@midscene/core 1.7.2 → 1.7.3-beta-20260410094527.0

package/dist/es/agent/tasks.mjs

@@ -266,16 +266,17 @@ class TaskExecutor {
                 const ifTypeRestricted = 'Query' !== type;
                 let demandInput = demand;
                 let keyOfResult = 'result';
+                const currentScreenshotConstraint = 'based on the current screenshot and its contents if provided, unless the user explicitly asks to compare with reference images';
                 if (ifTypeRestricted && ('Assert' === type || 'WaitFor' === type)) {
                     keyOfResult = 'StatementIsTruthy';
-                    const booleanPrompt = 'Assert' === type ? `Boolean, whether the following statement is true: ${demand}` : `Boolean, the user wants to do some 'wait for' operation, please check whether the following statement is true: ${demand}`;
+                    const booleanPrompt = 'Assert' === type ? `Boolean, ${currentScreenshotConstraint}, whether the following statement is true: ${demand}` : `Boolean, the user wants to do some 'wait for' operation. ${currentScreenshotConstraint}, please check whether the following statement is true: ${demand}`;
                     demandInput = {
                         [keyOfResult]: booleanPrompt
                     };
                 } else if (ifTypeRestricted) {
                     keyOfResult = type;
                     demandInput = {
-                        [keyOfResult]: `${type}, ${demand}`
+                        [keyOfResult]: `${type}, ${currentScreenshotConstraint}, ${demand}`
                     };
                 }
                 let extractResult;

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

@@ -1 +1 @@
-{"version":3,"file":"agent/tasks.mjs","sources":["../../../src/agent/tasks.ts"],"sourcesContent":["import {\n  AIResponseParseError,\n  ConversationHistory,\n  autoGLMPlanning,\n  plan,\n  uiTarsPlanning,\n} from '@/ai-model';\nimport { isAutoGLM, isUITars } from '@/ai-model/auto-glm/util';\nimport {\n  type TMultimodalPrompt,\n  type TUserPrompt,\n  getReadableTimeString,\n} from '@/common';\nimport type { AbstractInterface, FileChooserHandler } from '@/device';\nimport type Service from '@/service';\nimport type { TaskRunner } from '@/task-runner';\nimport { TaskExecutionError } from '@/task-runner';\nimport type {\n  DeepThinkOption,\n  DeviceAction,\n  ExecutionTaskApply,\n  ExecutionTaskInsightQueryApply,\n  ExecutionTaskPlanningApply,\n  ExecutionTaskProgressOptions,\n  InterfaceType,\n  MidsceneYamlFlowItem,\n  PlanningAIResponse,\n  PlanningAction,\n  PlanningActionParamWaitFor,\n  ServiceDump,\n  ServiceExtractOption,\n  ServiceExtractParam,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport { type IModelConfig, getCurrentTime } from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { ExecutionSession } from './execution-session';\nimport { TaskBuilder } from './task-builder';\nimport type { TaskCache } from './task-cache';\nexport { locatePlanForLocate } from './task-builder';\nimport { setTimingFieldOnce } from '@/task-timing';\nimport { descriptionOfTree } from '@midscene/shared/extractor';\nimport { taskTitleStr } from './ui-utils';\nimport { parsePrompt } from './utils';\n\ninterface ExecutionResult<OutputType = any> {\n  output: OutputType;\n  thought?: string;\n  runner: TaskRunner;\n}\n\ninterface TaskExecutorHooks {\n  onTaskUpdate?: (\n    runner: TaskRunner,\n    error?: TaskExecutionError,\n  ) => Promise<void> | void;\n}\n\nconst debug = getDebug('device-task-executor');\nconst maxErrorCountAllowedInOnePlanningLoop = 5;\n\nexport { TaskExecutionError };\n\nexport class TaskExecutor {\n  interface: AbstractInterface;\n\n  service: Service;\n\n  taskCache?: TaskCache;\n\n  private readonly providedActionSpace: DeviceAction[];\n\n  private readonly taskBuilder: TaskBuilder;\n\n  onTaskStartCallback?: ExecutionTaskProgressOptions['onTaskStart'];\n\n  private readonly hooks?: TaskExecutorHooks;\n\n  replanningCycleLimit?: number;\n\n  waitAfterAction?: number;\n\n  useDeviceTimestamp?: boolean;\n\n  // @deprecated use .interface instead\n  get page() {\n    return this.interface;\n  }\n\n  constructor(\n    interfaceInstance: AbstractInterface,\n    service: Service,\n    opts: {\n      taskCache?: TaskCache;\n      onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n      replanningCycleLimit?: number;\n      waitAfterAction?: number;\n      useDeviceTimestamp?: boolean;\n      hooks?: TaskExecutorHooks;\n      actionSpace: DeviceAction[];\n    },\n  ) {\n    this.interface = interfaceInstance;\n    this.service = service;\n    this.taskCache = opts.taskCache;\n    this.onTaskStartCallback = opts?.onTaskStart;\n    this.replanningCycleLimit = opts.replanningCycleLimit;\n    this.waitAfterAction = opts.waitAfterAction;\n    this.useDeviceTimestamp = opts.useDeviceTimestamp;\n    this.hooks = opts.hooks;\n    this.providedActionSpace = opts.actionSpace;\n    this.taskBuilder = new TaskBuilder({\n      interfaceInstance,\n      service,\n      taskCache: opts.taskCache,\n      actionSpace: this.getActionSpace(),\n      waitAfterAction: opts.waitAfterAction,\n    });\n  }\n\n  private createExecutionSession(\n    title: string,\n    options?: { tasks?: ExecutionTaskApply[] },\n  ) {\n    return new ExecutionSession(\n      title,\n      () => Promise.resolve(this.service.contextRetrieverFn()),\n      {\n        onTaskStart: this.onTaskStartCallback,\n        tasks: options?.tasks,\n        onTaskUpdate: this.hooks?.onTaskUpdate,\n      },\n    );\n  }\n\n  private getActionSpace(): DeviceAction[] {\n    return this.providedActionSpace;\n  }\n\n  /**\n   * Get a readable time string using device time when configured.\n   * This method respects the useDeviceTimestamp configuration.\n   * @param format - Optional format string\n   * @returns A formatted time string\n   */\n  private async getTimeString(format?: string): Promise<string> {\n    const timestamp = await getCurrentTime(\n      this.interface,\n      this.useDeviceTimestamp,\n    );\n    return getReadableTimeString(format, timestamp);\n  }\n\n  public async convertPlanToExecutable(\n    plans: PlanningAction[],\n    modelConfigForPlanning: IModelConfig,\n    modelConfigForDefaultIntent: IModelConfig,\n    options?: {\n      cacheable?: boolean;\n      deepLocate?: boolean;\n      abortSignal?: AbortSignal;\n    },\n  ) {\n    return this.taskBuilder.build(\n      plans,\n      modelConfigForPlanning,\n      modelConfigForDefaultIntent,\n      options,\n    );\n  }\n\n  async loadYamlFlowAsPlanning(userInstruction: string, yamlString: string) {\n    const session = this.createExecutionSession(\n      taskTitleStr('Act', userInstruction),\n    );\n\n    const task: ExecutionTaskPlanningApply = {\n      type: 'Planning',\n      subType: 'LoadYaml',\n      param: {\n        userInstruction,\n      },\n      executor: async (param, executorContext) => {\n        const { uiContext } = executorContext;\n        assert(uiContext, 'uiContext is required for Planning task');\n        return {\n          output: {\n            actions: [],\n            shouldContinuePlanning: false,\n            log: '',\n            yamlString,\n          },\n          cache: {\n            hit: true,\n          },\n          hitBy: {\n            from: 'Cache',\n            context: {\n              yamlString,\n            },\n          },\n        };\n      },\n    };\n    const runner = session.getRunner();\n    await session.appendAndRun(task);\n\n    return {\n      runner,\n    };\n  }\n\n  async runPlans(\n    title: string,\n    plans: PlanningAction[],\n    modelConfigForPlanning: IModelConfig,\n    modelConfigForDefaultIntent: IModelConfig,\n  ): Promise<ExecutionResult> {\n    const session = this.createExecutionSession(title);\n    const { tasks } = await this.convertPlanToExecutable(\n      plans,\n      modelConfigForPlanning,\n      modelConfigForDefaultIntent,\n    );\n    const runner = session.getRunner();\n    const result = await session.appendAndRun(tasks);\n    const { output } = result ?? {};\n    return {\n      output,\n      runner,\n    };\n  }\n\n  async action(\n    userPrompt: string,\n    modelConfigForPlanning: IModelConfig,\n    modelConfigForDefaultIntent: IModelConfig,\n    includeBboxInPlanning: boolean,\n    aiActContext?: string,\n    cacheable?: boolean,\n    replanningCycleLimitOverride?: number,\n    imagesIncludeCount?: number,\n    deepThink?: DeepThinkOption,\n    fileChooserAccept?: string[],\n    deepLocate?: boolean,\n    abortSignal?: AbortSignal,\n  ): Promise<\n    ExecutionResult<\n      | {\n          yamlFlow?: MidsceneYamlFlowItem[]; // for cache use\n          output?: string;\n        }\n      | undefined\n    >\n  > {\n    return withFileChooser(this.interface, fileChooserAccept, async () => {\n      return this.runAction(\n        userPrompt,\n        modelConfigForPlanning,\n        modelConfigForDefaultIntent,\n        includeBboxInPlanning,\n        aiActContext,\n        cacheable,\n        replanningCycleLimitOverride,\n        imagesIncludeCount,\n        deepThink,\n        deepLocate,\n        abortSignal,\n      );\n    });\n  }\n\n  private async runAction(\n    userPrompt: string,\n    modelConfigForPlanning: IModelConfig,\n    modelConfigForDefaultIntent: IModelConfig,\n    includeBboxInPlanning: boolean,\n    aiActContext?: string,\n    cacheable?: boolean,\n    replanningCycleLimitOverride?: number,\n    imagesIncludeCount?: number,\n    deepThink?: DeepThinkOption,\n    deepLocate?: boolean,\n    abortSignal?: AbortSignal,\n  ): Promise<\n    ExecutionResult<\n      | {\n          yamlFlow?: MidsceneYamlFlowItem[]; // for cache use\n          output?: string;\n        }\n      | undefined\n    >\n  > {\n    const conversationHistory = new ConversationHistory();\n\n    const session = this.createExecutionSession(\n      taskTitleStr('Act', userPrompt),\n    );\n    const runner = session.getRunner();\n\n    let replanCount = 0;\n    const yamlFlow: MidsceneYamlFlowItem[] = [];\n    const replanningCycleLimit =\n      replanningCycleLimitOverride ?? this.replanningCycleLimit;\n    assert(\n      replanningCycleLimit !== undefined,\n      'replanningCycleLimit is required for TaskExecutor.action',\n    );\n\n    let errorCountInOnePlanningLoop = 0; // count the number of errors in one planning loop\n    let outputString: string | undefined;\n\n    // Main planning loop - unified plan/replan logic\n    while (true) {\n      // Check abort signal before each planning cycle\n      if (abortSignal?.aborted) {\n        return session.appendErrorPlan(\n          `Task aborted: ${abortSignal.reason || 'abort signal received'}`,\n        );\n      }\n\n      // Get sub-goal status text if available\n      const subGoalStatus = conversationHistory.subGoalsToText() || undefined;\n\n      // Get memories text if available\n      const memoriesStatus = conversationHistory.memoriesToText() || undefined;\n\n      const result = await session.appendAndRun(\n        {\n          type: 'Planning',\n          subType: 'Plan',\n          param: {\n            userInstruction: userPrompt,\n            aiActContext,\n            imagesIncludeCount,\n            deepThink,\n            ...(subGoalStatus ? { subGoalStatus } : {}),\n            ...(memoriesStatus ? { memoriesStatus } : {}),\n          },\n          executor: async (param, executorContext) => {\n            const { uiContext } = executorContext;\n            assert(uiContext, 'uiContext is required for Planning task');\n            const { modelFamily } = modelConfigForPlanning;\n            const timing = executorContext.task.timing;\n\n            const actionSpace = this.getActionSpace();\n            debug(\n              'actionSpace for this interface is:',\n              actionSpace.map((action) => action.name).join(', '),\n            );\n            assert(Array.isArray(actionSpace), 'actionSpace must be an array');\n            if (actionSpace.length === 0) {\n              console.warn(\n                `ActionSpace for ${this.interface.interfaceType} is empty. This may lead to unexpected behavior.`,\n              );\n            }\n\n            const planImpl = isUITars(modelFamily)\n              ? uiTarsPlanning\n              : isAutoGLM(modelFamily)\n                ? autoGLMPlanning\n                : plan;\n\n            let planResult: Awaited<ReturnType<typeof planImpl>>;\n            try {\n              setTimingFieldOnce(timing, 'callAiStart');\n              planResult = await planImpl(param.userInstruction, {\n                context: uiContext,\n                actionContext: param.aiActContext,\n                interfaceType: this.interface.interfaceType as InterfaceType,\n                actionSpace,\n                modelConfig: modelConfigForPlanning,\n                conversationHistory,\n                includeBbox: includeBboxInPlanning,\n                imagesIncludeCount,\n                deepThink,\n                abortSignal,\n              });\n            } catch (planError) {\n              if (planError instanceof AIResponseParseError) {\n                // Record usage and rawResponse even when parsing fails\n                executorContext.task.usage = planError.usage;\n                executorContext.task.log = {\n                  ...(executorContext.task.log || {}),\n                  rawResponse: planError.rawResponse,\n                };\n              }\n              throw planError;\n            } finally {\n              setTimingFieldOnce(timing, 'callAiEnd');\n            }\n            debug('planResult', JSON.stringify(planResult, null, 2));\n\n            const {\n              actions,\n              thought,\n              log,\n              memory,\n              error,\n              usage,\n              rawResponse,\n              reasoning_content,\n              finalizeSuccess,\n              finalizeMessage,\n              updateSubGoals,\n              markFinishedIndexes,\n            } = planResult;\n            outputString = finalizeMessage;\n\n            executorContext.task.log = {\n              ...(executorContext.task.log || {}),\n              rawResponse,\n            };\n            executorContext.task.usage = usage;\n            executorContext.task.reasoning_content = reasoning_content;\n            executorContext.task.output = {\n              actions: actions || [],\n              log,\n              thought,\n              memory,\n              yamlFlow: planResult.yamlFlow,\n              output: finalizeMessage,\n              shouldContinuePlanning: planResult.shouldContinuePlanning,\n              updateSubGoals,\n              markFinishedIndexes,\n            };\n            executorContext.uiContext = uiContext;\n\n            assert(!error, `Failed to continue: ${error}\\n${log || ''}`);\n\n            // Check if task was finalized with failure\n            if (finalizeSuccess === false) {\n              assert(\n                false,\n                `Task failed: ${finalizeMessage || 'No error message provided'}\\n${log || ''}`,\n              );\n            }\n\n            return {\n              cache: {\n                hit: false,\n              },\n            } as any;\n          },\n        },\n        {\n          allowWhenError: true,\n        },\n      );\n\n      const planResult = result?.output as PlanningAIResponse | undefined;\n\n      // Execute planned actions\n      const plans = planResult?.actions || [];\n      yamlFlow.push(...(planResult?.yamlFlow || []));\n\n      let executables: Awaited<ReturnType<typeof this.convertPlanToExecutable>>;\n      try {\n        executables = await this.convertPlanToExecutable(\n          plans,\n          modelConfigForPlanning,\n          modelConfigForDefaultIntent,\n          {\n            cacheable,\n            deepLocate,\n            abortSignal,\n          },\n        );\n      } catch (error) {\n        return session.appendErrorPlan(\n          `Error converting plans to executable tasks: ${error}, plans: ${JSON.stringify(\n            plans,\n          )}`,\n        );\n      }\n      if (conversationHistory.pendingFeedbackMessage) {\n        console.warn(\n          'unconsumed pending feedback message detected, this may lead to unexpected planning result:',\n          conversationHistory.pendingFeedbackMessage,\n        );\n      }\n\n      // Set initial time context for the first planning call\n      const initialTimeString = await this.getTimeString();\n      conversationHistory.pendingFeedbackMessage += `Current time: ${initialTimeString}`;\n\n      try {\n        await session.appendAndRun(executables.tasks);\n      } catch (error: any) {\n        // errorFlag = true;\n        errorCountInOnePlanningLoop++;\n        const timeString = await this.getTimeString();\n        conversationHistory.pendingFeedbackMessage = `Time: ${timeString}, Error executing running tasks: ${error?.message || String(error)}`;\n        debug(\n          'error when executing running tasks, but continue to run if it is not too many errors:',\n          error instanceof Error ? error.message : String(error),\n          'current error count in one planning loop:',\n          errorCountInOnePlanningLoop,\n        );\n      }\n\n      if (errorCountInOnePlanningLoop > maxErrorCountAllowedInOnePlanningLoop) {\n        return session.appendErrorPlan('Too many errors in one planning loop');\n      }\n\n      // Check abort signal after executing actions\n      if (abortSignal?.aborted) {\n        return session.appendErrorPlan(\n          `Task aborted: ${abortSignal.reason || 'abort signal received'}`,\n        );\n      }\n\n      // // Check if task is complete\n      if (!planResult?.shouldContinuePlanning) {\n        break;\n      }\n\n      // Increment replan count for next iteration\n      ++replanCount;\n\n      if (replanCount > replanningCycleLimit) {\n        const errorMsg = `Replanned ${replanningCycleLimit} times, exceeding the limit. Please configure a larger value for replanningCycleLimit (or use MIDSCENE_REPLANNING_CYCLE_LIMIT) to handle more complex tasks.`;\n        return session.appendErrorPlan(errorMsg);\n      }\n\n      if (!conversationHistory.pendingFeedbackMessage) {\n        const timeString = await this.getTimeString();\n        conversationHistory.pendingFeedbackMessage = `Time: ${timeString}, I have finished the action previously planned.`;\n      }\n    }\n\n    return {\n      output: {\n        yamlFlow,\n        output: outputString,\n      },\n      runner,\n    };\n  }\n\n  private createTypeQueryTask(\n    type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert' | 'WaitFor',\n    demand: ServiceExtractParam,\n    modelConfig: IModelConfig,\n    opt?: ServiceExtractOption,\n    multimodalPrompt?: TMultimodalPrompt,\n  ) {\n    const queryTask: ExecutionTaskInsightQueryApply = {\n      type: 'Insight',\n      subType: type,\n      param: {\n        domIncluded: opt?.domIncluded,\n        dataDemand: multimodalPrompt\n          ? ({\n              demand,\n              multimodalPrompt,\n            } as never)\n          : demand, // for user param presentation in report right sidebar\n      },\n      executor: async (param, taskContext) => {\n        const { task } = taskContext;\n        let queryDump: ServiceDump | undefined;\n        const applyDump = (dump: ServiceDump) => {\n          queryDump = dump;\n          task.log = {\n            dump,\n            rawResponse: dump.taskInfo?.rawResponse,\n          };\n          task.usage = dump.taskInfo?.usage;\n          if (dump.taskInfo?.reasoning_content) {\n            task.reasoning_content = dump.taskInfo.reasoning_content;\n          }\n        };\n\n        // Get context for query operations\n        const uiContext = taskContext.uiContext;\n        assert(uiContext, 'uiContext is required for Query task');\n\n        const ifTypeRestricted = type !== 'Query';\n        let demandInput = demand;\n        let keyOfResult = 'result';\n        if (ifTypeRestricted && (type === 'Assert' || type === 'WaitFor')) {\n          keyOfResult = 'StatementIsTruthy';\n          const booleanPrompt =\n            type === 'Assert'\n              ? `Boolean, whether the following statement is true: ${demand}`\n              : `Boolean, the user wants to do some 'wait for' operation, please check whether the following statement is true: ${demand}`;\n          demandInput = {\n            [keyOfResult]: booleanPrompt,\n          };\n        } else if (ifTypeRestricted) {\n          keyOfResult = type;\n          demandInput = {\n            [keyOfResult]: `${type}, ${demand}`,\n          };\n        }\n\n        let extractResult;\n\n        let extraPageDescription = '';\n        if (opt?.domIncluded && this.interface.getElementsNodeTree) {\n          debug('appending tree info for page');\n          const tree = await this.interface.getElementsNodeTree();\n          extraPageDescription = await descriptionOfTree(\n            tree,\n            200,\n            false,\n            opt?.domIncluded === 'visible-only',\n          );\n        }\n\n        try {\n          extractResult = await this.service.extract<any>(\n            demandInput,\n            modelConfig,\n            opt,\n            extraPageDescription,\n            multimodalPrompt,\n            uiContext,\n          );\n        } catch (error) {\n          if (error instanceof ServiceError) {\n            applyDump(error.dump);\n          }\n          throw error;\n        }\n\n        const { data, thought, dump } = extractResult;\n        applyDump(dump);\n\n        let outputResult = data;\n        if (ifTypeRestricted) {\n          // If AI returned a plain string instead of structured format, use it directly\n          if (typeof data === 'string') {\n            outputResult = data;\n          } else if (type === 'WaitFor') {\n            if (data === null || data === undefined) {\n              outputResult = false;\n            } else {\n              outputResult = (data as any)[keyOfResult];\n            }\n          } else if (data === null || data === undefined) {\n            outputResult = null;\n          } else {\n            // AI model may return {result: ...} instead of {[keyOfResult]: ...}\n            if (data?.[keyOfResult] !== undefined) {\n              outputResult = (data as any)[keyOfResult];\n            } else if (data?.result !== undefined) {\n              outputResult = (data as any).result;\n            } else {\n              assert(false, 'No result in query data');\n            }\n          }\n        }\n\n        if (type === 'Assert' && !outputResult) {\n          task.thought = thought;\n          throw new Error(`Assertion failed: ${thought}`);\n        }\n\n        return {\n          output: outputResult,\n          log: queryDump,\n          thought,\n        };\n      },\n    };\n\n    return queryTask;\n  }\n  async createTypeQueryExecution<T>(\n    type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert',\n    demand: ServiceExtractParam,\n    modelConfig: IModelConfig,\n    opt?: ServiceExtractOption,\n    multimodalPrompt?: TMultimodalPrompt,\n  ): Promise<ExecutionResult<T>> {\n    const session = this.createExecutionSession(\n      taskTitleStr(\n        type,\n        typeof demand === 'string' ? demand : JSON.stringify(demand),\n      ),\n    );\n\n    const queryTask = await this.createTypeQueryTask(\n      type,\n      demand,\n      modelConfig,\n      opt,\n      multimodalPrompt,\n    );\n\n    const runner = session.getRunner();\n    const result = await session.appendAndRun(queryTask);\n\n    if (!result) {\n      throw new Error(\n        'result of taskExecutor.flush() is undefined in function createTypeQueryTask',\n      );\n    }\n\n    const { output, thought } = result;\n\n    return {\n      output,\n      thought,\n      runner,\n    };\n  }\n\n  async waitFor(\n    assertion: TUserPrompt,\n    opt: PlanningActionParamWaitFor,\n    modelConfig: IModelConfig,\n  ): Promise<ExecutionResult<void>> {\n    const { textPrompt, multimodalPrompt } = parsePrompt(assertion);\n\n    const description = `waitFor: ${textPrompt}`;\n    const session = this.createExecutionSession(\n      taskTitleStr('WaitFor', description),\n    );\n    const runner = session.getRunner();\n    const {\n      timeoutMs,\n      checkIntervalMs,\n      domIncluded,\n      screenshotIncluded,\n      ...restOpt\n    } = opt;\n    const serviceExtractOpt: ServiceExtractOption = {\n      domIncluded,\n      screenshotIncluded,\n      ...restOpt,\n    };\n\n    assert(assertion, 'No assertion for waitFor');\n    assert(timeoutMs, 'No timeoutMs for waitFor');\n    assert(checkIntervalMs, 'No checkIntervalMs for waitFor');\n\n    assert(\n      checkIntervalMs <= timeoutMs,\n      `wrong config for waitFor: checkIntervalMs must be less than timeoutMs, config: {checkIntervalMs: ${checkIntervalMs}, timeoutMs: ${timeoutMs}}`,\n    );\n\n    const overallStartTime = Date.now();\n    let lastCheckStart = overallStartTime;\n    let errorThought = '';\n    // Continue checking as long as the previous iteration began within the timeout window.\n    while (lastCheckStart - overallStartTime <= timeoutMs) {\n      const currentCheckStart = Date.now();\n      lastCheckStart = currentCheckStart;\n      const queryTask = await this.createTypeQueryTask(\n        'WaitFor',\n        textPrompt,\n        modelConfig,\n        serviceExtractOpt,\n        multimodalPrompt,\n      );\n\n      const result = (await session.appendAndRun(queryTask)) as\n        | {\n            output: boolean;\n            thought?: string;\n          }\n        | undefined;\n\n      if (result?.output) {\n        return {\n          output: undefined,\n          runner,\n        };\n      }\n\n      errorThought =\n        result?.thought ||\n        (!result && `No result from assertion: ${textPrompt}`) ||\n        `unknown error when waiting for assertion: ${textPrompt}`;\n      const now = Date.now();\n      if (now - currentCheckStart < checkIntervalMs) {\n        const elapsed = now - currentCheckStart;\n        const timeRemaining = checkIntervalMs - elapsed;\n        const thought = `Check interval is ${checkIntervalMs}ms, ${elapsed}ms elapsed since last check, sleeping for ${timeRemaining}ms`;\n        const { tasks: sleepTasks } = await this.convertPlanToExecutable(\n          [{ type: 'Sleep', param: { timeMs: timeRemaining }, thought }],\n          modelConfig,\n          modelConfig,\n        );\n        if (sleepTasks[0]) {\n          await session.appendAndRun(sleepTasks[0]);\n        }\n      }\n    }\n\n    return session.appendErrorPlan(`waitFor timeout: ${errorThought}`);\n  }\n}\n\nexport async function withFileChooser<T>(\n  interfaceInstance: AbstractInterface,\n  fileChooserAccept: string[] | undefined,\n  action: () => Promise<T>,\n): Promise<T> {\n  if (!fileChooserAccept?.length) {\n    return action();\n  }\n\n  if (!interfaceInstance.registerFileChooserListener) {\n    throw new Error(\n      `File upload is not supported on ${interfaceInstance.interfaceType}`,\n    );\n  }\n\n  const handler = async (chooser: FileChooserHandler) => {\n    await chooser.accept(fileChooserAccept);\n  };\n\n  const { dispose, getError } =\n    await interfaceInstance.registerFileChooserListener(handler);\n  try {\n    const result = await action();\n    // Check for errors that occurred during file chooser handling\n    const error = getError();\n    if (error) {\n      throw error;\n    }\n    return result;\n  } finally {\n    dispose();\n  }\n}\n"],"names":["debug","getDebug","maxErrorCountAllowedInOnePlanningLoop","TaskExecutor","title","options","ExecutionSession","Promise","format","timestamp","getCurrentTime","getReadableTimeString","plans","modelConfigForPlanning","modelConfigForDefaultIntent","userInstruction","yamlString","session","taskTitleStr","task","param","executorContext","uiContext","assert","runner","tasks","result","output","userPrompt","includeBboxInPlanning","aiActContext","cacheable","replanningCycleLimitOverride","imagesIncludeCount","deepThink","fileChooserAccept","deepLocate","abortSignal","withFileChooser","conversationHistory","ConversationHistory","replanCount","yamlFlow","replanningCycleLimit","undefined","errorCountInOnePlanningLoop","outputString","subGoalStatus","memoriesStatus","modelFamily","timing","actionSpace","action","Array","console","planImpl","isUITars","uiTarsPlanning","isAutoGLM","autoGLMPlanning","plan","planResult","setTimingFieldOnce","planError","AIResponseParseError","JSON","actions","thought","log","memory","error","usage","rawResponse","reasoning_content","finalizeSuccess","finalizeMessage","updateSubGoals","markFinishedIndexes","executables","initialTimeString","timeString","String","Error","errorMsg","type","demand","modelConfig","opt","multimodalPrompt","queryTask","taskContext","queryDump","applyDump","dump","ifTypeRestricted","demandInput","keyOfResult","booleanPrompt","extractResult","extraPageDescription","tree","descriptionOfTree","ServiceError","data","outputResult","assertion","textPrompt","parsePrompt","description","timeoutMs","checkIntervalMs","domIncluded","screenshotIncluded","restOpt","serviceExtractOpt","overallStartTime","Date","lastCheckStart","errorThought","currentCheckStart","now","elapsed","timeRemaining","sleepTasks","interfaceInstance","service","opts","TaskBuilder","handler","chooser","dispose","getError"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;AA2DA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,wCAAwC;AAIvC,MAAMC;IAsBX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IAiCQ,uBACNC,KAAa,EACbC,OAA0C,EAC1C;QACA,OAAO,IAAIC,iBACTF,OACA,IAAMG,QAAQ,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,kBAAkB,KACrD;YACE,aAAa,IAAI,CAAC,mBAAmB;YACrC,OAAOF,SAAS;YAChB,cAAc,IAAI,CAAC,KAAK,EAAE;QAC5B;IAEJ;IAEQ,iBAAiC;QACvC,OAAO,IAAI,CAAC,mBAAmB;IACjC;IAQA,MAAc,cAAcG,MAAe,EAAmB;QAC5D,MAAMC,YAAY,MAAMC,eACtB,IAAI,CAAC,SAAS,EACd,IAAI,CAAC,kBAAkB;QAEzB,OAAOC,sBAAsBH,QAAQC;IACvC;IAEA,MAAa,wBACXG,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACzCT,OAIC,EACD;QACA,OAAO,IAAI,CAAC,WAAW,CAAC,KAAK,CAC3BO,OACAC,wBACAC,6BACAT;IAEJ;IAEA,MAAM,uBAAuBU,eAAuB,EAAEC,UAAkB,EAAE;QACxE,MAAMC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,OAAOH;QAGtB,MAAMI,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,OAAO;gBACLJ;YACF;YACA,UAAU,OAAOK,OAAOC;gBACtB,MAAM,EAAEC,SAAS,EAAE,GAAGD;gBACtBE,OAAOD,WAAW;gBAClB,OAAO;oBACL,QAAQ;wBACN,SAAS,EAAE;wBACX,wBAAwB;wBACxB,KAAK;wBACLN;oBACF;oBACA,OAAO;wBACL,KAAK;oBACP;oBACA,OAAO;wBACL,MAAM;wBACN,SAAS;4BACPA;wBACF;oBACF;gBACF;YACF;QACF;QACA,MAAMQ,SAASP,QAAQ,SAAS;QAChC,MAAMA,QAAQ,YAAY,CAACE;QAE3B,OAAO;YACLK;QACF;IACF;IAEA,MAAM,SACJpB,KAAa,EACbQ,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACf;QAC1B,MAAMG,UAAU,IAAI,CAAC,sBAAsB,CAACb;QAC5C,MAAM,EAAEqB,KAAK,EAAE,GAAG,MAAM,IAAI,CAAC,uBAAuB,CAClDb,OACAC,wBACAC;QAEF,MAAMU,SAASP,QAAQ,SAAS;QAChC,MAAMS,SAAS,MAAMT,QAAQ,YAAY,CAACQ;QAC1C,MAAM,EAAEE,MAAM,EAAE,GAAGD,UAAU,CAAC;QAC9B,OAAO;YACLC;YACAH;QACF;IACF;IAEA,MAAM,OACJI,UAAkB,EAClBf,sBAAoC,EACpCC,2BAAyC,EACzCe,qBAA8B,EAC9BC,YAAqB,EACrBC,SAAmB,EACnBC,4BAAqC,EACrCC,kBAA2B,EAC3BC,SAA2B,EAC3BC,iBAA4B,EAC5BC,UAAoB,EACpBC,WAAyB,EASzB;QACA,OAAOC,gBAAgB,IAAI,CAAC,SAAS,EAAEH,mBAAmB,UACjD,IAAI,CAAC,SAAS,CACnBP,YACAf,wBACAC,6BACAe,uBACAC,cACAC,WACAC,8BACAC,oBACAC,WACAE,YACAC;IAGN;IAEA,MAAc,UACZT,UAAkB,EAClBf,sBAAoC,EACpCC,2BAAyC,EACzCe,qBAA8B,EAC9BC,YAAqB,EACrBC,SAAmB,EACnBC,4BAAqC,EACrCC,kBAA2B,EAC3BC,SAA2B,EAC3BE,UAAoB,EACpBC,WAAyB,EASzB;QACA,MAAME,sBAAsB,IAAIC;QAEhC,MAAMvB,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,OAAOU;QAEtB,MAAMJ,SAASP,QAAQ,SAAS;QAEhC,IAAIwB,cAAc;QAClB,MAAMC,WAAmC,EAAE;QAC3C,MAAMC,uBACJX,gCAAgC,IAAI,CAAC,oBAAoB;QAC3DT,OACEoB,AAAyBC,WAAzBD,sBACA;QAGF,IAAIE,8BAA8B;QAClC,IAAIC;QAGJ,MAAO,KAAM;YAEX,IAAIT,aAAa,SACf,OAAOpB,QAAQ,eAAe,CAC5B,CAAC,cAAc,EAAEoB,YAAY,MAAM,IAAI,yBAAyB;YAKpE,MAAMU,gBAAgBR,oBAAoB,cAAc,MAAMK;YAG9D,MAAMI,iBAAiBT,oBAAoB,cAAc,MAAMK;YAE/D,MAAMlB,SAAS,MAAMT,QAAQ,YAAY,CACvC;gBACE,MAAM;gBACN,SAAS;gBACT,OAAO;oBACL,iBAAiBW;oBACjBE;oBACAG;oBACAC;oBACA,GAAIa,gBAAgB;wBAAEA;oBAAc,IAAI,CAAC,CAAC;oBAC1C,GAAIC,iBAAiB;wBAAEA;oBAAe,IAAI,CAAC,CAAC;gBAC9C;gBACA,UAAU,OAAO5B,OAAOC;oBACtB,MAAM,EAAEC,SAAS,EAAE,GAAGD;oBACtBE,OAAOD,WAAW;oBAClB,MAAM,EAAE2B,WAAW,EAAE,GAAGpC;oBACxB,MAAMqC,SAAS7B,gBAAgB,IAAI,CAAC,MAAM;oBAE1C,MAAM8B,cAAc,IAAI,CAAC,cAAc;oBACvCnD,MACE,sCACAmD,YAAY,GAAG,CAAC,CAACC,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;oBAEhD7B,OAAO8B,MAAM,OAAO,CAACF,cAAc;oBACnC,IAAIA,AAAuB,MAAvBA,YAAY,MAAM,EACpBG,QAAQ,IAAI,CACV,CAAC,gBAAgB,EAAE,IAAI,CAAC,SAAS,CAAC,aAAa,CAAC,gDAAgD,CAAC;oBAIrG,MAAMC,WAAWC,SAASP,eACtBQ,iBACAC,UAAUT,eACRU,kBACAC;oBAEN,IAAIC;oBACJ,IAAI;wBACFC,mBAAmBZ,QAAQ;wBAC3BW,aAAa,MAAMN,SAASnC,MAAM,eAAe,EAAE;4BACjD,SAASE;4BACT,eAAeF,MAAM,YAAY;4BACjC,eAAe,IAAI,CAAC,SAAS,CAAC,aAAa;4BAC3C+B;4BACA,aAAatC;4BACb0B;4BACA,aAAaV;4BACbI;4BACAC;4BACAG;wBACF;oBACF,EAAE,OAAO0B,WAAW;wBAClB,IAAIA,qBAAqBC,sBAAsB;4BAE7C3C,gBAAgB,IAAI,CAAC,KAAK,GAAG0C,UAAU,KAAK;4BAC5C1C,gBAAgB,IAAI,CAAC,GAAG,GAAG;gCACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;gCAClC,aAAa0C,UAAU,WAAW;4BACpC;wBACF;wBACA,MAAMA;oBACR,SAAU;wBACRD,mBAAmBZ,QAAQ;oBAC7B;oBACAlD,MAAM,cAAciE,KAAK,SAAS,CAACJ,YAAY,MAAM;oBAErD,MAAM,EACJK,OAAO,EACPC,OAAO,EACPC,GAAG,EACHC,MAAM,EACNC,KAAK,EACLC,KAAK,EACLC,WAAW,EACXC,iBAAiB,EACjBC,eAAe,EACfC,eAAe,EACfC,cAAc,EACdC,mBAAmB,EACpB,GAAGhB;oBACJf,eAAe6B;oBAEftD,gBAAgB,IAAI,CAAC,GAAG,GAAG;wBACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;wBAClCmD;oBACF;oBACAnD,gBAAgB,IAAI,CAAC,KAAK,GAAGkD;oBAC7BlD,gBAAgB,IAAI,CAAC,iBAAiB,GAAGoD;oBACzCpD,gBAAgB,IAAI,CAAC,MAAM,GAAG;wBAC5B,SAAS6C,WAAW,EAAE;wBACtBE;wBACAD;wBACAE;wBACA,UAAUR,WAAW,QAAQ;wBAC7B,QAAQc;wBACR,wBAAwBd,WAAW,sBAAsB;wBACzDe;wBACAC;oBACF;oBACAxD,gBAAgB,SAAS,GAAGC;oBAE5BC,OAAO,CAAC+C,OAAO,CAAC,oBAAoB,EAAEA,MAAM,EAAE,EAAEF,OAAO,IAAI;oBAG3D,IAAIM,AAAoB,UAApBA,iBACFnD,OACE,OACA,CAAC,aAAa,EAAEoD,mBAAmB,4BAA4B,EAAE,EAAEP,OAAO,IAAI;oBAIlF,OAAO;wBACL,OAAO;4BACL,KAAK;wBACP;oBACF;gBACF;YACF,GACA;gBACE,gBAAgB;YAClB;YAGF,MAAMP,aAAanC,QAAQ;YAG3B,MAAMd,QAAQiD,YAAY,WAAW,EAAE;YACvCnB,SAAS,IAAI,IAAKmB,YAAY,YAAY,EAAE;YAE5C,IAAIiB;YACJ,IAAI;gBACFA,cAAc,MAAM,IAAI,CAAC,uBAAuB,CAC9ClE,OACAC,wBACAC,6BACA;oBACEiB;oBACAK;oBACAC;gBACF;YAEJ,EAAE,OAAOiC,OAAO;gBACd,OAAOrD,QAAQ,eAAe,CAC5B,CAAC,4CAA4C,EAAEqD,MAAM,SAAS,EAAEL,KAAK,SAAS,CAC5ErD,QACC;YAEP;YACA,IAAI2B,oBAAoB,sBAAsB,EAC5Ce,QAAQ,IAAI,CACV,8FACAf,oBAAoB,sBAAsB;YAK9C,MAAMwC,oBAAoB,MAAM,IAAI,CAAC,aAAa;YAClDxC,oBAAoB,sBAAsB,IAAI,CAAC,cAAc,EAAEwC,mBAAmB;YAElF,IAAI;gBACF,MAAM9D,QAAQ,YAAY,CAAC6D,YAAY,KAAK;YAC9C,EAAE,OAAOR,OAAY;gBAEnBzB;gBACA,MAAMmC,aAAa,MAAM,IAAI,CAAC,aAAa;gBAC3CzC,oBAAoB,sBAAsB,GAAG,CAAC,MAAM,EAAEyC,WAAW,iCAAiC,EAAEV,OAAO,WAAWW,OAAOX,QAAQ;gBACrItE,MACE,yFACAsE,iBAAiBY,QAAQZ,MAAM,OAAO,GAAGW,OAAOX,QAChD,6CACAzB;YAEJ;YAEA,IAAIA,8BAA8B3C,uCAChC,OAAOe,QAAQ,eAAe,CAAC;YAIjC,IAAIoB,aAAa,SACf,OAAOpB,QAAQ,eAAe,CAC5B,CAAC,cAAc,EAAEoB,YAAY,MAAM,IAAI,yBAAyB;YAKpE,IAAI,CAACwB,YAAY,wBACf;YAIF,EAAEpB;YAEF,IAAIA,cAAcE,sBAAsB;gBACtC,MAAMwC,WAAW,CAAC,UAAU,EAAExC,qBAAqB,4JAA4J,CAAC;gBAChN,OAAO1B,QAAQ,eAAe,CAACkE;YACjC;YAEA,IAAI,CAAC5C,oBAAoB,sBAAsB,EAAE;gBAC/C,MAAMyC,aAAa,MAAM,IAAI,CAAC,aAAa;gBAC3CzC,oBAAoB,sBAAsB,GAAG,CAAC,MAAM,EAAEyC,WAAW,gDAAgD,CAAC;YACpH;QACF;QAEA,OAAO;YACL,QAAQ;gBACNtC;gBACA,QAAQI;YACV;YACAtB;QACF;IACF;IAEQ,oBACN4D,IAAsE,EACtEC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACpC;QACA,MAAMC,YAA4C;YAChD,MAAM;YACN,SAASL;YACT,OAAO;gBACL,aAAaG,KAAK;gBAClB,YAAYC,mBACP;oBACCH;oBACAG;gBACF,IACAH;YACN;YACA,UAAU,OAAOjE,OAAOsE;gBACtB,MAAM,EAAEvE,IAAI,EAAE,GAAGuE;gBACjB,IAAIC;gBACJ,MAAMC,YAAY,CAACC;oBACjBF,YAAYE;oBACZ1E,KAAK,GAAG,GAAG;wBACT0E;wBACA,aAAaA,KAAK,QAAQ,EAAE;oBAC9B;oBACA1E,KAAK,KAAK,GAAG0E,KAAK,QAAQ,EAAE;oBAC5B,IAAIA,KAAK,QAAQ,EAAE,mBACjB1E,KAAK,iBAAiB,GAAG0E,KAAK,QAAQ,CAAC,iBAAiB;gBAE5D;gBAGA,MAAMvE,YAAYoE,YAAY,SAAS;gBACvCnE,OAAOD,WAAW;gBAElB,MAAMwE,mBAAmBV,AAAS,YAATA;gBACzB,IAAIW,cAAcV;gBAClB,IAAIW,cAAc;gBAClB,IAAIF,oBAAqBV,CAAAA,AAAS,aAATA,QAAqBA,AAAS,cAATA,IAAiB,GAAI;oBACjEY,cAAc;oBACd,MAAMC,gBACJb,AAAS,aAATA,OACI,CAAC,kDAAkD,EAAEC,QAAQ,GAC7D,CAAC,+GAA+G,EAAEA,QAAQ;oBAChIU,cAAc;wBACZ,CAACC,YAAY,EAAEC;oBACjB;gBACF,OAAO,IAAIH,kBAAkB;oBAC3BE,cAAcZ;oBACdW,cAAc;wBACZ,CAACC,YAAY,EAAE,GAAGZ,KAAK,EAAE,EAAEC,QAAQ;oBACrC;gBACF;gBAEA,IAAIa;gBAEJ,IAAIC,uBAAuB;gBAC3B,IAAIZ,KAAK,eAAe,IAAI,CAAC,SAAS,CAAC,mBAAmB,EAAE;oBAC1DvF,MAAM;oBACN,MAAMoG,OAAO,MAAM,IAAI,CAAC,SAAS,CAAC,mBAAmB;oBACrDD,uBAAuB,MAAME,kBAC3BD,MACA,KACA,OACAb,KAAK,gBAAgB;gBAEzB;gBAEA,IAAI;oBACFW,gBAAgB,MAAM,IAAI,CAAC,OAAO,CAAC,OAAO,CACxCH,aACAT,aACAC,KACAY,sBACAX,kBACAlE;gBAEJ,EAAE,OAAOgD,OAAO;oBACd,IAAIA,iBAAiBgC,cACnBV,UAAUtB,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAEA,MAAM,EAAEiC,IAAI,EAAEpC,OAAO,EAAE0B,IAAI,EAAE,GAAGK;gBAChCN,UAAUC;gBAEV,IAAIW,eAAeD;gBACnB,IAAIT,kBAEF,IAAI,AAAgB,YAAhB,OAAOS,MACTC,eAAeD;qBACV,IAAInB,AAAS,cAATA,MAEPoB,eADED,QAAAA,OACa,QAECA,IAAY,CAACP,YAAY;qBAEtC,IAAIO,QAAAA,MACTC,eAAe;qBAGf,IAAID,MAAM,CAACP,YAAY,KAAKpD,QAC1B4D,eAAgBD,IAAY,CAACP,YAAY;qBACpC,IAAIO,MAAM,WAAW3D,QAC1B4D,eAAgBD,KAAa,MAAM;qBAEnChF,OAAO,OAAO;gBAKpB,IAAI6D,AAAS,aAATA,QAAqB,CAACoB,cAAc;oBACtCrF,KAAK,OAAO,GAAGgD;oBACf,MAAM,IAAIe,MAAM,CAAC,kBAAkB,EAAEf,SAAS;gBAChD;gBAEA,OAAO;oBACL,QAAQqC;oBACR,KAAKb;oBACLxB;gBACF;YACF;QACF;QAEA,OAAOsB;IACT;IACA,MAAM,yBACJL,IAA0D,EAC1DC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACP;QAC7B,MAAMvE,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aACEkE,MACA,AAAkB,YAAlB,OAAOC,SAAsBA,SAASpB,KAAK,SAAS,CAACoB;QAIzD,MAAMI,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9CL,MACAC,QACAC,aACAC,KACAC;QAGF,MAAMhE,SAASP,QAAQ,SAAS;QAChC,MAAMS,SAAS,MAAMT,QAAQ,YAAY,CAACwE;QAE1C,IAAI,CAAC/D,QACH,MAAM,IAAIwD,MACR;QAIJ,MAAM,EAAEvD,MAAM,EAAEwC,OAAO,EAAE,GAAGzC;QAE5B,OAAO;YACLC;YACAwC;YACA3C;QACF;IACF;IAEA,MAAM,QACJiF,SAAsB,EACtBlB,GAA+B,EAC/BD,WAAyB,EACO;QAChC,MAAM,EAAEoB,UAAU,EAAElB,gBAAgB,EAAE,GAAGmB,YAAYF;QAErD,MAAMG,cAAc,CAAC,SAAS,EAAEF,YAAY;QAC5C,MAAMzF,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,WAAW0F;QAE1B,MAAMpF,SAASP,QAAQ,SAAS;QAChC,MAAM,EACJ4F,SAAS,EACTC,eAAe,EACfC,WAAW,EACXC,kBAAkB,EAClB,GAAGC,SACJ,GAAG1B;QACJ,MAAM2B,oBAA0C;YAC9CH;YACAC;YACA,GAAGC,OAAO;QACZ;QAEA1F,OAAOkF,WAAW;QAClBlF,OAAOsF,WAAW;QAClBtF,OAAOuF,iBAAiB;QAExBvF,OACEuF,mBAAmBD,WACnB,CAAC,iGAAiG,EAAEC,gBAAgB,aAAa,EAAED,UAAU,CAAC,CAAC;QAGjJ,MAAMM,mBAAmBC,KAAK,GAAG;QACjC,IAAIC,iBAAiBF;QACrB,IAAIG,eAAe;QAEnB,MAAOD,iBAAiBF,oBAAoBN,UAAW;YACrD,MAAMU,oBAAoBH,KAAK,GAAG;YAClCC,iBAAiBE;YACjB,MAAM9B,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9C,WACAiB,YACApB,aACA4B,mBACA1B;YAGF,MAAM9D,SAAU,MAAMT,QAAQ,YAAY,CAACwE;YAO3C,IAAI/D,QAAQ,QACV,OAAO;gBACL,QAAQkB;gBACRpB;YACF;YAGF8F,eACE5F,QAAQ,WACP,CAACA,UAAU,CAAC,0BAA0B,EAAEgF,YAAY,IACrD,CAAC,0CAA0C,EAAEA,YAAY;YAC3D,MAAMc,MAAMJ,KAAK,GAAG;YACpB,IAAII,MAAMD,oBAAoBT,iBAAiB;gBAC7C,MAAMW,UAAUD,MAAMD;gBACtB,MAAMG,gBAAgBZ,kBAAkBW;gBACxC,MAAMtD,UAAU,CAAC,kBAAkB,EAAE2C,gBAAgB,IAAI,EAAEW,QAAQ,0CAA0C,EAAEC,cAAc,EAAE,CAAC;gBAChI,MAAM,EAAE,OAAOC,UAAU,EAAE,GAAG,MAAM,IAAI,CAAC,uBAAuB,CAC9D;oBAAC;wBAAE,MAAM;wBAAS,OAAO;4BAAE,QAAQD;wBAAc;wBAAGvD;oBAAQ;iBAAE,EAC9DmB,aACAA;gBAEF,IAAIqC,UAAU,CAAC,EAAE,EACf,MAAM1G,QAAQ,YAAY,CAAC0G,UAAU,CAAC,EAAE;YAE5C;QACF;QAEA,OAAO1G,QAAQ,eAAe,CAAC,CAAC,iBAAiB,EAAEqG,cAAc;IACnE;IAjsBA,YACEM,iBAAoC,EACpCC,OAAgB,EAChBC,IAQC,CACD;QArCF;QAEA;QAEA;QAEA,uBAAiB,uBAAjB;QAEA,uBAAiB,eAAjB;QAEA;QAEA,uBAAiB,SAAjB;QAEA;QAEA;QAEA;QAoBE,IAAI,CAAC,SAAS,GAAGF;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC,KAAK,SAAS;QAC/B,IAAI,CAAC,mBAAmB,GAAGA,MAAM;QACjC,IAAI,CAAC,oBAAoB,GAAGA,KAAK,oBAAoB;QACrD,IAAI,CAAC,eAAe,GAAGA,KAAK,eAAe;QAC3C,IAAI,CAAC,kBAAkB,GAAGA,KAAK,kBAAkB;QACjD,IAAI,CAAC,KAAK,GAAGA,KAAK,KAAK;QACvB,IAAI,CAAC,mBAAmB,GAAGA,KAAK,WAAW;QAC3C,IAAI,CAAC,WAAW,GAAG,IAAIC,YAAY;YACjCH;YACAC;YACA,WAAWC,KAAK,SAAS;YACzB,aAAa,IAAI,CAAC,cAAc;YAChC,iBAAiBA,KAAK,eAAe;QACvC;IACF;AAqqBF;AAEO,eAAexF,gBACpBsF,iBAAoC,EACpCzF,iBAAuC,EACvCiB,MAAwB;IAExB,IAAI,CAACjB,mBAAmB,QACtB,OAAOiB;IAGT,IAAI,CAACwE,kBAAkB,2BAA2B,EAChD,MAAM,IAAI1C,MACR,CAAC,gCAAgC,EAAE0C,kBAAkB,aAAa,EAAE;IAIxE,MAAMI,UAAU,OAAOC;QACrB,MAAMA,QAAQ,MAAM,CAAC9F;IACvB;IAEA,MAAM,EAAE+F,OAAO,EAAEC,QAAQ,EAAE,GACzB,MAAMP,kBAAkB,2BAA2B,CAACI;IACtD,IAAI;QACF,MAAMtG,SAAS,MAAM0B;QAErB,MAAMkB,QAAQ6D;QACd,IAAI7D,OACF,MAAMA;QAER,OAAO5C;IACT,SAAU;QACRwG;IACF;AACF"}
+{"version":3,"file":"agent/tasks.mjs","sources":["../../../src/agent/tasks.ts"],"sourcesContent":["import {\n  AIResponseParseError,\n  ConversationHistory,\n  autoGLMPlanning,\n  plan,\n  uiTarsPlanning,\n} from '@/ai-model';\nimport { isAutoGLM, isUITars } from '@/ai-model/auto-glm/util';\nimport {\n  type TMultimodalPrompt,\n  type TUserPrompt,\n  getReadableTimeString,\n} from '@/common';\nimport type { AbstractInterface, FileChooserHandler } from '@/device';\nimport type Service from '@/service';\nimport type { TaskRunner } from '@/task-runner';\nimport { TaskExecutionError } from '@/task-runner';\nimport type {\n  DeepThinkOption,\n  DeviceAction,\n  ExecutionTaskApply,\n  ExecutionTaskInsightQueryApply,\n  ExecutionTaskPlanningApply,\n  ExecutionTaskProgressOptions,\n  InterfaceType,\n  MidsceneYamlFlowItem,\n  PlanningAIResponse,\n  PlanningAction,\n  PlanningActionParamWaitFor,\n  ServiceDump,\n  ServiceExtractOption,\n  ServiceExtractParam,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport { type IModelConfig, getCurrentTime } from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { ExecutionSession } from './execution-session';\nimport { TaskBuilder } from './task-builder';\nimport type { TaskCache } from './task-cache';\nexport { locatePlanForLocate } from './task-builder';\nimport { setTimingFieldOnce } from '@/task-timing';\nimport { descriptionOfTree } from '@midscene/shared/extractor';\nimport { taskTitleStr } from './ui-utils';\nimport { parsePrompt } from './utils';\n\ninterface ExecutionResult<OutputType = any> {\n  output: OutputType;\n  thought?: string;\n  runner: TaskRunner;\n}\n\ninterface TaskExecutorHooks {\n  onTaskUpdate?: (\n    runner: TaskRunner,\n    error?: TaskExecutionError,\n  ) => Promise<void> | void;\n}\n\nconst debug = getDebug('device-task-executor');\nconst maxErrorCountAllowedInOnePlanningLoop = 5;\n\nexport { TaskExecutionError };\n\nexport class TaskExecutor {\n  interface: AbstractInterface;\n\n  service: Service;\n\n  taskCache?: TaskCache;\n\n  private readonly providedActionSpace: DeviceAction[];\n\n  private readonly taskBuilder: TaskBuilder;\n\n  onTaskStartCallback?: ExecutionTaskProgressOptions['onTaskStart'];\n\n  private readonly hooks?: TaskExecutorHooks;\n\n  replanningCycleLimit?: number;\n\n  waitAfterAction?: number;\n\n  useDeviceTimestamp?: boolean;\n\n  // @deprecated use .interface instead\n  get page() {\n    return this.interface;\n  }\n\n  constructor(\n    interfaceInstance: AbstractInterface,\n    service: Service,\n    opts: {\n      taskCache?: TaskCache;\n      onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n      replanningCycleLimit?: number;\n      waitAfterAction?: number;\n      useDeviceTimestamp?: boolean;\n      hooks?: TaskExecutorHooks;\n      actionSpace: DeviceAction[];\n    },\n  ) {\n    this.interface = interfaceInstance;\n    this.service = service;\n    this.taskCache = opts.taskCache;\n    this.onTaskStartCallback = opts?.onTaskStart;\n    this.replanningCycleLimit = opts.replanningCycleLimit;\n    this.waitAfterAction = opts.waitAfterAction;\n    this.useDeviceTimestamp = opts.useDeviceTimestamp;\n    this.hooks = opts.hooks;\n    this.providedActionSpace = opts.actionSpace;\n    this.taskBuilder = new TaskBuilder({\n      interfaceInstance,\n      service,\n      taskCache: opts.taskCache,\n      actionSpace: this.getActionSpace(),\n      waitAfterAction: opts.waitAfterAction,\n    });\n  }\n\n  private createExecutionSession(\n    title: string,\n    options?: { tasks?: ExecutionTaskApply[] },\n  ) {\n    return new ExecutionSession(\n      title,\n      () => Promise.resolve(this.service.contextRetrieverFn()),\n      {\n        onTaskStart: this.onTaskStartCallback,\n        tasks: options?.tasks,\n        onTaskUpdate: this.hooks?.onTaskUpdate,\n      },\n    );\n  }\n\n  private getActionSpace(): DeviceAction[] {\n    return this.providedActionSpace;\n  }\n\n  /**\n   * Get a readable time string using device time when configured.\n   * This method respects the useDeviceTimestamp configuration.\n   * @param format - Optional format string\n   * @returns A formatted time string\n   */\n  private async getTimeString(format?: string): Promise<string> {\n    const timestamp = await getCurrentTime(\n      this.interface,\n      this.useDeviceTimestamp,\n    );\n    return getReadableTimeString(format, timestamp);\n  }\n\n  public async convertPlanToExecutable(\n    plans: PlanningAction[],\n    modelConfigForPlanning: IModelConfig,\n    modelConfigForDefaultIntent: IModelConfig,\n    options?: {\n      cacheable?: boolean;\n      deepLocate?: boolean;\n      abortSignal?: AbortSignal;\n    },\n  ) {\n    return this.taskBuilder.build(\n      plans,\n      modelConfigForPlanning,\n      modelConfigForDefaultIntent,\n      options,\n    );\n  }\n\n  async loadYamlFlowAsPlanning(userInstruction: string, yamlString: string) {\n    const session = this.createExecutionSession(\n      taskTitleStr('Act', userInstruction),\n    );\n\n    const task: ExecutionTaskPlanningApply = {\n      type: 'Planning',\n      subType: 'LoadYaml',\n      param: {\n        userInstruction,\n      },\n      executor: async (param, executorContext) => {\n        const { uiContext } = executorContext;\n        assert(uiContext, 'uiContext is required for Planning task');\n        return {\n          output: {\n            actions: [],\n            shouldContinuePlanning: false,\n            log: '',\n            yamlString,\n          },\n          cache: {\n            hit: true,\n          },\n          hitBy: {\n            from: 'Cache',\n            context: {\n              yamlString,\n            },\n          },\n        };\n      },\n    };\n    const runner = session.getRunner();\n    await session.appendAndRun(task);\n\n    return {\n      runner,\n    };\n  }\n\n  async runPlans(\n    title: string,\n    plans: PlanningAction[],\n    modelConfigForPlanning: IModelConfig,\n    modelConfigForDefaultIntent: IModelConfig,\n  ): Promise<ExecutionResult> {\n    const session = this.createExecutionSession(title);\n    const { tasks } = await this.convertPlanToExecutable(\n      plans,\n      modelConfigForPlanning,\n      modelConfigForDefaultIntent,\n    );\n    const runner = session.getRunner();\n    const result = await session.appendAndRun(tasks);\n    const { output } = result ?? {};\n    return {\n      output,\n      runner,\n    };\n  }\n\n  async action(\n    userPrompt: string,\n    modelConfigForPlanning: IModelConfig,\n    modelConfigForDefaultIntent: IModelConfig,\n    includeBboxInPlanning: boolean,\n    aiActContext?: string,\n    cacheable?: boolean,\n    replanningCycleLimitOverride?: number,\n    imagesIncludeCount?: number,\n    deepThink?: DeepThinkOption,\n    fileChooserAccept?: string[],\n    deepLocate?: boolean,\n    abortSignal?: AbortSignal,\n  ): Promise<\n    ExecutionResult<\n      | {\n          yamlFlow?: MidsceneYamlFlowItem[]; // for cache use\n          output?: string;\n        }\n      | undefined\n    >\n  > {\n    return withFileChooser(this.interface, fileChooserAccept, async () => {\n      return this.runAction(\n        userPrompt,\n        modelConfigForPlanning,\n        modelConfigForDefaultIntent,\n        includeBboxInPlanning,\n        aiActContext,\n        cacheable,\n        replanningCycleLimitOverride,\n        imagesIncludeCount,\n        deepThink,\n        deepLocate,\n        abortSignal,\n      );\n    });\n  }\n\n  private async runAction(\n    userPrompt: string,\n    modelConfigForPlanning: IModelConfig,\n    modelConfigForDefaultIntent: IModelConfig,\n    includeBboxInPlanning: boolean,\n    aiActContext?: string,\n    cacheable?: boolean,\n    replanningCycleLimitOverride?: number,\n    imagesIncludeCount?: number,\n    deepThink?: DeepThinkOption,\n    deepLocate?: boolean,\n    abortSignal?: AbortSignal,\n  ): Promise<\n    ExecutionResult<\n      | {\n          yamlFlow?: MidsceneYamlFlowItem[]; // for cache use\n          output?: string;\n        }\n      | undefined\n    >\n  > {\n    const conversationHistory = new ConversationHistory();\n\n    const session = this.createExecutionSession(\n      taskTitleStr('Act', userPrompt),\n    );\n    const runner = session.getRunner();\n\n    let replanCount = 0;\n    const yamlFlow: MidsceneYamlFlowItem[] = [];\n    const replanningCycleLimit =\n      replanningCycleLimitOverride ?? this.replanningCycleLimit;\n    assert(\n      replanningCycleLimit !== undefined,\n      'replanningCycleLimit is required for TaskExecutor.action',\n    );\n\n    let errorCountInOnePlanningLoop = 0; // count the number of errors in one planning loop\n    let outputString: string | undefined;\n\n    // Main planning loop - unified plan/replan logic\n    while (true) {\n      // Check abort signal before each planning cycle\n      if (abortSignal?.aborted) {\n        return session.appendErrorPlan(\n          `Task aborted: ${abortSignal.reason || 'abort signal received'}`,\n        );\n      }\n\n      // Get sub-goal status text if available\n      const subGoalStatus = conversationHistory.subGoalsToText() || undefined;\n\n      // Get memories text if available\n      const memoriesStatus = conversationHistory.memoriesToText() || undefined;\n\n      const result = await session.appendAndRun(\n        {\n          type: 'Planning',\n          subType: 'Plan',\n          param: {\n            userInstruction: userPrompt,\n            aiActContext,\n            imagesIncludeCount,\n            deepThink,\n            ...(subGoalStatus ? { subGoalStatus } : {}),\n            ...(memoriesStatus ? { memoriesStatus } : {}),\n          },\n          executor: async (param, executorContext) => {\n            const { uiContext } = executorContext;\n            assert(uiContext, 'uiContext is required for Planning task');\n            const { modelFamily } = modelConfigForPlanning;\n            const timing = executorContext.task.timing;\n\n            const actionSpace = this.getActionSpace();\n            debug(\n              'actionSpace for this interface is:',\n              actionSpace.map((action) => action.name).join(', '),\n            );\n            assert(Array.isArray(actionSpace), 'actionSpace must be an array');\n            if (actionSpace.length === 0) {\n              console.warn(\n                `ActionSpace for ${this.interface.interfaceType} is empty. This may lead to unexpected behavior.`,\n              );\n            }\n\n            const planImpl = isUITars(modelFamily)\n              ? uiTarsPlanning\n              : isAutoGLM(modelFamily)\n                ? autoGLMPlanning\n                : plan;\n\n            let planResult: Awaited<ReturnType<typeof planImpl>>;\n            try {\n              setTimingFieldOnce(timing, 'callAiStart');\n              planResult = await planImpl(param.userInstruction, {\n                context: uiContext,\n                actionContext: param.aiActContext,\n                interfaceType: this.interface.interfaceType as InterfaceType,\n                actionSpace,\n                modelConfig: modelConfigForPlanning,\n                conversationHistory,\n                includeBbox: includeBboxInPlanning,\n                imagesIncludeCount,\n                deepThink,\n                abortSignal,\n              });\n            } catch (planError) {\n              if (planError instanceof AIResponseParseError) {\n                // Record usage and rawResponse even when parsing fails\n                executorContext.task.usage = planError.usage;\n                executorContext.task.log = {\n                  ...(executorContext.task.log || {}),\n                  rawResponse: planError.rawResponse,\n                };\n              }\n              throw planError;\n            } finally {\n              setTimingFieldOnce(timing, 'callAiEnd');\n            }\n            debug('planResult', JSON.stringify(planResult, null, 2));\n\n            const {\n              actions,\n              thought,\n              log,\n              memory,\n              error,\n              usage,\n              rawResponse,\n              reasoning_content,\n              finalizeSuccess,\n              finalizeMessage,\n              updateSubGoals,\n              markFinishedIndexes,\n            } = planResult;\n            outputString = finalizeMessage;\n\n            executorContext.task.log = {\n              ...(executorContext.task.log || {}),\n              rawResponse,\n            };\n            executorContext.task.usage = usage;\n            executorContext.task.reasoning_content = reasoning_content;\n            executorContext.task.output = {\n              actions: actions || [],\n              log,\n              thought,\n              memory,\n              yamlFlow: planResult.yamlFlow,\n              output: finalizeMessage,\n              shouldContinuePlanning: planResult.shouldContinuePlanning,\n              updateSubGoals,\n              markFinishedIndexes,\n            };\n            executorContext.uiContext = uiContext;\n\n            assert(!error, `Failed to continue: ${error}\\n${log || ''}`);\n\n            // Check if task was finalized with failure\n            if (finalizeSuccess === false) {\n              assert(\n                false,\n                `Task failed: ${finalizeMessage || 'No error message provided'}\\n${log || ''}`,\n              );\n            }\n\n            return {\n              cache: {\n                hit: false,\n              },\n            } as any;\n          },\n        },\n        {\n          allowWhenError: true,\n        },\n      );\n\n      const planResult = result?.output as PlanningAIResponse | undefined;\n\n      // Execute planned actions\n      const plans = planResult?.actions || [];\n      yamlFlow.push(...(planResult?.yamlFlow || []));\n\n      let executables: Awaited<ReturnType<typeof this.convertPlanToExecutable>>;\n      try {\n        executables = await this.convertPlanToExecutable(\n          plans,\n          modelConfigForPlanning,\n          modelConfigForDefaultIntent,\n          {\n            cacheable,\n            deepLocate,\n            abortSignal,\n          },\n        );\n      } catch (error) {\n        return session.appendErrorPlan(\n          `Error converting plans to executable tasks: ${error}, plans: ${JSON.stringify(\n            plans,\n          )}`,\n        );\n      }\n      if (conversationHistory.pendingFeedbackMessage) {\n        console.warn(\n          'unconsumed pending feedback message detected, this may lead to unexpected planning result:',\n          conversationHistory.pendingFeedbackMessage,\n        );\n      }\n\n      // Set initial time context for the first planning call\n      const initialTimeString = await this.getTimeString();\n      conversationHistory.pendingFeedbackMessage += `Current time: ${initialTimeString}`;\n\n      try {\n        await session.appendAndRun(executables.tasks);\n      } catch (error: any) {\n        // errorFlag = true;\n        errorCountInOnePlanningLoop++;\n        const timeString = await this.getTimeString();\n        conversationHistory.pendingFeedbackMessage = `Time: ${timeString}, Error executing running tasks: ${error?.message || String(error)}`;\n        debug(\n          'error when executing running tasks, but continue to run if it is not too many errors:',\n          error instanceof Error ? error.message : String(error),\n          'current error count in one planning loop:',\n          errorCountInOnePlanningLoop,\n        );\n      }\n\n      if (errorCountInOnePlanningLoop > maxErrorCountAllowedInOnePlanningLoop) {\n        return session.appendErrorPlan('Too many errors in one planning loop');\n      }\n\n      // Check abort signal after executing actions\n      if (abortSignal?.aborted) {\n        return session.appendErrorPlan(\n          `Task aborted: ${abortSignal.reason || 'abort signal received'}`,\n        );\n      }\n\n      // // Check if task is complete\n      if (!planResult?.shouldContinuePlanning) {\n        break;\n      }\n\n      // Increment replan count for next iteration\n      ++replanCount;\n\n      if (replanCount > replanningCycleLimit) {\n        const errorMsg = `Replanned ${replanningCycleLimit} times, exceeding the limit. Please configure a larger value for replanningCycleLimit (or use MIDSCENE_REPLANNING_CYCLE_LIMIT) to handle more complex tasks.`;\n        return session.appendErrorPlan(errorMsg);\n      }\n\n      if (!conversationHistory.pendingFeedbackMessage) {\n        const timeString = await this.getTimeString();\n        conversationHistory.pendingFeedbackMessage = `Time: ${timeString}, I have finished the action previously planned.`;\n      }\n    }\n\n    return {\n      output: {\n        yamlFlow,\n        output: outputString,\n      },\n      runner,\n    };\n  }\n\n  private createTypeQueryTask(\n    type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert' | 'WaitFor',\n    demand: ServiceExtractParam,\n    modelConfig: IModelConfig,\n    opt?: ServiceExtractOption,\n    multimodalPrompt?: TMultimodalPrompt,\n  ) {\n    const queryTask: ExecutionTaskInsightQueryApply = {\n      type: 'Insight',\n      subType: type,\n      param: {\n        domIncluded: opt?.domIncluded,\n        dataDemand: multimodalPrompt\n          ? ({\n              demand,\n              multimodalPrompt,\n            } as never)\n          : demand, // for user param presentation in report right sidebar\n      },\n      executor: async (param, taskContext) => {\n        const { task } = taskContext;\n        let queryDump: ServiceDump | undefined;\n        const applyDump = (dump: ServiceDump) => {\n          queryDump = dump;\n          task.log = {\n            dump,\n            rawResponse: dump.taskInfo?.rawResponse,\n          };\n          task.usage = dump.taskInfo?.usage;\n          if (dump.taskInfo?.reasoning_content) {\n            task.reasoning_content = dump.taskInfo.reasoning_content;\n          }\n        };\n\n        // Get context for query operations\n        const uiContext = taskContext.uiContext;\n        assert(uiContext, 'uiContext is required for Query task');\n\n        const ifTypeRestricted = type !== 'Query';\n        let demandInput = demand;\n        let keyOfResult = 'result';\n        const currentScreenshotConstraint =\n          'based on the current screenshot and its contents if provided, unless the user explicitly asks to compare with reference images';\n        if (ifTypeRestricted && (type === 'Assert' || type === 'WaitFor')) {\n          keyOfResult = 'StatementIsTruthy';\n          const booleanPrompt =\n            type === 'Assert'\n              ? `Boolean, ${currentScreenshotConstraint}, whether the following statement is true: ${demand}`\n              : `Boolean, the user wants to do some 'wait for' operation. ${currentScreenshotConstraint}, please check whether the following statement is true: ${demand}`;\n          demandInput = {\n            [keyOfResult]: booleanPrompt,\n          };\n        } else if (ifTypeRestricted) {\n          keyOfResult = type;\n          demandInput = {\n            [keyOfResult]: `${type}, ${currentScreenshotConstraint}, ${demand}`,\n          };\n        }\n\n        let extractResult;\n\n        let extraPageDescription = '';\n        if (opt?.domIncluded && this.interface.getElementsNodeTree) {\n          debug('appending tree info for page');\n          const tree = await this.interface.getElementsNodeTree();\n          extraPageDescription = await descriptionOfTree(\n            tree,\n            200,\n            false,\n            opt?.domIncluded === 'visible-only',\n          );\n        }\n\n        try {\n          extractResult = await this.service.extract<any>(\n            demandInput,\n            modelConfig,\n            opt,\n            extraPageDescription,\n            multimodalPrompt,\n            uiContext,\n          );\n        } catch (error) {\n          if (error instanceof ServiceError) {\n            applyDump(error.dump);\n          }\n          throw error;\n        }\n\n        const { data, thought, dump } = extractResult;\n        applyDump(dump);\n\n        let outputResult = data;\n        if (ifTypeRestricted) {\n          // If AI returned a plain string instead of structured format, use it directly\n          if (typeof data === 'string') {\n            outputResult = data;\n          } else if (type === 'WaitFor') {\n            if (data === null || data === undefined) {\n              outputResult = false;\n            } else {\n              outputResult = (data as any)[keyOfResult];\n            }\n          } else if (data === null || data === undefined) {\n            outputResult = null;\n          } else {\n            // AI model may return {result: ...} instead of {[keyOfResult]: ...}\n            if (data?.[keyOfResult] !== undefined) {\n              outputResult = (data as any)[keyOfResult];\n            } else if (data?.result !== undefined) {\n              outputResult = (data as any).result;\n            } else {\n              assert(false, 'No result in query data');\n            }\n          }\n        }\n\n        if (type === 'Assert' && !outputResult) {\n          task.thought = thought;\n          throw new Error(`Assertion failed: ${thought}`);\n        }\n\n        return {\n          output: outputResult,\n          log: queryDump,\n          thought,\n        };\n      },\n    };\n\n    return queryTask;\n  }\n  async createTypeQueryExecution<T>(\n    type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert',\n    demand: ServiceExtractParam,\n    modelConfig: IModelConfig,\n    opt?: ServiceExtractOption,\n    multimodalPrompt?: TMultimodalPrompt,\n  ): Promise<ExecutionResult<T>> {\n    const session = this.createExecutionSession(\n      taskTitleStr(\n        type,\n        typeof demand === 'string' ? demand : JSON.stringify(demand),\n      ),\n    );\n\n    const queryTask = await this.createTypeQueryTask(\n      type,\n      demand,\n      modelConfig,\n      opt,\n      multimodalPrompt,\n    );\n\n    const runner = session.getRunner();\n    const result = await session.appendAndRun(queryTask);\n\n    if (!result) {\n      throw new Error(\n        'result of taskExecutor.flush() is undefined in function createTypeQueryTask',\n      );\n    }\n\n    const { output, thought } = result;\n\n    return {\n      output,\n      thought,\n      runner,\n    };\n  }\n\n  async waitFor(\n    assertion: TUserPrompt,\n    opt: PlanningActionParamWaitFor,\n    modelConfig: IModelConfig,\n  ): Promise<ExecutionResult<void>> {\n    const { textPrompt, multimodalPrompt } = parsePrompt(assertion);\n\n    const description = `waitFor: ${textPrompt}`;\n    const session = this.createExecutionSession(\n      taskTitleStr('WaitFor', description),\n    );\n    const runner = session.getRunner();\n    const {\n      timeoutMs,\n      checkIntervalMs,\n      domIncluded,\n      screenshotIncluded,\n      ...restOpt\n    } = opt;\n    const serviceExtractOpt: ServiceExtractOption = {\n      domIncluded,\n      screenshotIncluded,\n      ...restOpt,\n    };\n\n    assert(assertion, 'No assertion for waitFor');\n    assert(timeoutMs, 'No timeoutMs for waitFor');\n    assert(checkIntervalMs, 'No checkIntervalMs for waitFor');\n\n    assert(\n      checkIntervalMs <= timeoutMs,\n      `wrong config for waitFor: checkIntervalMs must be less than timeoutMs, config: {checkIntervalMs: ${checkIntervalMs}, timeoutMs: ${timeoutMs}}`,\n    );\n\n    const overallStartTime = Date.now();\n    let lastCheckStart = overallStartTime;\n    let errorThought = '';\n    // Continue checking as long as the previous iteration began within the timeout window.\n    while (lastCheckStart - overallStartTime <= timeoutMs) {\n      const currentCheckStart = Date.now();\n      lastCheckStart = currentCheckStart;\n      const queryTask = await this.createTypeQueryTask(\n        'WaitFor',\n        textPrompt,\n        modelConfig,\n        serviceExtractOpt,\n        multimodalPrompt,\n      );\n\n      const result = (await session.appendAndRun(queryTask)) as\n        | {\n            output: boolean;\n            thought?: string;\n          }\n        | undefined;\n\n      if (result?.output) {\n        return {\n          output: undefined,\n          runner,\n        };\n      }\n\n      errorThought =\n        result?.thought ||\n        (!result && `No result from assertion: ${textPrompt}`) ||\n        `unknown error when waiting for assertion: ${textPrompt}`;\n      const now = Date.now();\n      if (now - currentCheckStart < checkIntervalMs) {\n        const elapsed = now - currentCheckStart;\n        const timeRemaining = checkIntervalMs - elapsed;\n        const thought = `Check interval is ${checkIntervalMs}ms, ${elapsed}ms elapsed since last check, sleeping for ${timeRemaining}ms`;\n        const { tasks: sleepTasks } = await this.convertPlanToExecutable(\n          [{ type: 'Sleep', param: { timeMs: timeRemaining }, thought }],\n          modelConfig,\n          modelConfig,\n        );\n        if (sleepTasks[0]) {\n          await session.appendAndRun(sleepTasks[0]);\n        }\n      }\n    }\n\n    return session.appendErrorPlan(`waitFor timeout: ${errorThought}`);\n  }\n}\n\nexport async function withFileChooser<T>(\n  interfaceInstance: AbstractInterface,\n  fileChooserAccept: string[] | undefined,\n  action: () => Promise<T>,\n): Promise<T> {\n  if (!fileChooserAccept?.length) {\n    return action();\n  }\n\n  if (!interfaceInstance.registerFileChooserListener) {\n    throw new Error(\n      `File upload is not supported on ${interfaceInstance.interfaceType}`,\n    );\n  }\n\n  const handler = async (chooser: FileChooserHandler) => {\n    await chooser.accept(fileChooserAccept);\n  };\n\n  const { dispose, getError } =\n    await interfaceInstance.registerFileChooserListener(handler);\n  try {\n    const result = await action();\n    // Check for errors that occurred during file chooser handling\n    const error = getError();\n    if (error) {\n      throw error;\n    }\n    return result;\n  } finally {\n    dispose();\n  }\n}\n"],"names":["debug","getDebug","maxErrorCountAllowedInOnePlanningLoop","TaskExecutor","title","options","ExecutionSession","Promise","format","timestamp","getCurrentTime","getReadableTimeString","plans","modelConfigForPlanning","modelConfigForDefaultIntent","userInstruction","yamlString","session","taskTitleStr","task","param","executorContext","uiContext","assert","runner","tasks","result","output","userPrompt","includeBboxInPlanning","aiActContext","cacheable","replanningCycleLimitOverride","imagesIncludeCount","deepThink","fileChooserAccept","deepLocate","abortSignal","withFileChooser","conversationHistory","ConversationHistory","replanCount","yamlFlow","replanningCycleLimit","undefined","errorCountInOnePlanningLoop","outputString","subGoalStatus","memoriesStatus","modelFamily","timing","actionSpace","action","Array","console","planImpl","isUITars","uiTarsPlanning","isAutoGLM","autoGLMPlanning","plan","planResult","setTimingFieldOnce","planError","AIResponseParseError","JSON","actions","thought","log","memory","error","usage","rawResponse","reasoning_content","finalizeSuccess","finalizeMessage","updateSubGoals","markFinishedIndexes","executables","initialTimeString","timeString","String","Error","errorMsg","type","demand","modelConfig","opt","multimodalPrompt","queryTask","taskContext","queryDump","applyDump","dump","ifTypeRestricted","demandInput","keyOfResult","currentScreenshotConstraint","booleanPrompt","extractResult","extraPageDescription","tree","descriptionOfTree","ServiceError","data","outputResult","assertion","textPrompt","parsePrompt","description","timeoutMs","checkIntervalMs","domIncluded","screenshotIncluded","restOpt","serviceExtractOpt","overallStartTime","Date","lastCheckStart","errorThought","currentCheckStart","now","elapsed","timeRemaining","sleepTasks","interfaceInstance","service","opts","TaskBuilder","handler","chooser","dispose","getError"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;AA2DA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,wCAAwC;AAIvC,MAAMC;IAsBX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IAiCQ,uBACNC,KAAa,EACbC,OAA0C,EAC1C;QACA,OAAO,IAAIC,iBACTF,OACA,IAAMG,QAAQ,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,kBAAkB,KACrD;YACE,aAAa,IAAI,CAAC,mBAAmB;YACrC,OAAOF,SAAS;YAChB,cAAc,IAAI,CAAC,KAAK,EAAE;QAC5B;IAEJ;IAEQ,iBAAiC;QACvC,OAAO,IAAI,CAAC,mBAAmB;IACjC;IAQA,MAAc,cAAcG,MAAe,EAAmB;QAC5D,MAAMC,YAAY,MAAMC,eACtB,IAAI,CAAC,SAAS,EACd,IAAI,CAAC,kBAAkB;QAEzB,OAAOC,sBAAsBH,QAAQC;IACvC;IAEA,MAAa,wBACXG,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACzCT,OAIC,EACD;QACA,OAAO,IAAI,CAAC,WAAW,CAAC,KAAK,CAC3BO,OACAC,wBACAC,6BACAT;IAEJ;IAEA,MAAM,uBAAuBU,eAAuB,EAAEC,UAAkB,EAAE;QACxE,MAAMC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,OAAOH;QAGtB,MAAMI,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,OAAO;gBACLJ;YACF;YACA,UAAU,OAAOK,OAAOC;gBACtB,MAAM,EAAEC,SAAS,EAAE,GAAGD;gBACtBE,OAAOD,WAAW;gBAClB,OAAO;oBACL,QAAQ;wBACN,SAAS,EAAE;wBACX,wBAAwB;wBACxB,KAAK;wBACLN;oBACF;oBACA,OAAO;wBACL,KAAK;oBACP;oBACA,OAAO;wBACL,MAAM;wBACN,SAAS;4BACPA;wBACF;oBACF;gBACF;YACF;QACF;QACA,MAAMQ,SAASP,QAAQ,SAAS;QAChC,MAAMA,QAAQ,YAAY,CAACE;QAE3B,OAAO;YACLK;QACF;IACF;IAEA,MAAM,SACJpB,KAAa,EACbQ,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACf;QAC1B,MAAMG,UAAU,IAAI,CAAC,sBAAsB,CAACb;QAC5C,MAAM,EAAEqB,KAAK,EAAE,GAAG,MAAM,IAAI,CAAC,uBAAuB,CAClDb,OACAC,wBACAC;QAEF,MAAMU,SAASP,QAAQ,SAAS;QAChC,MAAMS,SAAS,MAAMT,QAAQ,YAAY,CAACQ;QAC1C,MAAM,EAAEE,MAAM,EAAE,GAAGD,UAAU,CAAC;QAC9B,OAAO;YACLC;YACAH;QACF;IACF;IAEA,MAAM,OACJI,UAAkB,EAClBf,sBAAoC,EACpCC,2BAAyC,EACzCe,qBAA8B,EAC9BC,YAAqB,EACrBC,SAAmB,EACnBC,4BAAqC,EACrCC,kBAA2B,EAC3BC,SAA2B,EAC3BC,iBAA4B,EAC5BC,UAAoB,EACpBC,WAAyB,EASzB;QACA,OAAOC,gBAAgB,IAAI,CAAC,SAAS,EAAEH,mBAAmB,UACjD,IAAI,CAAC,SAAS,CACnBP,YACAf,wBACAC,6BACAe,uBACAC,cACAC,WACAC,8BACAC,oBACAC,WACAE,YACAC;IAGN;IAEA,MAAc,UACZT,UAAkB,EAClBf,sBAAoC,EACpCC,2BAAyC,EACzCe,qBAA8B,EAC9BC,YAAqB,EACrBC,SAAmB,EACnBC,4BAAqC,EACrCC,kBAA2B,EAC3BC,SAA2B,EAC3BE,UAAoB,EACpBC,WAAyB,EASzB;QACA,MAAME,sBAAsB,IAAIC;QAEhC,MAAMvB,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,OAAOU;QAEtB,MAAMJ,SAASP,QAAQ,SAAS;QAEhC,IAAIwB,cAAc;QAClB,MAAMC,WAAmC,EAAE;QAC3C,MAAMC,uBACJX,gCAAgC,IAAI,CAAC,oBAAoB;QAC3DT,OACEoB,AAAyBC,WAAzBD,sBACA;QAGF,IAAIE,8BAA8B;QAClC,IAAIC;QAGJ,MAAO,KAAM;YAEX,IAAIT,aAAa,SACf,OAAOpB,QAAQ,eAAe,CAC5B,CAAC,cAAc,EAAEoB,YAAY,MAAM,IAAI,yBAAyB;YAKpE,MAAMU,gBAAgBR,oBAAoB,cAAc,MAAMK;YAG9D,MAAMI,iBAAiBT,oBAAoB,cAAc,MAAMK;YAE/D,MAAMlB,SAAS,MAAMT,QAAQ,YAAY,CACvC;gBACE,MAAM;gBACN,SAAS;gBACT,OAAO;oBACL,iBAAiBW;oBACjBE;oBACAG;oBACAC;oBACA,GAAIa,gBAAgB;wBAAEA;oBAAc,IAAI,CAAC,CAAC;oBAC1C,GAAIC,iBAAiB;wBAAEA;oBAAe,IAAI,CAAC,CAAC;gBAC9C;gBACA,UAAU,OAAO5B,OAAOC;oBACtB,MAAM,EAAEC,SAAS,EAAE,GAAGD;oBACtBE,OAAOD,WAAW;oBAClB,MAAM,EAAE2B,WAAW,EAAE,GAAGpC;oBACxB,MAAMqC,SAAS7B,gBAAgB,IAAI,CAAC,MAAM;oBAE1C,MAAM8B,cAAc,IAAI,CAAC,cAAc;oBACvCnD,MACE,sCACAmD,YAAY,GAAG,CAAC,CAACC,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;oBAEhD7B,OAAO8B,MAAM,OAAO,CAACF,cAAc;oBACnC,IAAIA,AAAuB,MAAvBA,YAAY,MAAM,EACpBG,QAAQ,IAAI,CACV,CAAC,gBAAgB,EAAE,IAAI,CAAC,SAAS,CAAC,aAAa,CAAC,gDAAgD,CAAC;oBAIrG,MAAMC,WAAWC,SAASP,eACtBQ,iBACAC,UAAUT,eACRU,kBACAC;oBAEN,IAAIC;oBACJ,IAAI;wBACFC,mBAAmBZ,QAAQ;wBAC3BW,aAAa,MAAMN,SAASnC,MAAM,eAAe,EAAE;4BACjD,SAASE;4BACT,eAAeF,MAAM,YAAY;4BACjC,eAAe,IAAI,CAAC,SAAS,CAAC,aAAa;4BAC3C+B;4BACA,aAAatC;4BACb0B;4BACA,aAAaV;4BACbI;4BACAC;4BACAG;wBACF;oBACF,EAAE,OAAO0B,WAAW;wBAClB,IAAIA,qBAAqBC,sBAAsB;4BAE7C3C,gBAAgB,IAAI,CAAC,KAAK,GAAG0C,UAAU,KAAK;4BAC5C1C,gBAAgB,IAAI,CAAC,GAAG,GAAG;gCACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;gCAClC,aAAa0C,UAAU,WAAW;4BACpC;wBACF;wBACA,MAAMA;oBACR,SAAU;wBACRD,mBAAmBZ,QAAQ;oBAC7B;oBACAlD,MAAM,cAAciE,KAAK,SAAS,CAACJ,YAAY,MAAM;oBAErD,MAAM,EACJK,OAAO,EACPC,OAAO,EACPC,GAAG,EACHC,MAAM,EACNC,KAAK,EACLC,KAAK,EACLC,WAAW,EACXC,iBAAiB,EACjBC,eAAe,EACfC,eAAe,EACfC,cAAc,EACdC,mBAAmB,EACpB,GAAGhB;oBACJf,eAAe6B;oBAEftD,gBAAgB,IAAI,CAAC,GAAG,GAAG;wBACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;wBAClCmD;oBACF;oBACAnD,gBAAgB,IAAI,CAAC,KAAK,GAAGkD;oBAC7BlD,gBAAgB,IAAI,CAAC,iBAAiB,GAAGoD;oBACzCpD,gBAAgB,IAAI,CAAC,MAAM,GAAG;wBAC5B,SAAS6C,WAAW,EAAE;wBACtBE;wBACAD;wBACAE;wBACA,UAAUR,WAAW,QAAQ;wBAC7B,QAAQc;wBACR,wBAAwBd,WAAW,sBAAsB;wBACzDe;wBACAC;oBACF;oBACAxD,gBAAgB,SAAS,GAAGC;oBAE5BC,OAAO,CAAC+C,OAAO,CAAC,oBAAoB,EAAEA,MAAM,EAAE,EAAEF,OAAO,IAAI;oBAG3D,IAAIM,AAAoB,UAApBA,iBACFnD,OACE,OACA,CAAC,aAAa,EAAEoD,mBAAmB,4BAA4B,EAAE,EAAEP,OAAO,IAAI;oBAIlF,OAAO;wBACL,OAAO;4BACL,KAAK;wBACP;oBACF;gBACF;YACF,GACA;gBACE,gBAAgB;YAClB;YAGF,MAAMP,aAAanC,QAAQ;YAG3B,MAAMd,QAAQiD,YAAY,WAAW,EAAE;YACvCnB,SAAS,IAAI,IAAKmB,YAAY,YAAY,EAAE;YAE5C,IAAIiB;YACJ,IAAI;gBACFA,cAAc,MAAM,IAAI,CAAC,uBAAuB,CAC9ClE,OACAC,wBACAC,6BACA;oBACEiB;oBACAK;oBACAC;gBACF;YAEJ,EAAE,OAAOiC,OAAO;gBACd,OAAOrD,QAAQ,eAAe,CAC5B,CAAC,4CAA4C,EAAEqD,MAAM,SAAS,EAAEL,KAAK,SAAS,CAC5ErD,QACC;YAEP;YACA,IAAI2B,oBAAoB,sBAAsB,EAC5Ce,QAAQ,IAAI,CACV,8FACAf,oBAAoB,sBAAsB;YAK9C,MAAMwC,oBAAoB,MAAM,IAAI,CAAC,aAAa;YAClDxC,oBAAoB,sBAAsB,IAAI,CAAC,cAAc,EAAEwC,mBAAmB;YAElF,IAAI;gBACF,MAAM9D,QAAQ,YAAY,CAAC6D,YAAY,KAAK;YAC9C,EAAE,OAAOR,OAAY;gBAEnBzB;gBACA,MAAMmC,aAAa,MAAM,IAAI,CAAC,aAAa;gBAC3CzC,oBAAoB,sBAAsB,GAAG,CAAC,MAAM,EAAEyC,WAAW,iCAAiC,EAAEV,OAAO,WAAWW,OAAOX,QAAQ;gBACrItE,MACE,yFACAsE,iBAAiBY,QAAQZ,MAAM,OAAO,GAAGW,OAAOX,QAChD,6CACAzB;YAEJ;YAEA,IAAIA,8BAA8B3C,uCAChC,OAAOe,QAAQ,eAAe,CAAC;YAIjC,IAAIoB,aAAa,SACf,OAAOpB,QAAQ,eAAe,CAC5B,CAAC,cAAc,EAAEoB,YAAY,MAAM,IAAI,yBAAyB;YAKpE,IAAI,CAACwB,YAAY,wBACf;YAIF,EAAEpB;YAEF,IAAIA,cAAcE,sBAAsB;gBACtC,MAAMwC,WAAW,CAAC,UAAU,EAAExC,qBAAqB,4JAA4J,CAAC;gBAChN,OAAO1B,QAAQ,eAAe,CAACkE;YACjC;YAEA,IAAI,CAAC5C,oBAAoB,sBAAsB,EAAE;gBAC/C,MAAMyC,aAAa,MAAM,IAAI,CAAC,aAAa;gBAC3CzC,oBAAoB,sBAAsB,GAAG,CAAC,MAAM,EAAEyC,WAAW,gDAAgD,CAAC;YACpH;QACF;QAEA,OAAO;YACL,QAAQ;gBACNtC;gBACA,QAAQI;YACV;YACAtB;QACF;IACF;IAEQ,oBACN4D,IAAsE,EACtEC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACpC;QACA,MAAMC,YAA4C;YAChD,MAAM;YACN,SAASL;YACT,OAAO;gBACL,aAAaG,KAAK;gBAClB,YAAYC,mBACP;oBACCH;oBACAG;gBACF,IACAH;YACN;YACA,UAAU,OAAOjE,OAAOsE;gBACtB,MAAM,EAAEvE,IAAI,EAAE,GAAGuE;gBACjB,IAAIC;gBACJ,MAAMC,YAAY,CAACC;oBACjBF,YAAYE;oBACZ1E,KAAK,GAAG,GAAG;wBACT0E;wBACA,aAAaA,KAAK,QAAQ,EAAE;oBAC9B;oBACA1E,KAAK,KAAK,GAAG0E,KAAK,QAAQ,EAAE;oBAC5B,IAAIA,KAAK,QAAQ,EAAE,mBACjB1E,KAAK,iBAAiB,GAAG0E,KAAK,QAAQ,CAAC,iBAAiB;gBAE5D;gBAGA,MAAMvE,YAAYoE,YAAY,SAAS;gBACvCnE,OAAOD,WAAW;gBAElB,MAAMwE,mBAAmBV,AAAS,YAATA;gBACzB,IAAIW,cAAcV;gBAClB,IAAIW,cAAc;gBAClB,MAAMC,8BACJ;gBACF,IAAIH,oBAAqBV,CAAAA,AAAS,aAATA,QAAqBA,AAAS,cAATA,IAAiB,GAAI;oBACjEY,cAAc;oBACd,MAAME,gBACJd,AAAS,aAATA,OACI,CAAC,SAAS,EAAEa,4BAA4B,2CAA2C,EAAEZ,QAAQ,GAC7F,CAAC,yDAAyD,EAAEY,4BAA4B,wDAAwD,EAAEZ,QAAQ;oBAChKU,cAAc;wBACZ,CAACC,YAAY,EAAEE;oBACjB;gBACF,OAAO,IAAIJ,kBAAkB;oBAC3BE,cAAcZ;oBACdW,cAAc;wBACZ,CAACC,YAAY,EAAE,GAAGZ,KAAK,EAAE,EAAEa,4BAA4B,EAAE,EAAEZ,QAAQ;oBACrE;gBACF;gBAEA,IAAIc;gBAEJ,IAAIC,uBAAuB;gBAC3B,IAAIb,KAAK,eAAe,IAAI,CAAC,SAAS,CAAC,mBAAmB,EAAE;oBAC1DvF,MAAM;oBACN,MAAMqG,OAAO,MAAM,IAAI,CAAC,SAAS,CAAC,mBAAmB;oBACrDD,uBAAuB,MAAME,kBAC3BD,MACA,KACA,OACAd,KAAK,gBAAgB;gBAEzB;gBAEA,IAAI;oBACFY,gBAAgB,MAAM,IAAI,CAAC,OAAO,CAAC,OAAO,CACxCJ,aACAT,aACAC,KACAa,sBACAZ,kBACAlE;gBAEJ,EAAE,OAAOgD,OAAO;oBACd,IAAIA,iBAAiBiC,cACnBX,UAAUtB,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAEA,MAAM,EAAEkC,IAAI,EAAErC,OAAO,EAAE0B,IAAI,EAAE,GAAGM;gBAChCP,UAAUC;gBAEV,IAAIY,eAAeD;gBACnB,IAAIV,kBAEF,IAAI,AAAgB,YAAhB,OAAOU,MACTC,eAAeD;qBACV,IAAIpB,AAAS,cAATA,MAEPqB,eADED,QAAAA,OACa,QAECA,IAAY,CAACR,YAAY;qBAEtC,IAAIQ,QAAAA,MACTC,eAAe;qBAGf,IAAID,MAAM,CAACR,YAAY,KAAKpD,QAC1B6D,eAAgBD,IAAY,CAACR,YAAY;qBACpC,IAAIQ,MAAM,WAAW5D,QAC1B6D,eAAgBD,KAAa,MAAM;qBAEnCjF,OAAO,OAAO;gBAKpB,IAAI6D,AAAS,aAATA,QAAqB,CAACqB,cAAc;oBACtCtF,KAAK,OAAO,GAAGgD;oBACf,MAAM,IAAIe,MAAM,CAAC,kBAAkB,EAAEf,SAAS;gBAChD;gBAEA,OAAO;oBACL,QAAQsC;oBACR,KAAKd;oBACLxB;gBACF;YACF;QACF;QAEA,OAAOsB;IACT;IACA,MAAM,yBACJL,IAA0D,EAC1DC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACP;QAC7B,MAAMvE,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aACEkE,MACA,AAAkB,YAAlB,OAAOC,SAAsBA,SAASpB,KAAK,SAAS,CAACoB;QAIzD,MAAMI,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9CL,MACAC,QACAC,aACAC,KACAC;QAGF,MAAMhE,SAASP,QAAQ,SAAS;QAChC,MAAMS,SAAS,MAAMT,QAAQ,YAAY,CAACwE;QAE1C,IAAI,CAAC/D,QACH,MAAM,IAAIwD,MACR;QAIJ,MAAM,EAAEvD,MAAM,EAAEwC,OAAO,EAAE,GAAGzC;QAE5B,OAAO;YACLC;YACAwC;YACA3C;QACF;IACF;IAEA,MAAM,QACJkF,SAAsB,EACtBnB,GAA+B,EAC/BD,WAAyB,EACO;QAChC,MAAM,EAAEqB,UAAU,EAAEnB,gBAAgB,EAAE,GAAGoB,YAAYF;QAErD,MAAMG,cAAc,CAAC,SAAS,EAAEF,YAAY;QAC5C,MAAM1F,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,WAAW2F;QAE1B,MAAMrF,SAASP,QAAQ,SAAS;QAChC,MAAM,EACJ6F,SAAS,EACTC,eAAe,EACfC,WAAW,EACXC,kBAAkB,EAClB,GAAGC,SACJ,GAAG3B;QACJ,MAAM4B,oBAA0C;YAC9CH;YACAC;YACA,GAAGC,OAAO;QACZ;QAEA3F,OAAOmF,WAAW;QAClBnF,OAAOuF,WAAW;QAClBvF,OAAOwF,iBAAiB;QAExBxF,OACEwF,mBAAmBD,WACnB,CAAC,iGAAiG,EAAEC,gBAAgB,aAAa,EAAED,UAAU,CAAC,CAAC;QAGjJ,MAAMM,mBAAmBC,KAAK,GAAG;QACjC,IAAIC,iBAAiBF;QACrB,IAAIG,eAAe;QAEnB,MAAOD,iBAAiBF,oBAAoBN,UAAW;YACrD,MAAMU,oBAAoBH,KAAK,GAAG;YAClCC,iBAAiBE;YACjB,MAAM/B,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9C,WACAkB,YACArB,aACA6B,mBACA3B;YAGF,MAAM9D,SAAU,MAAMT,QAAQ,YAAY,CAACwE;YAO3C,IAAI/D,QAAQ,QACV,OAAO;gBACL,QAAQkB;gBACRpB;YACF;YAGF+F,eACE7F,QAAQ,WACP,CAACA,UAAU,CAAC,0BAA0B,EAAEiF,YAAY,IACrD,CAAC,0CAA0C,EAAEA,YAAY;YAC3D,MAAMc,MAAMJ,KAAK,GAAG;YACpB,IAAII,MAAMD,oBAAoBT,iBAAiB;gBAC7C,MAAMW,UAAUD,MAAMD;gBACtB,MAAMG,gBAAgBZ,kBAAkBW;gBACxC,MAAMvD,UAAU,CAAC,kBAAkB,EAAE4C,gBAAgB,IAAI,EAAEW,QAAQ,0CAA0C,EAAEC,cAAc,EAAE,CAAC;gBAChI,MAAM,EAAE,OAAOC,UAAU,EAAE,GAAG,MAAM,IAAI,CAAC,uBAAuB,CAC9D;oBAAC;wBAAE,MAAM;wBAAS,OAAO;4BAAE,QAAQD;wBAAc;wBAAGxD;oBAAQ;iBAAE,EAC9DmB,aACAA;gBAEF,IAAIsC,UAAU,CAAC,EAAE,EACf,MAAM3G,QAAQ,YAAY,CAAC2G,UAAU,CAAC,EAAE;YAE5C;QACF;QAEA,OAAO3G,QAAQ,eAAe,CAAC,CAAC,iBAAiB,EAAEsG,cAAc;IACnE;IAnsBA,YACEM,iBAAoC,EACpCC,OAAgB,EAChBC,IAQC,CACD;QArCF;QAEA;QAEA;QAEA,uBAAiB,uBAAjB;QAEA,uBAAiB,eAAjB;QAEA;QAEA,uBAAiB,SAAjB;QAEA;QAEA;QAEA;QAoBE,IAAI,CAAC,SAAS,GAAGF;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC,KAAK,SAAS;QAC/B,IAAI,CAAC,mBAAmB,GAAGA,MAAM;QACjC,IAAI,CAAC,oBAAoB,GAAGA,KAAK,oBAAoB;QACrD,IAAI,CAAC,eAAe,GAAGA,KAAK,eAAe;QAC3C,IAAI,CAAC,kBAAkB,GAAGA,KAAK,kBAAkB;QACjD,IAAI,CAAC,KAAK,GAAGA,KAAK,KAAK;QACvB,IAAI,CAAC,mBAAmB,GAAGA,KAAK,WAAW;QAC3C,IAAI,CAAC,WAAW,GAAG,IAAIC,YAAY;YACjCH;YACAC;YACA,WAAWC,KAAK,SAAS;YACzB,aAAa,IAAI,CAAC,cAAc;YAChC,iBAAiBA,KAAK,eAAe;QACvC;IACF;AAuqBF;AAEO,eAAezF,gBACpBuF,iBAAoC,EACpC1F,iBAAuC,EACvCiB,MAAwB;IAExB,IAAI,CAACjB,mBAAmB,QACtB,OAAOiB;IAGT,IAAI,CAACyE,kBAAkB,2BAA2B,EAChD,MAAM,IAAI3C,MACR,CAAC,gCAAgC,EAAE2C,kBAAkB,aAAa,EAAE;IAIxE,MAAMI,UAAU,OAAOC;QACrB,MAAMA,QAAQ,MAAM,CAAC/F;IACvB;IAEA,MAAM,EAAEgG,OAAO,EAAEC,QAAQ,EAAE,GACzB,MAAMP,kBAAkB,2BAA2B,CAACI;IACtD,IAAI;QACF,MAAMvG,SAAS,MAAM0B;QAErB,MAAMkB,QAAQ8D;QACd,IAAI9D,OACF,MAAMA;QAER,OAAO5C;IACT,SAAU;QACRyG;IACF;AACF"}

package/dist/es/agent/utils.mjs

@@ -123,7 +123,7 @@ async function matchElementFromCache(context, cacheEntry, cachePrompt, cacheable
         return;
     }
 }
-const getMidsceneVersion = ()=>"1.7.2";
+const getMidsceneVersion = ()=>"1.7.3-beta-20260410094527.0";
 const parsePrompt = (prompt)=>{
     if ('string' == typeof prompt) return {
         textPrompt: prompt,

package/dist/es/ai-model/inspect.mjs

@@ -39,7 +39,7 @@ const promptsToChatParam = async (multimodalPrompt)=>{
             content: [
                 {
                     type: 'text',
-                    text: 'Next, I will provide all the reference images.'
+                    text: 'Next, I will provide all the reference images. These reference images are supporting context only, not the current screenshot being evaluated, unless the task explicitly asks for comparison or matching.'
                 }
             ]
         });
@@ -50,7 +50,7 @@ const promptsToChatParam = async (multimodalPrompt)=>{
                 content: [
                     {
                         type: 'text',
-                        text: `this is the reference image named '${item.name}':`
+                        text: `this is the reference image named '${item.name}'. It is a reference image, not the current screenshot:`
                     }
                 ]
             });
@@ -327,13 +327,19 @@ async function AiExtractElementInfo(options) {
     const screenshotBase64 = context.screenshot.base64;
     const extractDataPromptText = extractDataQueryPrompt(options.pageDescription || '', dataQuery);
     const userContent = [];
-    if (extractOption?.screenshotIncluded !== false) userContent.push({
-        type: 'image_url',
-        image_url: {
-            url: screenshotBase64,
-            detail: 'high'
-        }
-    });
+    if (extractOption?.screenshotIncluded !== false) {
+        userContent.push({
+            type: 'text',
+            text: 'This is the current screenshot to evaluate. Unless <DATA_DEMAND> explicitly asks for comparison or matching against reference images, base your answer on this screenshot and its contents when provided.'
+        });
+        userContent.push({
+            type: 'image_url',
+            image_url: {
+                url: screenshotBase64,
+                detail: 'high'
+            }
+        });
+    }
     userContent.push({
         type: 'text',
         text: extractDataPromptText

package/dist/es/ai-model/inspect.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/inspect.mjs","sources":["../../../src/ai-model/inspect.ts"],"sourcesContent":["import type {\n  AIDataExtractionResponse,\n  AIElementResponse,\n  AISectionLocatorResponse,\n  AIUsageInfo,\n  Rect,\n  ServiceExtractOption,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport {\n  generateElementByPoint,\n  generateElementByRect,\n} from '@midscene/shared/extractor/dom-util';\nimport {\n  cropByRect,\n  paddingToMatchBlockByBase64,\n  preProcessImageUrl,\n  scaleImage,\n} from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport type { LocateResultElement } from '@midscene/shared/types';\nimport { assert } from '@midscene/shared/utils';\nimport type {\n  ChatCompletionSystemMessageParam,\n  ChatCompletionUserMessageParam,\n} from 'openai/resources/index';\nimport type { TMultimodalPrompt, TUserPrompt } from '../common';\nimport { adaptBboxToRect, expandSearchArea, mergeRects } from '../common';\nimport { parseAutoGLMLocateResponse } from './auto-glm/parser';\nimport { getAutoGLMLocatePrompt } from './auto-glm/prompt';\nimport { isAutoGLM } from './auto-glm/util';\nimport {\n  extractDataQueryPrompt,\n  parseXMLExtractionResponse,\n  systemPromptToExtract,\n} from './prompt/extraction';\nimport {\n  findElementPrompt,\n  systemPromptToLocateElement,\n} from './prompt/llm-locator';\nimport {\n  sectionLocatorInstruction,\n  systemPromptToLocateSection,\n} from './prompt/llm-section-locator';\nimport {\n  orderSensitiveJudgePrompt,\n  systemPromptToJudgeOrderSensitive,\n} from './prompt/order-sensitive-judge';\nimport {\n  AIResponseParseError,\n  callAI,\n  callAIWithObjectResponse,\n  callAIWithStringResponse,\n} from './service-caller/index';\n\nexport type AIArgs = [\n  ChatCompletionSystemMessageParam,\n  ...ChatCompletionUserMessageParam[],\n];\n\nconst debugInspect = getDebug('ai:inspect');\nconst debugSection = getDebug('ai:section');\n\nexport async function buildSearchAreaConfig(options: {\n  context: UIContext;\n  baseRect: Rect;\n  modelFamily: IModelConfig['modelFamily'];\n}): Promise<{ rect: Rect; imageBase64: string; scale: number }> {\n  const { context, baseRect, modelFamily } = options;\n  const scaleRatio = 2;\n  const sectionRect = expandSearchArea(baseRect, context.shotSize);\n\n  const croppedResult = await cropByRect(\n    context.screenshot.base64,\n    sectionRect,\n    modelFamily === 'qwen2.5-vl',\n  );\n\n  const scaledResult = await scaleImage(croppedResult.imageBase64, scaleRatio);\n  sectionRect.width = scaledResult.width;\n  sectionRect.height = scaledResult.height;\n  return {\n    rect: sectionRect,\n    imageBase64: scaledResult.imageBase64,\n    scale: scaleRatio,\n  };\n}\n\nconst extraTextFromUserPrompt = (prompt: TUserPrompt): string => {\n  if (typeof prompt === 'string') {\n    return prompt;\n  } else {\n    return prompt.prompt;\n  }\n};\n\nconst promptsToChatParam = async (\n  multimodalPrompt: TMultimodalPrompt,\n): Promise<ChatCompletionUserMessageParam[]> => {\n  const msgs: ChatCompletionUserMessageParam[] = [];\n  if (multimodalPrompt?.images?.length) {\n    msgs.push({\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: 'Next, I will provide all the reference images.',\n        },\n      ],\n    });\n\n    for (const item of multimodalPrompt.images) {\n      const base64 = await preProcessImageUrl(\n        item.url,\n        !!multimodalPrompt.convertHttpImage2Base64,\n      );\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'text',\n            text: `this is the reference image named '${item.name}':`,\n          },\n        ],\n      });\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: base64,\n              detail: 'high',\n            },\n          },\n        ],\n      });\n    }\n  }\n  return msgs;\n};\n\nexport async function AiLocateElement(options: {\n  context: UIContext;\n  targetElementDescription: TUserPrompt;\n  searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;\n  modelConfig: IModelConfig;\n  abortSignal?: AbortSignal;\n}): Promise<{\n  parseResult: {\n    elements: LocateResultElement[];\n    errors?: string[];\n  };\n  rect?: Rect;\n  rawResponse: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}> {\n  const { context, targetElementDescription, modelConfig } = options;\n  const { modelFamily } = modelConfig;\n  const screenshotBase64 = context.screenshot.base64;\n\n  assert(\n    targetElementDescription,\n    'cannot find the target element description',\n  );\n  const targetElementDescriptionText = extraTextFromUserPrompt(\n    targetElementDescription,\n  );\n  const userInstructionPrompt = findElementPrompt(targetElementDescriptionText);\n  const systemPrompt = isAutoGLM(modelFamily)\n    ? getAutoGLMLocatePrompt(modelFamily)\n    : systemPromptToLocateElement(modelFamily);\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = context.shotSize.width;\n  let imageHeight = context.shotSize.height;\n  let originalImageWidth = imageWidth;\n  let originalImageHeight = imageHeight;\n\n  if (options.searchConfig) {\n    assert(\n      options.searchConfig.rect,\n      'searchArea is provided but its rect cannot be found. Failed to locate element',\n    );\n    assert(\n      options.searchConfig.imageBase64,\n      'searchArea is provided but its imageBase64 cannot be found. Failed to locate element',\n    );\n\n    imagePayload = options.searchConfig.imageBase64;\n    imageWidth = options.searchConfig.rect?.width;\n    imageHeight = options.searchConfig.rect?.height;\n    originalImageWidth = imageWidth;\n    originalImageHeight = imageHeight;\n  } else if (modelFamily === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  }\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: isAutoGLM(modelFamily)\n            ? `Tap: ${userInstructionPrompt}`\n            : userInstructionPrompt,\n        },\n      ],\n    },\n  ];\n\n  if (typeof targetElementDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: targetElementDescription.images,\n      convertHttpImage2Base64: targetElementDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  if (isAutoGLM(modelFamily)) {\n    const { content: rawResponseContent, usage } =\n      await callAIWithStringResponse(msgs, modelConfig, {\n        abortSignal: options.abortSignal,\n      });\n\n    debugInspect('auto-glm rawResponse:', rawResponseContent);\n\n    const parsed = parseAutoGLMLocateResponse(rawResponseContent);\n\n    debugInspect('auto-glm thinking:', parsed.think);\n    debugInspect('auto-glm coordinates:', parsed.coordinates);\n\n    let resRect: Rect | undefined;\n    let matchedElements: LocateResultElement[] = [];\n    let errors: string[] = [];\n\n    if (parsed.error || !parsed.coordinates) {\n      errors = [parsed.error || 'Failed to parse auto-glm response'];\n      debugInspect('auto-glm parse error:', errors[0]);\n    } else {\n      const { x, y } = parsed.coordinates;\n\n      debugInspect('auto-glm coordinates [0-999]:', { x, y });\n\n      // Convert auto-glm coordinates [0,999] to pixel bbox\n      // Map from [0,999] to pixel coordinates\n      const pixelX = Math.round((x * imageWidth) / 1000);\n      const pixelY = Math.round((y * imageHeight) / 1000);\n\n      debugInspect('auto-glm pixel coordinates:', { pixelX, pixelY });\n\n      // Apply offset if searching in a cropped area\n      let finalX = pixelX;\n      let finalY = pixelY;\n      if (options.searchConfig?.rect) {\n        finalX += options.searchConfig.rect.left;\n        finalY += options.searchConfig.rect.top;\n      }\n\n      const element: LocateResultElement = generateElementByPoint(\n        [finalX, finalY],\n        targetElementDescriptionText as string,\n      );\n\n      resRect = element.rect;\n      debugInspect('auto-glm resRect:', resRect);\n\n      if (element) {\n        matchedElements = [element];\n      }\n    }\n\n    return {\n      rect: resRect,\n      parseResult: {\n        elements: matchedElements,\n        errors,\n      },\n      rawResponse: rawResponseContent,\n      usage,\n      reasoning_content: parsed.think,\n    };\n  }\n\n  let res: Awaited<\n    ReturnType<\n      typeof callAIWithObjectResponse<AIElementResponse | [number, number]>\n    >\n  >;\n  try {\n    res = await callAIWithObjectResponse<AIElementResponse | [number, number]>(\n      msgs,\n      modelConfig,\n      { abortSignal: options.abortSignal },\n    );\n  } catch (callError) {\n    // Return error with usage and rawResponse if available\n    const errorMessage =\n      callError instanceof Error ? callError.message : String(callError);\n    const rawResponse =\n      callError instanceof AIResponseParseError\n        ? callError.rawResponse\n        : errorMessage;\n    const usage =\n      callError instanceof AIResponseParseError ? callError.usage : undefined;\n    return {\n      rect: undefined,\n      parseResult: {\n        elements: [],\n        errors: [`AI call error: ${errorMessage}`],\n      },\n      rawResponse,\n      usage,\n      reasoning_content: undefined,\n    };\n  }\n\n  const rawResponse = JSON.stringify(res.content);\n\n  let resRect: Rect | undefined;\n  let matchedElements: LocateResultElement[] = [];\n  let errors: string[] | undefined =\n    'errors' in res.content ? res.content.errors : [];\n  try {\n    if (\n      'bbox' in res.content &&\n      Array.isArray(res.content.bbox) &&\n      res.content.bbox.length >= 1\n    ) {\n      resRect = adaptBboxToRect(\n        res.content.bbox,\n        imageWidth,\n        imageHeight,\n        options.searchConfig?.rect?.left,\n        options.searchConfig?.rect?.top,\n        originalImageWidth,\n        originalImageHeight,\n        modelFamily,\n        options.searchConfig?.scale,\n      );\n\n      debugInspect('resRect', resRect);\n\n      const element: LocateResultElement = generateElementByRect(\n        resRect,\n        targetElementDescriptionText as string,\n      );\n      errors = [];\n\n      if (element) {\n        matchedElements = [element];\n      }\n    }\n  } catch (e) {\n    const msg =\n      e instanceof Error\n        ? `Failed to parse bbox: ${e.message}`\n        : 'unknown error in locate';\n    if (!errors || errors?.length === 0) {\n      errors = [msg];\n    } else {\n      errors.push(`(${msg})`);\n    }\n  }\n\n  return {\n    rect: resRect,\n    parseResult: {\n      elements: matchedElements as LocateResultElement[],\n      errors: errors as string[],\n    },\n    rawResponse,\n    usage: res.usage,\n    reasoning_content: res.reasoning_content,\n  };\n}\n\nexport async function AiLocateSection(options: {\n  context: UIContext;\n  sectionDescription: TUserPrompt;\n  modelConfig: IModelConfig;\n  abortSignal?: AbortSignal;\n}): Promise<{\n  rect?: Rect;\n  imageBase64?: string;\n  scale?: number;\n  error?: string;\n  rawResponse: string;\n  usage?: AIUsageInfo;\n}> {\n  const { context, sectionDescription, modelConfig } = options;\n  const { modelFamily } = modelConfig;\n  const screenshotBase64 = context.screenshot.base64;\n\n  const systemPrompt = systemPromptToLocateSection(modelFamily);\n  const sectionLocatorInstructionText = sectionLocatorInstruction(\n    extraTextFromUserPrompt(sectionDescription),\n  );\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'image_url',\n          image_url: {\n            url: screenshotBase64,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: sectionLocatorInstructionText,\n        },\n      ],\n    },\n  ];\n\n  if (typeof sectionDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: sectionDescription.images,\n      convertHttpImage2Base64: sectionDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  let result: Awaited<\n    ReturnType<typeof callAIWithObjectResponse<AISectionLocatorResponse>>\n  >;\n  try {\n    result = await callAIWithObjectResponse<AISectionLocatorResponse>(\n      msgs,\n      modelConfig,\n      { abortSignal: options.abortSignal },\n    );\n  } catch (callError) {\n    // Return error with usage and rawResponse if available\n    const errorMessage =\n      callError instanceof Error ? callError.message : String(callError);\n    const rawResponse =\n      callError instanceof AIResponseParseError\n        ? callError.rawResponse\n        : errorMessage;\n    const usage =\n      callError instanceof AIResponseParseError ? callError.usage : undefined;\n    return {\n      rect: undefined,\n      imageBase64: undefined,\n      error: `AI call error: ${errorMessage}`,\n      rawResponse,\n      usage,\n    };\n  }\n\n  let searchAreaConfig:\n    | Awaited<ReturnType<typeof buildSearchAreaConfig>>\n    | undefined;\n  const sectionBbox = result.content.bbox;\n  if (sectionBbox) {\n    const targetRect = adaptBboxToRect(\n      sectionBbox,\n      context.shotSize.width,\n      context.shotSize.height,\n      0,\n      0,\n      context.shotSize.width,\n      context.shotSize.height,\n      modelFamily,\n    );\n    debugSection('original targetRect %j', targetRect);\n\n    const referenceBboxList = result.content.references_bbox || [];\n    debugSection('referenceBboxList %j', referenceBboxList);\n\n    const referenceRects = referenceBboxList\n      .filter((bbox) => Array.isArray(bbox))\n      .map((bbox) => {\n        return adaptBboxToRect(\n          bbox,\n          context.shotSize.width,\n          context.shotSize.height,\n          0,\n          0,\n          context.shotSize.width,\n          context.shotSize.height,\n          modelFamily,\n        );\n      });\n    debugSection('referenceRects %j', referenceRects);\n\n    // merge the sectionRect and referenceRects\n    const mergedRect = mergeRects([targetRect, ...referenceRects]);\n    debugSection('mergedRect %j', mergedRect);\n\n    const expandedRect = expandSearchArea(mergedRect, context.shotSize);\n    const originalWidth = expandedRect.width;\n    const originalHeight = expandedRect.height;\n    debugSection('expanded sectionRect %j', expandedRect);\n\n    searchAreaConfig = await buildSearchAreaConfig({\n      context,\n      baseRect: mergedRect,\n      modelFamily,\n    });\n\n    debugSection(\n      'scaled sectionRect from %dx%d to %dx%d (scale=%d)',\n      originalWidth,\n      originalHeight,\n      searchAreaConfig.rect.width,\n      searchAreaConfig.rect.height,\n      searchAreaConfig.scale,\n    );\n  }\n\n  return {\n    rect: searchAreaConfig?.rect,\n    imageBase64: searchAreaConfig?.imageBase64,\n    scale: searchAreaConfig?.scale,\n    error: result.content.error,\n    rawResponse: JSON.stringify(result.content),\n    usage: result.usage,\n  };\n}\n\nexport async function AiExtractElementInfo<T>(options: {\n  dataQuery: string | Record<string, string>;\n  multimodalPrompt?: TMultimodalPrompt;\n  context: UIContext;\n  pageDescription?: string;\n  extractOption?: ServiceExtractOption;\n  modelConfig: IModelConfig;\n}) {\n  const { dataQuery, context, extractOption, multimodalPrompt, modelConfig } =\n    options;\n  const systemPrompt = systemPromptToExtract();\n  const screenshotBase64 = context.screenshot.base64;\n\n  const extractDataPromptText = extractDataQueryPrompt(\n    options.pageDescription || '',\n    dataQuery,\n  );\n\n  const userContent: ChatCompletionUserMessageParam['content'] = [];\n\n  if (extractOption?.screenshotIncluded !== false) {\n    userContent.push({\n      type: 'image_url',\n      image_url: {\n        url: screenshotBase64,\n        detail: 'high',\n      },\n    });\n  }\n\n  userContent.push({\n    type: 'text',\n    text: extractDataPromptText,\n  });\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: userContent,\n    },\n  ];\n\n  if (multimodalPrompt) {\n    const addOns = await promptsToChatParam({\n      images: multimodalPrompt.images,\n      convertHttpImage2Base64: multimodalPrompt.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  const {\n    content: rawResponse,\n    usage,\n    reasoning_content,\n  } = await callAI(msgs, modelConfig);\n\n  // Parse XML response to JSON object\n  let parseResult: AIDataExtractionResponse<T>;\n  try {\n    parseResult = parseXMLExtractionResponse<T>(rawResponse);\n  } catch (parseError) {\n    // Throw AIResponseParseError with usage and rawResponse preserved\n    const errorMessage =\n      parseError instanceof Error ? parseError.message : String(parseError);\n    throw new AIResponseParseError(\n      `XML parse error: ${errorMessage}`,\n      rawResponse,\n      usage,\n    );\n  }\n\n  return {\n    parseResult,\n    rawResponse,\n    usage,\n    reasoning_content,\n  };\n}\n\nexport async function AiJudgeOrderSensitive(\n  description: string,\n  callAIFn: typeof callAIWithObjectResponse<{ isOrderSensitive: boolean }>,\n  modelConfig: IModelConfig,\n): Promise<{\n  isOrderSensitive: boolean;\n  usage?: AIUsageInfo;\n}> {\n  const systemPrompt = systemPromptToJudgeOrderSensitive();\n  const userPrompt = orderSensitiveJudgePrompt(description);\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: userPrompt,\n    },\n  ];\n\n  debugInspect(\n    'AiJudgeOrderSensitive: deepThink=false, description=%s',\n    description,\n  );\n\n  const result = await callAIFn(msgs, modelConfig, {\n    deepThink: false,\n  });\n\n  return {\n    isOrderSensitive: result.content.isOrderSensitive ?? false,\n    usage: result.usage,\n  };\n}\n"],"names":["debugInspect","getDebug","debugSection","buildSearchAreaConfig","options","context","baseRect","modelFamily","scaleRatio","sectionRect","expandSearchArea","croppedResult","cropByRect","scaledResult","scaleImage","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","msgs","item","base64","preProcessImageUrl","AiLocateElement","targetElementDescription","modelConfig","screenshotBase64","assert","targetElementDescriptionText","userInstructionPrompt","findElementPrompt","systemPrompt","isAutoGLM","getAutoGLMLocatePrompt","systemPromptToLocateElement","imagePayload","imageWidth","imageHeight","originalImageWidth","originalImageHeight","paddedResult","paddingToMatchBlockByBase64","addOns","rawResponseContent","usage","callAIWithStringResponse","parsed","parseAutoGLMLocateResponse","resRect","matchedElements","errors","x","y","pixelX","Math","pixelY","finalX","finalY","element","generateElementByPoint","res","callAIWithObjectResponse","callError","errorMessage","Error","String","rawResponse","AIResponseParseError","undefined","JSON","Array","adaptBboxToRect","generateElementByRect","e","msg","AiLocateSection","sectionDescription","systemPromptToLocateSection","sectionLocatorInstructionText","sectionLocatorInstruction","result","searchAreaConfig","sectionBbox","targetRect","referenceBboxList","referenceRects","bbox","mergedRect","mergeRects","expandedRect","originalWidth","originalHeight","AiExtractElementInfo","dataQuery","extractOption","systemPromptToExtract","extractDataPromptText","extractDataQueryPrompt","userContent","reasoning_content","callAI","parseResult","parseXMLExtractionResponse","parseError","AiJudgeOrderSensitive","description","callAIFn","systemPromptToJudgeOrderSensitive","userPrompt","orderSensitiveJudgePrompt"],"mappings":";;;;;;;;;;;;;AA6DA,MAAMA,eAAeC,SAAS;AAC9B,MAAMC,eAAeD,SAAS;AAEvB,eAAeE,sBAAsBC,OAI3C;IACC,MAAM,EAAEC,OAAO,EAAEC,QAAQ,EAAEC,WAAW,EAAE,GAAGH;IAC3C,MAAMI,aAAa;IACnB,MAAMC,cAAcC,iBAAiBJ,UAAUD,QAAQ,QAAQ;IAE/D,MAAMM,gBAAgB,MAAMC,WAC1BP,QAAQ,UAAU,CAAC,MAAM,EACzBI,aACAF,AAAgB,iBAAhBA;IAGF,MAAMM,eAAe,MAAMC,WAAWH,cAAc,WAAW,EAAEH;IACjEC,YAAY,KAAK,GAAGI,aAAa,KAAK;IACtCJ,YAAY,MAAM,GAAGI,aAAa,MAAM;IACxC,OAAO;QACL,MAAMJ;QACN,aAAaI,aAAa,WAAW;QACrC,OAAOL;IACT;AACF;AAEA,MAAMO,0BAA0B,CAACC;IAC/B,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAEP,OAAOA,OAAO,MAAM;AAExB;AAEA,MAAMC,qBAAqB,OACzBC;IAEA,MAAMC,OAAyC,EAAE;IACjD,IAAID,kBAAkB,QAAQ,QAAQ;QACpCC,KAAK,IAAI,CAAC;YACR,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM;gBACR;aACD;QACH;QAEA,KAAK,MAAMC,QAAQF,iBAAiB,MAAM,CAAE;YAC1C,MAAMG,SAAS,MAAMC,mBACnBF,KAAK,GAAG,EACR,CAAC,CAACF,iBAAiB,uBAAuB;YAG5CC,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,MAAM,CAAC,mCAAmC,EAAEC,KAAK,IAAI,CAAC,EAAE,CAAC;oBAC3D;iBACD;YACH;YAEAD,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKE;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;QACF;IACF;IACA,OAAOF;AACT;AAEO,eAAeI,gBAAgBnB,OAMrC;IAUC,MAAM,EAAEC,OAAO,EAAEmB,wBAAwB,EAAEC,WAAW,EAAE,GAAGrB;IAC3D,MAAM,EAAEG,WAAW,EAAE,GAAGkB;IACxB,MAAMC,mBAAmBrB,QAAQ,UAAU,CAAC,MAAM;IAElDsB,OACEH,0BACA;IAEF,MAAMI,+BAA+Bb,wBACnCS;IAEF,MAAMK,wBAAwBC,kBAAkBF;IAChD,MAAMG,eAAeC,UAAUzB,eAC3B0B,uBAAuB1B,eACvB2B,4BAA4B3B;IAEhC,IAAI4B,eAAeT;IACnB,IAAIU,aAAa/B,QAAQ,QAAQ,CAAC,KAAK;IACvC,IAAIgC,cAAchC,QAAQ,QAAQ,CAAC,MAAM;IACzC,IAAIiC,qBAAqBF;IACzB,IAAIG,sBAAsBF;IAE1B,IAAIjC,QAAQ,YAAY,EAAE;QACxBuB,OACEvB,QAAQ,YAAY,CAAC,IAAI,EACzB;QAEFuB,OACEvB,QAAQ,YAAY,CAAC,WAAW,EAChC;QAGF+B,eAAe/B,QAAQ,YAAY,CAAC,WAAW;QAC/CgC,aAAahC,QAAQ,YAAY,CAAC,IAAI,EAAE;QACxCiC,cAAcjC,QAAQ,YAAY,CAAC,IAAI,EAAE;QACzCkC,qBAAqBF;QACrBG,sBAAsBF;IACxB,OAAO,IAAI9B,AAAgB,iBAAhBA,aAA8B;QACvC,MAAMiC,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAMrB,OAAe;QACnB;YAAE,MAAM;YAAU,SAASY;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKI;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMH,UAAUzB,eACZ,CAAC,KAAK,EAAEsB,uBAAuB,GAC/BA;gBACN;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAOL,0BAAuC;QAChD,MAAMkB,SAAS,MAAMzB,mBAAmB;YACtC,QAAQO,yBAAyB,MAAM;YACvC,yBAAyBA,yBAAyB,uBAAuB;QAC3E;QACAL,KAAK,IAAI,IAAIuB;IACf;IAEA,IAAIV,UAAUzB,cAAc;QAC1B,MAAM,EAAE,SAASoC,kBAAkB,EAAEC,KAAK,EAAE,GAC1C,MAAMC,yBAAyB1B,MAAMM,aAAa;YAChD,aAAarB,QAAQ,WAAW;QAClC;QAEFJ,aAAa,yBAAyB2C;QAEtC,MAAMG,SAASC,2BAA2BJ;QAE1C3C,aAAa,sBAAsB8C,OAAO,KAAK;QAC/C9C,aAAa,yBAAyB8C,OAAO,WAAW;QAExD,IAAIE;QACJ,IAAIC,kBAAyC,EAAE;QAC/C,IAAIC,SAAmB,EAAE;QAEzB,IAAIJ,OAAO,KAAK,IAAI,CAACA,OAAO,WAAW,EAAE;YACvCI,SAAS;gBAACJ,OAAO,KAAK,IAAI;aAAoC;YAC9D9C,aAAa,yBAAyBkD,MAAM,CAAC,EAAE;QACjD,OAAO;YACL,MAAM,EAAEC,CAAC,EAAEC,CAAC,EAAE,GAAGN,OAAO,WAAW;YAEnC9C,aAAa,iCAAiC;gBAAEmD;gBAAGC;YAAE;YAIrD,MAAMC,SAASC,KAAK,KAAK,CAAEH,IAAIf,aAAc;YAC7C,MAAMmB,SAASD,KAAK,KAAK,CAAEF,IAAIf,cAAe;YAE9CrC,aAAa,+BAA+B;gBAAEqD;gBAAQE;YAAO;YAG7D,IAAIC,SAASH;YACb,IAAII,SAASF;YACb,IAAInD,QAAQ,YAAY,EAAE,MAAM;gBAC9BoD,UAAUpD,QAAQ,YAAY,CAAC,IAAI,CAAC,IAAI;gBACxCqD,UAAUrD,QAAQ,YAAY,CAAC,IAAI,CAAC,GAAG;YACzC;YAEA,MAAMsD,UAA+BC,uBACnC;gBAACH;gBAAQC;aAAO,EAChB7B;YAGFoB,UAAUU,QAAQ,IAAI;YACtB1D,aAAa,qBAAqBgD;YAElC,IAAIU,SACFT,kBAAkB;gBAACS;aAAQ;QAE/B;QAEA,OAAO;YACL,MAAMV;YACN,aAAa;gBACX,UAAUC;gBACVC;YACF;YACA,aAAaP;YACbC;YACA,mBAAmBE,OAAO,KAAK;QACjC;IACF;IAEA,IAAIc;IAKJ,IAAI;QACFA,MAAM,MAAMC,yBACV1C,MACAM,aACA;YAAE,aAAarB,QAAQ,WAAW;QAAC;IAEvC,EAAE,OAAO0D,WAAW;QAElB,MAAMC,eACJD,qBAAqBE,QAAQF,UAAU,OAAO,GAAGG,OAAOH;QAC1D,MAAMI,cACJJ,qBAAqBK,uBACjBL,UAAU,WAAW,GACrBC;QACN,MAAMnB,QACJkB,qBAAqBK,uBAAuBL,UAAU,KAAK,GAAGM;QAChE,OAAO;YACL,MAAMA;YACN,aAAa;gBACX,UAAU,EAAE;gBACZ,QAAQ;oBAAC,CAAC,eAAe,EAAEL,cAAc;iBAAC;YAC5C;YACAG;YACAtB;YACA,mBAAmBwB;QACrB;IACF;IAEA,MAAMF,cAAcG,KAAK,SAAS,CAACT,IAAI,OAAO;IAE9C,IAAIZ;IACJ,IAAIC,kBAAyC,EAAE;IAC/C,IAAIC,SACF,YAAYU,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,MAAM,GAAG,EAAE;IACnD,IAAI;QACF,IACE,UAAUA,IAAI,OAAO,IACrBU,MAAM,OAAO,CAACV,IAAI,OAAO,CAAC,IAAI,KAC9BA,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,IAAI,GAC3B;YACAZ,UAAUuB,gBACRX,IAAI,OAAO,CAAC,IAAI,EAChBxB,YACAC,aACAjC,QAAQ,YAAY,EAAE,MAAM,MAC5BA,QAAQ,YAAY,EAAE,MAAM,KAC5BkC,oBACAC,qBACAhC,aACAH,QAAQ,YAAY,EAAE;YAGxBJ,aAAa,WAAWgD;YAExB,MAAMU,UAA+Bc,sBACnCxB,SACApB;YAEFsB,SAAS,EAAE;YAEX,IAAIQ,SACFT,kBAAkB;gBAACS;aAAQ;QAE/B;IACF,EAAE,OAAOe,GAAG;QACV,MAAMC,MACJD,aAAaT,QACT,CAAC,sBAAsB,EAAES,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACvB,UAAUA,QAAQ,WAAW,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAEwB,IAAI,CAAC,CAAC;aAFtBxB,SAAS;YAACwB;SAAI;IAIlB;IAEA,OAAO;QACL,MAAM1B;QACN,aAAa;YACX,UAAUC;YACV,QAAQC;QACV;QACAgB;QACA,OAAON,IAAI,KAAK;QAChB,mBAAmBA,IAAI,iBAAiB;IAC1C;AACF;AAEO,eAAee,gBAAgBvE,OAKrC;IAQC,MAAM,EAAEC,OAAO,EAAEuE,kBAAkB,EAAEnD,WAAW,EAAE,GAAGrB;IACrD,MAAM,EAAEG,WAAW,EAAE,GAAGkB;IACxB,MAAMC,mBAAmBrB,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM0B,eAAe8C,4BAA4BtE;IACjD,MAAMuE,gCAAgCC,0BACpChE,wBAAwB6D;IAE1B,MAAMzD,OAAe;QACnB;YAAE,MAAM;YAAU,SAASY;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKL;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMoD;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAMlC,SAAS,MAAMzB,mBAAmB;YACtC,QAAQ2D,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACAzD,KAAK,IAAI,IAAIuB;IACf;IAEA,IAAIsC;IAGJ,IAAI;QACFA,SAAS,MAAMnB,yBACb1C,MACAM,aACA;YAAE,aAAarB,QAAQ,WAAW;QAAC;IAEvC,EAAE,OAAO0D,WAAW;QAElB,MAAMC,eACJD,qBAAqBE,QAAQF,UAAU,OAAO,GAAGG,OAAOH;QAC1D,MAAMI,cACJJ,qBAAqBK,uBACjBL,UAAU,WAAW,GACrBC;QACN,MAAMnB,QACJkB,qBAAqBK,uBAAuBL,UAAU,KAAK,GAAGM;QAChE,OAAO;YACL,MAAMA;YACN,aAAaA;YACb,OAAO,CAAC,eAAe,EAAEL,cAAc;YACvCG;YACAtB;QACF;IACF;IAEA,IAAIqC;IAGJ,MAAMC,cAAcF,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIE,aAAa;QACf,MAAMC,aAAaZ,gBACjBW,aACA7E,QAAQ,QAAQ,CAAC,KAAK,EACtBA,QAAQ,QAAQ,CAAC,MAAM,EACvB,GACA,GACAA,QAAQ,QAAQ,CAAC,KAAK,EACtBA,QAAQ,QAAQ,CAAC,MAAM,EACvBE;QAEFL,aAAa,0BAA0BiF;QAEvC,MAAMC,oBAAoBJ,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9D9E,aAAa,wBAAwBkF;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAAShB,MAAM,OAAO,CAACgB,OAC/B,GAAG,CAAC,CAACA,OACGf,gBACLe,MACAjF,QAAQ,QAAQ,CAAC,KAAK,EACtBA,QAAQ,QAAQ,CAAC,MAAM,EACvB,GACA,GACAA,QAAQ,QAAQ,CAAC,KAAK,EACtBA,QAAQ,QAAQ,CAAC,MAAM,EACvBE;QAGNL,aAAa,qBAAqBmF;QAGlC,MAAME,aAAaC,WAAW;YAACL;eAAeE;SAAe;QAC7DnF,aAAa,iBAAiBqF;QAE9B,MAAME,eAAe/E,iBAAiB6E,YAAYlF,QAAQ,QAAQ;QAClE,MAAMqF,gBAAgBD,aAAa,KAAK;QACxC,MAAME,iBAAiBF,aAAa,MAAM;QAC1CvF,aAAa,2BAA2BuF;QAExCR,mBAAmB,MAAM9E,sBAAsB;YAC7CE;YACA,UAAUkF;YACVhF;QACF;QAEAL,aACE,qDACAwF,eACAC,gBACAV,iBAAiB,IAAI,CAAC,KAAK,EAC3BA,iBAAiB,IAAI,CAAC,MAAM,EAC5BA,iBAAiB,KAAK;IAE1B;IAEA,OAAO;QACL,MAAMA,kBAAkB;QACxB,aAAaA,kBAAkB;QAC/B,OAAOA,kBAAkB;QACzB,OAAOD,OAAO,OAAO,CAAC,KAAK;QAC3B,aAAaX,KAAK,SAAS,CAACW,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAeY,qBAAwBxF,OAO7C;IACC,MAAM,EAAEyF,SAAS,EAAExF,OAAO,EAAEyF,aAAa,EAAE5E,gBAAgB,EAAEO,WAAW,EAAE,GACxErB;IACF,MAAM2B,eAAegE;IACrB,MAAMrE,mBAAmBrB,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM2F,wBAAwBC,uBAC5B7F,QAAQ,eAAe,IAAI,IAC3ByF;IAGF,MAAMK,cAAyD,EAAE;IAEjE,IAAIJ,eAAe,uBAAuB,OACxCI,YAAY,IAAI,CAAC;QACf,MAAM;QACN,WAAW;YACT,KAAKxE;YACL,QAAQ;QACV;IACF;IAGFwE,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAM7E,OAAe;QACnB;YAAE,MAAM;YAAU,SAASY;QAAa;QACxC;YACE,MAAM;YACN,SAASmE;QACX;KACD;IAED,IAAIhF,kBAAkB;QACpB,MAAMwB,SAAS,MAAMzB,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAC,KAAK,IAAI,IAAIuB;IACf;IAEA,MAAM,EACJ,SAASwB,WAAW,EACpBtB,KAAK,EACLuD,iBAAiB,EAClB,GAAG,MAAMC,OAAOjF,MAAMM;IAGvB,IAAI4E;IACJ,IAAI;QACFA,cAAcC,2BAA8BpC;IAC9C,EAAE,OAAOqC,YAAY;QAEnB,MAAMxC,eACJwC,sBAAsBvC,QAAQuC,WAAW,OAAO,GAAGtC,OAAOsC;QAC5D,MAAM,IAAIpC,qBACR,CAAC,iBAAiB,EAAEJ,cAAc,EAClCG,aACAtB;IAEJ;IAEA,OAAO;QACLyD;QACAnC;QACAtB;QACAuD;IACF;AACF;AAEO,eAAeK,sBACpBC,WAAmB,EACnBC,QAAwE,EACxEjF,WAAyB;IAKzB,MAAMM,eAAe4E;IACrB,MAAMC,aAAaC,0BAA0BJ;IAE7C,MAAMtF,OAAe;QACnB;YAAE,MAAM;YAAU,SAASY;QAAa;QACxC;YACE,MAAM;YACN,SAAS6E;QACX;KACD;IAED5G,aACE,0DACAyG;IAGF,MAAMzB,SAAS,MAAM0B,SAASvF,MAAMM,aAAa;QAC/C,WAAW;IACb;IAEA,OAAO;QACL,kBAAkBuD,OAAO,OAAO,CAAC,gBAAgB,IAAI;QACrD,OAAOA,OAAO,KAAK;IACrB;AACF"}
+{"version":3,"file":"ai-model/inspect.mjs","sources":["../../../src/ai-model/inspect.ts"],"sourcesContent":["import type {\n  AIDataExtractionResponse,\n  AIElementResponse,\n  AISectionLocatorResponse,\n  AIUsageInfo,\n  Rect,\n  ServiceExtractOption,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport {\n  generateElementByPoint,\n  generateElementByRect,\n} from '@midscene/shared/extractor/dom-util';\nimport {\n  cropByRect,\n  paddingToMatchBlockByBase64,\n  preProcessImageUrl,\n  scaleImage,\n} from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport type { LocateResultElement } from '@midscene/shared/types';\nimport { assert } from '@midscene/shared/utils';\nimport type {\n  ChatCompletionSystemMessageParam,\n  ChatCompletionUserMessageParam,\n} from 'openai/resources/index';\nimport type { TMultimodalPrompt, TUserPrompt } from '../common';\nimport { adaptBboxToRect, expandSearchArea, mergeRects } from '../common';\nimport { parseAutoGLMLocateResponse } from './auto-glm/parser';\nimport { getAutoGLMLocatePrompt } from './auto-glm/prompt';\nimport { isAutoGLM } from './auto-glm/util';\nimport {\n  extractDataQueryPrompt,\n  parseXMLExtractionResponse,\n  systemPromptToExtract,\n} from './prompt/extraction';\nimport {\n  findElementPrompt,\n  systemPromptToLocateElement,\n} from './prompt/llm-locator';\nimport {\n  sectionLocatorInstruction,\n  systemPromptToLocateSection,\n} from './prompt/llm-section-locator';\nimport {\n  orderSensitiveJudgePrompt,\n  systemPromptToJudgeOrderSensitive,\n} from './prompt/order-sensitive-judge';\nimport {\n  AIResponseParseError,\n  callAI,\n  callAIWithObjectResponse,\n  callAIWithStringResponse,\n} from './service-caller/index';\n\nexport type AIArgs = [\n  ChatCompletionSystemMessageParam,\n  ...ChatCompletionUserMessageParam[],\n];\n\nconst debugInspect = getDebug('ai:inspect');\nconst debugSection = getDebug('ai:section');\n\nexport async function buildSearchAreaConfig(options: {\n  context: UIContext;\n  baseRect: Rect;\n  modelFamily: IModelConfig['modelFamily'];\n}): Promise<{ rect: Rect; imageBase64: string; scale: number }> {\n  const { context, baseRect, modelFamily } = options;\n  const scaleRatio = 2;\n  const sectionRect = expandSearchArea(baseRect, context.shotSize);\n\n  const croppedResult = await cropByRect(\n    context.screenshot.base64,\n    sectionRect,\n    modelFamily === 'qwen2.5-vl',\n  );\n\n  const scaledResult = await scaleImage(croppedResult.imageBase64, scaleRatio);\n  sectionRect.width = scaledResult.width;\n  sectionRect.height = scaledResult.height;\n  return {\n    rect: sectionRect,\n    imageBase64: scaledResult.imageBase64,\n    scale: scaleRatio,\n  };\n}\n\nconst extraTextFromUserPrompt = (prompt: TUserPrompt): string => {\n  if (typeof prompt === 'string') {\n    return prompt;\n  } else {\n    return prompt.prompt;\n  }\n};\n\nconst promptsToChatParam = async (\n  multimodalPrompt: TMultimodalPrompt,\n): Promise<ChatCompletionUserMessageParam[]> => {\n  const msgs: ChatCompletionUserMessageParam[] = [];\n  if (multimodalPrompt?.images?.length) {\n    msgs.push({\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: 'Next, I will provide all the reference images. These reference images are supporting context only, not the current screenshot being evaluated, unless the task explicitly asks for comparison or matching.',\n        },\n      ],\n    });\n\n    for (const item of multimodalPrompt.images) {\n      const base64 = await preProcessImageUrl(\n        item.url,\n        !!multimodalPrompt.convertHttpImage2Base64,\n      );\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'text',\n            text: `this is the reference image named '${item.name}'. It is a reference image, not the current screenshot:`,\n          },\n        ],\n      });\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: base64,\n              detail: 'high',\n            },\n          },\n        ],\n      });\n    }\n  }\n  return msgs;\n};\n\nexport async function AiLocateElement(options: {\n  context: UIContext;\n  targetElementDescription: TUserPrompt;\n  searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;\n  modelConfig: IModelConfig;\n  abortSignal?: AbortSignal;\n}): Promise<{\n  parseResult: {\n    elements: LocateResultElement[];\n    errors?: string[];\n  };\n  rect?: Rect;\n  rawResponse: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}> {\n  const { context, targetElementDescription, modelConfig } = options;\n  const { modelFamily } = modelConfig;\n  const screenshotBase64 = context.screenshot.base64;\n\n  assert(\n    targetElementDescription,\n    'cannot find the target element description',\n  );\n  const targetElementDescriptionText = extraTextFromUserPrompt(\n    targetElementDescription,\n  );\n  const userInstructionPrompt = findElementPrompt(targetElementDescriptionText);\n  const systemPrompt = isAutoGLM(modelFamily)\n    ? getAutoGLMLocatePrompt(modelFamily)\n    : systemPromptToLocateElement(modelFamily);\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = context.shotSize.width;\n  let imageHeight = context.shotSize.height;\n  let originalImageWidth = imageWidth;\n  let originalImageHeight = imageHeight;\n\n  if (options.searchConfig) {\n    assert(\n      options.searchConfig.rect,\n      'searchArea is provided but its rect cannot be found. Failed to locate element',\n    );\n    assert(\n      options.searchConfig.imageBase64,\n      'searchArea is provided but its imageBase64 cannot be found. Failed to locate element',\n    );\n\n    imagePayload = options.searchConfig.imageBase64;\n    imageWidth = options.searchConfig.rect?.width;\n    imageHeight = options.searchConfig.rect?.height;\n    originalImageWidth = imageWidth;\n    originalImageHeight = imageHeight;\n  } else if (modelFamily === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  }\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: isAutoGLM(modelFamily)\n            ? `Tap: ${userInstructionPrompt}`\n            : userInstructionPrompt,\n        },\n      ],\n    },\n  ];\n\n  if (typeof targetElementDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: targetElementDescription.images,\n      convertHttpImage2Base64: targetElementDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  if (isAutoGLM(modelFamily)) {\n    const { content: rawResponseContent, usage } =\n      await callAIWithStringResponse(msgs, modelConfig, {\n        abortSignal: options.abortSignal,\n      });\n\n    debugInspect('auto-glm rawResponse:', rawResponseContent);\n\n    const parsed = parseAutoGLMLocateResponse(rawResponseContent);\n\n    debugInspect('auto-glm thinking:', parsed.think);\n    debugInspect('auto-glm coordinates:', parsed.coordinates);\n\n    let resRect: Rect | undefined;\n    let matchedElements: LocateResultElement[] = [];\n    let errors: string[] = [];\n\n    if (parsed.error || !parsed.coordinates) {\n      errors = [parsed.error || 'Failed to parse auto-glm response'];\n      debugInspect('auto-glm parse error:', errors[0]);\n    } else {\n      const { x, y } = parsed.coordinates;\n\n      debugInspect('auto-glm coordinates [0-999]:', { x, y });\n\n      // Convert auto-glm coordinates [0,999] to pixel bbox\n      // Map from [0,999] to pixel coordinates\n      const pixelX = Math.round((x * imageWidth) / 1000);\n      const pixelY = Math.round((y * imageHeight) / 1000);\n\n      debugInspect('auto-glm pixel coordinates:', { pixelX, pixelY });\n\n      // Apply offset if searching in a cropped area\n      let finalX = pixelX;\n      let finalY = pixelY;\n      if (options.searchConfig?.rect) {\n        finalX += options.searchConfig.rect.left;\n        finalY += options.searchConfig.rect.top;\n      }\n\n      const element: LocateResultElement = generateElementByPoint(\n        [finalX, finalY],\n        targetElementDescriptionText as string,\n      );\n\n      resRect = element.rect;\n      debugInspect('auto-glm resRect:', resRect);\n\n      if (element) {\n        matchedElements = [element];\n      }\n    }\n\n    return {\n      rect: resRect,\n      parseResult: {\n        elements: matchedElements,\n        errors,\n      },\n      rawResponse: rawResponseContent,\n      usage,\n      reasoning_content: parsed.think,\n    };\n  }\n\n  let res: Awaited<\n    ReturnType<\n      typeof callAIWithObjectResponse<AIElementResponse | [number, number]>\n    >\n  >;\n  try {\n    res = await callAIWithObjectResponse<AIElementResponse | [number, number]>(\n      msgs,\n      modelConfig,\n      { abortSignal: options.abortSignal },\n    );\n  } catch (callError) {\n    // Return error with usage and rawResponse if available\n    const errorMessage =\n      callError instanceof Error ? callError.message : String(callError);\n    const rawResponse =\n      callError instanceof AIResponseParseError\n        ? callError.rawResponse\n        : errorMessage;\n    const usage =\n      callError instanceof AIResponseParseError ? callError.usage : undefined;\n    return {\n      rect: undefined,\n      parseResult: {\n        elements: [],\n        errors: [`AI call error: ${errorMessage}`],\n      },\n      rawResponse,\n      usage,\n      reasoning_content: undefined,\n    };\n  }\n\n  const rawResponse = JSON.stringify(res.content);\n\n  let resRect: Rect | undefined;\n  let matchedElements: LocateResultElement[] = [];\n  let errors: string[] | undefined =\n    'errors' in res.content ? res.content.errors : [];\n  try {\n    if (\n      'bbox' in res.content &&\n      Array.isArray(res.content.bbox) &&\n      res.content.bbox.length >= 1\n    ) {\n      resRect = adaptBboxToRect(\n        res.content.bbox,\n        imageWidth,\n        imageHeight,\n        options.searchConfig?.rect?.left,\n        options.searchConfig?.rect?.top,\n        originalImageWidth,\n        originalImageHeight,\n        modelFamily,\n        options.searchConfig?.scale,\n      );\n\n      debugInspect('resRect', resRect);\n\n      const element: LocateResultElement = generateElementByRect(\n        resRect,\n        targetElementDescriptionText as string,\n      );\n      errors = [];\n\n      if (element) {\n        matchedElements = [element];\n      }\n    }\n  } catch (e) {\n    const msg =\n      e instanceof Error\n        ? `Failed to parse bbox: ${e.message}`\n        : 'unknown error in locate';\n    if (!errors || errors?.length === 0) {\n      errors = [msg];\n    } else {\n      errors.push(`(${msg})`);\n    }\n  }\n\n  return {\n    rect: resRect,\n    parseResult: {\n      elements: matchedElements as LocateResultElement[],\n      errors: errors as string[],\n    },\n    rawResponse,\n    usage: res.usage,\n    reasoning_content: res.reasoning_content,\n  };\n}\n\nexport async function AiLocateSection(options: {\n  context: UIContext;\n  sectionDescription: TUserPrompt;\n  modelConfig: IModelConfig;\n  abortSignal?: AbortSignal;\n}): Promise<{\n  rect?: Rect;\n  imageBase64?: string;\n  scale?: number;\n  error?: string;\n  rawResponse: string;\n  usage?: AIUsageInfo;\n}> {\n  const { context, sectionDescription, modelConfig } = options;\n  const { modelFamily } = modelConfig;\n  const screenshotBase64 = context.screenshot.base64;\n\n  const systemPrompt = systemPromptToLocateSection(modelFamily);\n  const sectionLocatorInstructionText = sectionLocatorInstruction(\n    extraTextFromUserPrompt(sectionDescription),\n  );\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'image_url',\n          image_url: {\n            url: screenshotBase64,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: sectionLocatorInstructionText,\n        },\n      ],\n    },\n  ];\n\n  if (typeof sectionDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: sectionDescription.images,\n      convertHttpImage2Base64: sectionDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  let result: Awaited<\n    ReturnType<typeof callAIWithObjectResponse<AISectionLocatorResponse>>\n  >;\n  try {\n    result = await callAIWithObjectResponse<AISectionLocatorResponse>(\n      msgs,\n      modelConfig,\n      { abortSignal: options.abortSignal },\n    );\n  } catch (callError) {\n    // Return error with usage and rawResponse if available\n    const errorMessage =\n      callError instanceof Error ? callError.message : String(callError);\n    const rawResponse =\n      callError instanceof AIResponseParseError\n        ? callError.rawResponse\n        : errorMessage;\n    const usage =\n      callError instanceof AIResponseParseError ? callError.usage : undefined;\n    return {\n      rect: undefined,\n      imageBase64: undefined,\n      error: `AI call error: ${errorMessage}`,\n      rawResponse,\n      usage,\n    };\n  }\n\n  let searchAreaConfig:\n    | Awaited<ReturnType<typeof buildSearchAreaConfig>>\n    | undefined;\n  const sectionBbox = result.content.bbox;\n  if (sectionBbox) {\n    const targetRect = adaptBboxToRect(\n      sectionBbox,\n      context.shotSize.width,\n      context.shotSize.height,\n      0,\n      0,\n      context.shotSize.width,\n      context.shotSize.height,\n      modelFamily,\n    );\n    debugSection('original targetRect %j', targetRect);\n\n    const referenceBboxList = result.content.references_bbox || [];\n    debugSection('referenceBboxList %j', referenceBboxList);\n\n    const referenceRects = referenceBboxList\n      .filter((bbox) => Array.isArray(bbox))\n      .map((bbox) => {\n        return adaptBboxToRect(\n          bbox,\n          context.shotSize.width,\n          context.shotSize.height,\n          0,\n          0,\n          context.shotSize.width,\n          context.shotSize.height,\n          modelFamily,\n        );\n      });\n    debugSection('referenceRects %j', referenceRects);\n\n    // merge the sectionRect and referenceRects\n    const mergedRect = mergeRects([targetRect, ...referenceRects]);\n    debugSection('mergedRect %j', mergedRect);\n\n    const expandedRect = expandSearchArea(mergedRect, context.shotSize);\n    const originalWidth = expandedRect.width;\n    const originalHeight = expandedRect.height;\n    debugSection('expanded sectionRect %j', expandedRect);\n\n    searchAreaConfig = await buildSearchAreaConfig({\n      context,\n      baseRect: mergedRect,\n      modelFamily,\n    });\n\n    debugSection(\n      'scaled sectionRect from %dx%d to %dx%d (scale=%d)',\n      originalWidth,\n      originalHeight,\n      searchAreaConfig.rect.width,\n      searchAreaConfig.rect.height,\n      searchAreaConfig.scale,\n    );\n  }\n\n  return {\n    rect: searchAreaConfig?.rect,\n    imageBase64: searchAreaConfig?.imageBase64,\n    scale: searchAreaConfig?.scale,\n    error: result.content.error,\n    rawResponse: JSON.stringify(result.content),\n    usage: result.usage,\n  };\n}\n\nexport async function AiExtractElementInfo<T>(options: {\n  dataQuery: string | Record<string, string>;\n  multimodalPrompt?: TMultimodalPrompt;\n  context: UIContext;\n  pageDescription?: string;\n  extractOption?: ServiceExtractOption;\n  modelConfig: IModelConfig;\n}) {\n  const { dataQuery, context, extractOption, multimodalPrompt, modelConfig } =\n    options;\n  const systemPrompt = systemPromptToExtract();\n  const screenshotBase64 = context.screenshot.base64;\n\n  const extractDataPromptText = extractDataQueryPrompt(\n    options.pageDescription || '',\n    dataQuery,\n  );\n\n  const userContent: ChatCompletionUserMessageParam['content'] = [];\n\n  if (extractOption?.screenshotIncluded !== false) {\n    userContent.push({\n      type: 'text',\n      text: 'This is the current screenshot to evaluate. Unless <DATA_DEMAND> explicitly asks for comparison or matching against reference images, base your answer on this screenshot and its contents when provided.',\n    });\n\n    userContent.push({\n      type: 'image_url',\n      image_url: {\n        url: screenshotBase64,\n        detail: 'high',\n      },\n    });\n  }\n\n  userContent.push({\n    type: 'text',\n    text: extractDataPromptText,\n  });\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: userContent,\n    },\n  ];\n\n  if (multimodalPrompt) {\n    const addOns = await promptsToChatParam({\n      images: multimodalPrompt.images,\n      convertHttpImage2Base64: multimodalPrompt.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  const {\n    content: rawResponse,\n    usage,\n    reasoning_content,\n  } = await callAI(msgs, modelConfig);\n\n  // Parse XML response to JSON object\n  let parseResult: AIDataExtractionResponse<T>;\n  try {\n    parseResult = parseXMLExtractionResponse<T>(rawResponse);\n  } catch (parseError) {\n    // Throw AIResponseParseError with usage and rawResponse preserved\n    const errorMessage =\n      parseError instanceof Error ? parseError.message : String(parseError);\n    throw new AIResponseParseError(\n      `XML parse error: ${errorMessage}`,\n      rawResponse,\n      usage,\n    );\n  }\n\n  return {\n    parseResult,\n    rawResponse,\n    usage,\n    reasoning_content,\n  };\n}\n\nexport async function AiJudgeOrderSensitive(\n  description: string,\n  callAIFn: typeof callAIWithObjectResponse<{ isOrderSensitive: boolean }>,\n  modelConfig: IModelConfig,\n): Promise<{\n  isOrderSensitive: boolean;\n  usage?: AIUsageInfo;\n}> {\n  const systemPrompt = systemPromptToJudgeOrderSensitive();\n  const userPrompt = orderSensitiveJudgePrompt(description);\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: userPrompt,\n    },\n  ];\n\n  debugInspect(\n    'AiJudgeOrderSensitive: deepThink=false, description=%s',\n    description,\n  );\n\n  const result = await callAIFn(msgs, modelConfig, {\n    deepThink: false,\n  });\n\n  return {\n    isOrderSensitive: result.content.isOrderSensitive ?? false,\n    usage: result.usage,\n  };\n}\n"],"names":["debugInspect","getDebug","debugSection","buildSearchAreaConfig","options","context","baseRect","modelFamily","scaleRatio","sectionRect","expandSearchArea","croppedResult","cropByRect","scaledResult","scaleImage","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","msgs","item","base64","preProcessImageUrl","AiLocateElement","targetElementDescription","modelConfig","screenshotBase64","assert","targetElementDescriptionText","userInstructionPrompt","findElementPrompt","systemPrompt","isAutoGLM","getAutoGLMLocatePrompt","systemPromptToLocateElement","imagePayload","imageWidth","imageHeight","originalImageWidth","originalImageHeight","paddedResult","paddingToMatchBlockByBase64","addOns","rawResponseContent","usage","callAIWithStringResponse","parsed","parseAutoGLMLocateResponse","resRect","matchedElements","errors","x","y","pixelX","Math","pixelY","finalX","finalY","element","generateElementByPoint","res","callAIWithObjectResponse","callError","errorMessage","Error","String","rawResponse","AIResponseParseError","undefined","JSON","Array","adaptBboxToRect","generateElementByRect","e","msg","AiLocateSection","sectionDescription","systemPromptToLocateSection","sectionLocatorInstructionText","sectionLocatorInstruction","result","searchAreaConfig","sectionBbox","targetRect","referenceBboxList","referenceRects","bbox","mergedRect","mergeRects","expandedRect","originalWidth","originalHeight","AiExtractElementInfo","dataQuery","extractOption","systemPromptToExtract","extractDataPromptText","extractDataQueryPrompt","userContent","reasoning_content","callAI","parseResult","parseXMLExtractionResponse","parseError","AiJudgeOrderSensitive","description","callAIFn","systemPromptToJudgeOrderSensitive","userPrompt","orderSensitiveJudgePrompt"],"mappings":";;;;;;;;;;;;;AA6DA,MAAMA,eAAeC,SAAS;AAC9B,MAAMC,eAAeD,SAAS;AAEvB,eAAeE,sBAAsBC,OAI3C;IACC,MAAM,EAAEC,OAAO,EAAEC,QAAQ,EAAEC,WAAW,EAAE,GAAGH;IAC3C,MAAMI,aAAa;IACnB,MAAMC,cAAcC,iBAAiBJ,UAAUD,QAAQ,QAAQ;IAE/D,MAAMM,gBAAgB,MAAMC,WAC1BP,QAAQ,UAAU,CAAC,MAAM,EACzBI,aACAF,AAAgB,iBAAhBA;IAGF,MAAMM,eAAe,MAAMC,WAAWH,cAAc,WAAW,EAAEH;IACjEC,YAAY,KAAK,GAAGI,aAAa,KAAK;IACtCJ,YAAY,MAAM,GAAGI,aAAa,MAAM;IACxC,OAAO;QACL,MAAMJ;QACN,aAAaI,aAAa,WAAW;QACrC,OAAOL;IACT;AACF;AAEA,MAAMO,0BAA0B,CAACC;IAC/B,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAEP,OAAOA,OAAO,MAAM;AAExB;AAEA,MAAMC,qBAAqB,OACzBC;IAEA,MAAMC,OAAyC,EAAE;IACjD,IAAID,kBAAkB,QAAQ,QAAQ;QACpCC,KAAK,IAAI,CAAC;YACR,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM;gBACR;aACD;QACH;QAEA,KAAK,MAAMC,QAAQF,iBAAiB,MAAM,CAAE;YAC1C,MAAMG,SAAS,MAAMC,mBACnBF,KAAK,GAAG,EACR,CAAC,CAACF,iBAAiB,uBAAuB;YAG5CC,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,MAAM,CAAC,mCAAmC,EAAEC,KAAK,IAAI,CAAC,uDAAuD,CAAC;oBAChH;iBACD;YACH;YAEAD,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKE;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;QACF;IACF;IACA,OAAOF;AACT;AAEO,eAAeI,gBAAgBnB,OAMrC;IAUC,MAAM,EAAEC,OAAO,EAAEmB,wBAAwB,EAAEC,WAAW,EAAE,GAAGrB;IAC3D,MAAM,EAAEG,WAAW,EAAE,GAAGkB;IACxB,MAAMC,mBAAmBrB,QAAQ,UAAU,CAAC,MAAM;IAElDsB,OACEH,0BACA;IAEF,MAAMI,+BAA+Bb,wBACnCS;IAEF,MAAMK,wBAAwBC,kBAAkBF;IAChD,MAAMG,eAAeC,UAAUzB,eAC3B0B,uBAAuB1B,eACvB2B,4BAA4B3B;IAEhC,IAAI4B,eAAeT;IACnB,IAAIU,aAAa/B,QAAQ,QAAQ,CAAC,KAAK;IACvC,IAAIgC,cAAchC,QAAQ,QAAQ,CAAC,MAAM;IACzC,IAAIiC,qBAAqBF;IACzB,IAAIG,sBAAsBF;IAE1B,IAAIjC,QAAQ,YAAY,EAAE;QACxBuB,OACEvB,QAAQ,YAAY,CAAC,IAAI,EACzB;QAEFuB,OACEvB,QAAQ,YAAY,CAAC,WAAW,EAChC;QAGF+B,eAAe/B,QAAQ,YAAY,CAAC,WAAW;QAC/CgC,aAAahC,QAAQ,YAAY,CAAC,IAAI,EAAE;QACxCiC,cAAcjC,QAAQ,YAAY,CAAC,IAAI,EAAE;QACzCkC,qBAAqBF;QACrBG,sBAAsBF;IACxB,OAAO,IAAI9B,AAAgB,iBAAhBA,aAA8B;QACvC,MAAMiC,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAMrB,OAAe;QACnB;YAAE,MAAM;YAAU,SAASY;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKI;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMH,UAAUzB,eACZ,CAAC,KAAK,EAAEsB,uBAAuB,GAC/BA;gBACN;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAOL,0BAAuC;QAChD,MAAMkB,SAAS,MAAMzB,mBAAmB;YACtC,QAAQO,yBAAyB,MAAM;YACvC,yBAAyBA,yBAAyB,uBAAuB;QAC3E;QACAL,KAAK,IAAI,IAAIuB;IACf;IAEA,IAAIV,UAAUzB,cAAc;QAC1B,MAAM,EAAE,SAASoC,kBAAkB,EAAEC,KAAK,EAAE,GAC1C,MAAMC,yBAAyB1B,MAAMM,aAAa;YAChD,aAAarB,QAAQ,WAAW;QAClC;QAEFJ,aAAa,yBAAyB2C;QAEtC,MAAMG,SAASC,2BAA2BJ;QAE1C3C,aAAa,sBAAsB8C,OAAO,KAAK;QAC/C9C,aAAa,yBAAyB8C,OAAO,WAAW;QAExD,IAAIE;QACJ,IAAIC,kBAAyC,EAAE;QAC/C,IAAIC,SAAmB,EAAE;QAEzB,IAAIJ,OAAO,KAAK,IAAI,CAACA,OAAO,WAAW,EAAE;YACvCI,SAAS;gBAACJ,OAAO,KAAK,IAAI;aAAoC;YAC9D9C,aAAa,yBAAyBkD,MAAM,CAAC,EAAE;QACjD,OAAO;YACL,MAAM,EAAEC,CAAC,EAAEC,CAAC,EAAE,GAAGN,OAAO,WAAW;YAEnC9C,aAAa,iCAAiC;gBAAEmD;gBAAGC;YAAE;YAIrD,MAAMC,SAASC,KAAK,KAAK,CAAEH,IAAIf,aAAc;YAC7C,MAAMmB,SAASD,KAAK,KAAK,CAAEF,IAAIf,cAAe;YAE9CrC,aAAa,+BAA+B;gBAAEqD;gBAAQE;YAAO;YAG7D,IAAIC,SAASH;YACb,IAAII,SAASF;YACb,IAAInD,QAAQ,YAAY,EAAE,MAAM;gBAC9BoD,UAAUpD,QAAQ,YAAY,CAAC,IAAI,CAAC,IAAI;gBACxCqD,UAAUrD,QAAQ,YAAY,CAAC,IAAI,CAAC,GAAG;YACzC;YAEA,MAAMsD,UAA+BC,uBACnC;gBAACH;gBAAQC;aAAO,EAChB7B;YAGFoB,UAAUU,QAAQ,IAAI;YACtB1D,aAAa,qBAAqBgD;YAElC,IAAIU,SACFT,kBAAkB;gBAACS;aAAQ;QAE/B;QAEA,OAAO;YACL,MAAMV;YACN,aAAa;gBACX,UAAUC;gBACVC;YACF;YACA,aAAaP;YACbC;YACA,mBAAmBE,OAAO,KAAK;QACjC;IACF;IAEA,IAAIc;IAKJ,IAAI;QACFA,MAAM,MAAMC,yBACV1C,MACAM,aACA;YAAE,aAAarB,QAAQ,WAAW;QAAC;IAEvC,EAAE,OAAO0D,WAAW;QAElB,MAAMC,eACJD,qBAAqBE,QAAQF,UAAU,OAAO,GAAGG,OAAOH;QAC1D,MAAMI,cACJJ,qBAAqBK,uBACjBL,UAAU,WAAW,GACrBC;QACN,MAAMnB,QACJkB,qBAAqBK,uBAAuBL,UAAU,KAAK,GAAGM;QAChE,OAAO;YACL,MAAMA;YACN,aAAa;gBACX,UAAU,EAAE;gBACZ,QAAQ;oBAAC,CAAC,eAAe,EAAEL,cAAc;iBAAC;YAC5C;YACAG;YACAtB;YACA,mBAAmBwB;QACrB;IACF;IAEA,MAAMF,cAAcG,KAAK,SAAS,CAACT,IAAI,OAAO;IAE9C,IAAIZ;IACJ,IAAIC,kBAAyC,EAAE;IAC/C,IAAIC,SACF,YAAYU,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,MAAM,GAAG,EAAE;IACnD,IAAI;QACF,IACE,UAAUA,IAAI,OAAO,IACrBU,MAAM,OAAO,CAACV,IAAI,OAAO,CAAC,IAAI,KAC9BA,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,IAAI,GAC3B;YACAZ,UAAUuB,gBACRX,IAAI,OAAO,CAAC,IAAI,EAChBxB,YACAC,aACAjC,QAAQ,YAAY,EAAE,MAAM,MAC5BA,QAAQ,YAAY,EAAE,MAAM,KAC5BkC,oBACAC,qBACAhC,aACAH,QAAQ,YAAY,EAAE;YAGxBJ,aAAa,WAAWgD;YAExB,MAAMU,UAA+Bc,sBACnCxB,SACApB;YAEFsB,SAAS,EAAE;YAEX,IAAIQ,SACFT,kBAAkB;gBAACS;aAAQ;QAE/B;IACF,EAAE,OAAOe,GAAG;QACV,MAAMC,MACJD,aAAaT,QACT,CAAC,sBAAsB,EAAES,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACvB,UAAUA,QAAQ,WAAW,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAEwB,IAAI,CAAC,CAAC;aAFtBxB,SAAS;YAACwB;SAAI;IAIlB;IAEA,OAAO;QACL,MAAM1B;QACN,aAAa;YACX,UAAUC;YACV,QAAQC;QACV;QACAgB;QACA,OAAON,IAAI,KAAK;QAChB,mBAAmBA,IAAI,iBAAiB;IAC1C;AACF;AAEO,eAAee,gBAAgBvE,OAKrC;IAQC,MAAM,EAAEC,OAAO,EAAEuE,kBAAkB,EAAEnD,WAAW,EAAE,GAAGrB;IACrD,MAAM,EAAEG,WAAW,EAAE,GAAGkB;IACxB,MAAMC,mBAAmBrB,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM0B,eAAe8C,4BAA4BtE;IACjD,MAAMuE,gCAAgCC,0BACpChE,wBAAwB6D;IAE1B,MAAMzD,OAAe;QACnB;YAAE,MAAM;YAAU,SAASY;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKL;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMoD;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAMlC,SAAS,MAAMzB,mBAAmB;YACtC,QAAQ2D,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACAzD,KAAK,IAAI,IAAIuB;IACf;IAEA,IAAIsC;IAGJ,IAAI;QACFA,SAAS,MAAMnB,yBACb1C,MACAM,aACA;YAAE,aAAarB,QAAQ,WAAW;QAAC;IAEvC,EAAE,OAAO0D,WAAW;QAElB,MAAMC,eACJD,qBAAqBE,QAAQF,UAAU,OAAO,GAAGG,OAAOH;QAC1D,MAAMI,cACJJ,qBAAqBK,uBACjBL,UAAU,WAAW,GACrBC;QACN,MAAMnB,QACJkB,qBAAqBK,uBAAuBL,UAAU,KAAK,GAAGM;QAChE,OAAO;YACL,MAAMA;YACN,aAAaA;YACb,OAAO,CAAC,eAAe,EAAEL,cAAc;YACvCG;YACAtB;QACF;IACF;IAEA,IAAIqC;IAGJ,MAAMC,cAAcF,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIE,aAAa;QACf,MAAMC,aAAaZ,gBACjBW,aACA7E,QAAQ,QAAQ,CAAC,KAAK,EACtBA,QAAQ,QAAQ,CAAC,MAAM,EACvB,GACA,GACAA,QAAQ,QAAQ,CAAC,KAAK,EACtBA,QAAQ,QAAQ,CAAC,MAAM,EACvBE;QAEFL,aAAa,0BAA0BiF;QAEvC,MAAMC,oBAAoBJ,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9D9E,aAAa,wBAAwBkF;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAAShB,MAAM,OAAO,CAACgB,OAC/B,GAAG,CAAC,CAACA,OACGf,gBACLe,MACAjF,QAAQ,QAAQ,CAAC,KAAK,EACtBA,QAAQ,QAAQ,CAAC,MAAM,EACvB,GACA,GACAA,QAAQ,QAAQ,CAAC,KAAK,EACtBA,QAAQ,QAAQ,CAAC,MAAM,EACvBE;QAGNL,aAAa,qBAAqBmF;QAGlC,MAAME,aAAaC,WAAW;YAACL;eAAeE;SAAe;QAC7DnF,aAAa,iBAAiBqF;QAE9B,MAAME,eAAe/E,iBAAiB6E,YAAYlF,QAAQ,QAAQ;QAClE,MAAMqF,gBAAgBD,aAAa,KAAK;QACxC,MAAME,iBAAiBF,aAAa,MAAM;QAC1CvF,aAAa,2BAA2BuF;QAExCR,mBAAmB,MAAM9E,sBAAsB;YAC7CE;YACA,UAAUkF;YACVhF;QACF;QAEAL,aACE,qDACAwF,eACAC,gBACAV,iBAAiB,IAAI,CAAC,KAAK,EAC3BA,iBAAiB,IAAI,CAAC,MAAM,EAC5BA,iBAAiB,KAAK;IAE1B;IAEA,OAAO;QACL,MAAMA,kBAAkB;QACxB,aAAaA,kBAAkB;QAC/B,OAAOA,kBAAkB;QACzB,OAAOD,OAAO,OAAO,CAAC,KAAK;QAC3B,aAAaX,KAAK,SAAS,CAACW,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAeY,qBAAwBxF,OAO7C;IACC,MAAM,EAAEyF,SAAS,EAAExF,OAAO,EAAEyF,aAAa,EAAE5E,gBAAgB,EAAEO,WAAW,EAAE,GACxErB;IACF,MAAM2B,eAAegE;IACrB,MAAMrE,mBAAmBrB,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM2F,wBAAwBC,uBAC5B7F,QAAQ,eAAe,IAAI,IAC3ByF;IAGF,MAAMK,cAAyD,EAAE;IAEjE,IAAIJ,eAAe,uBAAuB,OAAO;QAC/CI,YAAY,IAAI,CAAC;YACf,MAAM;YACN,MAAM;QACR;QAEAA,YAAY,IAAI,CAAC;YACf,MAAM;YACN,WAAW;gBACT,KAAKxE;gBACL,QAAQ;YACV;QACF;IACF;IAEAwE,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAM7E,OAAe;QACnB;YAAE,MAAM;YAAU,SAASY;QAAa;QACxC;YACE,MAAM;YACN,SAASmE;QACX;KACD;IAED,IAAIhF,kBAAkB;QACpB,MAAMwB,SAAS,MAAMzB,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAC,KAAK,IAAI,IAAIuB;IACf;IAEA,MAAM,EACJ,SAASwB,WAAW,EACpBtB,KAAK,EACLuD,iBAAiB,EAClB,GAAG,MAAMC,OAAOjF,MAAMM;IAGvB,IAAI4E;IACJ,IAAI;QACFA,cAAcC,2BAA8BpC;IAC9C,EAAE,OAAOqC,YAAY;QAEnB,MAAMxC,eACJwC,sBAAsBvC,QAAQuC,WAAW,OAAO,GAAGtC,OAAOsC;QAC5D,MAAM,IAAIpC,qBACR,CAAC,iBAAiB,EAAEJ,cAAc,EAClCG,aACAtB;IAEJ;IAEA,OAAO;QACLyD;QACAnC;QACAtB;QACAuD;IACF;AACF;AAEO,eAAeK,sBACpBC,WAAmB,EACnBC,QAAwE,EACxEjF,WAAyB;IAKzB,MAAMM,eAAe4E;IACrB,MAAMC,aAAaC,0BAA0BJ;IAE7C,MAAMtF,OAAe;QACnB;YAAE,MAAM;YAAU,SAASY;QAAa;QACxC;YACE,MAAM;YACN,SAAS6E;QACX;KACD;IAED5G,aACE,0DACAyG;IAGF,MAAMzB,SAAS,MAAM0B,SAASvF,MAAMM,aAAa;QAC/C,WAAW;IACb;IAEA,OAAO;QACL,kBAAkBuD,OAAO,OAAO,CAAC,gBAAgB,IAAI;QACrD,OAAOA,OAAO,KAAK;IACrB;AACF"}

package/dist/es/ai-model/prompt/extraction.mjs

@@ -32,14 +32,18 @@ function systemPromptToExtract() {
     return `
 You are a versatile professional in software UI design and testing. Your outstanding contributions will impact the user experience of billions of users.
 
-The user will give you a screenshot, the contents of it (optional), and some data requirements in <DATA_DEMAND>. You need to understand the user's requirements and extract the data satisfying the <DATA_DEMAND>.
+The user will give you a current screenshot to evaluate, the contents of it (optional), and some data requirements in <DATA_DEMAND>. You need to understand the user's requirements and extract the data satisfying the <DATA_DEMAND>.
+
+Base your answer on the current screenshot, and on the contents of it when provided. Treat the current screenshot and its contents as the primary source of truth for what is currently visible or true in the current state.
+
+If reference images are provided, use them only as supporting context unless <DATA_DEMAND> explicitly asks you to compare against them, match against them, or reason about them directly.
+
+Do not conclude that something exists in the current screenshot solely because it appears in a reference image. When the current screenshot or its contents conflict with a reference image, trust the current screenshot and its contents about the current state.
 
 If a key specifies a JSON data type (such as Number, String, Boolean, Object, Array), ensure the returned value strictly matches that data type.
 
 When DATA_DEMAND is a JSON object, the keys in your response must exactly match the keys in DATA_DEMAND. Do not rename, translate, or substitute any key.
 
-If the user provides multiple reference images, please carefully review the reference images with the screenshot and provide the correct answer for <DATA_DEMAND>.
-
 
 Return in the following XML format:
 <thought>the thinking process of the extraction, less than 300 words. Use ${preferredLanguage} in this field.</thought>

package/dist/es/ai-model/prompt/extraction.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/prompt/extraction.mjs","sources":["../../../../src/ai-model/prompt/extraction.ts"],"sourcesContent":["import type { AIDataExtractionResponse } from '@/types';\nimport { getPreferredLanguage } from '@midscene/shared/env';\nimport type { ResponseFormatJSONSchema } from 'openai/resources/index';\nimport { safeParseJson } from '../service-caller/index';\nimport { extractXMLTag } from './util';\n\n/**\n * Parse XML response from LLM and convert to AIDataExtractionResponse\n */\nexport function parseXMLExtractionResponse<T>(\n  xmlString: string,\n): AIDataExtractionResponse<T> {\n  const thought = extractXMLTag(xmlString, 'thought');\n  const dataJsonStr = extractXMLTag(xmlString, 'data-json');\n  const errorsStr = extractXMLTag(xmlString, 'errors');\n\n  // Parse data-json (required)\n  if (!dataJsonStr) {\n    throw new Error('Missing required field: data-json');\n  }\n\n  let data: T;\n  try {\n    data = safeParseJson(dataJsonStr, undefined) as T;\n  } catch (e) {\n    throw new Error(`Failed to parse data-json: ${e}`);\n  }\n\n  // Parse errors (optional)\n  let errors: string[] | undefined;\n  if (errorsStr) {\n    try {\n      const parsedErrors = safeParseJson(errorsStr, undefined);\n      if (Array.isArray(parsedErrors)) {\n        errors = parsedErrors;\n      }\n    } catch (e) {\n      // If errors parsing fails, just ignore it\n    }\n  }\n\n  return {\n    ...(thought ? { thought } : {}),\n    data,\n    ...(errors && errors.length > 0 ? { errors } : {}),\n  };\n}\n\nexport function systemPromptToExtract() {\n  const preferredLanguage = getPreferredLanguage();\n\n  return `\nYou are a versatile professional in software UI design and testing. Your outstanding contributions will impact the user experience of billions of users.\n\nThe user will give you a screenshot, the contents of it (optional), and some data requirements in <DATA_DEMAND>. You need to understand the user's requirements and extract the data satisfying the <DATA_DEMAND>.\n\nIf a key specifies a JSON data type (such as Number, String, Boolean, Object, Array), ensure the returned value strictly matches that data type.\n\nWhen DATA_DEMAND is a JSON object, the keys in your response must exactly match the keys in DATA_DEMAND. Do not rename, translate, or substitute any key.\n\nIf the user provides multiple reference images, please carefully review the reference images with the screenshot and provide the correct answer for <DATA_DEMAND>.\n\n\nReturn in the following XML format:\n<thought>the thinking process of the extraction, less than 300 words. Use ${preferredLanguage} in this field.</thought>\n<data-json>the extracted data as JSON. Make sure both the value and scheme meet the DATA_DEMAND. If you want to write some description in this field, use the same language as the DATA_DEMAND.</data-json>\n<errors>optional error messages as JSON array, e.g., [\"error1\", \"error2\"]</errors>\n\n# Example 1\nFor example, if the DATA_DEMAND is:\n\n<DATA_DEMAND>\n{\n  \"name\": \"name shows on the left panel, string\",\n  \"age\": \"age shows on the right panel, number\",\n  \"isAdmin\": \"if the user is admin, boolean\"\n}\n</DATA_DEMAND>\n\nBy viewing the screenshot and page contents, you can extract the following data:\n\n<thought>According to the screenshot, i can see ...</thought>\n<data-json>\n{\n  \"name\": \"John\",\n  \"age\": 30,\n  \"isAdmin\": true\n}\n</data-json>\n\n# Example 2\nIf the DATA_DEMAND is:\n\n<DATA_DEMAND>\nthe todo items list, string[]\n</DATA_DEMAND>\n\nBy viewing the screenshot and page contents, you can extract the following data:\n\n<thought>According to the screenshot, i can see ...</thought>\n<data-json>\n[\"todo 1\", \"todo 2\", \"todo 3\"]\n</data-json>\n\n# Example 3\nIf the DATA_DEMAND is:\n\n<DATA_DEMAND>\nthe page title, string\n</DATA_DEMAND>\n\nBy viewing the screenshot and page contents, you can extract the following data:\n\n<thought>According to the screenshot, i can see ...</thought>\n<data-json>\n\"todo list\"\n</data-json>\n\n# Example 4\nIf the DATA_DEMAND is:\n\n<DATA_DEMAND>\n{\n  \"StatementIsTruthy\": \"Boolean, is it currently the SMS page?\"\n}\n</DATA_DEMAND>\n\nBy viewing the screenshot and page contents, you can extract the following data:\n\n<thought>According to the screenshot, i can see ...</thought>\n<data-json>\n{ \"StatementIsTruthy\": true }\n</data-json>\n`;\n}\n\nexport const extractDataQueryPrompt = (\n  pageDescription: string,\n  dataQuery: string | Record<string, string>,\n) => {\n  let dataQueryText = '';\n  if (typeof dataQuery === 'string') {\n    dataQueryText = dataQuery;\n  } else {\n    dataQueryText = JSON.stringify(dataQuery, null, 2);\n  }\n\n  return `\n<PageDescription>\n${pageDescription}\n</PageDescription>\n\n<DATA_DEMAND>\n${dataQueryText}\n</DATA_DEMAND>\n  `;\n};\n"],"names":["parseXMLExtractionResponse","xmlString","thought","extractXMLTag","dataJsonStr","errorsStr","Error","data","safeParseJson","undefined","e","errors","parsedErrors","Array","systemPromptToExtract","preferredLanguage","getPreferredLanguage","extractDataQueryPrompt","pageDescription","dataQuery","dataQueryText","JSON"],"mappings":";;;AASO,SAASA,2BACdC,SAAiB;IAEjB,MAAMC,UAAUC,cAAcF,WAAW;IACzC,MAAMG,cAAcD,cAAcF,WAAW;IAC7C,MAAMI,YAAYF,cAAcF,WAAW;IAG3C,IAAI,CAACG,aACH,MAAM,IAAIE,MAAM;IAGlB,IAAIC;IACJ,IAAI;QACFA,OAAOC,cAAcJ,aAAaK;IACpC,EAAE,OAAOC,GAAG;QACV,MAAM,IAAIJ,MAAM,CAAC,2BAA2B,EAAEI,GAAG;IACnD;IAGA,IAAIC;IACJ,IAAIN,WACF,IAAI;QACF,MAAMO,eAAeJ,cAAcH,WAAWI;QAC9C,IAAII,MAAM,OAAO,CAACD,eAChBD,SAASC;IAEb,EAAE,OAAOF,GAAG,CAEZ;IAGF,OAAO;QACL,GAAIR,UAAU;YAAEA;QAAQ,IAAI,CAAC,CAAC;QAC9BK;QACA,GAAII,UAAUA,OAAO,MAAM,GAAG,IAAI;YAAEA;QAAO,IAAI,CAAC,CAAC;IACnD;AACF;AAEO,SAASG;IACd,MAAMC,oBAAoBC;IAE1B,OAAO,CAAC;;;;;;;;;;;;;0EAagE,EAAED,kBAAkB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAqE9F,CAAC;AACD;AAEO,MAAME,yBAAyB,CACpCC,iBACAC;IAEA,IAAIC,gBAAgB;IAElBA,gBADE,AAAqB,YAArB,OAAOD,YACOA,YAEAE,KAAK,SAAS,CAACF,WAAW,MAAM;IAGlD,OAAO,CAAC;;AAEV,EAAED,gBAAgB;;;;AAIlB,EAAEE,cAAc;;EAEd,CAAC;AACH"}
+{"version":3,"file":"ai-model/prompt/extraction.mjs","sources":["../../../../src/ai-model/prompt/extraction.ts"],"sourcesContent":["import type { AIDataExtractionResponse } from '@/types';\nimport { getPreferredLanguage } from '@midscene/shared/env';\nimport type { ResponseFormatJSONSchema } from 'openai/resources/index';\nimport { safeParseJson } from '../service-caller/index';\nimport { extractXMLTag } from './util';\n\n/**\n * Parse XML response from LLM and convert to AIDataExtractionResponse\n */\nexport function parseXMLExtractionResponse<T>(\n  xmlString: string,\n): AIDataExtractionResponse<T> {\n  const thought = extractXMLTag(xmlString, 'thought');\n  const dataJsonStr = extractXMLTag(xmlString, 'data-json');\n  const errorsStr = extractXMLTag(xmlString, 'errors');\n\n  // Parse data-json (required)\n  if (!dataJsonStr) {\n    throw new Error('Missing required field: data-json');\n  }\n\n  let data: T;\n  try {\n    data = safeParseJson(dataJsonStr, undefined) as T;\n  } catch (e) {\n    throw new Error(`Failed to parse data-json: ${e}`);\n  }\n\n  // Parse errors (optional)\n  let errors: string[] | undefined;\n  if (errorsStr) {\n    try {\n      const parsedErrors = safeParseJson(errorsStr, undefined);\n      if (Array.isArray(parsedErrors)) {\n        errors = parsedErrors;\n      }\n    } catch (e) {\n      // If errors parsing fails, just ignore it\n    }\n  }\n\n  return {\n    ...(thought ? { thought } : {}),\n    data,\n    ...(errors && errors.length > 0 ? { errors } : {}),\n  };\n}\n\nexport function systemPromptToExtract() {\n  const preferredLanguage = getPreferredLanguage();\n\n  return `\nYou are a versatile professional in software UI design and testing. Your outstanding contributions will impact the user experience of billions of users.\n\nThe user will give you a current screenshot to evaluate, the contents of it (optional), and some data requirements in <DATA_DEMAND>. You need to understand the user's requirements and extract the data satisfying the <DATA_DEMAND>.\n\nBase your answer on the current screenshot, and on the contents of it when provided. Treat the current screenshot and its contents as the primary source of truth for what is currently visible or true in the current state.\n\nIf reference images are provided, use them only as supporting context unless <DATA_DEMAND> explicitly asks you to compare against them, match against them, or reason about them directly.\n\nDo not conclude that something exists in the current screenshot solely because it appears in a reference image. When the current screenshot or its contents conflict with a reference image, trust the current screenshot and its contents about the current state.\n\nIf a key specifies a JSON data type (such as Number, String, Boolean, Object, Array), ensure the returned value strictly matches that data type.\n\nWhen DATA_DEMAND is a JSON object, the keys in your response must exactly match the keys in DATA_DEMAND. Do not rename, translate, or substitute any key.\n\n\nReturn in the following XML format:\n<thought>the thinking process of the extraction, less than 300 words. Use ${preferredLanguage} in this field.</thought>\n<data-json>the extracted data as JSON. Make sure both the value and scheme meet the DATA_DEMAND. If you want to write some description in this field, use the same language as the DATA_DEMAND.</data-json>\n<errors>optional error messages as JSON array, e.g., [\"error1\", \"error2\"]</errors>\n\n# Example 1\nFor example, if the DATA_DEMAND is:\n\n<DATA_DEMAND>\n{\n  \"name\": \"name shows on the left panel, string\",\n  \"age\": \"age shows on the right panel, number\",\n  \"isAdmin\": \"if the user is admin, boolean\"\n}\n</DATA_DEMAND>\n\nBy viewing the screenshot and page contents, you can extract the following data:\n\n<thought>According to the screenshot, i can see ...</thought>\n<data-json>\n{\n  \"name\": \"John\",\n  \"age\": 30,\n  \"isAdmin\": true\n}\n</data-json>\n\n# Example 2\nIf the DATA_DEMAND is:\n\n<DATA_DEMAND>\nthe todo items list, string[]\n</DATA_DEMAND>\n\nBy viewing the screenshot and page contents, you can extract the following data:\n\n<thought>According to the screenshot, i can see ...</thought>\n<data-json>\n[\"todo 1\", \"todo 2\", \"todo 3\"]\n</data-json>\n\n# Example 3\nIf the DATA_DEMAND is:\n\n<DATA_DEMAND>\nthe page title, string\n</DATA_DEMAND>\n\nBy viewing the screenshot and page contents, you can extract the following data:\n\n<thought>According to the screenshot, i can see ...</thought>\n<data-json>\n\"todo list\"\n</data-json>\n\n# Example 4\nIf the DATA_DEMAND is:\n\n<DATA_DEMAND>\n{\n  \"StatementIsTruthy\": \"Boolean, is it currently the SMS page?\"\n}\n</DATA_DEMAND>\n\nBy viewing the screenshot and page contents, you can extract the following data:\n\n<thought>According to the screenshot, i can see ...</thought>\n<data-json>\n{ \"StatementIsTruthy\": true }\n</data-json>\n`;\n}\n\nexport const extractDataQueryPrompt = (\n  pageDescription: string,\n  dataQuery: string | Record<string, string>,\n) => {\n  let dataQueryText = '';\n  if (typeof dataQuery === 'string') {\n    dataQueryText = dataQuery;\n  } else {\n    dataQueryText = JSON.stringify(dataQuery, null, 2);\n  }\n\n  return `\n<PageDescription>\n${pageDescription}\n</PageDescription>\n\n<DATA_DEMAND>\n${dataQueryText}\n</DATA_DEMAND>\n  `;\n};\n"],"names":["parseXMLExtractionResponse","xmlString","thought","extractXMLTag","dataJsonStr","errorsStr","Error","data","safeParseJson","undefined","e","errors","parsedErrors","Array","systemPromptToExtract","preferredLanguage","getPreferredLanguage","extractDataQueryPrompt","pageDescription","dataQuery","dataQueryText","JSON"],"mappings":";;;AASO,SAASA,2BACdC,SAAiB;IAEjB,MAAMC,UAAUC,cAAcF,WAAW;IACzC,MAAMG,cAAcD,cAAcF,WAAW;IAC7C,MAAMI,YAAYF,cAAcF,WAAW;IAG3C,IAAI,CAACG,aACH,MAAM,IAAIE,MAAM;IAGlB,IAAIC;IACJ,IAAI;QACFA,OAAOC,cAAcJ,aAAaK;IACpC,EAAE,OAAOC,GAAG;QACV,MAAM,IAAIJ,MAAM,CAAC,2BAA2B,EAAEI,GAAG;IACnD;IAGA,IAAIC;IACJ,IAAIN,WACF,IAAI;QACF,MAAMO,eAAeJ,cAAcH,WAAWI;QAC9C,IAAII,MAAM,OAAO,CAACD,eAChBD,SAASC;IAEb,EAAE,OAAOF,GAAG,CAEZ;IAGF,OAAO;QACL,GAAIR,UAAU;YAAEA;QAAQ,IAAI,CAAC,CAAC;QAC9BK;QACA,GAAII,UAAUA,OAAO,MAAM,GAAG,IAAI;YAAEA;QAAO,IAAI,CAAC,CAAC;IACnD;AACF;AAEO,SAASG;IACd,MAAMC,oBAAoBC;IAE1B,OAAO,CAAC;;;;;;;;;;;;;;;;;0EAiBgE,EAAED,kBAAkB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAqE9F,CAAC;AACD;AAEO,MAAME,yBAAyB,CACpCC,iBACAC;IAEA,IAAIC,gBAAgB;IAElBA,gBADE,AAAqB,YAArB,OAAOD,YACOA,YAEAE,KAAK,SAAS,CAACF,WAAW,MAAM;IAGlD,OAAO,CAAC;;AAEV,EAAED,gBAAgB;;;;AAIlB,EAAEE,cAAc;;EAEd,CAAC;AACH"}