/llm-planning.mjs

@@ -1,15 +1,6 @@
 import { getPreferredLanguage } from "@midscene/shared/env";
 import { getZodDescription, getZodTypeName } from "@midscene/shared/zod-schema-utils";
 import { bboxDescription } from "./common.mjs";
-const buildCommonOutputFields = (includeThought, preferredLanguage)=>{
-    const fields = [
-        `"note"?: string, // some important notes to finish the follow-up action should be written here, and the agent executing the subsequent steps will focus on this information. For example, the data extracted from the current screenshot which will be used in the follow-up action. Use ${preferredLanguage}.`,
-        `"log": string, // a brief preamble to the user explaining what you're about to do. Use ${preferredLanguage}.`,
-        `"error"?: string, // Error messages about unexpected situations, if any. Only think it is an error when the situation is not foreseeable according to the instruction. Use ${preferredLanguage}.`
-    ];
-    if (includeThought) fields.unshift('"thought": string, // your thought process about the next action');
-    return fields.join('\n  ');
-};
 const vlLocateParam = (modelFamily)=>{
     if (modelFamily) return `{bbox: [number, number, number, number], prompt: string } // ${bboxDescription(modelFamily)}`;
     return "{ prompt: string /* description of the target element */ }";
@@ -77,6 +68,20 @@ async function systemPromptToTaskPlanning({ actionSpace, modelFamily, includeBbo
     if (includeBbox && !modelFamily) throw new Error('modelFamily cannot be undefined when includeBbox is true. A valid modelFamily is required for bbox-based location.');
     const actionDescriptionList = actionSpace.map((action)=>descriptionForAction(action, vlLocateParam(includeBbox ? modelFamily : void 0)));
     const actionList = actionDescriptionList.join('\n');
+    const thoughtFieldInstruction = `
+## About the \`thought\` field (reasoning process)
+
+The \`thought\` field captures your structured reasoning about the next action. It should include:
+
+1. **Overall task**: What is the overall instruction/goal?
+2. **Current observation**: What do I see on the current screen?
+3. **Progress**: What steps have been completed so far?
+4. **Next step**: What should the next step be and why?
+5. **Action selection**: Which Action Type should be used to accomplish this step?
+
+**Example thought structure:**
+"The overall task is to [task]. On the current screen, I can see [observations]. We have completed [steps]. The next step should be [next action] because [reason]. I will use [Action Type] to accomplish this."
+`;
     const logFieldInstruction = `
 ## About the \`log\` field (preamble message)
 
@@ -84,7 +89,7 @@ The \`log\` field is a brief preamble message to the user explaining what you're
 
 - **Use ${preferredLanguage}**
 - **Keep it concise**: be no more than 1-2 sentences, focused on immediate, tangible next steps. (8–12 words or Chinese characters for quick updates).
-- **Build on prior context**: if this is not the first action to be done, use the preamble message to connect the dots with what’s been done so far and create a sense of momentum and clarity for the user to understand your next actions.
+- **Build on prior context**: if this is not the first action to be done, use the preamble message to connect the dots with what's been done so far and create a sense of momentum and clarity for the user to understand your next actions.
 - **Keep your tone light, friendly and curious**: add small touches of personality in preambles feel collaborative and engaging.
 
 **Examples:**
@@ -94,11 +99,6 @@ The \`log\` field is a brief preamble message to the user explaining what you're
 - "Go back to find the login button"
 `;
     const shouldIncludeThought = includeThought ?? true;
-    const commonOutputFields = buildCommonOutputFields(shouldIncludeThought, preferredLanguage);
-    const exampleThoughtLine = shouldIncludeThought ? `  "thought": "The form has already been filled, I need to click the login button to login",
-` : '';
-    const exampleThoughtLineWithNote = shouldIncludeThought ? `  "thought": "I need to note the titles in the current screenshot for further processing and scroll to find more titles",
-` : '';
     return `
 Target: User will give you an instruction, some screenshots and previous logs indicating what have been done. Your task is to plan the next one action according to current situation to accomplish the instruction.
 
@@ -116,51 +116,45 @@ Please tell what the next one action is (or null if no action should be done) to
 ## Supporting actions
 ${actionList}
 
+${shouldIncludeThought ? thoughtFieldInstruction : ''}
+
 ${logFieldInstruction}
 
 ## Return format
 
-Return in JSON format:
-{
-  ${commonOutputFields}
-  "action": 
-    {
-      "type": string, // the type of the action
-      "param"?: { // The parameter of the action, if any
-         // k-v style parameter fields
-      }, 
-    } | null
-}
+Return in XML format with the following structure:
+${shouldIncludeThought ? '<thought>Structured reasoning: overall task, current observation, progress, next step, action selection</thought>' : ''}
+<note>CRITICAL: If any information from the current screenshot will be needed in follow-up actions, you MUST record it here completely. The current screenshot will NOT be available in subsequent steps, so this note is your only way to preserve essential information for later use. Examples: extracted data, element states, content that needs to be referenced. Leave empty if no follow-up information is needed.</note>
+<log>a brief preamble to the user</log>
+<error>error messages (optional)</error>
+<action-type>the type of the action, or null if no action</action-type>
+<action-param-json>JSON object containing the action parameters</action-param-json>
 
 For example, if the instruction is to login and the form has already been filled, this is a valid return value:
 
+${shouldIncludeThought ? '<thought>The overall task is to login. On the current screen, I can see a login form with username and password fields already filled, and a login button. We have completed filling in the credentials. The next step should be to submit the login form by clicking the login button. I will use Tap action to click it.</thought>' : ''}<log>Click the login button</log>
+<action-type>Tap</action-type>
+<action-param-json>
 {
-${exampleThoughtLine}  "log": "Click the login button",
-  "action": {
-    "type": "Tap",
-    "param": {
-      "locate": { 
-        "prompt": "The login button"${modelFamily && includeBbox ? ', "bbox": [100, 200, 300, 400]' : ''}
-      }
-    }
+  "locate": {
+    "prompt": "The login button"${modelFamily && includeBbox ? ', "bbox": [100, 200, 300, 400]' : ''}
   }
 }
+</action-param-json>
 
 For example, if the instruction is to find out every title in the screenshot, the return value should be:
 
+${shouldIncludeThought ? '<thought>The overall task is to find out every title in the screenshot. On the current screen, I can see several article titles: \'Hello, world!\', \'Midscene 101\', and \'Model strategy\'. We have just started and observed the visible titles. The next step should be to scroll down to find more titles that may not be currently visible. I will use Scroll action to reveal more content.</thought>' : ''}<note>The titles in the current screenshot are: 'Hello, world!', 'Midscene 101', 'Model strategy'</note>
+<log>Scroll to find more titles</log>
+<action-type>Scroll</action-type>
+<action-param-json>
 {
-${exampleThoughtLineWithNote}  "note": "The titles in the current screenshot are: 'Hello, world!', 'Midscene 101', 'Model strategy'",
-  "log": "Scroll to find more titles",
-  "action": {
-    "type": "Scroll",
-    "param": {
-      "locate": {
-        "prompt": "The page content area"
-      },
-      "direction": "down"
-    }
-  }
+  "locate": {
+    "prompt": "The page content area"
+  },
+  "direction": "down"
 }
+</action-param-json>
 `;
 }
 export { descriptionForAction, systemPromptToTaskPlanning };