@midscene/core 0.27.7-beta-20250908115945.0 → 0.27.7-beta-20250909031245.0

package/dist/es/agent/agent.mjs

@@ -525,6 +525,7 @@ class Agent {
         this.reportFileName = (null == opts ? void 0 : opts.reportFileName) || getReportFileName((null == opts ? void 0 : opts.testId) || this.interface.interfaceType || 'web');
     }
 }
-export { Agent };
+const createAgent = (interfaceInstance, opts)=>new Agent(interfaceInstance, opts);
+export { Agent, createAgent };
 
 //# sourceMappingURL=agent.mjs.map

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

@@ -1 +1 @@
-{"version":3,"file":"agent/agent.mjs","sources":["webpack://@midscene/core/./src/agent/agent.ts"],"sourcesContent":["import {\n  type AgentAssertOpt,\n  type AgentDescribeElementAtPointResult,\n  type AgentWaitForOpt,\n  type DeviceAction,\n  type ExecutionDump,\n  type ExecutionRecorderItem,\n  type ExecutionTask,\n  type ExecutionTaskLog,\n  type Executor,\n  type GroupedActionDump,\n  Insight,\n  type InsightAction,\n  type InsightExtractOption,\n  type InsightExtractParam,\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  type ScrollParam,\n  type TUserPrompt,\n  type UIContext,\n} from '../index';\n\nimport yaml from 'js-yaml';\n\nimport {\n  groupedActionDumpFileExt,\n  reportHTMLContent,\n  stringifyDumpData,\n  writeLogFile,\n} from '@/utils';\nimport {\n  ScriptPlayer,\n  buildDetailedLocateParam,\n  parseYamlScript,\n} from '../yaml/index';\n\nimport type { AbstractInterface } from '@/device';\nimport {\n  type IModelPreferences,\n  MIDSCENE_CACHE,\n  type TModelConfigFn,\n  globalConfigManager,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\n// import type { AndroidDeviceInputOpt } from '../device';\nimport { TaskCache } from './task-cache';\nimport { TaskExecutor, locatePlanForLocate } from './tasks';\nimport { locateParamStr, paramStr, taskTitleStr, typeStr } from './ui-utils';\nimport {\n  commonContextParser,\n  getReportFileName,\n  parsePrompt,\n  printReportMsg,\n} from './utils';\nimport { trimContextByViewport } 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 defaultInsightExtractOption: InsightExtractOption = {\n  domIncluded: false,\n  screenshotIncluded: true,\n};\n\nexport interface AgentOpt {\n  testId?: string;\n  cacheId?: string;\n  groupName?: string;\n  groupDescription?: string;\n  /* if auto generate report, default true */\n  generateReport?: boolean;\n  /* if auto print report msg, default true */\n  autoPrintReportMsg?: boolean;\n  onTaskStartTip?: OnTaskStartTip;\n  aiActionContext?: string;\n  /* custom report file name */\n  reportFileName?: string;\n  modelConfig?: TModelConfigFn;\n}\n\nexport class Agent<\n  InterfaceType extends AbstractInterface = AbstractInterface,\n> {\n  interface: InterfaceType;\n\n  insight: Insight;\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  onDumpUpdate?: (dump: string) => void;\n\n  destroyed = false;\n\n  /**\n   * Frozen page context for consistent AI operations\n   */\n  private frozenUIContext?: UIContext;\n\n  // @deprecated use .interface instead\n  get page() {\n    return this.interface;\n  }\n\n  constructor(interfaceInstance: InterfaceType, opts?: AgentOpt) {\n    this.interface = interfaceInstance;\n    this.opts = Object.assign(\n      {\n        generateReport: true,\n        autoPrintReportMsg: true,\n        groupName: 'Midscene Report',\n        groupDescription: '',\n      },\n      opts || {},\n    );\n    if (opts?.modelConfig && typeof opts?.modelConfig !== 'function') {\n      throw new Error(\n        `opts.modelConfig must be one of function or undefined, but got ${typeof opts?.modelConfig}`,\n      );\n    }\n\n    globalConfigManager.init(opts?.modelConfig);\n\n    this.onTaskStartTip = this.opts.onTaskStartTip;\n\n    this.insight = new Insight(async (action: InsightAction) => {\n      return this.getUIContext(action);\n    });\n\n    if (opts?.cacheId) {\n      this.taskCache = new TaskCache(\n        opts.cacheId,\n        globalConfigManager.getEnvConfigInBoolean(MIDSCENE_CACHE), // if we should use cache to match the element\n      );\n    }\n\n    this.taskExecutor = new TaskExecutor(this.interface, this.insight, {\n      taskCache: this.taskCache,\n      onTaskStart: this.callbackOnTaskStartTip.bind(this),\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.interface.actionSpace();\n  }\n\n  async getUIContext(action?: InsightAction): Promise<UIContext> {\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    if (this.interface.getContext) {\n      debug('Using page.getContext for action:', action);\n      return await this.interface.getContext();\n    } else {\n      debug('Using commonContextParser for action:', action);\n      return await commonContextParser(this.interface);\n    }\n  }\n\n  async _snapshotContext(): Promise<UIContext> {\n    return await this.getUIContext('locate');\n  }\n\n  async setAIActionContext(prompt: string) {\n    this.opts.aiActionContext = prompt;\n  }\n\n  resetDump() {\n    this.dump = {\n      groupName: this.opts.groupName!,\n      groupDescription: this.opts.groupDescription,\n      executions: [],\n      modelBriefs: [],\n    };\n\n    return this.dump;\n  }\n\n  appendExecutionDump(execution: ExecutionDump) {\n    // use trimContextByViewport to process execution\n    const trimmedExecution = trimContextByViewport(execution);\n    const currentDump = this.dump;\n    currentDump.executions.push(trimmedExecution);\n  }\n\n  dumpDataString() {\n    // update dump info\n    this.dump.groupName = this.opts.groupName!;\n    this.dump.groupDescription = this.opts.groupDescription;\n    return stringifyDumpData(this.dump);\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)} - ${param}` : typeStr(task);\n\n    if (this.onTaskStartTip) {\n      await this.onTaskStartTip(tip);\n    }\n  }\n\n  private async afterTaskRunning(executor: Executor, doNotThrowError = false) {\n    this.appendExecutionDump(executor.dump());\n\n    try {\n      if (this.onDumpUpdate) {\n        this.onDumpUpdate(this.dumpDataString());\n      }\n    } catch (error) {\n      console.error('Error in onDumpUpdate', error);\n    }\n\n    this.writeOutActionDumps();\n\n    if (executor.isInErrorState() && !doNotThrowError) {\n      const errorTask = executor.latestErrorTask();\n      throw new Error(`${errorTask?.errorMessage}\\n${errorTask?.errorStack}`, {\n        cause: errorTask?.error,\n      });\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, ',', 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    const { output, executor } = await this.taskExecutor.runPlans(title, plans);\n    await this.afterTaskRunning(executor);\n    return output;\n  }\n\n  async aiTap(locatePrompt: TUserPrompt, opt?: LocateOption) {\n    assert(locatePrompt, 'missing locate prompt for tap');\n\n    const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n    return this.callActionInActionSpace('Tap', {\n      locate: detailedLocateParam,\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 }, // AndroidDeviceInputOpt &\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,\n    locatePrompt: TUserPrompt,\n    opt?: LocateOption, // AndroidDeviceInputOpt &\n  ): Promise<any>;\n\n  // Implementation\n  async aiInput(\n    locatePromptOrValue: TUserPrompt | string,\n    locatePromptOrOpt:\n      | TUserPrompt\n      | (LocateOption & { value: string }) // AndroidDeviceInputOpt &\n      | undefined,\n    optOrUndefined?: LocateOption, // AndroidDeviceInputOpt &\n  ) {\n    let value: string;\n    let locatePrompt: TUserPrompt;\n    let opt:\n      | (LocateOption & { value: string }) // 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;\n      };\n      value = optWithValue.value;\n      opt = optWithValue;\n    } else {\n      // Legacy signature: aiInput(value, locatePrompt, opt)\n      value = locatePromptOrValue as string;\n      locatePrompt = locatePromptOrOpt as TUserPrompt;\n      opt = {\n        ...optOrUndefined,\n        value,\n      };\n    }\n\n    assert(\n      typeof value === 'string',\n      'input value must be a string, 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    return this.callActionInActionSpace('Input', {\n      ...(opt || {}),\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    const detailedLocateParam = buildDetailedLocateParam(\n      locatePrompt || '',\n      opt,\n    );\n\n    return this.callActionInActionSpace('Scroll', {\n      ...(opt || {}),\n      locate: detailedLocateParam,\n    });\n  }\n\n  // async aiLongPress(locatePrompt: TUserPrompt, opt?: LocateOption & PlanningActionParamLongPress) {\n  //   const detailedLocateParam = buildDetailedLocateParam(\n  //     locatePrompt,\n  //     opt,\n  //   );\n\n  //   return this.callActionInActionSpace('LongPress', {\n  //     ...(opt || {}),\n  //     locate: detailedLocateParam,\n  //   });\n  // }\n\n  // async aiSwipe(\n  //   locatePrompt?: TUserPrompt,\n  //   opt?: LocateOption & PlanningActionParamSwipe,\n  // ) {\n  //   const detailedLocateParam = locatePrompt\n  //     ? buildDetailedLocateParam(locatePrompt, opt)\n  //     : undefined;\n\n  //   return this.callActionInActionSpace('Swipe', {\n  //     ...(opt || {}),\n  //     locate: detailedLocateParam,\n  //   });\n  // }\n\n  async aiAction(\n    taskPrompt: string,\n    opt?: {\n      cacheable?: boolean;\n    },\n  ) {\n    const modelPreferences: IModelPreferences = { intent: 'planning' };\n    const cacheable = opt?.cacheable;\n    // if vlm-ui-tars, plan cache is not used\n    const isVlmUiTars = vlLocateMode(modelPreferences) === 'vlm-ui-tars';\n    const matchedCache =\n      isVlmUiTars || cacheable === false\n        ? undefined\n        : this.taskCache?.matchPlanCache(taskPrompt);\n    if (matchedCache && this.taskCache?.isCacheResultUsed) {\n      // log into report file\n      const { executor } = await this.taskExecutor.loadYamlFlowAsPlanning(\n        taskPrompt,\n        matchedCache.cacheContent?.yamlWorkflow,\n      );\n\n      await this.afterTaskRunning(executor);\n\n      debug('matched cache, will call .runYaml to run the action');\n      const yaml = matchedCache.cacheContent?.yamlWorkflow;\n      return this.runYaml(yaml);\n    }\n\n    const { output, executor } = await (isVlmUiTars\n      ? this.taskExecutor.actionToGoal(taskPrompt)\n      : this.taskExecutor.action(taskPrompt, this.opts.aiActionContext));\n\n    // update cache\n    if (this.taskCache && output?.yamlFlow && cacheable !== false) {\n      const yamlContent: MidsceneYamlScript = {\n        tasks: [\n          {\n            name: taskPrompt,\n            flow: output.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    await this.afterTaskRunning(executor);\n    return output;\n  }\n\n  async aiQuery<ReturnType = any>(\n    demand: InsightExtractParam,\n    opt: InsightExtractOption = defaultInsightExtractOption,\n  ): Promise<ReturnType> {\n    const { output, executor } =\n      await this.taskExecutor.createTypeQueryExecution('Query', demand, opt);\n    await this.afterTaskRunning(executor);\n    return output as ReturnType;\n  }\n\n  async aiBoolean(\n    prompt: TUserPrompt,\n    opt: InsightExtractOption = defaultInsightExtractOption,\n  ): Promise<boolean> {\n    const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n    const { output, executor } =\n      await this.taskExecutor.createTypeQueryExecution(\n        'Boolean',\n        textPrompt,\n        opt,\n        multimodalPrompt,\n      );\n    await this.afterTaskRunning(executor);\n    return output as boolean;\n  }\n\n  async aiNumber(\n    prompt: TUserPrompt,\n    opt: InsightExtractOption = defaultInsightExtractOption,\n  ): Promise<number> {\n    const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n    const { output, executor } =\n      await this.taskExecutor.createTypeQueryExecution(\n        'Number',\n        textPrompt,\n        opt,\n        multimodalPrompt,\n      );\n    await this.afterTaskRunning(executor);\n    return output as number;\n  }\n\n  async aiString(\n    prompt: TUserPrompt,\n    opt: InsightExtractOption = defaultInsightExtractOption,\n  ): Promise<string> {\n    const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n    const { output, executor } =\n      await this.taskExecutor.createTypeQueryExecution(\n        'String',\n        textPrompt,\n        opt,\n        multimodalPrompt,\n      );\n    await this.afterTaskRunning(executor);\n    return output as string;\n  }\n\n  async aiAsk(\n    prompt: TUserPrompt,\n    opt: InsightExtractOption = defaultInsightExtractOption,\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      const text = await this.insight.describe(center, { deepThink });\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 { executor, output } = await this.taskExecutor.runPlans(\n      taskTitleStr('Locate', locateParamStr(locateParam)),\n      plans,\n    );\n    await this.afterTaskRunning(executor);\n\n    const { element } = output;\n\n    return {\n      rect: element?.rect,\n      center: element?.center,\n      scale: (await this.interface.size()).dpr,\n    } as Pick<LocateResultElement, 'rect' | 'center'> & {\n      scale: number;\n    };\n  }\n\n  async aiAssert(\n    assertion: TUserPrompt,\n    msg?: string,\n    opt?: AgentAssertOpt & InsightExtractOption,\n  ) {\n    const insightOpt: InsightExtractOption = {\n      domIncluded: opt?.domIncluded ?? defaultInsightExtractOption.domIncluded,\n      screenshotIncluded:\n        opt?.screenshotIncluded ??\n        defaultInsightExtractOption.screenshotIncluded,\n      returnThought: opt?.returnThought ?? true,\n      isWaitForAssert: opt?.isWaitForAssert,\n      doNotThrowError: opt?.doNotThrowError,\n    };\n\n    const { output, executor, thought } = await this.taskExecutor.assert(\n      assertion,\n      insightOpt,\n    );\n    await this.afterTaskRunning(executor, true);\n\n    const message = output\n      ? undefined\n      : `Assertion failed: ${msg || (typeof assertion === 'string' ? assertion : assertion.prompt)}\\nReason: ${\n          thought || executor.latestErrorTask()?.error || '(no_reason)'\n        }`;\n\n    if (opt?.keepRawResponse) {\n      return {\n        pass: output,\n        thought,\n        message,\n      };\n    }\n\n    if (!output) {\n      throw new Error(message);\n    }\n  }\n\n  async aiWaitFor(assertion: TUserPrompt, opt?: AgentWaitForOpt) {\n    const { executor } = await this.taskExecutor.waitFor(assertion, {\n      timeoutMs: opt?.timeoutMs || 15 * 1000,\n      checkIntervalMs: opt?.checkIntervalMs || 3 * 1000,\n    });\n    await this.afterTaskRunning(executor, true);\n\n    if (executor.isInErrorState()) {\n      const errorTask = executor.latestErrorTask();\n      throw new Error(`${errorTask?.error}\\n${errorTask?.errorStack}`);\n    }\n  }\n\n  async ai(taskPrompt: string, type = 'action') {\n    if (type === 'action') {\n      return this.aiAction(taskPrompt);\n    }\n    if (type === 'query') {\n      return this.aiQuery(taskPrompt);\n    }\n\n    if (type === 'assert') {\n      return this.aiAssert(taskPrompt);\n    }\n\n    if (type === 'tap') {\n      return this.aiTap(taskPrompt);\n    }\n\n    if (type === 'rightClick') {\n      return this.aiRightClick(taskPrompt);\n    }\n\n    if (type === 'doubleClick') {\n      return this.aiDoubleClick(taskPrompt);\n    }\n\n    throw new Error(\n      `Unknown type: ${type}, only support 'action', 'query', 'assert', 'tap', 'rightClick', 'doubleClick'`,\n    );\n  }\n\n  async runYaml(yamlScriptContent: string): Promise<{\n    result: Record<string, any>;\n  }> {\n    const script = parseYamlScript(yamlScriptContent, 'yaml', true);\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  async destroy() {\n    await this.interface.destroy?.();\n    this.resetDump(); // reset dump to release memory\n    this.destroyed = true;\n  }\n\n  async logScreenshot(\n    title?: string,\n    opt?: {\n      content: string;\n    },\n  ) {\n    // 1. screenshot\n    const base64 = await this.interface.screenshotBase64();\n    const now = Date.now();\n    // 2. build recorder\n    const recorder: ExecutionRecorderItem[] = [\n      {\n        type: 'screenshot',\n        ts: now,\n        screenshot: base64,\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: ExecutionDump = {\n      sdkVersion: '',\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    try {\n      this.onDumpUpdate?.(this.dumpDataString());\n    } catch (error) {\n      console.error('Failed to update dump', error);\n    }\n\n    this.writeOutActionDumps();\n  }\n\n  _unstableLogContent() {\n    const { groupName, groupDescription, executions } = this.dump;\n    const newExecutions = Array.isArray(executions)\n      ? executions.map((execution: any) => {\n          const { tasks, ...restExecution } = execution;\n          let newTasks = tasks;\n          if (Array.isArray(tasks)) {\n            newTasks = tasks.map((task: any) => {\n              // only remove uiContext and log from task\n              const { uiContext, log, ...restTask } = task;\n              return restTask;\n            });\n          }\n          return { ...restExecution, ...(newTasks ? { tasks: newTasks } : {}) };\n        })\n      : [];\n    return {\n      groupName,\n      groupDescription,\n      executions: newExecutions,\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"],"names":["debug","getDebug","distanceOfTwoPoints","p1","p2","x1","y1","x2","y2","Math","includedInRect","point","rect","x","y","left","top","width","height","defaultInsightExtractOption","Agent","action","commonContextParser","prompt","execution","trimmedExecution","trimContextByViewport","currentDump","stringifyDumpData","reportHTMLContent","Error","generateReport","autoPrintReportMsg","writeLogFile","groupedActionDumpFileExt","printReportMsg","task","param","paramStr","tip","typeStr","executor","doNotThrowError","error","console","errorTask","type","opt","actionPlan","plans","Boolean","title","taskTitleStr","locateParamStr","output","locatePrompt","assert","detailedLocateParam","buildDetailedLocateParam","locatePromptOrValue","locatePromptOrOpt","optOrUndefined","value","optWithValue","locatePromptOrKeyName","keyName","undefined","locatePromptOrScrollParam","scrollParam","taskPrompt","_this_taskCache","_this_taskCache1","modelPreferences","cacheable","isVlmUiTars","vlLocateMode","matchedCache","_matchedCache_cacheContent","_matchedCache_cacheContent1","yaml","yamlContent","yamlFlowStr","demand","textPrompt","multimodalPrompt","parsePrompt","center","verifyPrompt","retryLimit","success","retryCount","resultPrompt","deepThink","verifyResult","text","locateOpt","expectCenter","verifyLocateOption","verifyCenter","verifyRect","distance","included","pass","locateParam","locatePlan","locatePlanForLocate","element","assertion","msg","_executor_latestErrorTask","insightOpt","thought","message","yamlScriptContent","script","parseYamlScript","player","ScriptPlayer","errors","_task_error","_this_interface","base64","now","Date","recorder","executionDump","_this","groupName","groupDescription","executions","newExecutions","Array","tasks","restExecution","newTasks","uiContext","log","restTask","context","interfaceInstance","opts","Object","globalConfigManager","Insight","TaskCache","MIDSCENE_CACHE","TaskExecutor","getReportFileName"],"mappings":";;;;;;;;;;;;;;;;;;;;;AAgEA,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;AAkBO,MAAMC;IAoCX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IA4CA,MAAM,iBAA0C;QAC9C,OAAO,IAAI,CAAC,SAAS,CAAC,WAAW;IACnC;IAEA,MAAM,aAAaC,MAAsB,EAAsB;QAE7D,IAAI,IAAI,CAAC,eAAe,EAAE;YACxBrB,MAAM,yCAAyCqB;YAC/C,OAAO,IAAI,CAAC,eAAe;QAC7B;QAEA,IAAI,IAAI,CAAC,SAAS,CAAC,UAAU,EAAE;YAC7BrB,MAAM,qCAAqCqB;YAC3C,OAAO,MAAM,IAAI,CAAC,SAAS,CAAC,UAAU;QACxC;QACErB,MAAM,yCAAyCqB;QAC/C,OAAO,MAAMC,oBAAoB,IAAI,CAAC,SAAS;IAEnD;IAEA,MAAM,mBAAuC;QAC3C,OAAO,MAAM,IAAI,CAAC,YAAY,CAAC;IACjC;IAEA,MAAM,mBAAmBC,MAAc,EAAE;QACvC,IAAI,CAAC,IAAI,CAAC,eAAe,GAAGA;IAC9B;IAEA,YAAY;QACV,IAAI,CAAC,IAAI,GAAG;YACV,WAAW,IAAI,CAAC,IAAI,CAAC,SAAS;YAC9B,kBAAkB,IAAI,CAAC,IAAI,CAAC,gBAAgB;YAC5C,YAAY,EAAE;YACd,aAAa,EAAE;QACjB;QAEA,OAAO,IAAI,CAAC,IAAI;IAClB;IAEA,oBAAoBC,SAAwB,EAAE;QAE5C,MAAMC,mBAAmBC,sBAAsBF;QAC/C,MAAMG,cAAc,IAAI,CAAC,IAAI;QAC7BA,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,OAAOG,kBAAkB,IAAI,CAAC,IAAI;IACpC;IAEA,mBAAmB;QACjB,OAAOC,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;QACA/B,MAAM,uBAAuB,IAAI,CAAC,UAAU;QAC5C,IAAI+B,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,MAAc,iBAAiBE,QAAkB,EAAEC,kBAAkB,KAAK,EAAE;QAC1E,IAAI,CAAC,mBAAmB,CAACD,SAAS,IAAI;QAEtC,IAAI;YACF,IAAI,IAAI,CAAC,YAAY,EACnB,IAAI,CAAC,YAAY,CAAC,IAAI,CAAC,cAAc;QAEzC,EAAE,OAAOE,OAAO;YACdC,QAAQ,KAAK,CAAC,yBAAyBD;QACzC;QAEA,IAAI,CAAC,mBAAmB;QAExB,IAAIF,SAAS,cAAc,MAAM,CAACC,iBAAiB;YACjD,MAAMG,YAAYJ,SAAS,eAAe;YAC1C,MAAM,IAAIX,MAAM,GAAGe,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,YAAY,CAAC,EAAE,EAAEA,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,UAAU,EAAE,EAAE;gBACtE,OAAOA,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,KAAK;YACzB;QACF;IACF;IAEA,MAAM,wBACJC,IAAY,EACZC,GAAO,EACP;QACA/C,MAAM,2BAA2B8C,MAAM,KAAKC,KAAK,KAAKA;QAEtD,MAAMC,aAAgC;YACpC,MAAMF;YACN,OAAQC,OAAe,CAAC;YACxB,SAAS;QACX;QACA/C,MAAM,cAAcgD;QAEpB,MAAMC,QAA0B;YAACD;SAAW,CAAC,MAAM,CACjDE;QAGF,MAAMC,QAAQC,aACZN,MACAO,eAAe,AAACN,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAa,MAAM,AAAD,KAAK,CAAC;QAG1C,MAAM,EAAEO,MAAM,EAAEb,QAAQ,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,CAACU,OAAOF;QACrE,MAAM,IAAI,CAAC,gBAAgB,CAACR;QAC5B,OAAOa;IACT;IAEA,MAAM,MAAMC,YAAyB,EAAER,GAAkB,EAAE;QACzDS,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcR;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,OAAO;YACzC,QAAQU;QACV;IACF;IAEA,MAAM,aAAaF,YAAyB,EAAER,GAAkB,EAAE;QAChES,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcR;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,cAAc;YAChD,QAAQU;QACV;IACF;IAEA,MAAM,cAAcF,YAAyB,EAAER,GAAkB,EAAE;QACjES,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcR;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,eAAe;YACjD,QAAQU;QACV;IACF;IAEA,MAAM,QAAQF,YAAyB,EAAER,GAAkB,EAAE;QAC3DS,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcR;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,SAAS;YAC3C,QAAQU;QACV;IACF;IAmBA,MAAM,QACJE,mBAAyC,EACzCC,iBAGa,EACbC,cAA6B,EAC7B;QACA,IAAIC;QACJ,IAAIP;QACJ,IAAIR;QAKJ,IACE,AAA6B,YAA7B,OAAOa,qBACPA,AAAsB,SAAtBA,qBACA,WAAWA,mBACX;YAEAL,eAAeI;YACf,MAAMI,eAAeH;YAIrBE,QAAQC,aAAa,KAAK;YAC1BhB,MAAMgB;QACR,OAAO;YAELD,QAAQH;YACRJ,eAAeK;YACfb,MAAM;gBACJ,GAAGc,cAAc;gBACjBC;YACF;QACF;QAEAN,OACE,AAAiB,YAAjB,OAAOM,OACP;QAEFN,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcR;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,SAAS;YAC3C,GAAIA,OAAO,CAAC,CAAC;YACb,QAAQU;QACV;IACF;IAmBA,MAAM,gBACJO,qBAA2C,EAC3CJ,iBAGa,EACbC,cAA6B,EAC7B;QACA,IAAII;QACJ,IAAIV;QACJ,IAAIR;QAGJ,IACE,AAA6B,YAA7B,OAAOa,qBACPA,AAAsB,SAAtBA,qBACA,aAAaA,mBACb;YAEAL,eAAeS;YACfjB,MAAMa;QAGR,OAAO;YAELK,UAAUD;YACVT,eAAeK;YACfb,MAAM;gBACJ,GAAIc,kBAAkB,CAAC,CAAC;gBACxBI;YACF;QACF;QAEAT,OAAOT,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,EAAE;QAErB,MAAMU,sBAAsBF,eACxBG,yBAAyBH,cAAcR,OACvCmB;QAEJ,OAAO,IAAI,CAAC,uBAAuB,CAAC,iBAAiB;YACnD,GAAInB,OAAO,CAAC,CAAC;YACb,QAAQU;QACV;IACF;IAmBA,MAAM,SACJU,yBAAgE,EAChEP,iBAAyE,EACzEC,cAA6B,EAC7B;QACA,IAAIO;QACJ,IAAIb;QACJ,IAAIR;QAGJ,IACE,AAA6B,YAA7B,OAAOa,qBACN,gBAAeA,qBACd,gBAAgBA,qBAChB,cAAcA,iBAAgB,GAChC;YAEAL,eAAeY;YACfpB,MAAMa;QACR,OAAO;YAELQ,cAAcD;YACdZ,eAAeK;YACfb,MAAM;gBACJ,GAAIc,kBAAkB,CAAC,CAAC;gBACxB,GAAIO,eAAe,CAAC,CAAC;YACvB;QACF;QAEA,MAAMX,sBAAsBC,yBAC1BH,gBAAgB,IAChBR;QAGF,OAAO,IAAI,CAAC,uBAAuB,CAAC,UAAU;YAC5C,GAAIA,OAAO,CAAC,CAAC;YACb,QAAQU;QACV;IACF;IA4BA,MAAM,SACJY,UAAkB,EAClBtB,GAEC,EACD;YAQMuB,iBACcC;QARpB,MAAMC,mBAAsC;YAAE,QAAQ;QAAW;QACjE,MAAMC,YAAY1B,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS;QAEhC,MAAM2B,cAAcC,AAAmC,kBAAnCA,aAAaH;QACjC,MAAMI,eACJF,eAAeD,AAAc,UAAdA,YACXP,SAAAA,QACAI,CAAAA,kBAAAA,IAAI,CAAC,SAAS,AAAD,IAAbA,KAAAA,IAAAA,gBAAgB,cAAc,CAACD;QACrC,IAAIO,gBAAAA,SAAgBL,CAAAA,mBAAAA,IAAI,CAAC,SAAS,AAAD,IAAbA,KAAAA,IAAAA,iBAAgB,iBAAiB,AAAD,GAAG;gBAInDM,4BAMWC;YARb,MAAM,EAAErC,QAAQ,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,sBAAsB,CACjE4B,YAAAA,QACAQ,CAAAA,6BAAAA,aAAa,YAAY,AAAD,IAAxBA,KAAAA,IAAAA,2BAA2B,YAAY;YAGzC,MAAM,IAAI,CAAC,gBAAgB,CAACpC;YAE5BzC,MAAM;YACN,MAAM+E,OAAO,QAAAD,CAAAA,8BAAAA,aAAa,YAAY,AAAD,IAAxBA,KAAAA,IAAAA,4BAA2B,YAAY;YACpD,OAAO,IAAI,CAAC,OAAO,CAACC;QACtB;QAEA,MAAM,EAAEzB,MAAM,EAAEb,QAAQ,EAAE,GAAG,MAAOiC,CAAAA,cAChC,IAAI,CAAC,YAAY,CAAC,YAAY,CAACL,cAC/B,IAAI,CAAC,YAAY,CAAC,MAAM,CAACA,YAAY,IAAI,CAAC,IAAI,CAAC,eAAe;QAGlE,IAAI,IAAI,CAAC,SAAS,IAAIf,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,QAAQ,AAAD,KAAKmB,AAAc,UAAdA,WAAqB;YAC7D,MAAMO,cAAkC;gBACtC,OAAO;oBACL;wBACE,MAAMX;wBACN,MAAMf,OAAO,QAAQ;oBACvB;iBACD;YACH;YACA,MAAM2B,cAAcF,QAAAA,IAAS,CAACC;YAC9B,IAAI,CAAC,SAAS,CAAC,yBAAyB,CACtC;gBACE,MAAM;gBACN,QAAQX;gBACR,cAAcY;YAChB,GACAL;QAEJ;QAEA,MAAM,IAAI,CAAC,gBAAgB,CAACnC;QAC5B,OAAOa;IACT;IAEA,MAAM,QACJ4B,MAA2B,EAC3BnC,MAA4B5B,2BAA2B,EAClC;QACrB,MAAM,EAAEmC,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAAC,SAASyC,QAAQnC;QACpE,MAAM,IAAI,CAAC,gBAAgB,CAACN;QAC5B,OAAOa;IACT;IAEA,MAAM,UACJ/B,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACrC;QAClB,MAAM,EAAEgE,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAY9D;QACrD,MAAM,EAAE+B,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,WACA0C,YACApC,KACAqC;QAEJ,MAAM,IAAI,CAAC,gBAAgB,CAAC3C;QAC5B,OAAOa;IACT;IAEA,MAAM,SACJ/B,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACtC;QACjB,MAAM,EAAEgE,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAY9D;QACrD,MAAM,EAAE+B,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,UACA0C,YACApC,KACAqC;QAEJ,MAAM,IAAI,CAAC,gBAAgB,CAAC3C;QAC5B,OAAOa;IACT;IAEA,MAAM,SACJ/B,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACtC;QACjB,MAAM,EAAEgE,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAY9D;QACrD,MAAM,EAAE+B,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,UACA0C,YACApC,KACAqC;QAEJ,MAAM,IAAI,CAAC,gBAAgB,CAAC3C;QAC5B,OAAOa;IACT;IAEA,MAAM,MACJ/B,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACtC;QACjB,OAAO,IAAI,CAAC,QAAQ,CAACI,QAAQwB;IAC/B;IAEA,MAAM,uBACJuC,MAAwB,EACxBvC,GAI0B,EACkB;QAC5C,MAAM,EAAEwC,eAAe,IAAI,EAAEC,aAAa,CAAC,EAAE,GAAGzC,OAAO,CAAC;QAExD,IAAI0C,UAAU;QACd,IAAIC,aAAa;QACjB,IAAIC,eAAe;QACnB,IAAIC,YAAY7C,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,AAAD,KAAK;QAClC,IAAI8C;QAEJ,MAAO,CAACJ,WAAWC,aAAaF,WAAY;YAC1C,IAAIE,cAAc,GAChBE,YAAY;YAEd5F,MACE,cACAsF,QACA,gBACAC,cACA,cACAG,YACA,aACAE;YAEF,MAAME,OAAO,MAAM,IAAI,CAAC,OAAO,CAAC,QAAQ,CAACR,QAAQ;gBAAEM;YAAU;YAC7D5F,MAAM,mBAAmB8F;YACzBtC,OAAOsC,KAAK,WAAW,EAAE,CAAC,+BAA+B,EAAER,OAAO,CAAC,CAAC;YACpEK,eAAeG,KAAK,WAAW;YAE/BD,eAAe,MAAM,IAAI,CAAC,aAAa,CACrCF,cACAC,YAAY;gBAAE,WAAW;YAAK,IAAI1B,QAClCoB,QACAvC;YAEF,IAAI8C,aAAa,IAAI,EACnBJ,UAAU;iBAEVC;QAEJ;QAEA,OAAO;YACL,QAAQC;YACRC;YACAC;QACF;IACF;IAEA,MAAM,cACJtE,MAAc,EACdwE,SAAmC,EACnCC,YAA8B,EAC9BC,kBAA2C,EACX;QAChCjG,MAAM,iBAAiBuB,QAAQwE,WAAWC,cAAcC;QAExD,MAAM,EAAE,QAAQC,YAAY,EAAE,MAAMC,UAAU,EAAE,GAAG,MAAM,IAAI,CAAC,QAAQ,CACpE5E,QACAwE;QAEF,MAAMK,WAAWlG,oBAAoB8F,cAAcE;QACnD,MAAMG,WAAW3F,eAAesF,cAAcG;QAC9C,MAAMG,OACJF,YAAaH,CAAAA,CAAAA,QAAAA,qBAAAA,KAAAA,IAAAA,mBAAoB,uBAAuB,AAAD,KAAK,EAAC,KAC7DI;QACF,MAAMR,eAAe;YACnBS;YACA,MAAMH;YACN,QAAQD;YACR,gBAAgBE;QAClB;QACApG,MAAM,2BAA2B6F;QACjC,OAAOA;IACT;IAEA,MAAM,SAAStE,MAAmB,EAAEwB,GAAkB,EAAE;QACtD,MAAMwD,cAAc7C,yBAAyBnC,QAAQwB;QACrDS,OAAO+C,aAAa;QACpB,MAAMC,aAAaC,oBAAoBF;QACvC,MAAMtD,QAAQ;YAACuD;SAAW;QAC1B,MAAM,EAAE/D,QAAQ,EAAEa,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,CAC3DF,aAAa,UAAUC,eAAekD,eACtCtD;QAEF,MAAM,IAAI,CAAC,gBAAgB,CAACR;QAE5B,MAAM,EAAEiE,OAAO,EAAE,GAAGpD;QAEpB,OAAO;YACL,MAAMoD,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,IAAI;YACnB,QAAQA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM;YACvB,OAAQ,OAAM,IAAI,CAAC,SAAS,CAAC,IAAI,EAAC,EAAG,GAAG;QAC1C;IAGF;IAEA,MAAM,SACJC,SAAsB,EACtBC,GAAY,EACZ7D,GAA2C,EAC3C;YAoBiB8D;QAnBjB,MAAMC,aAAmC;YACvC,aAAa/D,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,KAAK5B,4BAA4B,WAAW;YACxE,oBACE4B,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,kBAAkB,AAAD,KACtB5B,4BAA4B,kBAAkB;YAChD,eAAe4B,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,aAAa,AAAD,KAAK;YACrC,iBAAiBA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe;YACrC,iBAAiBA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe;QACvC;QAEA,MAAM,EAAEO,MAAM,EAAEb,QAAQ,EAAEsE,OAAO,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,MAAM,CAClEJ,WACAG;QAEF,MAAM,IAAI,CAAC,gBAAgB,CAACrE,UAAU;QAEtC,MAAMuE,UAAU1D,SACZY,SACA,CAAC,kBAAkB,EAAE0C,OAAQ,CAAqB,YAArB,OAAOD,YAAyBA,YAAYA,UAAU,MAAK,EAAG,UAAU,EACnGI,WAAAA,SAAWF,CAAAA,4BAAAA,SAAS,eAAe,EAAC,IAAzBA,KAAAA,IAAAA,0BAA4B,KAAK,AAAD,KAAK,eAChD;QAEN,IAAI9D,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,EACtB,OAAO;YACL,MAAMO;YACNyD;YACAC;QACF;QAGF,IAAI,CAAC1D,QACH,MAAM,IAAIxB,MAAMkF;IAEpB;IAEA,MAAM,UAAUL,SAAsB,EAAE5D,GAAqB,EAAE;QAC7D,MAAM,EAAEN,QAAQ,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,OAAO,CAACkE,WAAW;YAC9D,WAAW5D,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,AAAD,KAAK;YAC7B,iBAAiBA,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,AAAD,KAAK;QAC3C;QACA,MAAM,IAAI,CAAC,gBAAgB,CAACN,UAAU;QAEtC,IAAIA,SAAS,cAAc,IAAI;YAC7B,MAAMI,YAAYJ,SAAS,eAAe;YAC1C,MAAM,IAAIX,MAAM,GAAGe,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,KAAK,CAAC,EAAE,EAAEA,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,UAAU,EAAE;QACjE;IACF;IAEA,MAAM,GAAGwB,UAAkB,EAAEvB,OAAO,QAAQ,EAAE;QAC5C,IAAIA,AAAS,aAATA,MACF,OAAO,IAAI,CAAC,QAAQ,CAACuB;QAEvB,IAAIvB,AAAS,YAATA,MACF,OAAO,IAAI,CAAC,OAAO,CAACuB;QAGtB,IAAIvB,AAAS,aAATA,MACF,OAAO,IAAI,CAAC,QAAQ,CAACuB;QAGvB,IAAIvB,AAAS,UAATA,MACF,OAAO,IAAI,CAAC,KAAK,CAACuB;QAGpB,IAAIvB,AAAS,iBAATA,MACF,OAAO,IAAI,CAAC,YAAY,CAACuB;QAG3B,IAAIvB,AAAS,kBAATA,MACF,OAAO,IAAI,CAAC,aAAa,CAACuB;QAG5B,MAAM,IAAIvC,MACR,CAAC,cAAc,EAAEgB,KAAK,8EAA8E,CAAC;IAEzG;IAEA,MAAM,QAAQmE,iBAAyB,EAEpC;QACD,MAAMC,SAASC,gBAAgBF,mBAAmB,QAAQ;QAC1D,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,CAAChF,OAASA,AAAgB,YAAhBA,KAAK,MAAM,EAC5B,GAAG,CAAC,CAACA;oBAC2BmF;gBAA/B,OAAO,CAAC,OAAO,EAAEnF,KAAK,IAAI,CAAC,EAAE,EAAE,QAAAmF,CAAAA,cAAAA,KAAK,KAAK,AAAD,IAATA,KAAAA,IAAAA,YAAY,OAAO,EAAE;YACtD,GACC,IAAI,CAAC;YACR,MAAM,IAAIzF,MAAM,CAAC,2CAA2C,EAAEwF,QAAQ;QACxE;QAEA,OAAO;YACL,QAAQF,OAAO,MAAM;QACvB;IACF;IAEA,MAAM,mBAAmBF,MAAc,EAAE;QACvC1D,OACE,IAAI,CAAC,SAAS,CAAC,kBAAkB,EACjC;QAEF,OAAO,IAAI,CAAC,SAAS,CAAC,kBAAkB,CAAC0D;IAC3C;IAEA,MAAM,UAAU;YACRM,yBAAAA;QAAN,eAAMA,CAAAA,0BAAAA,AAAAA,CAAAA,kBAAAA,IAAI,CAAC,SAAS,AAAD,EAAE,OAAO,AAAD,IAArBA,KAAAA,IAAAA,wBAAAA,IAAAA,CAAAA,gBAAAA;QACN,IAAI,CAAC,SAAS;QACd,IAAI,CAAC,SAAS,GAAG;IACnB;IAEA,MAAM,cACJrE,KAAc,EACdJ,GAEC,EACD;QAEA,MAAM0E,SAAS,MAAM,IAAI,CAAC,SAAS,CAAC,gBAAgB;QACpD,MAAMC,MAAMC,KAAK,GAAG;QAEpB,MAAMC,WAAoC;YACxC;gBACE,MAAM;gBACN,IAAIF;gBACJ,YAAYD;YACd;SACD;QAED,MAAMrF,OAAyB;YAC7B,MAAM;YACN,SAAS;YACT,QAAQ;YACRwF;YACA,QAAQ;gBACN,OAAOF;gBACP,KAAKA;gBACL,MAAM;YACR;YACA,OAAO;gBACL,SAAS3E,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAK;YAC3B;YACA,UAAU,WAAa;QACzB;QAEA,MAAM8E,gBAA+B;YACnC,YAAY;YACZ,SAASH;YACT,MAAM,CAAC,MAAM,EAAEvE,SAAS,YAAY;YACpC,aAAaJ,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAK;YAC7B,OAAO;gBAACX;aAAK;QACf;QAEA,IAAI,CAAC,mBAAmB,CAACyF;QAEzB,IAAI;gBACFC,oBAAAA;oBAAAA,CAAAA,qBAAAA,AAAAA,CAAAA,QAAAA,IAAI,AAAD,EAAE,YAAY,AAAD,KAAhBA,mBAAAA,IAAAA,CAAAA,OAAoB,IAAI,CAAC,cAAc;QACzC,EAAE,OAAOnF,OAAO;YACdC,QAAQ,KAAK,CAAC,yBAAyBD;QACzC;QAEA,IAAI,CAAC,mBAAmB;IAC1B;IAEA,sBAAsB;QACpB,MAAM,EAAEoF,SAAS,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAG,IAAI,CAAC,IAAI;QAC7D,MAAMC,gBAAgBC,MAAM,OAAO,CAACF,cAChCA,WAAW,GAAG,CAAC,CAACzG;YACd,MAAM,EAAE4G,KAAK,EAAE,GAAGC,eAAe,GAAG7G;YACpC,IAAI8G,WAAWF;YACf,IAAID,MAAM,OAAO,CAACC,QAChBE,WAAWF,MAAM,GAAG,CAAC,CAAChG;gBAEpB,MAAM,EAAEmG,SAAS,EAAEC,GAAG,EAAE,GAAGC,UAAU,GAAGrG;gBACxC,OAAOqG;YACT;YAEF,OAAO;gBAAE,GAAGJ,aAAa;gBAAE,GAAIC,WAAW;oBAAE,OAAOA;gBAAS,IAAI,CAAC,CAAC;YAAE;QACtE,KACA,EAAE;QACN,OAAO;YACLP;YACAC;YACA,YAAYE;QACd;IACF;IAMA,MAAM,oBAAmC;QACvClI,MAAM;QACN,MAAM0I,UAAU,MAAM,IAAI,CAAC,gBAAgB;QAE3CA,QAAQ,SAAS,GAAG;QACpB,IAAI,CAAC,eAAe,GAAGA;QACvB1I,MAAM;IACR;IAKA,MAAM,sBAAqC;QACzCA,MAAM;QACN,IAAI,CAAC,eAAe,GAAGkE;QACvBlE,MAAM;IACR;IAx2BA,YAAY2I,iBAAgC,EAAEC,IAAe,CAAE;QArC/D;QAEA;QAEA;QAEA;QAEA;QAEA;QAEA;QAKA,kCAAU;QAEV;QAEA;QAEA;QAEA,oCAAY;QAKZ,uBAAQ,mBAAR;QAQE,IAAI,CAAC,SAAS,GAAGD;QACjB,IAAI,CAAC,IAAI,GAAGE,OAAO,MAAM,CACvB;YACE,gBAAgB;YAChB,oBAAoB;YACpB,WAAW;YACX,kBAAkB;QACpB,GACAD,QAAQ,CAAC;QAEX,IAAIA,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW,AAAD,KAAK,AAA6B,cAA7B,OAAOA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW,AAAD,GAC9C,MAAM,IAAI9G,MACR,CAAC,+DAA+D,EAAE,OAAO8G,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW,AAAD,GAAG;QAIhGE,oBAAoB,IAAI,CAACF,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW;QAE1C,IAAI,CAAC,cAAc,GAAG,IAAI,CAAC,IAAI,CAAC,cAAc;QAE9C,IAAI,CAAC,OAAO,GAAG,IAAIG,QAAQ,OAAO1H,SACzB,IAAI,CAAC,YAAY,CAACA;QAG3B,IAAIuH,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,OAAO,EACf,IAAI,CAAC,SAAS,GAAG,IAAII,UACnBJ,KAAK,OAAO,EACZE,oBAAoB,qBAAqB,CAACG;QAI9C,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;QACpD;QACA,IAAI,CAAC,IAAI,GAAG,IAAI,CAAC,SAAS;QAC1B,IAAI,CAAC,cAAc,GACjBN,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,cAAc,AAAD,KACnBO,kBAAkBP,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,MAAM,AAAD,KAAK,IAAI,CAAC,SAAS,CAAC,aAAa,IAAI;IACtE;AAi0BF"}
+{"version":3,"file":"agent/agent.mjs","sources":["webpack://@midscene/core/./src/agent/agent.ts"],"sourcesContent":["import {\n  type AgentAssertOpt,\n  type AgentDescribeElementAtPointResult,\n  type AgentWaitForOpt,\n  type DeviceAction,\n  type ExecutionDump,\n  type ExecutionRecorderItem,\n  type ExecutionTask,\n  type ExecutionTaskLog,\n  type Executor,\n  type GroupedActionDump,\n  Insight,\n  type InsightAction,\n  type InsightExtractOption,\n  type InsightExtractParam,\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  type ScrollParam,\n  type TUserPrompt,\n  type UIContext,\n} from '../index';\n\nimport yaml from 'js-yaml';\n\nimport {\n  groupedActionDumpFileExt,\n  reportHTMLContent,\n  stringifyDumpData,\n  writeLogFile,\n} from '@/utils';\nimport {\n  ScriptPlayer,\n  buildDetailedLocateParam,\n  parseYamlScript,\n} from '../yaml/index';\n\nimport type { AbstractInterface } from '@/device';\nimport type { AgentOpt } from '@/types';\nimport {\n  type IModelPreferences,\n  MIDSCENE_CACHE,\n  type TModelConfigFn,\n  globalConfigManager,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\n// import type { AndroidDeviceInputOpt } from '../device';\nimport { TaskCache } from './task-cache';\nimport { TaskExecutor, locatePlanForLocate } from './tasks';\nimport { locateParamStr, paramStr, taskTitleStr, typeStr } from './ui-utils';\nimport {\n  commonContextParser,\n  getReportFileName,\n  parsePrompt,\n  printReportMsg,\n} from './utils';\nimport { trimContextByViewport } 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 defaultInsightExtractOption: InsightExtractOption = {\n  domIncluded: false,\n  screenshotIncluded: true,\n};\n\nexport class Agent<\n  InterfaceType extends AbstractInterface = AbstractInterface,\n> {\n  interface: InterfaceType;\n\n  insight: Insight;\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  onDumpUpdate?: (dump: string) => void;\n\n  destroyed = false;\n\n  /**\n   * Frozen page context for consistent AI operations\n   */\n  private frozenUIContext?: UIContext;\n\n  // @deprecated use .interface instead\n  get page() {\n    return this.interface;\n  }\n\n  constructor(interfaceInstance: InterfaceType, opts?: AgentOpt) {\n    this.interface = interfaceInstance;\n    this.opts = Object.assign(\n      {\n        generateReport: true,\n        autoPrintReportMsg: true,\n        groupName: 'Midscene Report',\n        groupDescription: '',\n      },\n      opts || {},\n    );\n    if (opts?.modelConfig && typeof opts?.modelConfig !== 'function') {\n      throw new Error(\n        `opts.modelConfig must be one of function or undefined, but got ${typeof opts?.modelConfig}`,\n      );\n    }\n\n    globalConfigManager.init(opts?.modelConfig);\n\n    this.onTaskStartTip = this.opts.onTaskStartTip;\n\n    this.insight = new Insight(async (action: InsightAction) => {\n      return this.getUIContext(action);\n    });\n\n    if (opts?.cacheId) {\n      this.taskCache = new TaskCache(\n        opts.cacheId,\n        globalConfigManager.getEnvConfigInBoolean(MIDSCENE_CACHE), // if we should use cache to match the element\n      );\n    }\n\n    this.taskExecutor = new TaskExecutor(this.interface, this.insight, {\n      taskCache: this.taskCache,\n      onTaskStart: this.callbackOnTaskStartTip.bind(this),\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.interface.actionSpace();\n  }\n\n  async getUIContext(action?: InsightAction): Promise<UIContext> {\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    if (this.interface.getContext) {\n      debug('Using page.getContext for action:', action);\n      return await this.interface.getContext();\n    } else {\n      debug('Using commonContextParser for action:', action);\n      return await commonContextParser(this.interface);\n    }\n  }\n\n  async _snapshotContext(): Promise<UIContext> {\n    return await this.getUIContext('locate');\n  }\n\n  async setAIActionContext(prompt: string) {\n    this.opts.aiActionContext = prompt;\n  }\n\n  resetDump() {\n    this.dump = {\n      groupName: this.opts.groupName!,\n      groupDescription: this.opts.groupDescription,\n      executions: [],\n      modelBriefs: [],\n    };\n\n    return this.dump;\n  }\n\n  appendExecutionDump(execution: ExecutionDump) {\n    // use trimContextByViewport to process execution\n    const trimmedExecution = trimContextByViewport(execution);\n    const currentDump = this.dump;\n    currentDump.executions.push(trimmedExecution);\n  }\n\n  dumpDataString() {\n    // update dump info\n    this.dump.groupName = this.opts.groupName!;\n    this.dump.groupDescription = this.opts.groupDescription;\n    return stringifyDumpData(this.dump);\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)} - ${param}` : typeStr(task);\n\n    if (this.onTaskStartTip) {\n      await this.onTaskStartTip(tip);\n    }\n  }\n\n  private async afterTaskRunning(executor: Executor, doNotThrowError = false) {\n    this.appendExecutionDump(executor.dump());\n\n    try {\n      if (this.onDumpUpdate) {\n        this.onDumpUpdate(this.dumpDataString());\n      }\n    } catch (error) {\n      console.error('Error in onDumpUpdate', error);\n    }\n\n    this.writeOutActionDumps();\n\n    if (executor.isInErrorState() && !doNotThrowError) {\n      const errorTask = executor.latestErrorTask();\n      throw new Error(`${errorTask?.errorMessage}\\n${errorTask?.errorStack}`, {\n        cause: errorTask?.error,\n      });\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, ',', 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    const { output, executor } = await this.taskExecutor.runPlans(title, plans);\n    await this.afterTaskRunning(executor);\n    return output;\n  }\n\n  async aiTap(locatePrompt: TUserPrompt, opt?: LocateOption) {\n    assert(locatePrompt, 'missing locate prompt for tap');\n\n    const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n    return this.callActionInActionSpace('Tap', {\n      locate: detailedLocateParam,\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 }, // AndroidDeviceInputOpt &\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,\n    locatePrompt: TUserPrompt,\n    opt?: LocateOption, // AndroidDeviceInputOpt &\n  ): Promise<any>;\n\n  // Implementation\n  async aiInput(\n    locatePromptOrValue: TUserPrompt | string,\n    locatePromptOrOpt:\n      | TUserPrompt\n      | (LocateOption & { value: string }) // AndroidDeviceInputOpt &\n      | undefined,\n    optOrUndefined?: LocateOption, // AndroidDeviceInputOpt &\n  ) {\n    let value: string;\n    let locatePrompt: TUserPrompt;\n    let opt:\n      | (LocateOption & { value: string }) // 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;\n      };\n      value = optWithValue.value;\n      opt = optWithValue;\n    } else {\n      // Legacy signature: aiInput(value, locatePrompt, opt)\n      value = locatePromptOrValue as string;\n      locatePrompt = locatePromptOrOpt as TUserPrompt;\n      opt = {\n        ...optOrUndefined,\n        value,\n      };\n    }\n\n    assert(\n      typeof value === 'string',\n      'input value must be a string, 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    return this.callActionInActionSpace('Input', {\n      ...(opt || {}),\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    const detailedLocateParam = buildDetailedLocateParam(\n      locatePrompt || '',\n      opt,\n    );\n\n    return this.callActionInActionSpace('Scroll', {\n      ...(opt || {}),\n      locate: detailedLocateParam,\n    });\n  }\n\n  // async aiLongPress(locatePrompt: TUserPrompt, opt?: LocateOption & PlanningActionParamLongPress) {\n  //   const detailedLocateParam = buildDetailedLocateParam(\n  //     locatePrompt,\n  //     opt,\n  //   );\n\n  //   return this.callActionInActionSpace('LongPress', {\n  //     ...(opt || {}),\n  //     locate: detailedLocateParam,\n  //   });\n  // }\n\n  // async aiSwipe(\n  //   locatePrompt?: TUserPrompt,\n  //   opt?: LocateOption & PlanningActionParamSwipe,\n  // ) {\n  //   const detailedLocateParam = locatePrompt\n  //     ? buildDetailedLocateParam(locatePrompt, opt)\n  //     : undefined;\n\n  //   return this.callActionInActionSpace('Swipe', {\n  //     ...(opt || {}),\n  //     locate: detailedLocateParam,\n  //   });\n  // }\n\n  async aiAction(\n    taskPrompt: string,\n    opt?: {\n      cacheable?: boolean;\n    },\n  ) {\n    const modelPreferences: IModelPreferences = { intent: 'planning' };\n    const cacheable = opt?.cacheable;\n    // if vlm-ui-tars, plan cache is not used\n    const isVlmUiTars = vlLocateMode(modelPreferences) === 'vlm-ui-tars';\n    const matchedCache =\n      isVlmUiTars || cacheable === false\n        ? undefined\n        : this.taskCache?.matchPlanCache(taskPrompt);\n    if (matchedCache && this.taskCache?.isCacheResultUsed) {\n      // log into report file\n      const { executor } = await this.taskExecutor.loadYamlFlowAsPlanning(\n        taskPrompt,\n        matchedCache.cacheContent?.yamlWorkflow,\n      );\n\n      await this.afterTaskRunning(executor);\n\n      debug('matched cache, will call .runYaml to run the action');\n      const yaml = matchedCache.cacheContent?.yamlWorkflow;\n      return this.runYaml(yaml);\n    }\n\n    const { output, executor } = await (isVlmUiTars\n      ? this.taskExecutor.actionToGoal(taskPrompt)\n      : this.taskExecutor.action(taskPrompt, this.opts.aiActionContext));\n\n    // update cache\n    if (this.taskCache && output?.yamlFlow && cacheable !== false) {\n      const yamlContent: MidsceneYamlScript = {\n        tasks: [\n          {\n            name: taskPrompt,\n            flow: output.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    await this.afterTaskRunning(executor);\n    return output;\n  }\n\n  async aiQuery<ReturnType = any>(\n    demand: InsightExtractParam,\n    opt: InsightExtractOption = defaultInsightExtractOption,\n  ): Promise<ReturnType> {\n    const { output, executor } =\n      await this.taskExecutor.createTypeQueryExecution('Query', demand, opt);\n    await this.afterTaskRunning(executor);\n    return output as ReturnType;\n  }\n\n  async aiBoolean(\n    prompt: TUserPrompt,\n    opt: InsightExtractOption = defaultInsightExtractOption,\n  ): Promise<boolean> {\n    const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n    const { output, executor } =\n      await this.taskExecutor.createTypeQueryExecution(\n        'Boolean',\n        textPrompt,\n        opt,\n        multimodalPrompt,\n      );\n    await this.afterTaskRunning(executor);\n    return output as boolean;\n  }\n\n  async aiNumber(\n    prompt: TUserPrompt,\n    opt: InsightExtractOption = defaultInsightExtractOption,\n  ): Promise<number> {\n    const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n    const { output, executor } =\n      await this.taskExecutor.createTypeQueryExecution(\n        'Number',\n        textPrompt,\n        opt,\n        multimodalPrompt,\n      );\n    await this.afterTaskRunning(executor);\n    return output as number;\n  }\n\n  async aiString(\n    prompt: TUserPrompt,\n    opt: InsightExtractOption = defaultInsightExtractOption,\n  ): Promise<string> {\n    const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n    const { output, executor } =\n      await this.taskExecutor.createTypeQueryExecution(\n        'String',\n        textPrompt,\n        opt,\n        multimodalPrompt,\n      );\n    await this.afterTaskRunning(executor);\n    return output as string;\n  }\n\n  async aiAsk(\n    prompt: TUserPrompt,\n    opt: InsightExtractOption = defaultInsightExtractOption,\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      const text = await this.insight.describe(center, { deepThink });\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 { executor, output } = await this.taskExecutor.runPlans(\n      taskTitleStr('Locate', locateParamStr(locateParam)),\n      plans,\n    );\n    await this.afterTaskRunning(executor);\n\n    const { element } = output;\n\n    return {\n      rect: element?.rect,\n      center: element?.center,\n      scale: (await this.interface.size()).dpr,\n    } as Pick<LocateResultElement, 'rect' | 'center'> & {\n      scale: number;\n    };\n  }\n\n  async aiAssert(\n    assertion: TUserPrompt,\n    msg?: string,\n    opt?: AgentAssertOpt & InsightExtractOption,\n  ) {\n    const insightOpt: InsightExtractOption = {\n      domIncluded: opt?.domIncluded ?? defaultInsightExtractOption.domIncluded,\n      screenshotIncluded:\n        opt?.screenshotIncluded ??\n        defaultInsightExtractOption.screenshotIncluded,\n      returnThought: opt?.returnThought ?? true,\n      isWaitForAssert: opt?.isWaitForAssert,\n      doNotThrowError: opt?.doNotThrowError,\n    };\n\n    const { output, executor, thought } = await this.taskExecutor.assert(\n      assertion,\n      insightOpt,\n    );\n    await this.afterTaskRunning(executor, true);\n\n    const message = output\n      ? undefined\n      : `Assertion failed: ${msg || (typeof assertion === 'string' ? assertion : assertion.prompt)}\\nReason: ${\n          thought || executor.latestErrorTask()?.error || '(no_reason)'\n        }`;\n\n    if (opt?.keepRawResponse) {\n      return {\n        pass: output,\n        thought,\n        message,\n      };\n    }\n\n    if (!output) {\n      throw new Error(message);\n    }\n  }\n\n  async aiWaitFor(assertion: TUserPrompt, opt?: AgentWaitForOpt) {\n    const { executor } = await this.taskExecutor.waitFor(assertion, {\n      timeoutMs: opt?.timeoutMs || 15 * 1000,\n      checkIntervalMs: opt?.checkIntervalMs || 3 * 1000,\n    });\n    await this.afterTaskRunning(executor, true);\n\n    if (executor.isInErrorState()) {\n      const errorTask = executor.latestErrorTask();\n      throw new Error(`${errorTask?.error}\\n${errorTask?.errorStack}`);\n    }\n  }\n\n  async ai(taskPrompt: string, type = 'action') {\n    if (type === 'action') {\n      return this.aiAction(taskPrompt);\n    }\n    if (type === 'query') {\n      return this.aiQuery(taskPrompt);\n    }\n\n    if (type === 'assert') {\n      return this.aiAssert(taskPrompt);\n    }\n\n    if (type === 'tap') {\n      return this.aiTap(taskPrompt);\n    }\n\n    if (type === 'rightClick') {\n      return this.aiRightClick(taskPrompt);\n    }\n\n    if (type === 'doubleClick') {\n      return this.aiDoubleClick(taskPrompt);\n    }\n\n    throw new Error(\n      `Unknown type: ${type}, only support 'action', 'query', 'assert', 'tap', 'rightClick', 'doubleClick'`,\n    );\n  }\n\n  async runYaml(yamlScriptContent: string): Promise<{\n    result: Record<string, any>;\n  }> {\n    const script = parseYamlScript(yamlScriptContent, 'yaml', true);\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  async destroy() {\n    await this.interface.destroy?.();\n    this.resetDump(); // reset dump to release memory\n    this.destroyed = true;\n  }\n\n  async logScreenshot(\n    title?: string,\n    opt?: {\n      content: string;\n    },\n  ) {\n    // 1. screenshot\n    const base64 = await this.interface.screenshotBase64();\n    const now = Date.now();\n    // 2. build recorder\n    const recorder: ExecutionRecorderItem[] = [\n      {\n        type: 'screenshot',\n        ts: now,\n        screenshot: base64,\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: ExecutionDump = {\n      sdkVersion: '',\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    try {\n      this.onDumpUpdate?.(this.dumpDataString());\n    } catch (error) {\n      console.error('Failed to update dump', error);\n    }\n\n    this.writeOutActionDumps();\n  }\n\n  _unstableLogContent() {\n    const { groupName, groupDescription, executions } = this.dump;\n    const newExecutions = Array.isArray(executions)\n      ? executions.map((execution: any) => {\n          const { tasks, ...restExecution } = execution;\n          let newTasks = tasks;\n          if (Array.isArray(tasks)) {\n            newTasks = tasks.map((task: any) => {\n              // only remove uiContext and log from task\n              const { uiContext, log, ...restTask } = task;\n              return restTask;\n            });\n          }\n          return { ...restExecution, ...(newTasks ? { tasks: newTasks } : {}) };\n        })\n      : [];\n    return {\n      groupName,\n      groupDescription,\n      executions: newExecutions,\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\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","defaultInsightExtractOption","Agent","action","commonContextParser","prompt","execution","trimmedExecution","trimContextByViewport","currentDump","stringifyDumpData","reportHTMLContent","Error","generateReport","autoPrintReportMsg","writeLogFile","groupedActionDumpFileExt","printReportMsg","task","param","paramStr","tip","typeStr","executor","doNotThrowError","error","console","errorTask","type","opt","actionPlan","plans","Boolean","title","taskTitleStr","locateParamStr","output","locatePrompt","assert","detailedLocateParam","buildDetailedLocateParam","locatePromptOrValue","locatePromptOrOpt","optOrUndefined","value","optWithValue","locatePromptOrKeyName","keyName","undefined","locatePromptOrScrollParam","scrollParam","taskPrompt","_this_taskCache","_this_taskCache1","modelPreferences","cacheable","isVlmUiTars","vlLocateMode","matchedCache","_matchedCache_cacheContent","_matchedCache_cacheContent1","yaml","yamlContent","yamlFlowStr","demand","textPrompt","multimodalPrompt","parsePrompt","center","verifyPrompt","retryLimit","success","retryCount","resultPrompt","deepThink","verifyResult","text","locateOpt","expectCenter","verifyLocateOption","verifyCenter","verifyRect","distance","included","pass","locateParam","locatePlan","locatePlanForLocate","element","assertion","msg","_executor_latestErrorTask","insightOpt","thought","message","yamlScriptContent","script","parseYamlScript","player","ScriptPlayer","errors","_task_error","_this_interface","base64","now","Date","recorder","executionDump","_this","groupName","groupDescription","executions","newExecutions","Array","tasks","restExecution","newTasks","uiContext","log","restTask","context","interfaceInstance","opts","Object","globalConfigManager","Insight","TaskCache","MIDSCENE_CACHE","TaskExecutor","getReportFileName","createAgent"],"mappings":";;;;;;;;;;;;;;;;;;;;;AAiEA,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;AAEO,MAAMC;IAoCX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IA4CA,MAAM,iBAA0C;QAC9C,OAAO,IAAI,CAAC,SAAS,CAAC,WAAW;IACnC;IAEA,MAAM,aAAaC,MAAsB,EAAsB;QAE7D,IAAI,IAAI,CAAC,eAAe,EAAE;YACxBrB,MAAM,yCAAyCqB;YAC/C,OAAO,IAAI,CAAC,eAAe;QAC7B;QAEA,IAAI,IAAI,CAAC,SAAS,CAAC,UAAU,EAAE;YAC7BrB,MAAM,qCAAqCqB;YAC3C,OAAO,MAAM,IAAI,CAAC,SAAS,CAAC,UAAU;QACxC;QACErB,MAAM,yCAAyCqB;QAC/C,OAAO,MAAMC,oBAAoB,IAAI,CAAC,SAAS;IAEnD;IAEA,MAAM,mBAAuC;QAC3C,OAAO,MAAM,IAAI,CAAC,YAAY,CAAC;IACjC;IAEA,MAAM,mBAAmBC,MAAc,EAAE;QACvC,IAAI,CAAC,IAAI,CAAC,eAAe,GAAGA;IAC9B;IAEA,YAAY;QACV,IAAI,CAAC,IAAI,GAAG;YACV,WAAW,IAAI,CAAC,IAAI,CAAC,SAAS;YAC9B,kBAAkB,IAAI,CAAC,IAAI,CAAC,gBAAgB;YAC5C,YAAY,EAAE;YACd,aAAa,EAAE;QACjB;QAEA,OAAO,IAAI,CAAC,IAAI;IAClB;IAEA,oBAAoBC,SAAwB,EAAE;QAE5C,MAAMC,mBAAmBC,sBAAsBF;QAC/C,MAAMG,cAAc,IAAI,CAAC,IAAI;QAC7BA,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,OAAOG,kBAAkB,IAAI,CAAC,IAAI;IACpC;IAEA,mBAAmB;QACjB,OAAOC,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;QACA/B,MAAM,uBAAuB,IAAI,CAAC,UAAU;QAC5C,IAAI+B,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,MAAc,iBAAiBE,QAAkB,EAAEC,kBAAkB,KAAK,EAAE;QAC1E,IAAI,CAAC,mBAAmB,CAACD,SAAS,IAAI;QAEtC,IAAI;YACF,IAAI,IAAI,CAAC,YAAY,EACnB,IAAI,CAAC,YAAY,CAAC,IAAI,CAAC,cAAc;QAEzC,EAAE,OAAOE,OAAO;YACdC,QAAQ,KAAK,CAAC,yBAAyBD;QACzC;QAEA,IAAI,CAAC,mBAAmB;QAExB,IAAIF,SAAS,cAAc,MAAM,CAACC,iBAAiB;YACjD,MAAMG,YAAYJ,SAAS,eAAe;YAC1C,MAAM,IAAIX,MAAM,GAAGe,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,YAAY,CAAC,EAAE,EAAEA,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,UAAU,EAAE,EAAE;gBACtE,OAAOA,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,KAAK;YACzB;QACF;IACF;IAEA,MAAM,wBACJC,IAAY,EACZC,GAAO,EACP;QACA/C,MAAM,2BAA2B8C,MAAM,KAAKC,KAAK,KAAKA;QAEtD,MAAMC,aAAgC;YACpC,MAAMF;YACN,OAAQC,OAAe,CAAC;YACxB,SAAS;QACX;QACA/C,MAAM,cAAcgD;QAEpB,MAAMC,QAA0B;YAACD;SAAW,CAAC,MAAM,CACjDE;QAGF,MAAMC,QAAQC,aACZN,MACAO,eAAe,AAACN,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAa,MAAM,AAAD,KAAK,CAAC;QAG1C,MAAM,EAAEO,MAAM,EAAEb,QAAQ,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,CAACU,OAAOF;QACrE,MAAM,IAAI,CAAC,gBAAgB,CAACR;QAC5B,OAAOa;IACT;IAEA,MAAM,MAAMC,YAAyB,EAAER,GAAkB,EAAE;QACzDS,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcR;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,OAAO;YACzC,QAAQU;QACV;IACF;IAEA,MAAM,aAAaF,YAAyB,EAAER,GAAkB,EAAE;QAChES,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcR;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,cAAc;YAChD,QAAQU;QACV;IACF;IAEA,MAAM,cAAcF,YAAyB,EAAER,GAAkB,EAAE;QACjES,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcR;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,eAAe;YACjD,QAAQU;QACV;IACF;IAEA,MAAM,QAAQF,YAAyB,EAAER,GAAkB,EAAE;QAC3DS,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcR;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,SAAS;YAC3C,QAAQU;QACV;IACF;IAmBA,MAAM,QACJE,mBAAyC,EACzCC,iBAGa,EACbC,cAA6B,EAC7B;QACA,IAAIC;QACJ,IAAIP;QACJ,IAAIR;QAKJ,IACE,AAA6B,YAA7B,OAAOa,qBACPA,AAAsB,SAAtBA,qBACA,WAAWA,mBACX;YAEAL,eAAeI;YACf,MAAMI,eAAeH;YAIrBE,QAAQC,aAAa,KAAK;YAC1BhB,MAAMgB;QACR,OAAO;YAELD,QAAQH;YACRJ,eAAeK;YACfb,MAAM;gBACJ,GAAGc,cAAc;gBACjBC;YACF;QACF;QAEAN,OACE,AAAiB,YAAjB,OAAOM,OACP;QAEFN,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcR;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,SAAS;YAC3C,GAAIA,OAAO,CAAC,CAAC;YACb,QAAQU;QACV;IACF;IAmBA,MAAM,gBACJO,qBAA2C,EAC3CJ,iBAGa,EACbC,cAA6B,EAC7B;QACA,IAAII;QACJ,IAAIV;QACJ,IAAIR;QAGJ,IACE,AAA6B,YAA7B,OAAOa,qBACPA,AAAsB,SAAtBA,qBACA,aAAaA,mBACb;YAEAL,eAAeS;YACfjB,MAAMa;QAGR,OAAO;YAELK,UAAUD;YACVT,eAAeK;YACfb,MAAM;gBACJ,GAAIc,kBAAkB,CAAC,CAAC;gBACxBI;YACF;QACF;QAEAT,OAAOT,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,EAAE;QAErB,MAAMU,sBAAsBF,eACxBG,yBAAyBH,cAAcR,OACvCmB;QAEJ,OAAO,IAAI,CAAC,uBAAuB,CAAC,iBAAiB;YACnD,GAAInB,OAAO,CAAC,CAAC;YACb,QAAQU;QACV;IACF;IAmBA,MAAM,SACJU,yBAAgE,EAChEP,iBAAyE,EACzEC,cAA6B,EAC7B;QACA,IAAIO;QACJ,IAAIb;QACJ,IAAIR;QAGJ,IACE,AAA6B,YAA7B,OAAOa,qBACN,gBAAeA,qBACd,gBAAgBA,qBAChB,cAAcA,iBAAgB,GAChC;YAEAL,eAAeY;YACfpB,MAAMa;QACR,OAAO;YAELQ,cAAcD;YACdZ,eAAeK;YACfb,MAAM;gBACJ,GAAIc,kBAAkB,CAAC,CAAC;gBACxB,GAAIO,eAAe,CAAC,CAAC;YACvB;QACF;QAEA,MAAMX,sBAAsBC,yBAC1BH,gBAAgB,IAChBR;QAGF,OAAO,IAAI,CAAC,uBAAuB,CAAC,UAAU;YAC5C,GAAIA,OAAO,CAAC,CAAC;YACb,QAAQU;QACV;IACF;IA4BA,MAAM,SACJY,UAAkB,EAClBtB,GAEC,EACD;YAQMuB,iBACcC;QARpB,MAAMC,mBAAsC;YAAE,QAAQ;QAAW;QACjE,MAAMC,YAAY1B,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS;QAEhC,MAAM2B,cAAcC,AAAmC,kBAAnCA,aAAaH;QACjC,MAAMI,eACJF,eAAeD,AAAc,UAAdA,YACXP,SAAAA,QACAI,CAAAA,kBAAAA,IAAI,CAAC,SAAS,AAAD,IAAbA,KAAAA,IAAAA,gBAAgB,cAAc,CAACD;QACrC,IAAIO,gBAAAA,SAAgBL,CAAAA,mBAAAA,IAAI,CAAC,SAAS,AAAD,IAAbA,KAAAA,IAAAA,iBAAgB,iBAAiB,AAAD,GAAG;gBAInDM,4BAMWC;YARb,MAAM,EAAErC,QAAQ,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,sBAAsB,CACjE4B,YAAAA,QACAQ,CAAAA,6BAAAA,aAAa,YAAY,AAAD,IAAxBA,KAAAA,IAAAA,2BAA2B,YAAY;YAGzC,MAAM,IAAI,CAAC,gBAAgB,CAACpC;YAE5BzC,MAAM;YACN,MAAM+E,OAAO,QAAAD,CAAAA,8BAAAA,aAAa,YAAY,AAAD,IAAxBA,KAAAA,IAAAA,4BAA2B,YAAY;YACpD,OAAO,IAAI,CAAC,OAAO,CAACC;QACtB;QAEA,MAAM,EAAEzB,MAAM,EAAEb,QAAQ,EAAE,GAAG,MAAOiC,CAAAA,cAChC,IAAI,CAAC,YAAY,CAAC,YAAY,CAACL,cAC/B,IAAI,CAAC,YAAY,CAAC,MAAM,CAACA,YAAY,IAAI,CAAC,IAAI,CAAC,eAAe;QAGlE,IAAI,IAAI,CAAC,SAAS,IAAIf,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,QAAQ,AAAD,KAAKmB,AAAc,UAAdA,WAAqB;YAC7D,MAAMO,cAAkC;gBACtC,OAAO;oBACL;wBACE,MAAMX;wBACN,MAAMf,OAAO,QAAQ;oBACvB;iBACD;YACH;YACA,MAAM2B,cAAcF,QAAAA,IAAS,CAACC;YAC9B,IAAI,CAAC,SAAS,CAAC,yBAAyB,CACtC;gBACE,MAAM;gBACN,QAAQX;gBACR,cAAcY;YAChB,GACAL;QAEJ;QAEA,MAAM,IAAI,CAAC,gBAAgB,CAACnC;QAC5B,OAAOa;IACT;IAEA,MAAM,QACJ4B,MAA2B,EAC3BnC,MAA4B5B,2BAA2B,EAClC;QACrB,MAAM,EAAEmC,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAAC,SAASyC,QAAQnC;QACpE,MAAM,IAAI,CAAC,gBAAgB,CAACN;QAC5B,OAAOa;IACT;IAEA,MAAM,UACJ/B,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACrC;QAClB,MAAM,EAAEgE,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAY9D;QACrD,MAAM,EAAE+B,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,WACA0C,YACApC,KACAqC;QAEJ,MAAM,IAAI,CAAC,gBAAgB,CAAC3C;QAC5B,OAAOa;IACT;IAEA,MAAM,SACJ/B,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACtC;QACjB,MAAM,EAAEgE,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAY9D;QACrD,MAAM,EAAE+B,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,UACA0C,YACApC,KACAqC;QAEJ,MAAM,IAAI,CAAC,gBAAgB,CAAC3C;QAC5B,OAAOa;IACT;IAEA,MAAM,SACJ/B,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACtC;QACjB,MAAM,EAAEgE,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAY9D;QACrD,MAAM,EAAE+B,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,UACA0C,YACApC,KACAqC;QAEJ,MAAM,IAAI,CAAC,gBAAgB,CAAC3C;QAC5B,OAAOa;IACT;IAEA,MAAM,MACJ/B,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACtC;QACjB,OAAO,IAAI,CAAC,QAAQ,CAACI,QAAQwB;IAC/B;IAEA,MAAM,uBACJuC,MAAwB,EACxBvC,GAI0B,EACkB;QAC5C,MAAM,EAAEwC,eAAe,IAAI,EAAEC,aAAa,CAAC,EAAE,GAAGzC,OAAO,CAAC;QAExD,IAAI0C,UAAU;QACd,IAAIC,aAAa;QACjB,IAAIC,eAAe;QACnB,IAAIC,YAAY7C,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,AAAD,KAAK;QAClC,IAAI8C;QAEJ,MAAO,CAACJ,WAAWC,aAAaF,WAAY;YAC1C,IAAIE,cAAc,GAChBE,YAAY;YAEd5F,MACE,cACAsF,QACA,gBACAC,cACA,cACAG,YACA,aACAE;YAEF,MAAME,OAAO,MAAM,IAAI,CAAC,OAAO,CAAC,QAAQ,CAACR,QAAQ;gBAAEM;YAAU;YAC7D5F,MAAM,mBAAmB8F;YACzBtC,OAAOsC,KAAK,WAAW,EAAE,CAAC,+BAA+B,EAAER,OAAO,CAAC,CAAC;YACpEK,eAAeG,KAAK,WAAW;YAE/BD,eAAe,MAAM,IAAI,CAAC,aAAa,CACrCF,cACAC,YAAY;gBAAE,WAAW;YAAK,IAAI1B,QAClCoB,QACAvC;YAEF,IAAI8C,aAAa,IAAI,EACnBJ,UAAU;iBAEVC;QAEJ;QAEA,OAAO;YACL,QAAQC;YACRC;YACAC;QACF;IACF;IAEA,MAAM,cACJtE,MAAc,EACdwE,SAAmC,EACnCC,YAA8B,EAC9BC,kBAA2C,EACX;QAChCjG,MAAM,iBAAiBuB,QAAQwE,WAAWC,cAAcC;QAExD,MAAM,EAAE,QAAQC,YAAY,EAAE,MAAMC,UAAU,EAAE,GAAG,MAAM,IAAI,CAAC,QAAQ,CACpE5E,QACAwE;QAEF,MAAMK,WAAWlG,oBAAoB8F,cAAcE;QACnD,MAAMG,WAAW3F,eAAesF,cAAcG;QAC9C,MAAMG,OACJF,YAAaH,CAAAA,CAAAA,QAAAA,qBAAAA,KAAAA,IAAAA,mBAAoB,uBAAuB,AAAD,KAAK,EAAC,KAC7DI;QACF,MAAMR,eAAe;YACnBS;YACA,MAAMH;YACN,QAAQD;YACR,gBAAgBE;QAClB;QACApG,MAAM,2BAA2B6F;QACjC,OAAOA;IACT;IAEA,MAAM,SAAStE,MAAmB,EAAEwB,GAAkB,EAAE;QACtD,MAAMwD,cAAc7C,yBAAyBnC,QAAQwB;QACrDS,OAAO+C,aAAa;QACpB,MAAMC,aAAaC,oBAAoBF;QACvC,MAAMtD,QAAQ;YAACuD;SAAW;QAC1B,MAAM,EAAE/D,QAAQ,EAAEa,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,CAC3DF,aAAa,UAAUC,eAAekD,eACtCtD;QAEF,MAAM,IAAI,CAAC,gBAAgB,CAACR;QAE5B,MAAM,EAAEiE,OAAO,EAAE,GAAGpD;QAEpB,OAAO;YACL,MAAMoD,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,IAAI;YACnB,QAAQA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM;YACvB,OAAQ,OAAM,IAAI,CAAC,SAAS,CAAC,IAAI,EAAC,EAAG,GAAG;QAC1C;IAGF;IAEA,MAAM,SACJC,SAAsB,EACtBC,GAAY,EACZ7D,GAA2C,EAC3C;YAoBiB8D;QAnBjB,MAAMC,aAAmC;YACvC,aAAa/D,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,KAAK5B,4BAA4B,WAAW;YACxE,oBACE4B,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,kBAAkB,AAAD,KACtB5B,4BAA4B,kBAAkB;YAChD,eAAe4B,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,aAAa,AAAD,KAAK;YACrC,iBAAiBA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe;YACrC,iBAAiBA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe;QACvC;QAEA,MAAM,EAAEO,MAAM,EAAEb,QAAQ,EAAEsE,OAAO,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,MAAM,CAClEJ,WACAG;QAEF,MAAM,IAAI,CAAC,gBAAgB,CAACrE,UAAU;QAEtC,MAAMuE,UAAU1D,SACZY,SACA,CAAC,kBAAkB,EAAE0C,OAAQ,CAAqB,YAArB,OAAOD,YAAyBA,YAAYA,UAAU,MAAK,EAAG,UAAU,EACnGI,WAAAA,SAAWF,CAAAA,4BAAAA,SAAS,eAAe,EAAC,IAAzBA,KAAAA,IAAAA,0BAA4B,KAAK,AAAD,KAAK,eAChD;QAEN,IAAI9D,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,EACtB,OAAO;YACL,MAAMO;YACNyD;YACAC;QACF;QAGF,IAAI,CAAC1D,QACH,MAAM,IAAIxB,MAAMkF;IAEpB;IAEA,MAAM,UAAUL,SAAsB,EAAE5D,GAAqB,EAAE;QAC7D,MAAM,EAAEN,QAAQ,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,OAAO,CAACkE,WAAW;YAC9D,WAAW5D,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,AAAD,KAAK;YAC7B,iBAAiBA,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,AAAD,KAAK;QAC3C;QACA,MAAM,IAAI,CAAC,gBAAgB,CAACN,UAAU;QAEtC,IAAIA,SAAS,cAAc,IAAI;YAC7B,MAAMI,YAAYJ,SAAS,eAAe;YAC1C,MAAM,IAAIX,MAAM,GAAGe,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,KAAK,CAAC,EAAE,EAAEA,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,UAAU,EAAE;QACjE;IACF;IAEA,MAAM,GAAGwB,UAAkB,EAAEvB,OAAO,QAAQ,EAAE;QAC5C,IAAIA,AAAS,aAATA,MACF,OAAO,IAAI,CAAC,QAAQ,CAACuB;QAEvB,IAAIvB,AAAS,YAATA,MACF,OAAO,IAAI,CAAC,OAAO,CAACuB;QAGtB,IAAIvB,AAAS,aAATA,MACF,OAAO,IAAI,CAAC,QAAQ,CAACuB;QAGvB,IAAIvB,AAAS,UAATA,MACF,OAAO,IAAI,CAAC,KAAK,CAACuB;QAGpB,IAAIvB,AAAS,iBAATA,MACF,OAAO,IAAI,CAAC,YAAY,CAACuB;QAG3B,IAAIvB,AAAS,kBAATA,MACF,OAAO,IAAI,CAAC,aAAa,CAACuB;QAG5B,MAAM,IAAIvC,MACR,CAAC,cAAc,EAAEgB,KAAK,8EAA8E,CAAC;IAEzG;IAEA,MAAM,QAAQmE,iBAAyB,EAEpC;QACD,MAAMC,SAASC,gBAAgBF,mBAAmB,QAAQ;QAC1D,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,CAAChF,OAASA,AAAgB,YAAhBA,KAAK,MAAM,EAC5B,GAAG,CAAC,CAACA;oBAC2BmF;gBAA/B,OAAO,CAAC,OAAO,EAAEnF,KAAK,IAAI,CAAC,EAAE,EAAE,QAAAmF,CAAAA,cAAAA,KAAK,KAAK,AAAD,IAATA,KAAAA,IAAAA,YAAY,OAAO,EAAE;YACtD,GACC,IAAI,CAAC;YACR,MAAM,IAAIzF,MAAM,CAAC,2CAA2C,EAAEwF,QAAQ;QACxE;QAEA,OAAO;YACL,QAAQF,OAAO,MAAM;QACvB;IACF;IAEA,MAAM,mBAAmBF,MAAc,EAAE;QACvC1D,OACE,IAAI,CAAC,SAAS,CAAC,kBAAkB,EACjC;QAEF,OAAO,IAAI,CAAC,SAAS,CAAC,kBAAkB,CAAC0D;IAC3C;IAEA,MAAM,UAAU;YACRM,yBAAAA;QAAN,eAAMA,CAAAA,0BAAAA,AAAAA,CAAAA,kBAAAA,IAAI,CAAC,SAAS,AAAD,EAAE,OAAO,AAAD,IAArBA,KAAAA,IAAAA,wBAAAA,IAAAA,CAAAA,gBAAAA;QACN,IAAI,CAAC,SAAS;QACd,IAAI,CAAC,SAAS,GAAG;IACnB;IAEA,MAAM,cACJrE,KAAc,EACdJ,GAEC,EACD;QAEA,MAAM0E,SAAS,MAAM,IAAI,CAAC,SAAS,CAAC,gBAAgB;QACpD,MAAMC,MAAMC,KAAK,GAAG;QAEpB,MAAMC,WAAoC;YACxC;gBACE,MAAM;gBACN,IAAIF;gBACJ,YAAYD;YACd;SACD;QAED,MAAMrF,OAAyB;YAC7B,MAAM;YACN,SAAS;YACT,QAAQ;YACRwF;YACA,QAAQ;gBACN,OAAOF;gBACP,KAAKA;gBACL,MAAM;YACR;YACA,OAAO;gBACL,SAAS3E,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAK;YAC3B;YACA,UAAU,WAAa;QACzB;QAEA,MAAM8E,gBAA+B;YACnC,YAAY;YACZ,SAASH;YACT,MAAM,CAAC,MAAM,EAAEvE,SAAS,YAAY;YACpC,aAAaJ,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAK;YAC7B,OAAO;gBAACX;aAAK;QACf;QAEA,IAAI,CAAC,mBAAmB,CAACyF;QAEzB,IAAI;gBACFC,oBAAAA;oBAAAA,CAAAA,qBAAAA,AAAAA,CAAAA,QAAAA,IAAI,AAAD,EAAE,YAAY,AAAD,KAAhBA,mBAAAA,IAAAA,CAAAA,OAAoB,IAAI,CAAC,cAAc;QACzC,EAAE,OAAOnF,OAAO;YACdC,QAAQ,KAAK,CAAC,yBAAyBD;QACzC;QAEA,IAAI,CAAC,mBAAmB;IAC1B;IAEA,sBAAsB;QACpB,MAAM,EAAEoF,SAAS,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAG,IAAI,CAAC,IAAI;QAC7D,MAAMC,gBAAgBC,MAAM,OAAO,CAACF,cAChCA,WAAW,GAAG,CAAC,CAACzG;YACd,MAAM,EAAE4G,KAAK,EAAE,GAAGC,eAAe,GAAG7G;YACpC,IAAI8G,WAAWF;YACf,IAAID,MAAM,OAAO,CAACC,QAChBE,WAAWF,MAAM,GAAG,CAAC,CAAChG;gBAEpB,MAAM,EAAEmG,SAAS,EAAEC,GAAG,EAAE,GAAGC,UAAU,GAAGrG;gBACxC,OAAOqG;YACT;YAEF,OAAO;gBAAE,GAAGJ,aAAa;gBAAE,GAAIC,WAAW;oBAAE,OAAOA;gBAAS,IAAI,CAAC,CAAC;YAAE;QACtE,KACA,EAAE;QACN,OAAO;YACLP;YACAC;YACA,YAAYE;QACd;IACF;IAMA,MAAM,oBAAmC;QACvClI,MAAM;QACN,MAAM0I,UAAU,MAAM,IAAI,CAAC,gBAAgB;QAE3CA,QAAQ,SAAS,GAAG;QACpB,IAAI,CAAC,eAAe,GAAGA;QACvB1I,MAAM;IACR;IAKA,MAAM,sBAAqC;QACzCA,MAAM;QACN,IAAI,CAAC,eAAe,GAAGkE;QACvBlE,MAAM;IACR;IAx2BA,YAAY2I,iBAAgC,EAAEC,IAAe,CAAE;QArC/D;QAEA;QAEA;QAEA;QAEA;QAEA;QAEA;QAKA,kCAAU;QAEV;QAEA;QAEA;QAEA,oCAAY;QAKZ,uBAAQ,mBAAR;QAQE,IAAI,CAAC,SAAS,GAAGD;QACjB,IAAI,CAAC,IAAI,GAAGE,OAAO,MAAM,CACvB;YACE,gBAAgB;YAChB,oBAAoB;YACpB,WAAW;YACX,kBAAkB;QACpB,GACAD,QAAQ,CAAC;QAEX,IAAIA,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW,AAAD,KAAK,AAA6B,cAA7B,OAAOA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW,AAAD,GAC9C,MAAM,IAAI9G,MACR,CAAC,+DAA+D,EAAE,OAAO8G,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW,AAAD,GAAG;QAIhGE,oBAAoB,IAAI,CAACF,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW;QAE1C,IAAI,CAAC,cAAc,GAAG,IAAI,CAAC,IAAI,CAAC,cAAc;QAE9C,IAAI,CAAC,OAAO,GAAG,IAAIG,QAAQ,OAAO1H,SACzB,IAAI,CAAC,YAAY,CAACA;QAG3B,IAAIuH,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,OAAO,EACf,IAAI,CAAC,SAAS,GAAG,IAAII,UACnBJ,KAAK,OAAO,EACZE,oBAAoB,qBAAqB,CAACG;QAI9C,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;QACpD;QACA,IAAI,CAAC,IAAI,GAAG,IAAI,CAAC,SAAS;QAC1B,IAAI,CAAC,cAAc,GACjBN,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,cAAc,AAAD,KACnBO,kBAAkBP,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,MAAM,AAAD,KAAK,IAAI,CAAC,SAAS,CAAC,aAAa,IAAI;IACtE;AAi0BF;AAEO,MAAMQ,cAAc,CACzBT,mBACAC,OAEO,IAAIxH,MAAMuH,mBAAmBC"}

package/dist/es/agent/index.mjs

@@ -1,6 +1,6 @@
-import { Agent } from "./agent.mjs";
+import { Agent, createAgent } from "./agent.mjs";
 import { commonContextParser, getCurrentExecutionFile, getReportFileName, printReportMsg, trimContextByViewport } from "./utils.mjs";
 import { locateParamStr, paramStr, taskTitleStr, typeStr } from "./ui-utils.mjs";
 import { TaskCache, cacheFileExt } from "./task-cache.mjs";
 import { TaskExecutor } from "./tasks.mjs";
-export { Agent, TaskCache, TaskExecutor, cacheFileExt, commonContextParser, getCurrentExecutionFile, getReportFileName, locateParamStr, paramStr, printReportMsg, taskTitleStr, trimContextByViewport, typeStr };
+export { Agent, TaskCache, TaskExecutor, cacheFileExt, commonContextParser, createAgent, getCurrentExecutionFile, getReportFileName, locateParamStr, paramStr, printReportMsg, taskTitleStr, trimContextByViewport, typeStr };

package/dist/es/agent/utils.mjs

@@ -138,7 +138,7 @@ function trimContextByViewport(execution) {
         }) : execution.tasks
     };
 }
-const getMidsceneVersion = ()=>"0.27.7-beta-20250908115945.0";
+const getMidsceneVersion = ()=>"0.27.7-beta-20250909031245.0";
 const parsePrompt = (prompt)=>{
     if ('string' == typeof prompt) return {
         textPrompt: prompt,

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

@@ -1,6 +1,5 @@
 import node_assert from "node:assert";
 import { PromptTemplate } from "@langchain/core/prompts";
-import { z } from "zod";
 import { ifMidsceneLocatorField } from "../common.mjs";
 import { bboxDescription } from "./common.mjs";
 const vlCoTLog = '"what_the_user_wants_to_do_next_by_instruction": string, // What the user wants to do according to the instruction and previous logs. ';
@@ -15,7 +14,7 @@ const descriptionForAction = (action, locatorSchemaTypeDescription)=>{
     const fields = [];
     fields.push(`- type: "${action.name}"`);
     if (action.paramSchema) {
-        node_assert(action.paramSchema instanceof z.ZodObject, 'paramSchema must be a zod object');
+        node_assert('ZodObject' === action.paramSchema.constructor.name, 'paramSchema must be a zod object');
         const shape = action.paramSchema.shape;
         const paramLines = [];
         const getTypeName = (field)=>{

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

@@ -1 +1 @@
-{"version":3,"file":"ai-model/prompt/llm-planning.mjs","sources":["webpack://@midscene/core/./src/ai-model/prompt/llm-planning.ts"],"sourcesContent":["import assert from 'node:assert';\nimport type { DeviceAction } from '@/types';\nimport { PromptTemplate } from '@langchain/core/prompts';\nimport type { vlLocateMode } from '@midscene/shared/env';\nimport type { ResponseFormatJSONSchema } from 'openai/resources/index';\nimport { z } from 'zod';\nimport { ifMidsceneLocatorField } from '../common';\nimport { bboxDescription } from './common';\n\n// Note: put the log field first to trigger the CoT\nconst vlCoTLog = `\"what_the_user_wants_to_do_next_by_instruction\": string, // What the user wants to do according to the instruction and previous logs. `;\nconst vlCurrentLog = `\"log\": string, // Log what the next one action (ONLY ONE!) you can do according to the screenshot and the instruction. The typical log looks like \"Now i want to use action '{ action-type }' to do .. first\". If no action should be done, log the reason. \". Use the same language as the user's instruction.`;\nconst llmCurrentLog = `\"log\": string, // Log what the next actions you can do according to the screenshot and the instruction. The typical log looks like \"Now i want to use action '{ action-type }' to do ..\". If no action should be done, log the reason. \". Use the same language as the user's instruction.`;\n\nconst commonOutputFields = `\"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 the same language as the user's instruction.\n  \"more_actions_needed_by_instruction\": boolean, // Consider if there is still more action(s) to do after the action in \"Log\" is done, according to the instruction. If so, set this field to true. Otherwise, set it to false.`;\nconst vlLocateParam = () =>\n  '{bbox: [number, number, number, number], prompt: string }';\nconst llmLocateParam = () => '{\"id\": string, \"prompt\": string}';\n\nexport const descriptionForAction = (\n  action: DeviceAction<any>,\n  locatorSchemaTypeDescription: string,\n) => {\n  const tab = '  ';\n  const fields: string[] = [];\n\n  // Add the action type field\n  fields.push(`- type: \"${action.name}\"`);\n\n  // Handle paramSchema if it exists\n  if (action.paramSchema) {\n    assert(\n      action.paramSchema instanceof z.ZodObject,\n      'paramSchema must be a zod object',\n    );\n    // Try to extract parameter information from the zod schema\n    // For zod object schemas, extract type information and descriptions\n    const shape = action.paramSchema.shape;\n    const paramLines: string[] = [];\n\n    // Helper function to get type name from zod schema\n    const getTypeName = (field: any): string => {\n      // Recursively unwrap optional, nullable, and other wrapper types to get the actual inner type\n      const unwrapField = (f: any): any => {\n        if (!f._def) return f;\n\n        const typeName = f._def.typeName;\n\n        // Handle wrapper types that have innerType\n        if (\n          typeName === 'ZodOptional' ||\n          typeName === 'ZodNullable' ||\n          typeName === 'ZodDefault'\n        ) {\n          return unwrapField(f._def.innerType);\n        }\n\n        return f;\n      };\n\n      const actualField = unwrapField(field);\n      const fieldTypeName = actualField._def?.typeName;\n\n      if (fieldTypeName === 'ZodString') return 'string';\n      if (fieldTypeName === 'ZodNumber') return 'number';\n      if (fieldTypeName === 'ZodBoolean') return 'boolean';\n      if (fieldTypeName === 'ZodArray') return 'array';\n      if (fieldTypeName === 'ZodObject') {\n        // Check if this is a passthrough object (like MidsceneLocation)\n        if (ifMidsceneLocatorField(actualField)) {\n          return locatorSchemaTypeDescription;\n        }\n        return 'object';\n      }\n      if (fieldTypeName === 'ZodEnum') {\n        const values =\n          (actualField._def?.values as unknown[] | undefined)\n            ?.map((option: unknown) => String(`'${option}'`))\n            .join(', ') ?? 'enum';\n\n        return `enum(${values})`;\n      }\n\n      console.warn(\n        'failed to parse Zod type. This may lead to wrong params from the LLM.\\n',\n        actualField._def,\n      );\n      return actualField.toString();\n    };\n\n    // Helper function to get description from zod schema\n    const getDescription = (field: z.ZodTypeAny): string | null => {\n      // Recursively unwrap optional, nullable, and other wrapper types to get the actual inner type\n      const unwrapField = (f: any): any => {\n        if (!f._def) return f;\n\n        const typeName = f._def.typeName;\n\n        // Handle wrapper types that have innerType\n        if (\n          typeName === 'ZodOptional' ||\n          typeName === 'ZodNullable' ||\n          typeName === 'ZodDefault'\n        ) {\n          return unwrapField(f._def.innerType);\n        }\n\n        return f;\n      };\n\n      // Check for direct description on the original field (wrapper may have description)\n      if ('description' in field) {\n        return field.description || null;\n      }\n\n      const actualField = unwrapField(field);\n\n      // Check for description on the unwrapped field\n      if ('description' in actualField) {\n        return actualField.description || null;\n      }\n\n      // Check for MidsceneLocation fields and add description\n      if (actualField._def?.typeName === 'ZodObject') {\n        if ('midscene_location_field_flag' in actualField._def.shape()) {\n          return 'Location information for the target element';\n        }\n      }\n\n      return null;\n    };\n\n    for (const [key, field] of Object.entries(shape)) {\n      if (field && typeof field === 'object') {\n        // Check if field is optional\n        const isOptional =\n          typeof (field as any).isOptional === 'function' &&\n          (field as any).isOptional();\n        const keyWithOptional = isOptional ? `${key}?` : key;\n\n        // Get the type name\n        const typeName = getTypeName(field);\n\n        // Get description\n        const description = getDescription(field as z.ZodTypeAny);\n\n        // Build param line for this field\n        let paramLine = `${keyWithOptional}: ${typeName}`;\n        if (description) {\n          paramLine += ` // ${description}`;\n        }\n\n        paramLines.push(paramLine);\n      }\n    }\n\n    if (paramLines.length > 0) {\n      fields.push('- param:');\n      for (const paramLine of paramLines) {\n        fields.push(`  - ${paramLine}`);\n      }\n    }\n  }\n\n  return `- ${action.name}, ${action.description || 'No description provided'}\n${tab}${fields.join(`\\n${tab}`)}\n`.trim();\n};\n\nconst systemTemplateOfVLPlanning = ({\n  actionSpace,\n  vlMode,\n}: {\n  actionSpace: DeviceAction<any>[];\n  vlMode: ReturnType<typeof vlLocateMode>;\n}) => {\n  const actionNameList = actionSpace.map((action) => action.name).join(', ');\n  const actionDescriptionList = actionSpace.map((action) => {\n    return descriptionForAction(action, vlLocateParam());\n  });\n  const actionList = actionDescriptionList.join('\\n');\n\n  return `\nTarget: User will give you a screenshot, an instruction and some previous logs indicating what have been done. Please tell what the next one action is (or null if no action should be done) to do the tasks the instruction requires. \n\nRestriction:\n- Don't give extra actions or plans beyond the instruction. ONLY plan for what the instruction requires. For example, don't try to submit the form if the instruction is only to fill something.\n- Always give ONLY ONE action in \\`log\\` field (or null if no action should be done), instead of multiple actions. Supported actions are ${actionNameList}.\n- Don't repeat actions in the previous logs.\n- Bbox is the bounding box of the element to be located. It's an array of 4 numbers, representing ${bboxDescription(vlMode)}.\n\nSupporting actions:\n${actionList}\n\nField description:\n* The \\`prompt\\` field inside the \\`locate\\` field is a short description that could be used to locate the element.\n\nReturn in JSON format:\n{\n  ${vlCoTLog}\n  ${vlCurrentLog}\n  ${commonOutputFields}\n  \"action\": \n    {\n      // one of the supporting actions\n    } | null,\n  ,\n  \"sleep\"?: number, // The sleep time after the action, in milliseconds.\n}\n\nFor example, when the instruction is \"click 'Confirm' button, and click 'Yes' in popup\" and the log is \"I will use action Tap to click 'Confirm' button\", by viewing the screenshot and previous logs, you should consider: We have already clicked the 'Confirm' button, so next we should find and click 'Yes' in popup.\n\nthis and output the JSON:\n\n{\n  \"what_the_user_wants_to_do_next_by_instruction\": \"We have already clicked the 'Confirm' button, so next we should find and click 'Yes' in popup\",\n  \"log\": \"I will use action Tap to click 'Yes' in popup\",\n  \"more_actions_needed_by_instruction\": false,\n  \"action\": {\n    \"type\": \"Tap\",\n    \"param\": {\n      \"locate\": {\n        \"bbox\": [100, 100, 200, 200],\n        \"prompt\": \"The 'Yes' button in popup\"\n      }\n    }\n  }\n}\n`;\n};\n\nconst systemTemplateOfLLM = ({\n  actionSpace,\n}: { actionSpace: DeviceAction<any>[] }) => {\n  const actionNameList = actionSpace.map((action) => action.name).join(' / ');\n  const actionDescriptionList = actionSpace.map((action) => {\n    return descriptionForAction(action, llmLocateParam());\n  });\n  const actionList = actionDescriptionList.join('\\n');\n\n  return `\n## Role\n\nYou are a versatile professional in software UI automation. Your outstanding contributions will impact the user experience of billions of users.\n\n## Objective\n\n- Decompose the instruction user asked into a series of actions\n- Locate the target element if possible\n- If the instruction cannot be accomplished, give a further plan.\n\n## Workflow\n\n1. Receive the screenshot, element description of screenshot(if any), user's instruction and previous logs.\n2. Decompose the user's task into a sequence of feasible actions, and place it in the \\`actions\\` field. There are different types of actions (${actionNameList}). The \"About the action\" section below will give you more details.\n3. Consider whether the user's instruction will be accomplished after the actions you composed.\n- If the instruction is accomplished, set \\`more_actions_needed_by_instruction\\` to false.\n- If more actions are needed, set \\`more_actions_needed_by_instruction\\` to true. Get ready to hand over to the next talent people like you. Carefully log what have been done in the \\`log\\` field, he or she will continue the task according to your logs.\n4. If the task is not feasible on this page, set \\`error\\` field to the reason.\n\n## Constraints\n\n- All the actions you composed MUST be feasible, which means all the action fields can be filled with the page context information you get. If not, don't plan this action.\n- Trust the \"What have been done\" field about the task (if any), don't repeat actions in it.\n- Respond only with valid JSON. Do not write an introduction or summary or markdown prefix like \\`\\`\\`json\\`\\`\\`.\n- If the screenshot and the instruction are totally irrelevant, set reason in the \\`error\\` field.\n\n## About the \\`actions\\` field\n\nThe \\`locate\\` param is commonly used in the \\`param\\` field of the action, means to locate the target element to perform the action, it conforms to the following scheme:\n\ntype LocateParam = {\n  \"id\": string, // the id of the element found. It should either be the id marked with a rectangle in the screenshot or the id described in the description.\n  \"prompt\"?: string // the description of the element to find. It can only be omitted when locate is null.\n} | null // If it's not on the page, the LocateParam should be null\n\n## Supported actions\n\nEach action has a \\`type\\` and corresponding \\`param\\`. To be detailed:\n${actionList}\n\n`.trim();\n};\n\nconst outputTemplate = `\n## Output JSON Format:\n\nThe JSON format is as follows:\n\n{\n  \"actions\": [\n    // ... some actions\n  ],\n  ${llmCurrentLog}\n  ${commonOutputFields}\n}\n\n## Examples\n\n### Example: Decompose a task\n\nWhen you received the following information:\n\n* Instruction: 'Click the language switch button, wait 1s, click \"English\"'\n* Logs: null\n* Page Context (screenshot and description) shows: There is a language switch button, and the \"English\" option is not shown in the screenshot now.\n\nBy viewing the page screenshot and description, you should consider this and output the JSON:\n\n* The user intent is: tap the switch button, sleep, and tap the 'English' option\n* The language switch button is shown in the screenshot, and can be located by the page description or the id marked with a rectangle. So we can plan a Tap action to do this.\n* Plan a Sleep action to wait for 1 second to ensure the language options are displayed.\n* The \"English\" option button is not shown in the screenshot now, it means it may only show after the previous actions are finished. So don't plan any action to do this.\n* Log what these action do: Click the language switch button to open the language options. Wait for 1 second.\n* The task cannot be accomplished (because the last tapping action is not finished yet), so the \\`more_actions_needed_by_instruction\\` field is true. The \\`error\\` field is null.\n\n{\n  \"actions\":[\n    {\n      \"thought\": \"Click the language switch button to open the language options.\",\n      \"type\": \"Tap\", \n      \"param\": {\n        \"locate\": { id: \"c81c4e9a33\", prompt: \"The language switch button\" }\n      }\n    },\n    {\n      \"thought\": \"Wait for 1 second to ensure the language options are displayed.\",\n      \"type\": \"Sleep\",\n      \"param\": { \"timeMs\": 1000 },\n    }\n  ],\n  \"error\": null,\n  \"more_actions_needed_by_instruction\": true,\n  \"log\": \"Click the language switch button to open the language options. Wait for 1 second\",\n}\n\n### Example: What NOT to do\nWrong output:\n{\n  \"actions\":[\n    {\n      \"thought\": \"Click the language switch button to open the language options.\",\n      \"type\": \"Tap\",\n      \"param\": {\n        \"locate\": { \"id\": \"c81c4e9a33\" } // WRONG: prompt is missing, this is not a valid LocateParam\n      }\n    },\n    {\n      \"thought\": \"Click the English option\",\n      \"type\": \"Tap\", \n      \"param\": {\n        \"locate\": null // WRONG: if the element is not on the page, you should not plan this action\n      }\n    }\n  ],\n  \"more_actions_needed_by_instruction\": false, // WRONG: should be true\n  \"log\": \"Click the language switch button to open the language options\",\n}\n`;\n\nexport async function systemPromptToTaskPlanning({\n  actionSpace,\n  vlMode,\n}: {\n  actionSpace: DeviceAction<any>[];\n  vlMode: ReturnType<typeof vlLocateMode>;\n}) {\n  if (vlMode) {\n    return systemTemplateOfVLPlanning({ actionSpace, vlMode });\n  }\n\n  return `${systemTemplateOfLLM({ actionSpace })}\\n\\n${outputTemplate}`;\n}\n\nexport const planSchema: ResponseFormatJSONSchema = {\n  type: 'json_schema',\n  json_schema: {\n    name: 'action_items',\n    strict: false,\n    schema: {\n      type: 'object',\n      strict: false,\n      properties: {\n        actions: {\n          type: 'array',\n          items: {\n            type: 'object',\n            strict: false,\n            properties: {\n              thought: {\n                type: 'string',\n                description:\n                  'Reasons for generating this task, and why this task is feasible on this page',\n              },\n              type: {\n                type: 'string',\n                description: 'Type of action',\n              },\n              param: {\n                anyOf: [\n                  { type: 'null' },\n                  {\n                    type: 'object',\n                    additionalProperties: true,\n                  },\n                ],\n                description: 'Parameter of the action',\n              },\n              locate: {\n                type: ['object', 'null'],\n                properties: {\n                  id: { type: 'string' },\n                  prompt: { type: 'string' },\n                },\n                required: ['id', 'prompt'],\n                additionalProperties: false,\n                description: 'Location information for the target element',\n              },\n            },\n            required: ['thought', 'type', 'param', 'locate'],\n            additionalProperties: false,\n          },\n          description: 'List of actions to be performed',\n        },\n        more_actions_needed_by_instruction: {\n          type: 'boolean',\n          description:\n            'If all the actions described in the instruction have been covered by this action and logs, set this field to false.',\n        },\n        log: {\n          type: 'string',\n          description:\n            'Log what these planned actions do. Do not include further actions that have not been planned.',\n        },\n        error: {\n          type: ['string', 'null'],\n          description: 'Error messages about unexpected situations',\n        },\n      },\n      required: [\n        'actions',\n        'more_actions_needed_by_instruction',\n        'log',\n        'error',\n      ],\n      additionalProperties: false,\n    },\n  },\n};\n\nexport const generateTaskBackgroundContext = (\n  userInstruction: string,\n  log?: string,\n  userActionContext?: string,\n) => {\n  if (log) {\n    return `\nHere is the user's instruction:\n\n<instruction>\n  <high_priority_knowledge>\n    ${userActionContext}\n  </high_priority_knowledge>\n\n  ${userInstruction}\n</instruction>\n\nThese are the logs from previous executions, which indicate what was done in the previous actions.\nDo NOT repeat these actions.\n<previous_logs>\n${log}\n</previous_logs>\n`;\n  }\n\n  return `\nHere is the user's instruction:\n<instruction>\n  <high_priority_knowledge>\n    ${userActionContext}\n  </high_priority_knowledge>\n\n  ${userInstruction}\n</instruction>\n`;\n};\n\nexport const automationUserPrompt = (\n  vlMode: ReturnType<typeof vlLocateMode>,\n) => {\n  if (vlMode) {\n    return new PromptTemplate({\n      template: '{taskBackgroundContext}',\n      inputVariables: ['taskBackgroundContext'],\n    });\n  }\n\n  return new PromptTemplate({\n    template: `\npageDescription:\n=====================================\n{pageDescription}\n=====================================\n\n{taskBackgroundContext}`,\n    inputVariables: ['pageDescription', 'taskBackgroundContext'],\n  });\n};\n"],"names":["vlCoTLog","vlCurrentLog","llmCurrentLog","commonOutputFields","vlLocateParam","llmLocateParam","descriptionForAction","action","locatorSchemaTypeDescription","tab","fields","assert","z","shape","paramLines","getTypeName","field","_actualField__def","unwrapField","f","typeName","actualField","fieldTypeName","ifMidsceneLocatorField","_actualField__def_values","values","option","String","console","getDescription","key","Object","isOptional","keyWithOptional","description","paramLine","systemTemplateOfVLPlanning","actionSpace","vlMode","actionNameList","actionDescriptionList","actionList","bboxDescription","systemTemplateOfLLM","outputTemplate","systemPromptToTaskPlanning","planSchema","generateTaskBackgroundContext","userInstruction","log","userActionContext","automationUserPrompt","PromptTemplate"],"mappings":";;;;;AAUA,MAAMA,WAAW;AACjB,MAAMC,eAAe;AACrB,MAAMC,gBAAgB;AAEtB,MAAMC,qBAAqB,CAAC;+NACmM,CAAC;AAChO,MAAMC,gBAAgB,IACpB;AACF,MAAMC,iBAAiB,IAAM;AAEtB,MAAMC,uBAAuB,CAClCC,QACAC;IAEA,MAAMC,MAAM;IACZ,MAAMC,SAAmB,EAAE;IAG3BA,OAAO,IAAI,CAAC,CAAC,SAAS,EAAEH,OAAO,IAAI,CAAC,CAAC,CAAC;IAGtC,IAAIA,OAAO,WAAW,EAAE;QACtBI,YACEJ,OAAO,WAAW,YAAYK,EAAE,SAAS,EACzC;QAIF,MAAMC,QAAQN,OAAO,WAAW,CAAC,KAAK;QACtC,MAAMO,aAAuB,EAAE;QAG/B,MAAMC,cAAc,CAACC;gBAoBGC;YAlBtB,MAAMC,cAAc,CAACC;gBACnB,IAAI,CAACA,EAAE,IAAI,EAAE,OAAOA;gBAEpB,MAAMC,WAAWD,EAAE,IAAI,CAAC,QAAQ;gBAGhC,IACEC,AAAa,kBAAbA,YACAA,AAAa,kBAAbA,YACAA,AAAa,iBAAbA,UAEA,OAAOF,YAAYC,EAAE,IAAI,CAAC,SAAS;gBAGrC,OAAOA;YACT;YAEA,MAAME,cAAcH,YAAYF;YAChC,MAAMM,gBAAgB,QAAAL,CAAAA,oBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,kBAAkB,QAAQ;YAEhD,IAAIK,AAAkB,gBAAlBA,eAA+B,OAAO;YAC1C,IAAIA,AAAkB,gBAAlBA,eAA+B,OAAO;YAC1C,IAAIA,AAAkB,iBAAlBA,eAAgC,OAAO;YAC3C,IAAIA,AAAkB,eAAlBA,eAA8B,OAAO;YACzC,IAAIA,AAAkB,gBAAlBA,eAA+B;gBAEjC,IAAIC,uBAAuBF,cACzB,OAAOb;gBAET,OAAO;YACT;YACA,IAAIc,AAAkB,cAAlBA,eAA6B;oBAE5BE,0BAAAA;gBADH,MAAMC,SACJ,AAAC,SAAAD,CAAAA,qBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,QAAAA,CAAAA,2BAAAA,mBAAkB,MAAM,AAAD,IAAvBA,KAAAA,IAAAA,yBACG,GAAG,CAAC,CAACE,SAAoBC,OAAO,CAAC,CAAC,EAAED,OAAO,CAAC,CAAC,GAC9C,IAAI,CAAC,KAAI,KAAK;gBAEnB,OAAO,CAAC,KAAK,EAAED,OAAO,CAAC,CAAC;YAC1B;YAEAG,QAAQ,IAAI,CACV,2EACAP,YAAY,IAAI;YAElB,OAAOA,YAAY,QAAQ;QAC7B;QAGA,MAAMQ,iBAAiB,CAACb;gBAgClBC;YA9BJ,MAAMC,cAAc,CAACC;gBACnB,IAAI,CAACA,EAAE,IAAI,EAAE,OAAOA;gBAEpB,MAAMC,WAAWD,EAAE,IAAI,CAAC,QAAQ;gBAGhC,IACEC,AAAa,kBAAbA,YACAA,AAAa,kBAAbA,YACAA,AAAa,iBAAbA,UAEA,OAAOF,YAAYC,EAAE,IAAI,CAAC,SAAS;gBAGrC,OAAOA;YACT;YAGA,IAAI,iBAAiBH,OACnB,OAAOA,MAAM,WAAW,IAAI;YAG9B,MAAMK,cAAcH,YAAYF;YAGhC,IAAI,iBAAiBK,aACnB,OAAOA,YAAY,WAAW,IAAI;YAIpC,IAAIJ,AAAAA,SAAAA,CAAAA,oBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,kBAAkB,QAAQ,AAAD,MAAM,aACjC;gBAAA,IAAI,kCAAkCI,YAAY,IAAI,CAAC,KAAK,IAC1D,OAAO;YACT;YAGF,OAAO;QACT;QAEA,KAAK,MAAM,CAACS,KAAKd,MAAM,IAAIe,OAAO,OAAO,CAAClB,OACxC,IAAIG,SAAS,AAAiB,YAAjB,OAAOA,OAAoB;YAEtC,MAAMgB,aACJ,AAAqC,cAArC,OAAQhB,MAAc,UAAU,IAC/BA,MAAc,UAAU;YAC3B,MAAMiB,kBAAkBD,aAAa,GAAGF,IAAI,CAAC,CAAC,GAAGA;YAGjD,MAAMV,WAAWL,YAAYC;YAG7B,MAAMkB,cAAcL,eAAeb;YAGnC,IAAImB,YAAY,GAAGF,gBAAgB,EAAE,EAAEb,UAAU;YACjD,IAAIc,aACFC,aAAa,CAAC,IAAI,EAAED,aAAa;YAGnCpB,WAAW,IAAI,CAACqB;QAClB;QAGF,IAAIrB,WAAW,MAAM,GAAG,GAAG;YACzBJ,OAAO,IAAI,CAAC;YACZ,KAAK,MAAMyB,aAAarB,WACtBJ,OAAO,IAAI,CAAC,CAAC,IAAI,EAAEyB,WAAW;QAElC;IACF;IAEA,OAAO,CAAC,EAAE,EAAE5B,OAAO,IAAI,CAAC,EAAE,EAAEA,OAAO,WAAW,IAAI,0BAA0B;AAC9E,EAAEE,MAAMC,OAAO,IAAI,CAAC,CAAC,EAAE,EAAED,KAAK,EAAE;AAChC,CAAC,CAAC,IAAI;AACN;AAEA,MAAM2B,6BAA6B,CAAC,EAClCC,WAAW,EACXC,MAAM,EAIP;IACC,MAAMC,iBAAiBF,YAAY,GAAG,CAAC,CAAC9B,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;IACrE,MAAMiC,wBAAwBH,YAAY,GAAG,CAAC,CAAC9B,SACtCD,qBAAqBC,QAAQH;IAEtC,MAAMqC,aAAaD,sBAAsB,IAAI,CAAC;IAE9C,OAAO,CAAC;;;;;yIAK+H,EAAED,eAAe;;kGAExD,EAAEG,gBAAgBJ,QAAQ;;;AAG5H,EAAEG,WAAW;;;;;;;EAOX,EAAEzC,SAAS;EACX,EAAEC,aAAa;EACf,EAAEE,mBAAmB;;;;;;;;;;;;;;;;;;;;;;;;;;;AA2BvB,CAAC;AACD;AAEA,MAAMwC,sBAAsB,CAAC,EAC3BN,WAAW,EAC0B;IACrC,MAAME,iBAAiBF,YAAY,GAAG,CAAC,CAAC9B,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;IACrE,MAAMiC,wBAAwBH,YAAY,GAAG,CAAC,CAAC9B,SACtCD,qBAAqBC,QAAQF;IAEtC,MAAMoC,aAAaD,sBAAsB,IAAI,CAAC;IAE9C,OAAO,CAAC;;;;;;;;;;;;;;+IAcqI,EAAED,eAAe;;;;;;;;;;;;;;;;;;;;;;;;;AAyBhK,EAAEE,WAAW;;AAEb,CAAC,CAAC,IAAI;AACN;AAEA,MAAMG,iBAAiB,CAAC;;;;;;;;;EAStB,EAAE1C,cAAc;EAChB,EAAEC,mBAAmB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAgEvB,CAAC;AAEM,eAAe0C,2BAA2B,EAC/CR,WAAW,EACXC,MAAM,EAIP;IACC,IAAIA,QACF,OAAOF,2BAA2B;QAAEC;QAAaC;IAAO;IAG1D,OAAO,GAAGK,oBAAoB;QAAEN;IAAY,GAAG,IAAI,EAAEO,gBAAgB;AACvE;AAEO,MAAME,aAAuC;IAClD,MAAM;IACN,aAAa;QACX,MAAM;QACN,QAAQ;QACR,QAAQ;YACN,MAAM;YACN,QAAQ;YACR,YAAY;gBACV,SAAS;oBACP,MAAM;oBACN,OAAO;wBACL,MAAM;wBACN,QAAQ;wBACR,YAAY;4BACV,SAAS;gCACP,MAAM;gCACN,aACE;4BACJ;4BACA,MAAM;gCACJ,MAAM;gCACN,aAAa;4BACf;4BACA,OAAO;gCACL,OAAO;oCACL;wCAAE,MAAM;oCAAO;oCACf;wCACE,MAAM;wCACN,sBAAsB;oCACxB;iCACD;gCACD,aAAa;4BACf;4BACA,QAAQ;gCACN,MAAM;oCAAC;oCAAU;iCAAO;gCACxB,YAAY;oCACV,IAAI;wCAAE,MAAM;oCAAS;oCACrB,QAAQ;wCAAE,MAAM;oCAAS;gCAC3B;gCACA,UAAU;oCAAC;oCAAM;iCAAS;gCAC1B,sBAAsB;gCACtB,aAAa;4BACf;wBACF;wBACA,UAAU;4BAAC;4BAAW;4BAAQ;4BAAS;yBAAS;wBAChD,sBAAsB;oBACxB;oBACA,aAAa;gBACf;gBACA,oCAAoC;oBAClC,MAAM;oBACN,aACE;gBACJ;gBACA,KAAK;oBACH,MAAM;oBACN,aACE;gBACJ;gBACA,OAAO;oBACL,MAAM;wBAAC;wBAAU;qBAAO;oBACxB,aAAa;gBACf;YACF;YACA,UAAU;gBACR;gBACA;gBACA;gBACA;aACD;YACD,sBAAsB;QACxB;IACF;AACF;AAEO,MAAMC,gCAAgC,CAC3CC,iBACAC,KACAC;IAEA,IAAID,KACF,OAAO,CAAC;;;;;IAKR,EAAEC,kBAAkB;;;EAGtB,EAAEF,gBAAgB;;;;;;AAMpB,EAAEC,IAAI;;AAEN,CAAC;IAGC,OAAO,CAAC;;;;IAIN,EAAEC,kBAAkB;;;EAGtB,EAAEF,gBAAgB;;AAEpB,CAAC;AACD;AAEO,MAAMG,uBAAuB,CAClCb;IAEA,IAAIA,QACF,OAAO,IAAIc,eAAe;QACxB,UAAU;QACV,gBAAgB;YAAC;SAAwB;IAC3C;IAGF,OAAO,IAAIA,eAAe;QACxB,UAAU,CAAC;;;;;;uBAMQ,CAAC;QACpB,gBAAgB;YAAC;YAAmB;SAAwB;IAC9D;AACF"}
+{"version":3,"file":"ai-model/prompt/llm-planning.mjs","sources":["webpack://@midscene/core/./src/ai-model/prompt/llm-planning.ts"],"sourcesContent":["import assert from 'node:assert';\nimport type { DeviceAction } from '@/types';\nimport { PromptTemplate } from '@langchain/core/prompts';\nimport type { vlLocateMode } from '@midscene/shared/env';\nimport type { ResponseFormatJSONSchema } from 'openai/resources/index';\nimport type { ZodObject, z } from 'zod';\nimport { ifMidsceneLocatorField } from '../common';\nimport { bboxDescription } from './common';\n\n// Note: put the log field first to trigger the CoT\nconst vlCoTLog = `\"what_the_user_wants_to_do_next_by_instruction\": string, // What the user wants to do according to the instruction and previous logs. `;\nconst vlCurrentLog = `\"log\": string, // Log what the next one action (ONLY ONE!) you can do according to the screenshot and the instruction. The typical log looks like \"Now i want to use action '{ action-type }' to do .. first\". If no action should be done, log the reason. \". Use the same language as the user's instruction.`;\nconst llmCurrentLog = `\"log\": string, // Log what the next actions you can do according to the screenshot and the instruction. The typical log looks like \"Now i want to use action '{ action-type }' to do ..\". If no action should be done, log the reason. \". Use the same language as the user's instruction.`;\n\nconst commonOutputFields = `\"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 the same language as the user's instruction.\n  \"more_actions_needed_by_instruction\": boolean, // Consider if there is still more action(s) to do after the action in \"Log\" is done, according to the instruction. If so, set this field to true. Otherwise, set it to false.`;\nconst vlLocateParam = () =>\n  '{bbox: [number, number, number, number], prompt: string }';\nconst llmLocateParam = () => '{\"id\": string, \"prompt\": string}';\n\nexport const descriptionForAction = (\n  action: DeviceAction<any>,\n  locatorSchemaTypeDescription: string,\n) => {\n  const tab = '  ';\n  const fields: string[] = [];\n\n  // Add the action type field\n  fields.push(`- type: \"${action.name}\"`);\n\n  // Handle paramSchema if it exists\n  if (action.paramSchema) {\n    assert(\n      action.paramSchema.constructor.name === 'ZodObject',\n      'paramSchema must be a zod object',\n    );\n    // Try to extract parameter information from the zod schema\n    // For zod object schemas, extract type information and descriptions\n    const shape = (action.paramSchema as ZodObject<any>).shape;\n    const paramLines: string[] = [];\n\n    // Helper function to get type name from zod schema\n    const getTypeName = (field: any): string => {\n      // Recursively unwrap optional, nullable, and other wrapper types to get the actual inner type\n      const unwrapField = (f: any): any => {\n        if (!f._def) return f;\n\n        const typeName = f._def.typeName;\n\n        // Handle wrapper types that have innerType\n        if (\n          typeName === 'ZodOptional' ||\n          typeName === 'ZodNullable' ||\n          typeName === 'ZodDefault'\n        ) {\n          return unwrapField(f._def.innerType);\n        }\n\n        return f;\n      };\n\n      const actualField = unwrapField(field);\n      const fieldTypeName = actualField._def?.typeName;\n\n      if (fieldTypeName === 'ZodString') return 'string';\n      if (fieldTypeName === 'ZodNumber') return 'number';\n      if (fieldTypeName === 'ZodBoolean') return 'boolean';\n      if (fieldTypeName === 'ZodArray') return 'array';\n      if (fieldTypeName === 'ZodObject') {\n        // Check if this is a passthrough object (like MidsceneLocation)\n        if (ifMidsceneLocatorField(actualField)) {\n          return locatorSchemaTypeDescription;\n        }\n        return 'object';\n      }\n      if (fieldTypeName === 'ZodEnum') {\n        const values =\n          (actualField._def?.values as unknown[] | undefined)\n            ?.map((option: unknown) => String(`'${option}'`))\n            .join(', ') ?? 'enum';\n\n        return `enum(${values})`;\n      }\n\n      console.warn(\n        'failed to parse Zod type. This may lead to wrong params from the LLM.\\n',\n        actualField._def,\n      );\n      return actualField.toString();\n    };\n\n    // Helper function to get description from zod schema\n    const getDescription = (field: z.ZodTypeAny): string | null => {\n      // Recursively unwrap optional, nullable, and other wrapper types to get the actual inner type\n      const unwrapField = (f: any): any => {\n        if (!f._def) return f;\n\n        const typeName = f._def.typeName;\n\n        // Handle wrapper types that have innerType\n        if (\n          typeName === 'ZodOptional' ||\n          typeName === 'ZodNullable' ||\n          typeName === 'ZodDefault'\n        ) {\n          return unwrapField(f._def.innerType);\n        }\n\n        return f;\n      };\n\n      // Check for direct description on the original field (wrapper may have description)\n      if ('description' in field) {\n        return field.description || null;\n      }\n\n      const actualField = unwrapField(field);\n\n      // Check for description on the unwrapped field\n      if ('description' in actualField) {\n        return actualField.description || null;\n      }\n\n      // Check for MidsceneLocation fields and add description\n      if (actualField._def?.typeName === 'ZodObject') {\n        if ('midscene_location_field_flag' in actualField._def.shape()) {\n          return 'Location information for the target element';\n        }\n      }\n\n      return null;\n    };\n\n    for (const [key, field] of Object.entries(shape)) {\n      if (field && typeof field === 'object') {\n        // Check if field is optional\n        const isOptional =\n          typeof (field as any).isOptional === 'function' &&\n          (field as any).isOptional();\n        const keyWithOptional = isOptional ? `${key}?` : key;\n\n        // Get the type name\n        const typeName = getTypeName(field);\n\n        // Get description\n        const description = getDescription(field as z.ZodTypeAny);\n\n        // Build param line for this field\n        let paramLine = `${keyWithOptional}: ${typeName}`;\n        if (description) {\n          paramLine += ` // ${description}`;\n        }\n\n        paramLines.push(paramLine);\n      }\n    }\n\n    if (paramLines.length > 0) {\n      fields.push('- param:');\n      for (const paramLine of paramLines) {\n        fields.push(`  - ${paramLine}`);\n      }\n    }\n  }\n\n  return `- ${action.name}, ${action.description || 'No description provided'}\n${tab}${fields.join(`\\n${tab}`)}\n`.trim();\n};\n\nconst systemTemplateOfVLPlanning = ({\n  actionSpace,\n  vlMode,\n}: {\n  actionSpace: DeviceAction<any>[];\n  vlMode: ReturnType<typeof vlLocateMode>;\n}) => {\n  const actionNameList = actionSpace.map((action) => action.name).join(', ');\n  const actionDescriptionList = actionSpace.map((action) => {\n    return descriptionForAction(action, vlLocateParam());\n  });\n  const actionList = actionDescriptionList.join('\\n');\n\n  return `\nTarget: User will give you a screenshot, an instruction and some previous logs indicating what have been done. Please tell what the next one action is (or null if no action should be done) to do the tasks the instruction requires. \n\nRestriction:\n- Don't give extra actions or plans beyond the instruction. ONLY plan for what the instruction requires. For example, don't try to submit the form if the instruction is only to fill something.\n- Always give ONLY ONE action in \\`log\\` field (or null if no action should be done), instead of multiple actions. Supported actions are ${actionNameList}.\n- Don't repeat actions in the previous logs.\n- Bbox is the bounding box of the element to be located. It's an array of 4 numbers, representing ${bboxDescription(vlMode)}.\n\nSupporting actions:\n${actionList}\n\nField description:\n* The \\`prompt\\` field inside the \\`locate\\` field is a short description that could be used to locate the element.\n\nReturn in JSON format:\n{\n  ${vlCoTLog}\n  ${vlCurrentLog}\n  ${commonOutputFields}\n  \"action\": \n    {\n      // one of the supporting actions\n    } | null,\n  ,\n  \"sleep\"?: number, // The sleep time after the action, in milliseconds.\n}\n\nFor example, when the instruction is \"click 'Confirm' button, and click 'Yes' in popup\" and the log is \"I will use action Tap to click 'Confirm' button\", by viewing the screenshot and previous logs, you should consider: We have already clicked the 'Confirm' button, so next we should find and click 'Yes' in popup.\n\nthis and output the JSON:\n\n{\n  \"what_the_user_wants_to_do_next_by_instruction\": \"We have already clicked the 'Confirm' button, so next we should find and click 'Yes' in popup\",\n  \"log\": \"I will use action Tap to click 'Yes' in popup\",\n  \"more_actions_needed_by_instruction\": false,\n  \"action\": {\n    \"type\": \"Tap\",\n    \"param\": {\n      \"locate\": {\n        \"bbox\": [100, 100, 200, 200],\n        \"prompt\": \"The 'Yes' button in popup\"\n      }\n    }\n  }\n}\n`;\n};\n\nconst systemTemplateOfLLM = ({\n  actionSpace,\n}: { actionSpace: DeviceAction<any>[] }) => {\n  const actionNameList = actionSpace.map((action) => action.name).join(' / ');\n  const actionDescriptionList = actionSpace.map((action) => {\n    return descriptionForAction(action, llmLocateParam());\n  });\n  const actionList = actionDescriptionList.join('\\n');\n\n  return `\n## Role\n\nYou are a versatile professional in software UI automation. Your outstanding contributions will impact the user experience of billions of users.\n\n## Objective\n\n- Decompose the instruction user asked into a series of actions\n- Locate the target element if possible\n- If the instruction cannot be accomplished, give a further plan.\n\n## Workflow\n\n1. Receive the screenshot, element description of screenshot(if any), user's instruction and previous logs.\n2. Decompose the user's task into a sequence of feasible actions, and place it in the \\`actions\\` field. There are different types of actions (${actionNameList}). The \"About the action\" section below will give you more details.\n3. Consider whether the user's instruction will be accomplished after the actions you composed.\n- If the instruction is accomplished, set \\`more_actions_needed_by_instruction\\` to false.\n- If more actions are needed, set \\`more_actions_needed_by_instruction\\` to true. Get ready to hand over to the next talent people like you. Carefully log what have been done in the \\`log\\` field, he or she will continue the task according to your logs.\n4. If the task is not feasible on this page, set \\`error\\` field to the reason.\n\n## Constraints\n\n- All the actions you composed MUST be feasible, which means all the action fields can be filled with the page context information you get. If not, don't plan this action.\n- Trust the \"What have been done\" field about the task (if any), don't repeat actions in it.\n- Respond only with valid JSON. Do not write an introduction or summary or markdown prefix like \\`\\`\\`json\\`\\`\\`.\n- If the screenshot and the instruction are totally irrelevant, set reason in the \\`error\\` field.\n\n## About the \\`actions\\` field\n\nThe \\`locate\\` param is commonly used in the \\`param\\` field of the action, means to locate the target element to perform the action, it conforms to the following scheme:\n\ntype LocateParam = {\n  \"id\": string, // the id of the element found. It should either be the id marked with a rectangle in the screenshot or the id described in the description.\n  \"prompt\"?: string // the description of the element to find. It can only be omitted when locate is null.\n} | null // If it's not on the page, the LocateParam should be null\n\n## Supported actions\n\nEach action has a \\`type\\` and corresponding \\`param\\`. To be detailed:\n${actionList}\n\n`.trim();\n};\n\nconst outputTemplate = `\n## Output JSON Format:\n\nThe JSON format is as follows:\n\n{\n  \"actions\": [\n    // ... some actions\n  ],\n  ${llmCurrentLog}\n  ${commonOutputFields}\n}\n\n## Examples\n\n### Example: Decompose a task\n\nWhen you received the following information:\n\n* Instruction: 'Click the language switch button, wait 1s, click \"English\"'\n* Logs: null\n* Page Context (screenshot and description) shows: There is a language switch button, and the \"English\" option is not shown in the screenshot now.\n\nBy viewing the page screenshot and description, you should consider this and output the JSON:\n\n* The user intent is: tap the switch button, sleep, and tap the 'English' option\n* The language switch button is shown in the screenshot, and can be located by the page description or the id marked with a rectangle. So we can plan a Tap action to do this.\n* Plan a Sleep action to wait for 1 second to ensure the language options are displayed.\n* The \"English\" option button is not shown in the screenshot now, it means it may only show after the previous actions are finished. So don't plan any action to do this.\n* Log what these action do: Click the language switch button to open the language options. Wait for 1 second.\n* The task cannot be accomplished (because the last tapping action is not finished yet), so the \\`more_actions_needed_by_instruction\\` field is true. The \\`error\\` field is null.\n\n{\n  \"actions\":[\n    {\n      \"thought\": \"Click the language switch button to open the language options.\",\n      \"type\": \"Tap\", \n      \"param\": {\n        \"locate\": { id: \"c81c4e9a33\", prompt: \"The language switch button\" }\n      }\n    },\n    {\n      \"thought\": \"Wait for 1 second to ensure the language options are displayed.\",\n      \"type\": \"Sleep\",\n      \"param\": { \"timeMs\": 1000 },\n    }\n  ],\n  \"error\": null,\n  \"more_actions_needed_by_instruction\": true,\n  \"log\": \"Click the language switch button to open the language options. Wait for 1 second\",\n}\n\n### Example: What NOT to do\nWrong output:\n{\n  \"actions\":[\n    {\n      \"thought\": \"Click the language switch button to open the language options.\",\n      \"type\": \"Tap\",\n      \"param\": {\n        \"locate\": { \"id\": \"c81c4e9a33\" } // WRONG: prompt is missing, this is not a valid LocateParam\n      }\n    },\n    {\n      \"thought\": \"Click the English option\",\n      \"type\": \"Tap\", \n      \"param\": {\n        \"locate\": null // WRONG: if the element is not on the page, you should not plan this action\n      }\n    }\n  ],\n  \"more_actions_needed_by_instruction\": false, // WRONG: should be true\n  \"log\": \"Click the language switch button to open the language options\",\n}\n`;\n\nexport async function systemPromptToTaskPlanning({\n  actionSpace,\n  vlMode,\n}: {\n  actionSpace: DeviceAction<any>[];\n  vlMode: ReturnType<typeof vlLocateMode>;\n}) {\n  if (vlMode) {\n    return systemTemplateOfVLPlanning({ actionSpace, vlMode });\n  }\n\n  return `${systemTemplateOfLLM({ actionSpace })}\\n\\n${outputTemplate}`;\n}\n\nexport const planSchema: ResponseFormatJSONSchema = {\n  type: 'json_schema',\n  json_schema: {\n    name: 'action_items',\n    strict: false,\n    schema: {\n      type: 'object',\n      strict: false,\n      properties: {\n        actions: {\n          type: 'array',\n          items: {\n            type: 'object',\n            strict: false,\n            properties: {\n              thought: {\n                type: 'string',\n                description:\n                  'Reasons for generating this task, and why this task is feasible on this page',\n              },\n              type: {\n                type: 'string',\n                description: 'Type of action',\n              },\n              param: {\n                anyOf: [\n                  { type: 'null' },\n                  {\n                    type: 'object',\n                    additionalProperties: true,\n                  },\n                ],\n                description: 'Parameter of the action',\n              },\n              locate: {\n                type: ['object', 'null'],\n                properties: {\n                  id: { type: 'string' },\n                  prompt: { type: 'string' },\n                },\n                required: ['id', 'prompt'],\n                additionalProperties: false,\n                description: 'Location information for the target element',\n              },\n            },\n            required: ['thought', 'type', 'param', 'locate'],\n            additionalProperties: false,\n          },\n          description: 'List of actions to be performed',\n        },\n        more_actions_needed_by_instruction: {\n          type: 'boolean',\n          description:\n            'If all the actions described in the instruction have been covered by this action and logs, set this field to false.',\n        },\n        log: {\n          type: 'string',\n          description:\n            'Log what these planned actions do. Do not include further actions that have not been planned.',\n        },\n        error: {\n          type: ['string', 'null'],\n          description: 'Error messages about unexpected situations',\n        },\n      },\n      required: [\n        'actions',\n        'more_actions_needed_by_instruction',\n        'log',\n        'error',\n      ],\n      additionalProperties: false,\n    },\n  },\n};\n\nexport const generateTaskBackgroundContext = (\n  userInstruction: string,\n  log?: string,\n  userActionContext?: string,\n) => {\n  if (log) {\n    return `\nHere is the user's instruction:\n\n<instruction>\n  <high_priority_knowledge>\n    ${userActionContext}\n  </high_priority_knowledge>\n\n  ${userInstruction}\n</instruction>\n\nThese are the logs from previous executions, which indicate what was done in the previous actions.\nDo NOT repeat these actions.\n<previous_logs>\n${log}\n</previous_logs>\n`;\n  }\n\n  return `\nHere is the user's instruction:\n<instruction>\n  <high_priority_knowledge>\n    ${userActionContext}\n  </high_priority_knowledge>\n\n  ${userInstruction}\n</instruction>\n`;\n};\n\nexport const automationUserPrompt = (\n  vlMode: ReturnType<typeof vlLocateMode>,\n) => {\n  if (vlMode) {\n    return new PromptTemplate({\n      template: '{taskBackgroundContext}',\n      inputVariables: ['taskBackgroundContext'],\n    });\n  }\n\n  return new PromptTemplate({\n    template: `\npageDescription:\n=====================================\n{pageDescription}\n=====================================\n\n{taskBackgroundContext}`,\n    inputVariables: ['pageDescription', 'taskBackgroundContext'],\n  });\n};\n"],"names":["vlCoTLog","vlCurrentLog","llmCurrentLog","commonOutputFields","vlLocateParam","llmLocateParam","descriptionForAction","action","locatorSchemaTypeDescription","tab","fields","assert","shape","paramLines","getTypeName","field","_actualField__def","unwrapField","f","typeName","actualField","fieldTypeName","ifMidsceneLocatorField","_actualField__def_values","values","option","String","console","getDescription","key","Object","isOptional","keyWithOptional","description","paramLine","systemTemplateOfVLPlanning","actionSpace","vlMode","actionNameList","actionDescriptionList","actionList","bboxDescription","systemTemplateOfLLM","outputTemplate","systemPromptToTaskPlanning","planSchema","generateTaskBackgroundContext","userInstruction","log","userActionContext","automationUserPrompt","PromptTemplate"],"mappings":";;;;AAUA,MAAMA,WAAW;AACjB,MAAMC,eAAe;AACrB,MAAMC,gBAAgB;AAEtB,MAAMC,qBAAqB,CAAC;+NACmM,CAAC;AAChO,MAAMC,gBAAgB,IACpB;AACF,MAAMC,iBAAiB,IAAM;AAEtB,MAAMC,uBAAuB,CAClCC,QACAC;IAEA,MAAMC,MAAM;IACZ,MAAMC,SAAmB,EAAE;IAG3BA,OAAO,IAAI,CAAC,CAAC,SAAS,EAAEH,OAAO,IAAI,CAAC,CAAC,CAAC;IAGtC,IAAIA,OAAO,WAAW,EAAE;QACtBI,YACEJ,AAAwC,gBAAxCA,OAAO,WAAW,CAAC,WAAW,CAAC,IAAI,EACnC;QAIF,MAAMK,QAASL,OAAO,WAAW,CAAoB,KAAK;QAC1D,MAAMM,aAAuB,EAAE;QAG/B,MAAMC,cAAc,CAACC;gBAoBGC;YAlBtB,MAAMC,cAAc,CAACC;gBACnB,IAAI,CAACA,EAAE,IAAI,EAAE,OAAOA;gBAEpB,MAAMC,WAAWD,EAAE,IAAI,CAAC,QAAQ;gBAGhC,IACEC,AAAa,kBAAbA,YACAA,AAAa,kBAAbA,YACAA,AAAa,iBAAbA,UAEA,OAAOF,YAAYC,EAAE,IAAI,CAAC,SAAS;gBAGrC,OAAOA;YACT;YAEA,MAAME,cAAcH,YAAYF;YAChC,MAAMM,gBAAgB,QAAAL,CAAAA,oBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,kBAAkB,QAAQ;YAEhD,IAAIK,AAAkB,gBAAlBA,eAA+B,OAAO;YAC1C,IAAIA,AAAkB,gBAAlBA,eAA+B,OAAO;YAC1C,IAAIA,AAAkB,iBAAlBA,eAAgC,OAAO;YAC3C,IAAIA,AAAkB,eAAlBA,eAA8B,OAAO;YACzC,IAAIA,AAAkB,gBAAlBA,eAA+B;gBAEjC,IAAIC,uBAAuBF,cACzB,OAAOZ;gBAET,OAAO;YACT;YACA,IAAIa,AAAkB,cAAlBA,eAA6B;oBAE5BE,0BAAAA;gBADH,MAAMC,SACJ,AAAC,SAAAD,CAAAA,qBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,QAAAA,CAAAA,2BAAAA,mBAAkB,MAAM,AAAD,IAAvBA,KAAAA,IAAAA,yBACG,GAAG,CAAC,CAACE,SAAoBC,OAAO,CAAC,CAAC,EAAED,OAAO,CAAC,CAAC,GAC9C,IAAI,CAAC,KAAI,KAAK;gBAEnB,OAAO,CAAC,KAAK,EAAED,OAAO,CAAC,CAAC;YAC1B;YAEAG,QAAQ,IAAI,CACV,2EACAP,YAAY,IAAI;YAElB,OAAOA,YAAY,QAAQ;QAC7B;QAGA,MAAMQ,iBAAiB,CAACb;gBAgClBC;YA9BJ,MAAMC,cAAc,CAACC;gBACnB,IAAI,CAACA,EAAE,IAAI,EAAE,OAAOA;gBAEpB,MAAMC,WAAWD,EAAE,IAAI,CAAC,QAAQ;gBAGhC,IACEC,AAAa,kBAAbA,YACAA,AAAa,kBAAbA,YACAA,AAAa,iBAAbA,UAEA,OAAOF,YAAYC,EAAE,IAAI,CAAC,SAAS;gBAGrC,OAAOA;YACT;YAGA,IAAI,iBAAiBH,OACnB,OAAOA,MAAM,WAAW,IAAI;YAG9B,MAAMK,cAAcH,YAAYF;YAGhC,IAAI,iBAAiBK,aACnB,OAAOA,YAAY,WAAW,IAAI;YAIpC,IAAIJ,AAAAA,SAAAA,CAAAA,oBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,kBAAkB,QAAQ,AAAD,MAAM,aACjC;gBAAA,IAAI,kCAAkCI,YAAY,IAAI,CAAC,KAAK,IAC1D,OAAO;YACT;YAGF,OAAO;QACT;QAEA,KAAK,MAAM,CAACS,KAAKd,MAAM,IAAIe,OAAO,OAAO,CAAClB,OACxC,IAAIG,SAAS,AAAiB,YAAjB,OAAOA,OAAoB;YAEtC,MAAMgB,aACJ,AAAqC,cAArC,OAAQhB,MAAc,UAAU,IAC/BA,MAAc,UAAU;YAC3B,MAAMiB,kBAAkBD,aAAa,GAAGF,IAAI,CAAC,CAAC,GAAGA;YAGjD,MAAMV,WAAWL,YAAYC;YAG7B,MAAMkB,cAAcL,eAAeb;YAGnC,IAAImB,YAAY,GAAGF,gBAAgB,EAAE,EAAEb,UAAU;YACjD,IAAIc,aACFC,aAAa,CAAC,IAAI,EAAED,aAAa;YAGnCpB,WAAW,IAAI,CAACqB;QAClB;QAGF,IAAIrB,WAAW,MAAM,GAAG,GAAG;YACzBH,OAAO,IAAI,CAAC;YACZ,KAAK,MAAMwB,aAAarB,WACtBH,OAAO,IAAI,CAAC,CAAC,IAAI,EAAEwB,WAAW;QAElC;IACF;IAEA,OAAO,CAAC,EAAE,EAAE3B,OAAO,IAAI,CAAC,EAAE,EAAEA,OAAO,WAAW,IAAI,0BAA0B;AAC9E,EAAEE,MAAMC,OAAO,IAAI,CAAC,CAAC,EAAE,EAAED,KAAK,EAAE;AAChC,CAAC,CAAC,IAAI;AACN;AAEA,MAAM0B,6BAA6B,CAAC,EAClCC,WAAW,EACXC,MAAM,EAIP;IACC,MAAMC,iBAAiBF,YAAY,GAAG,CAAC,CAAC7B,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;IACrE,MAAMgC,wBAAwBH,YAAY,GAAG,CAAC,CAAC7B,SACtCD,qBAAqBC,QAAQH;IAEtC,MAAMoC,aAAaD,sBAAsB,IAAI,CAAC;IAE9C,OAAO,CAAC;;;;;yIAK+H,EAAED,eAAe;;kGAExD,EAAEG,gBAAgBJ,QAAQ;;;AAG5H,EAAEG,WAAW;;;;;;;EAOX,EAAExC,SAAS;EACX,EAAEC,aAAa;EACf,EAAEE,mBAAmB;;;;;;;;;;;;;;;;;;;;;;;;;;;AA2BvB,CAAC;AACD;AAEA,MAAMuC,sBAAsB,CAAC,EAC3BN,WAAW,EAC0B;IACrC,MAAME,iBAAiBF,YAAY,GAAG,CAAC,CAAC7B,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;IACrE,MAAMgC,wBAAwBH,YAAY,GAAG,CAAC,CAAC7B,SACtCD,qBAAqBC,QAAQF;IAEtC,MAAMmC,aAAaD,sBAAsB,IAAI,CAAC;IAE9C,OAAO,CAAC;;;;;;;;;;;;;;+IAcqI,EAAED,eAAe;;;;;;;;;;;;;;;;;;;;;;;;;AAyBhK,EAAEE,WAAW;;AAEb,CAAC,CAAC,IAAI;AACN;AAEA,MAAMG,iBAAiB,CAAC;;;;;;;;;EAStB,EAAEzC,cAAc;EAChB,EAAEC,mBAAmB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAgEvB,CAAC;AAEM,eAAeyC,2BAA2B,EAC/CR,WAAW,EACXC,MAAM,EAIP;IACC,IAAIA,QACF,OAAOF,2BAA2B;QAAEC;QAAaC;IAAO;IAG1D,OAAO,GAAGK,oBAAoB;QAAEN;IAAY,GAAG,IAAI,EAAEO,gBAAgB;AACvE;AAEO,MAAME,aAAuC;IAClD,MAAM;IACN,aAAa;QACX,MAAM;QACN,QAAQ;QACR,QAAQ;YACN,MAAM;YACN,QAAQ;YACR,YAAY;gBACV,SAAS;oBACP,MAAM;oBACN,OAAO;wBACL,MAAM;wBACN,QAAQ;wBACR,YAAY;4BACV,SAAS;gCACP,MAAM;gCACN,aACE;4BACJ;4BACA,MAAM;gCACJ,MAAM;gCACN,aAAa;4BACf;4BACA,OAAO;gCACL,OAAO;oCACL;wCAAE,MAAM;oCAAO;oCACf;wCACE,MAAM;wCACN,sBAAsB;oCACxB;iCACD;gCACD,aAAa;4BACf;4BACA,QAAQ;gCACN,MAAM;oCAAC;oCAAU;iCAAO;gCACxB,YAAY;oCACV,IAAI;wCAAE,MAAM;oCAAS;oCACrB,QAAQ;wCAAE,MAAM;oCAAS;gCAC3B;gCACA,UAAU;oCAAC;oCAAM;iCAAS;gCAC1B,sBAAsB;gCACtB,aAAa;4BACf;wBACF;wBACA,UAAU;4BAAC;4BAAW;4BAAQ;4BAAS;yBAAS;wBAChD,sBAAsB;oBACxB;oBACA,aAAa;gBACf;gBACA,oCAAoC;oBAClC,MAAM;oBACN,aACE;gBACJ;gBACA,KAAK;oBACH,MAAM;oBACN,aACE;gBACJ;gBACA,OAAO;oBACL,MAAM;wBAAC;wBAAU;qBAAO;oBACxB,aAAa;gBACf;YACF;YACA,UAAU;gBACR;gBACA;gBACA;gBACA;aACD;YACD,sBAAsB;QACxB;IACF;AACF;AAEO,MAAMC,gCAAgC,CAC3CC,iBACAC,KACAC;IAEA,IAAID,KACF,OAAO,CAAC;;;;;IAKR,EAAEC,kBAAkB;;;EAGtB,EAAEF,gBAAgB;;;;;;AAMpB,EAAEC,IAAI;;AAEN,CAAC;IAGC,OAAO,CAAC;;;;IAIN,EAAEC,kBAAkB;;;EAGtB,EAAEF,gBAAgB;;AAEpB,CAAC;AACD;AAEO,MAAMG,uBAAuB,CAClCb;IAEA,IAAIA,QACF,OAAO,IAAIc,eAAe;QACxB,UAAU;QACV,gBAAgB;YAAC;SAAwB;IAC3C;IAGF,OAAO,IAAIA,eAAe;QACxB,UAAU,CAAC;;;;;;uBAMQ,CAAC;QACpB,gBAAgB;YAAC;YAAmB;SAAwB;IAC9D;AACF"}

package/dist/es/index.mjs

@@ -4,8 +4,8 @@ import insight from "./insight/index.mjs";
 import { getVersion } from "./utils.mjs";
 import { AiLocateElement, PointSchema, RectSchema, SizeSchema, TMultimodalPromptSchema, TUserPromptSchema, describeUserPage, getMidsceneLocationSchema, plan } from "./ai-model/index.mjs";
 import { MIDSCENE_MODEL_NAME } from "@midscene/shared/env";
-import { Agent } from "./agent/index.mjs";
+import { Agent, createAgent } from "./agent/index.mjs";
 const src = insight;
-export { Agent, AiLocateElement, Executor, insight as Insight, MIDSCENE_MODEL_NAME, PointSchema, RectSchema, SizeSchema, TMultimodalPromptSchema, TUserPromptSchema, src as default, describeUserPage, getMidsceneLocationSchema, getVersion, plan, z };
+export { Agent, AiLocateElement, Executor, insight as Insight, MIDSCENE_MODEL_NAME, PointSchema, RectSchema, SizeSchema, TMultimodalPromptSchema, TUserPromptSchema, createAgent, src as default, describeUserPage, getMidsceneLocationSchema, getVersion, plan, z };
 
 //# sourceMappingURL=index.mjs.map

package/dist/es/index.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"index.mjs","sources":["webpack://@midscene/core/./src/index.ts"],"sourcesContent":["import { z } from 'zod';\nimport { Executor } from './ai-model/action-executor';\nimport Insight from './insight/index';\nimport { getVersion } from './utils';\n\nexport {\n  plan,\n  describeUserPage,\n  AiLocateElement,\n  getMidsceneLocationSchema,\n  type MidsceneLocationResultType,\n  PointSchema,\n  SizeSchema,\n  RectSchema,\n  TMultimodalPromptSchema,\n  TUserPromptSchema,\n  type TMultimodalPrompt,\n  type TUserPrompt,\n} from './ai-model/index';\n\nexport { MIDSCENE_MODEL_NAME } from '@midscene/shared/env';\n\nexport type * from './types';\n\nexport { z };\n\nexport default Insight;\nexport { Executor, Insight, getVersion };\n\nexport type {\n  MidsceneYamlScript,\n  MidsceneYamlTask,\n  MidsceneYamlFlowItem,\n  MidsceneYamlFlowItemAIRightClick,\n  MidsceneYamlFlowItemAIDoubleClick,\n  MidsceneYamlConfigResult,\n  LocateOption,\n  DetailedLocateParam,\n} from './yaml';\n\nexport { Agent, type AgentOpt } from './agent';\n"],"names":["Insight"],"mappings":";;;;;;;AA0BA,YAAeA"}
+{"version":3,"file":"index.mjs","sources":["webpack://@midscene/core/./src/index.ts"],"sourcesContent":["import { z } from 'zod';\nimport { Executor } from './ai-model/action-executor';\nimport Insight from './insight/index';\nimport { getVersion } from './utils';\n\nexport {\n  plan,\n  describeUserPage,\n  AiLocateElement,\n  getMidsceneLocationSchema,\n  type MidsceneLocationResultType,\n  PointSchema,\n  SizeSchema,\n  RectSchema,\n  TMultimodalPromptSchema,\n  TUserPromptSchema,\n  type TMultimodalPrompt,\n  type TUserPrompt,\n} from './ai-model/index';\n\nexport { MIDSCENE_MODEL_NAME } from '@midscene/shared/env';\n\nexport type * from './types';\n\nexport { z };\n\nexport default Insight;\nexport { Executor, Insight, getVersion };\n\nexport type {\n  MidsceneYamlScript,\n  MidsceneYamlTask,\n  MidsceneYamlFlowItem,\n  MidsceneYamlFlowItemAIRightClick,\n  MidsceneYamlFlowItemAIDoubleClick,\n  MidsceneYamlConfigResult,\n  LocateOption,\n  DetailedLocateParam,\n} from './yaml';\n\nexport { Agent, type AgentOpt, createAgent } from './agent';\n"],"names":["Insight"],"mappings":";;;;;;;AA0BA,YAAeA"}

package/dist/es/types.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"types.mjs","sources":["webpack://@midscene/core/./src/types.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n\nimport type { NodeType } from '@midscene/shared/constants';\nimport type {\n  BaseElement,\n  ElementTreeNode,\n  Rect,\n  Size,\n} from '@midscene/shared/types';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { z } from 'zod';\nimport type { TUserPrompt } from './ai-model/common';\nimport type { DetailedLocateParam, MidsceneYamlFlowItem } from './yaml';\n\nexport type {\n  ElementTreeNode,\n  BaseElement,\n  Rect,\n  Size,\n  Point,\n} from '@midscene/shared/types';\nexport * from './yaml';\n\nexport type AIUsageInfo = Record<string, any> & {\n  prompt_tokens: number | undefined;\n  completion_tokens: number | undefined;\n  total_tokens: number | undefined;\n  time_cost: number | undefined;\n  model_name: string | undefined;\n  model_description: string | undefined;\n  intent: string | undefined;\n};\n\n/**\n * openai\n *\n */\nexport enum AIResponseFormat {\n  JSON = 'json_object',\n  TEXT = 'text',\n}\n\nexport type AISingleElementResponseById = {\n  id: string;\n  reason?: string;\n  text?: string;\n  xpaths?: string[];\n};\n\nexport type AISingleElementResponseByPosition = {\n  position?: {\n    x: number;\n    y: number;\n  };\n  bbox?: [number, number, number, number];\n  reason: string;\n  text: string;\n};\n\nexport type AISingleElementResponse = AISingleElementResponseById;\nexport interface AIElementLocatorResponse {\n  elements: {\n    id: string;\n    reason?: string;\n    text?: string;\n    xpaths?: string[];\n  }[];\n  bbox?: [number, number, number, number];\n  isOrderSensitive?: boolean;\n  errors?: string[];\n}\n\nexport interface AIElementCoordinatesResponse {\n  bbox: [number, number, number, number];\n  isOrderSensitive?: boolean;\n  errors?: string[];\n}\n\nexport type AIElementResponse =\n  | AIElementLocatorResponse\n  | AIElementCoordinatesResponse;\n\nexport interface AIDataExtractionResponse<DataDemand> {\n  data: DataDemand;\n  errors?: string[];\n  thought?: string;\n}\n\nexport interface AISectionLocatorResponse {\n  bbox: [number, number, number, number];\n  references_bbox?: [number, number, number, number][];\n  error?: string;\n}\n\nexport interface AIAssertionResponse {\n  pass: boolean;\n  thought: string;\n}\n\nexport interface AIDescribeElementResponse {\n  description: string;\n  error?: string;\n}\n\nexport interface LocatorValidatorOption {\n  centerDistanceThreshold?: number;\n}\n\nexport interface LocateValidatorResult {\n  pass: boolean;\n  rect: Rect;\n  center: [number, number];\n  centerDistance?: number;\n}\n\nexport interface AgentDescribeElementAtPointResult {\n  prompt: string;\n  deepThink: boolean;\n  verifyResult?: LocateValidatorResult;\n}\n\n/**\n * context\n */\n\nexport abstract class UIContext<ElementType extends BaseElement = BaseElement> {\n  abstract screenshotBase64: string;\n\n  abstract tree: ElementTreeNode<ElementType>;\n\n  abstract size: Size;\n\n  abstract _isFrozen?: boolean;\n}\n\n/**\n * insight\n */\n\nexport type CallAIFn = <T>(\n  messages: ChatCompletionMessageParam[],\n) => Promise<T>;\n\nexport interface InsightOptions {\n  taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n  aiVendorFn?: CallAIFn;\n}\n\nexport type EnsureObject<T> = { [K in keyof T]: any };\n\nexport type InsightAction = 'locate' | 'extract' | 'assert' | 'describe';\n\nexport type InsightExtractParam = string | Record<string, string>;\n\nexport type LocateResultElement = {\n  center: [number, number];\n  rect: Rect;\n  id: string;\n  indexId?: number;\n  xpaths: string[];\n  attributes: {\n    nodeType: NodeType;\n    [key: string]: string;\n  };\n  isOrderSensitive?: boolean;\n};\n\nexport interface LocateResult {\n  element: LocateResultElement | null;\n  rect?: Rect;\n}\n\nexport interface InsightTaskInfo {\n  durationMs: number;\n  formatResponse?: string;\n  rawResponse?: string;\n  usage?: AIUsageInfo;\n  searchArea?: Rect;\n  searchAreaRawResponse?: string;\n  searchAreaUsage?: AIUsageInfo;\n}\n\nexport interface DumpMeta {\n  sdkVersion: string;\n  logTime: number;\n}\n\nexport interface ReportDumpWithAttributes {\n  dumpString: string;\n  attributes?: Record<string, any>;\n}\n\nexport interface InsightDump extends DumpMeta {\n  type: 'locate' | 'extract' | 'assert';\n  logId: string;\n  userQuery: {\n    element?: TUserPrompt;\n    dataDemand?: InsightExtractParam;\n    assertion?: TUserPrompt;\n  };\n  matchedElement: BaseElement[];\n  matchedRect?: Rect;\n  deepThink?: boolean;\n  data: any;\n  assertionPass?: boolean;\n  assertionThought?: string;\n  taskInfo: InsightTaskInfo;\n  error?: string;\n  output?: any;\n}\n\nexport type PartialInsightDumpFromSDK = Omit<\n  InsightDump,\n  'sdkVersion' | 'logTime' | 'logId' | 'model_name'\n>;\n\nexport type DumpSubscriber = (dump: InsightDump) => Promise<void> | void;\n\n// intermediate variables to optimize the return value by AI\nexport interface LiteUISection {\n  name: string;\n  description: string;\n  sectionCharacteristics: string;\n  textIds: string[];\n}\n\nexport type ElementById = (id: string) => BaseElement | null;\n\nexport type InsightAssertionResponse = AIAssertionResponse & {\n  usage?: AIUsageInfo;\n};\n\n/**\n * agent\n */\n\nexport type OnTaskStartTip = (tip: string) => Promise<void> | void;\n\nexport interface AgentWaitForOpt {\n  checkIntervalMs?: number;\n  timeoutMs?: number;\n}\n\nexport interface AgentAssertOpt {\n  keepRawResponse?: boolean;\n}\n\n/**\n * planning\n *\n */\n\nexport interface PlanningLocateParam extends DetailedLocateParam {\n  id?: string;\n  bbox?: [number, number, number, number];\n}\n\nexport interface PlanningAction<ParamType = any> {\n  thought?: string;\n  type: string;\n  param: ParamType;\n  locate?: PlanningLocateParam | null;\n}\n\nexport interface PlanningAIResponse {\n  action?: PlanningAction; // this is the qwen mode\n  actions?: PlanningAction[];\n  more_actions_needed_by_instruction: boolean;\n  log: string;\n  sleep?: number;\n  error?: string;\n  usage?: AIUsageInfo;\n  rawResponse?: string;\n  yamlFlow?: MidsceneYamlFlowItem[];\n  yamlString?: string;\n}\n\nexport type PlanningActionParamTap = null;\nexport type PlanningActionParamHover = null;\nexport type PlanningActionParamRightClick = null;\n\nexport interface PlanningActionParamInputOrKeyPress {\n  value: string;\n  autoDismissKeyboard?: boolean;\n}\n\nexport interface PlanningActionParamSleep {\n  timeMs: number;\n}\n\nexport interface PlanningActionParamError {\n  thought: string;\n}\n\nexport type PlanningActionParamWaitFor = AgentWaitForOpt & {};\n\nexport interface AndroidLongPressParam {\n  duration?: number;\n}\n\nexport interface AndroidPullParam {\n  direction: 'up' | 'down';\n  distance?: number;\n  duration?: number;\n}\n/**\n * misc\n */\n\nexport interface Color {\n  name: string;\n  hex: string;\n}\n\nexport interface BaseAgentParserOpt {\n  selector?: string;\n}\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PuppeteerParserOpt extends BaseAgentParserOpt {}\n\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PlaywrightParserOpt extends BaseAgentParserOpt {}\n\n/*\naction\n*/\nexport interface ExecutionTaskProgressOptions {\n  onTaskStart?: (task: ExecutionTask) => Promise<void> | void;\n}\n\nexport interface ExecutionRecorderItem {\n  type: 'screenshot';\n  ts: number;\n  screenshot?: string;\n  timing?: string;\n}\n\nexport type ExecutionTaskType =\n  | 'Planning'\n  | 'Insight'\n  | 'Action'\n  | 'Assertion'\n  | 'Log';\n\nexport interface ExecutorContext {\n  task: ExecutionTask;\n  element?: LocateResultElement | null;\n}\n\nexport interface ExecutionTaskApply<\n  Type extends ExecutionTaskType = any,\n  TaskParam = any,\n  TaskOutput = any,\n  TaskLog = any,\n> {\n  type: Type;\n  subType?: string;\n  param?: TaskParam;\n  thought?: string;\n  locate?: PlanningLocateParam | null;\n  uiContext?: UIContext;\n  executor: (\n    param: TaskParam,\n    context: ExecutorContext,\n  ) => // biome-ignore lint/suspicious/noConfusingVoidType: <explanation>\n    | Promise<ExecutionTaskReturn<TaskOutput, TaskLog> | undefined | void>\n    | undefined\n    | void;\n}\n\nexport interface ExecutionTaskHitBy {\n  from: string;\n  context: Record<string, any>;\n}\n\nexport interface ExecutionTaskReturn<TaskOutput = unknown, TaskLog = unknown> {\n  output?: TaskOutput;\n  log?: TaskLog;\n  recorder?: ExecutionRecorderItem[];\n  hitBy?: ExecutionTaskHitBy;\n}\n\nexport type ExecutionTask<\n  E extends ExecutionTaskApply<any, any, any> = ExecutionTaskApply<\n    any,\n    any,\n    any\n  >,\n> = E &\n  ExecutionTaskReturn<\n    E extends ExecutionTaskApply<any, any, infer TaskOutput, any>\n      ? TaskOutput\n      : unknown,\n    E extends ExecutionTaskApply<any, any, any, infer TaskLog>\n      ? TaskLog\n      : unknown\n  > & {\n    status: 'pending' | 'running' | 'finished' | 'failed' | 'cancelled';\n    error?: Error;\n    errorMessage?: string;\n    errorStack?: string;\n    timing?: {\n      start: number;\n      end?: number;\n      cost?: number;\n    };\n    usage?: AIUsageInfo;\n  };\n\nexport interface ExecutionDump extends DumpMeta {\n  name: string;\n  description?: string;\n  tasks: ExecutionTask[];\n}\n\n/*\ntask - insight-locate\n*/\nexport type ExecutionTaskInsightLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskInsightLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport interface ExecutionTaskInsightDumpLog {\n  dump?: InsightDump;\n}\n\nexport type ExecutionTaskInsightLocateApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightLocateParam,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightLocate =\n  ExecutionTask<ExecutionTaskInsightLocateApply>;\n\n/*\ntask - insight-query\n*/\nexport interface ExecutionTaskInsightQueryParam {\n  dataDemand: InsightExtractParam;\n}\n\nexport interface ExecutionTaskInsightQueryOutput {\n  data: any;\n}\n\nexport type ExecutionTaskInsightQueryApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightQueryParam,\n  any,\n  ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightQuery =\n  ExecutionTask<ExecutionTaskInsightQueryApply>;\n\n/*\ntask - assertion\n*/\nexport interface ExecutionTaskInsightAssertionParam {\n  assertion: string;\n}\n\nexport type ExecutionTaskInsightAssertionApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightAssertionParam,\n  InsightAssertionResponse,\n  ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightAssertion =\n  ExecutionTask<ExecutionTaskInsightAssertionApply>;\n\n/*\ntask - action (i.e. interact) \n*/\nexport type ExecutionTaskActionApply<ActionParam = any> = ExecutionTaskApply<\n  'Action',\n  ActionParam,\n  void,\n  void\n>;\n\nexport type ExecutionTaskAction = ExecutionTask<ExecutionTaskActionApply>;\n\n/*\ntask - Log\n*/\n\nexport type ExecutionTaskLogApply<\n  LogParam = {\n    content: string;\n  },\n> = ExecutionTaskApply<'Log', LogParam, void, void>;\n\nexport type ExecutionTaskLog = ExecutionTask<ExecutionTaskLogApply>;\n\n/*\ntask - planning\n*/\n\nexport type ExecutionTaskPlanningApply = ExecutionTaskApply<\n  'Planning',\n  {\n    userInstruction: string;\n    log?: string;\n  },\n  PlanningAIResponse\n>;\n\nexport type ExecutionTaskPlanning = ExecutionTask<ExecutionTaskPlanningApply>;\n\n/*\nGrouped dump\n*/\nexport interface GroupedActionDump {\n  groupName: string;\n  groupDescription?: string;\n  modelBriefs: string[];\n  executions: ExecutionDump[];\n}\n\nexport type InterfaceType =\n  | 'puppeteer'\n  | 'playwright'\n  | 'static'\n  | 'chrome-extension-proxy'\n  | 'android'\n  | string;\n\nexport interface StreamingCodeGenerationOptions {\n  /** Whether to enable streaming output */\n  stream?: boolean;\n  /** Callback function to handle streaming chunks */\n  onChunk?: StreamingCallback;\n  /** Callback function to handle streaming completion */\n  onComplete?: (finalCode: string) => void;\n  /** Callback function to handle streaming errors */\n  onError?: (error: Error) => void;\n}\n\nexport type StreamingCallback = (chunk: CodeGenerationChunk) => void;\n\nexport interface CodeGenerationChunk {\n  /** The incremental content chunk */\n  content: string;\n  /** The reasoning content */\n  reasoning_content: string;\n  /** The accumulated content so far */\n  accumulated: string;\n  /** Whether this is the final chunk */\n  isComplete: boolean;\n  /** Token usage information if available */\n  usage?: AIUsageInfo;\n}\n\nexport interface StreamingAIResponse {\n  /** The final accumulated content */\n  content: string;\n  /** Token usage information */\n  usage?: AIUsageInfo;\n  /** Whether the response was streamed */\n  isStreamed: boolean;\n}\n\nexport interface DeviceAction<T = any> {\n  name: string;\n  description?: string;\n  interfaceAlias?: string;\n  paramSchema?: z.ZodType<T>;\n  call: (param: T, context: ExecutorContext) => Promise<void> | void;\n}\n\n/**\n * Web-specific types\n */\nexport interface WebElementInfo extends BaseElement {\n  id: string;\n  attributes: {\n    nodeType: NodeType;\n    [key: string]: string;\n  };\n}\n\nexport type WebUIContext = UIContext<WebElementInfo>;\n"],"names":["AIResponseFormat","UIContext"],"mappings":";AAqCO,IAAKA,yBAAgBA,WAAAA,GAAAA,SAAhBA,gBAAgB;;;WAAhBA;;AAwFL,MAAeC;AAQtB"}
+{"version":3,"file":"types.mjs","sources":["webpack://@midscene/core/./src/types.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n\nimport type { NodeType } from '@midscene/shared/constants';\nimport type { TModelConfigFn } from '@midscene/shared/env';\nimport type {\n  BaseElement,\n  ElementTreeNode,\n  Rect,\n  Size,\n} from '@midscene/shared/types';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { z } from 'zod';\nimport type { TUserPrompt } from './ai-model/common';\nimport type { DetailedLocateParam, MidsceneYamlFlowItem } from './yaml';\n\nexport type {\n  ElementTreeNode,\n  BaseElement,\n  Rect,\n  Size,\n  Point,\n} from '@midscene/shared/types';\nexport * from './yaml';\n\nexport type AIUsageInfo = Record<string, any> & {\n  prompt_tokens: number | undefined;\n  completion_tokens: number | undefined;\n  total_tokens: number | undefined;\n  time_cost: number | undefined;\n  model_name: string | undefined;\n  model_description: string | undefined;\n  intent: string | undefined;\n};\n\n/**\n * openai\n *\n */\nexport enum AIResponseFormat {\n  JSON = 'json_object',\n  TEXT = 'text',\n}\n\nexport type AISingleElementResponseById = {\n  id: string;\n  reason?: string;\n  text?: string;\n  xpaths?: string[];\n};\n\nexport type AISingleElementResponseByPosition = {\n  position?: {\n    x: number;\n    y: number;\n  };\n  bbox?: [number, number, number, number];\n  reason: string;\n  text: string;\n};\n\nexport type AISingleElementResponse = AISingleElementResponseById;\nexport interface AIElementLocatorResponse {\n  elements: {\n    id: string;\n    reason?: string;\n    text?: string;\n    xpaths?: string[];\n  }[];\n  bbox?: [number, number, number, number];\n  isOrderSensitive?: boolean;\n  errors?: string[];\n}\n\nexport interface AIElementCoordinatesResponse {\n  bbox: [number, number, number, number];\n  isOrderSensitive?: boolean;\n  errors?: string[];\n}\n\nexport type AIElementResponse =\n  | AIElementLocatorResponse\n  | AIElementCoordinatesResponse;\n\nexport interface AIDataExtractionResponse<DataDemand> {\n  data: DataDemand;\n  errors?: string[];\n  thought?: string;\n}\n\nexport interface AISectionLocatorResponse {\n  bbox: [number, number, number, number];\n  references_bbox?: [number, number, number, number][];\n  error?: string;\n}\n\nexport interface AIAssertionResponse {\n  pass: boolean;\n  thought: string;\n}\n\nexport interface AIDescribeElementResponse {\n  description: string;\n  error?: string;\n}\n\nexport interface LocatorValidatorOption {\n  centerDistanceThreshold?: number;\n}\n\nexport interface LocateValidatorResult {\n  pass: boolean;\n  rect: Rect;\n  center: [number, number];\n  centerDistance?: number;\n}\n\nexport interface AgentDescribeElementAtPointResult {\n  prompt: string;\n  deepThink: boolean;\n  verifyResult?: LocateValidatorResult;\n}\n\n/**\n * context\n */\n\nexport abstract class UIContext<ElementType extends BaseElement = BaseElement> {\n  abstract screenshotBase64: string;\n\n  abstract tree: ElementTreeNode<ElementType>;\n\n  abstract size: Size;\n\n  abstract _isFrozen?: boolean;\n}\n\n/**\n * insight\n */\n\nexport type CallAIFn = <T>(\n  messages: ChatCompletionMessageParam[],\n) => Promise<T>;\n\nexport interface InsightOptions {\n  taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n  aiVendorFn?: CallAIFn;\n}\n\nexport type EnsureObject<T> = { [K in keyof T]: any };\n\nexport type InsightAction = 'locate' | 'extract' | 'assert' | 'describe';\n\nexport type InsightExtractParam = string | Record<string, string>;\n\nexport type LocateResultElement = {\n  center: [number, number];\n  rect: Rect;\n  id: string;\n  indexId?: number;\n  xpaths: string[];\n  attributes: {\n    nodeType: NodeType;\n    [key: string]: string;\n  };\n  isOrderSensitive?: boolean;\n};\n\nexport interface LocateResult {\n  element: LocateResultElement | null;\n  rect?: Rect;\n}\n\nexport interface InsightTaskInfo {\n  durationMs: number;\n  formatResponse?: string;\n  rawResponse?: string;\n  usage?: AIUsageInfo;\n  searchArea?: Rect;\n  searchAreaRawResponse?: string;\n  searchAreaUsage?: AIUsageInfo;\n}\n\nexport interface DumpMeta {\n  sdkVersion: string;\n  logTime: number;\n}\n\nexport interface ReportDumpWithAttributes {\n  dumpString: string;\n  attributes?: Record<string, any>;\n}\n\nexport interface InsightDump extends DumpMeta {\n  type: 'locate' | 'extract' | 'assert';\n  logId: string;\n  userQuery: {\n    element?: TUserPrompt;\n    dataDemand?: InsightExtractParam;\n    assertion?: TUserPrompt;\n  };\n  matchedElement: BaseElement[];\n  matchedRect?: Rect;\n  deepThink?: boolean;\n  data: any;\n  assertionPass?: boolean;\n  assertionThought?: string;\n  taskInfo: InsightTaskInfo;\n  error?: string;\n  output?: any;\n}\n\nexport type PartialInsightDumpFromSDK = Omit<\n  InsightDump,\n  'sdkVersion' | 'logTime' | 'logId' | 'model_name'\n>;\n\nexport type DumpSubscriber = (dump: InsightDump) => Promise<void> | void;\n\n// intermediate variables to optimize the return value by AI\nexport interface LiteUISection {\n  name: string;\n  description: string;\n  sectionCharacteristics: string;\n  textIds: string[];\n}\n\nexport type ElementById = (id: string) => BaseElement | null;\n\nexport type InsightAssertionResponse = AIAssertionResponse & {\n  usage?: AIUsageInfo;\n};\n\n/**\n * agent\n */\n\nexport type OnTaskStartTip = (tip: string) => Promise<void> | void;\n\nexport interface AgentWaitForOpt {\n  checkIntervalMs?: number;\n  timeoutMs?: number;\n}\n\nexport interface AgentAssertOpt {\n  keepRawResponse?: boolean;\n}\n\n/**\n * planning\n *\n */\n\nexport interface PlanningLocateParam extends DetailedLocateParam {\n  id?: string;\n  bbox?: [number, number, number, number];\n}\n\nexport interface PlanningAction<ParamType = any> {\n  thought?: string;\n  type: string;\n  param: ParamType;\n  locate?: PlanningLocateParam | null;\n}\n\nexport interface PlanningAIResponse {\n  action?: PlanningAction; // this is the qwen mode\n  actions?: PlanningAction[];\n  more_actions_needed_by_instruction: boolean;\n  log: string;\n  sleep?: number;\n  error?: string;\n  usage?: AIUsageInfo;\n  rawResponse?: string;\n  yamlFlow?: MidsceneYamlFlowItem[];\n  yamlString?: string;\n}\n\nexport type PlanningActionParamTap = null;\nexport type PlanningActionParamHover = null;\nexport type PlanningActionParamRightClick = null;\n\nexport interface PlanningActionParamInputOrKeyPress {\n  value: string;\n  autoDismissKeyboard?: boolean;\n}\n\nexport interface PlanningActionParamSleep {\n  timeMs: number;\n}\n\nexport interface PlanningActionParamError {\n  thought: string;\n}\n\nexport type PlanningActionParamWaitFor = AgentWaitForOpt & {};\n\nexport interface AndroidLongPressParam {\n  duration?: number;\n}\n\nexport interface AndroidPullParam {\n  direction: 'up' | 'down';\n  distance?: number;\n  duration?: number;\n}\n/**\n * misc\n */\n\nexport interface Color {\n  name: string;\n  hex: string;\n}\n\nexport interface BaseAgentParserOpt {\n  selector?: string;\n}\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PuppeteerParserOpt extends BaseAgentParserOpt {}\n\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PlaywrightParserOpt extends BaseAgentParserOpt {}\n\n/*\naction\n*/\nexport interface ExecutionTaskProgressOptions {\n  onTaskStart?: (task: ExecutionTask) => Promise<void> | void;\n}\n\nexport interface ExecutionRecorderItem {\n  type: 'screenshot';\n  ts: number;\n  screenshot?: string;\n  timing?: string;\n}\n\nexport type ExecutionTaskType =\n  | 'Planning'\n  | 'Insight'\n  | 'Action'\n  | 'Assertion'\n  | 'Log';\n\nexport interface ExecutorContext {\n  task: ExecutionTask;\n  element?: LocateResultElement | null;\n}\n\nexport interface ExecutionTaskApply<\n  Type extends ExecutionTaskType = any,\n  TaskParam = any,\n  TaskOutput = any,\n  TaskLog = any,\n> {\n  type: Type;\n  subType?: string;\n  param?: TaskParam;\n  thought?: string;\n  locate?: PlanningLocateParam | null;\n  uiContext?: UIContext;\n  executor: (\n    param: TaskParam,\n    context: ExecutorContext,\n  ) => // biome-ignore lint/suspicious/noConfusingVoidType: <explanation>\n    | Promise<ExecutionTaskReturn<TaskOutput, TaskLog> | undefined | void>\n    | undefined\n    | void;\n}\n\nexport interface ExecutionTaskHitBy {\n  from: string;\n  context: Record<string, any>;\n}\n\nexport interface ExecutionTaskReturn<TaskOutput = unknown, TaskLog = unknown> {\n  output?: TaskOutput;\n  log?: TaskLog;\n  recorder?: ExecutionRecorderItem[];\n  hitBy?: ExecutionTaskHitBy;\n}\n\nexport type ExecutionTask<\n  E extends ExecutionTaskApply<any, any, any> = ExecutionTaskApply<\n    any,\n    any,\n    any\n  >,\n> = E &\n  ExecutionTaskReturn<\n    E extends ExecutionTaskApply<any, any, infer TaskOutput, any>\n      ? TaskOutput\n      : unknown,\n    E extends ExecutionTaskApply<any, any, any, infer TaskLog>\n      ? TaskLog\n      : unknown\n  > & {\n    status: 'pending' | 'running' | 'finished' | 'failed' | 'cancelled';\n    error?: Error;\n    errorMessage?: string;\n    errorStack?: string;\n    timing?: {\n      start: number;\n      end?: number;\n      cost?: number;\n    };\n    usage?: AIUsageInfo;\n  };\n\nexport interface ExecutionDump extends DumpMeta {\n  name: string;\n  description?: string;\n  tasks: ExecutionTask[];\n}\n\n/*\ntask - insight-locate\n*/\nexport type ExecutionTaskInsightLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskInsightLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport interface ExecutionTaskInsightDumpLog {\n  dump?: InsightDump;\n}\n\nexport type ExecutionTaskInsightLocateApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightLocateParam,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightLocate =\n  ExecutionTask<ExecutionTaskInsightLocateApply>;\n\n/*\ntask - insight-query\n*/\nexport interface ExecutionTaskInsightQueryParam {\n  dataDemand: InsightExtractParam;\n}\n\nexport interface ExecutionTaskInsightQueryOutput {\n  data: any;\n}\n\nexport type ExecutionTaskInsightQueryApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightQueryParam,\n  any,\n  ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightQuery =\n  ExecutionTask<ExecutionTaskInsightQueryApply>;\n\n/*\ntask - assertion\n*/\nexport interface ExecutionTaskInsightAssertionParam {\n  assertion: string;\n}\n\nexport type ExecutionTaskInsightAssertionApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightAssertionParam,\n  InsightAssertionResponse,\n  ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightAssertion =\n  ExecutionTask<ExecutionTaskInsightAssertionApply>;\n\n/*\ntask - action (i.e. interact) \n*/\nexport type ExecutionTaskActionApply<ActionParam = any> = ExecutionTaskApply<\n  'Action',\n  ActionParam,\n  void,\n  void\n>;\n\nexport type ExecutionTaskAction = ExecutionTask<ExecutionTaskActionApply>;\n\n/*\ntask - Log\n*/\n\nexport type ExecutionTaskLogApply<\n  LogParam = {\n    content: string;\n  },\n> = ExecutionTaskApply<'Log', LogParam, void, void>;\n\nexport type ExecutionTaskLog = ExecutionTask<ExecutionTaskLogApply>;\n\n/*\ntask - planning\n*/\n\nexport type ExecutionTaskPlanningApply = ExecutionTaskApply<\n  'Planning',\n  {\n    userInstruction: string;\n    log?: string;\n  },\n  PlanningAIResponse\n>;\n\nexport type ExecutionTaskPlanning = ExecutionTask<ExecutionTaskPlanningApply>;\n\n/*\nGrouped dump\n*/\nexport interface GroupedActionDump {\n  groupName: string;\n  groupDescription?: string;\n  modelBriefs: string[];\n  executions: ExecutionDump[];\n}\n\nexport type InterfaceType =\n  | 'puppeteer'\n  | 'playwright'\n  | 'static'\n  | 'chrome-extension-proxy'\n  | 'android'\n  | string;\n\nexport interface StreamingCodeGenerationOptions {\n  /** Whether to enable streaming output */\n  stream?: boolean;\n  /** Callback function to handle streaming chunks */\n  onChunk?: StreamingCallback;\n  /** Callback function to handle streaming completion */\n  onComplete?: (finalCode: string) => void;\n  /** Callback function to handle streaming errors */\n  onError?: (error: Error) => void;\n}\n\nexport type StreamingCallback = (chunk: CodeGenerationChunk) => void;\n\nexport interface CodeGenerationChunk {\n  /** The incremental content chunk */\n  content: string;\n  /** The reasoning content */\n  reasoning_content: string;\n  /** The accumulated content so far */\n  accumulated: string;\n  /** Whether this is the final chunk */\n  isComplete: boolean;\n  /** Token usage information if available */\n  usage?: AIUsageInfo;\n}\n\nexport interface StreamingAIResponse {\n  /** The final accumulated content */\n  content: string;\n  /** Token usage information */\n  usage?: AIUsageInfo;\n  /** Whether the response was streamed */\n  isStreamed: boolean;\n}\n\nexport interface DeviceAction<T = any> {\n  name: string;\n  description?: string;\n  interfaceAlias?: string;\n  paramSchema?: z.ZodType<T>;\n  call: (param: T, context: ExecutorContext) => Promise<void> | void;\n}\n\n/**\n * Web-specific types\n */\nexport interface WebElementInfo extends BaseElement {\n  id: string;\n  attributes: {\n    nodeType: NodeType;\n    [key: string]: string;\n  };\n}\n\nexport type WebUIContext = UIContext<WebElementInfo>;\n\n/**\n * Agent\n */\n\nexport interface AgentOpt {\n  testId?: string;\n  cacheId?: string;\n  groupName?: string;\n  groupDescription?: string;\n  /* if auto generate report, default true */\n  generateReport?: boolean;\n  /* if auto print report msg, default true */\n  autoPrintReportMsg?: boolean;\n  onTaskStartTip?: OnTaskStartTip;\n  aiActionContext?: string;\n  /* custom report file name */\n  reportFileName?: string;\n  modelConfig?: TModelConfigFn;\n}\n"],"names":["AIResponseFormat","UIContext"],"mappings":";AAsCO,IAAKA,yBAAgBA,WAAAA,GAAAA,SAAhBA,gBAAgB;;;WAAhBA;;AAwFL,MAAeC;AAQtB"}