@midscene/core 1.2.2-beta-20260120033218.0 → 1.2.3-beta-20260120082504.0

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

@@ -1 +1 @@
-{"version":3,"file":"agent/tasks.mjs","sources":["../../../src/agent/tasks.ts"],"sourcesContent":["import {\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 } 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 { 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  private conversationHistory: ConversationHistory;\n\n  onTaskStartCallback?: ExecutionTaskProgressOptions['onTaskStart'];\n\n  private readonly hooks?: TaskExecutorHooks;\n\n  replanningCycleLimit?: number;\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      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.hooks = opts.hooks;\n    this.conversationHistory = new ConversationHistory();\n    this.providedActionSpace = opts.actionSpace;\n    this.taskBuilder = new TaskBuilder({\n      interfaceInstance,\n      service,\n      taskCache: opts.taskCache,\n      actionSpace: this.getActionSpace(),\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  public async convertPlanToExecutable(\n    plans: PlanningAction[],\n    modelConfigForPlanning: IModelConfig,\n    modelConfigForDefaultIntent: IModelConfig,\n    options?: {\n      cacheable?: boolean;\n      subTask?: boolean;\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('Action', 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  ): 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      );\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  ): Promise<\n    ExecutionResult<\n      | {\n          yamlFlow?: MidsceneYamlFlowItem[]; // for cache use\n          output?: string;\n        }\n      | undefined\n    >\n  > {\n    this.conversationHistory.reset();\n\n    const session = this.createExecutionSession(\n      taskTitleStr('Action', 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      const result = await session.appendAndRun(\n        {\n          type: 'Planning',\n          subType: 'Plan',\n          param: {\n            userInstruction: userPrompt,\n            aiActContext,\n            imagesIncludeCount,\n            deepThink,\n          },\n          executor: async (param, executorContext) => {\n            const { uiContext } = executorContext;\n            assert(uiContext, 'uiContext is required for Planning task');\n            const { modelFamily } = modelConfigForPlanning;\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            const 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: this.conversationHistory,\n              includeBbox: includeBboxInPlanning,\n              imagesIncludeCount,\n              deepThink,\n            });\n            debug('planResult', JSON.stringify(planResult, null, 2));\n\n            const {\n              actions,\n              thought,\n              log,\n              note,\n              error,\n              usage,\n              rawResponse,\n              reasoning_content,\n            } = planResult;\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              note,\n              yamlFlow: planResult.yamlFlow,\n              output: outputString,\n              shouldContinuePlanning: planResult.shouldContinuePlanning,\n            };\n            executorContext.uiContext = uiContext;\n\n            assert(!error, `Failed to continue: ${error}\\n${log || ''}`);\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            subTask: true,\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 (this.conversationHistory.pendingFeedbackMessage) {\n        console.warn(\n          'unconsumed pending feedback message detected, this may lead to unexpected planning result:',\n          this.conversationHistory.pendingFeedbackMessage,\n        );\n      }\n      // todo: set time string\n\n      try {\n        const result = await session.appendAndRun(executables.tasks);\n        outputString = result?.output;\n      } catch (error: any) {\n        // errorFlag = true;\n        errorCountInOnePlanningLoop++;\n        this.conversationHistory.pendingFeedbackMessage = `Time: ${getReadableTimeString()}, 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 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 (!this.conversationHistory.pendingFeedbackMessage) {\n        this.conversationHistory.pendingFeedbackMessage = `Time: ${getReadableTimeString()}, 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        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          };\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: ${demand}`;\n          demandInput = {\n            [keyOfResult]: booleanPrompt,\n          };\n        } else if (ifTypeRestricted) {\n          demandInput = {\n            [keyOfResult]: `${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          );\n        } catch (error) {\n          if (error instanceof ServiceError) {\n            applyDump(error.dump);\n          }\n          throw error;\n        }\n\n        const { data, usage, thought, dump, reasoning_content } = extractResult;\n        applyDump(dump);\n        task.reasoning_content = reasoning_content;\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            assert(\n              data?.[keyOfResult] !== undefined,\n              'No result in query data',\n            );\n            outputResult = (data as any)[keyOfResult];\n          }\n        }\n\n        if (type === 'Assert' && !outputResult) {\n          task.usage = usage;\n          task.thought = thought;\n          throw new Error(`Assertion failed: ${thought}`);\n        }\n\n        return {\n          output: outputResult,\n          log: queryDump,\n          usage,\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 taskForSleep(timeMs: number, _modelConfig: IModelConfig) {\n    return this.taskBuilder.createSleepTask({\n      timeMs,\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 timeRemaining = checkIntervalMs - (now - currentCheckStart);\n        const sleepTask = this.taskBuilder.createSleepTask({\n          timeMs: timeRemaining,\n        });\n        await session.append(sleepTask);\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","plans","modelConfigForPlanning","modelConfigForDefaultIntent","userInstruction","yamlString","session","taskTitleStr","task","param","executorContext","uiContext","assert","runner","tasks","result","output","userPrompt","includeBboxInPlanning","aiActContext","cacheable","replanningCycleLimitOverride","imagesIncludeCount","deepThink","fileChooserAccept","withFileChooser","replanCount","yamlFlow","replanningCycleLimit","undefined","errorCountInOnePlanningLoop","outputString","modelFamily","actionSpace","action","Array","console","planImpl","isUITars","uiTarsPlanning","isAutoGLM","autoGLMPlanning","plan","planResult","JSON","actions","thought","log","note","error","usage","rawResponse","reasoning_content","executables","getReadableTimeString","String","Error","errorMsg","type","demand","modelConfig","opt","multimodalPrompt","queryTask","taskContext","queryDump","applyDump","dump","ifTypeRestricted","demandInput","keyOfResult","booleanPrompt","extractResult","extraPageDescription","tree","descriptionOfTree","ServiceError","data","outputResult","timeMs","_modelConfig","assertion","textPrompt","parsePrompt","description","timeoutMs","checkIntervalMs","domIncluded","screenshotIncluded","restOpt","serviceExtractOpt","overallStartTime","Date","lastCheckStart","errorThought","currentCheckStart","now","timeRemaining","sleepTask","interfaceInstance","service","opts","ConversationHistory","TaskBuilder","handler","chooser","dispose","getError"],"mappings":";;;;;;;;;;;;;;;;;;;;;;AAyDA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,wCAAwC;AAIvC,MAAMC;IAoBX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IA6BQ,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;IAEA,MAAa,wBACXG,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACzCL,OAGC,EACD;QACA,OAAO,IAAI,CAAC,WAAW,CAAC,KAAK,CAC3BG,OACAC,wBACAC,6BACAL;IAEJ;IAEA,MAAM,uBAAuBM,eAAuB,EAAEC,UAAkB,EAAE;QACxE,MAAMC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUH;QAGzB,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,SACJhB,KAAa,EACbI,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACf;QAC1B,MAAMG,UAAU,IAAI,CAAC,sBAAsB,CAACT;QAC5C,MAAM,EAAEiB,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,EAS5B;QACA,OAAOC,gBAAgB,IAAI,CAAC,SAAS,EAAED,mBAAmB,UACjD,IAAI,CAAC,SAAS,CACnBP,YACAf,wBACAC,6BACAe,uBACAC,cACAC,WACAC,8BACAC,oBACAC;IAGN;IAEA,MAAc,UACZN,UAAkB,EAClBf,sBAAoC,EACpCC,2BAAyC,EACzCe,qBAA8B,EAC9BC,YAAqB,EACrBC,SAAmB,EACnBC,4BAAqC,EACrCC,kBAA2B,EAC3BC,SAA2B,EAS3B;QACA,IAAI,CAAC,mBAAmB,CAAC,KAAK;QAE9B,MAAMjB,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUU;QAEzB,MAAMJ,SAASP,QAAQ,SAAS;QAEhC,IAAIoB,cAAc;QAClB,MAAMC,WAAmC,EAAE;QAC3C,MAAMC,uBACJP,gCAAgC,IAAI,CAAC,oBAAoB;QAC3DT,OACEgB,AAAyBC,WAAzBD,sBACA;QAGF,IAAIE,8BAA8B;QAClC,IAAIC;QAGJ,MAAO,KAAM;YACX,MAAMhB,SAAS,MAAMT,QAAQ,YAAY,CACvC;gBACE,MAAM;gBACN,SAAS;gBACT,OAAO;oBACL,iBAAiBW;oBACjBE;oBACAG;oBACAC;gBACF;gBACA,UAAU,OAAOd,OAAOC;oBACtB,MAAM,EAAEC,SAAS,EAAE,GAAGD;oBACtBE,OAAOD,WAAW;oBAClB,MAAM,EAAEqB,WAAW,EAAE,GAAG9B;oBAExB,MAAM+B,cAAc,IAAI,CAAC,cAAc;oBACvCxC,MACE,sCACAwC,YAAY,GAAG,CAAC,CAACC,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;oBAEhDtB,OAAOuB,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,SAASN,eACtBO,iBACAC,UAAUR,eACRS,kBACAC;oBAEN,MAAMC,aAAa,MAAMN,SAAS5B,MAAM,eAAe,EAAE;wBACvD,SAASE;wBACT,eAAeF,MAAM,YAAY;wBACjC,eAAe,IAAI,CAAC,SAAS,CAAC,aAAa;wBAC3CwB;wBACA,aAAa/B;wBACb,qBAAqB,IAAI,CAAC,mBAAmB;wBAC7C,aAAagB;wBACbI;wBACAC;oBACF;oBACA9B,MAAM,cAAcmD,KAAK,SAAS,CAACD,YAAY,MAAM;oBAErD,MAAM,EACJE,OAAO,EACPC,OAAO,EACPC,GAAG,EACHC,IAAI,EACJC,KAAK,EACLC,KAAK,EACLC,WAAW,EACXC,iBAAiB,EAClB,GAAGT;oBAEJjC,gBAAgB,IAAI,CAAC,GAAG,GAAG;wBACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;wBAClCyC;oBACF;oBACAzC,gBAAgB,IAAI,CAAC,KAAK,GAAGwC;oBAC7BxC,gBAAgB,IAAI,CAAC,iBAAiB,GAAG0C;oBACzC1C,gBAAgB,IAAI,CAAC,MAAM,GAAG;wBAC5B,SAASmC,WAAW,EAAE;wBACtBE;wBACAD;wBACAE;wBACA,UAAUL,WAAW,QAAQ;wBAC7B,QAAQZ;wBACR,wBAAwBY,WAAW,sBAAsB;oBAC3D;oBACAjC,gBAAgB,SAAS,GAAGC;oBAE5BC,OAAO,CAACqC,OAAO,CAAC,oBAAoB,EAAEA,MAAM,EAAE,EAAEF,OAAO,IAAI;oBAE3D,OAAO;wBACL,OAAO;4BACL,KAAK;wBACP;oBACF;gBACF;YACF,GACA;gBACE,gBAAgB;YAClB;YAGF,MAAMJ,aAAa5B,QAAQ;YAG3B,MAAMd,QAAQ0C,YAAY,WAAW,EAAE;YACvChB,SAAS,IAAI,IAAKgB,YAAY,YAAY,EAAE;YAE5C,IAAIU;YACJ,IAAI;gBACFA,cAAc,MAAM,IAAI,CAAC,uBAAuB,CAC9CpD,OACAC,wBACAC,6BACA;oBACEiB;oBACA,SAAS;gBACX;YAEJ,EAAE,OAAO6B,OAAO;gBACd,OAAO3C,QAAQ,eAAe,CAC5B,CAAC,4CAA4C,EAAE2C,MAAM,SAAS,EAAEL,KAAK,SAAS,CAC5E3C,QACC;YAEP;YACA,IAAI,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,EACjDmC,QAAQ,IAAI,CACV,8FACA,IAAI,CAAC,mBAAmB,CAAC,sBAAsB;YAKnD,IAAI;gBACF,MAAMrB,SAAS,MAAMT,QAAQ,YAAY,CAAC+C,YAAY,KAAK;gBAC3DtB,eAAehB,QAAQ;YACzB,EAAE,OAAOkC,OAAY;gBAEnBnB;gBACA,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,GAAG,CAAC,MAAM,EAAEwB,wBAAwB,iCAAiC,EAAEL,OAAO,WAAWM,OAAON,QAAQ;gBACvJxD,MACE,yFACAwD,iBAAiBO,QAAQP,MAAM,OAAO,GAAGM,OAAON,QAChD,6CACAnB;YAEJ;YAEA,IAAIA,8BAA8BnC,uCAChC,OAAOW,QAAQ,eAAe,CAAC;YAIjC,IAAI,CAACqC,YAAY,wBACf;YAIF,EAAEjB;YAEF,IAAIA,cAAcE,sBAAsB;gBACtC,MAAM6B,WAAW,CAAC,UAAU,EAAE7B,qBAAqB,4JAA4J,CAAC;gBAChN,OAAOtB,QAAQ,eAAe,CAACmD;YACjC;YAEA,IAAI,CAAC,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,EAClD,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,GAAG,CAAC,MAAM,EAAEH,wBAAwB,gDAAgD,CAAC;QAExI;QAEA,OAAO;YACL,QAAQ;gBACN3B;gBACA,QAAQI;YACV;YACAlB;QACF;IACF;IAEQ,oBACN6C,IAAsE,EACtEC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACpC;QACA,MAAMC,YAA4C;YAChD,MAAM;YACN,SAASL;YACT,OAAO;gBACL,YAAYI,mBACP;oBACCH;oBACAG;gBACF,IACAH;YACN;YACA,UAAU,OAAOlD,OAAOuD;gBACtB,MAAM,EAAExD,IAAI,EAAE,GAAGwD;gBACjB,IAAIC;gBACJ,MAAMC,YAAY,CAACC;oBACjBF,YAAYE;oBACZ3D,KAAK,GAAG,GAAG;wBACT2D;oBACF;gBACF;gBAGA,MAAMxD,YAAYqD,YAAY,SAAS;gBACvCpD,OAAOD,WAAW;gBAElB,MAAMyD,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,kBACTC,cAAc;oBACZ,CAACC,YAAY,EAAE,GAAGZ,KAAK,EAAE,EAAEC,QAAQ;gBACrC;gBAGF,IAAIa;gBAEJ,IAAIC,uBAAuB;gBAC3B,IAAIZ,KAAK,eAAe,IAAI,CAAC,SAAS,CAAC,mBAAmB,EAAE;oBAC1DpE,MAAM;oBACN,MAAMiF,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;gBAEJ,EAAE,OAAOb,OAAO;oBACd,IAAIA,iBAAiB2B,cACnBV,UAAUjB,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAEA,MAAM,EAAE4B,IAAI,EAAE3B,KAAK,EAAEJ,OAAO,EAAEqB,IAAI,EAAEf,iBAAiB,EAAE,GAAGoB;gBAC1DN,UAAUC;gBACV3D,KAAK,iBAAiB,GAAG4C;gBAEzB,IAAI0B,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;qBACV;oBACLlE,OACEiE,MAAM,CAACP,YAAY,KAAKzC,QACxB;oBAEFiD,eAAgBD,IAAY,CAACP,YAAY;gBAC3C;gBAGF,IAAIZ,AAAS,aAATA,QAAqB,CAACoB,cAAc;oBACtCtE,KAAK,KAAK,GAAG0C;oBACb1C,KAAK,OAAO,GAAGsC;oBACf,MAAM,IAAIU,MAAM,CAAC,kBAAkB,EAAEV,SAAS;gBAChD;gBAEA,OAAO;oBACL,QAAQgC;oBACR,KAAKb;oBACLf;oBACAJ;gBACF;YACF;QACF;QAEA,OAAOiB;IACT;IACA,MAAM,yBACJL,IAA0D,EAC1DC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACP;QAC7B,MAAMxD,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aACEmD,MACA,AAAkB,YAAlB,OAAOC,SAAsBA,SAASf,KAAK,SAAS,CAACe;QAIzD,MAAMI,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9CL,MACAC,QACAC,aACAC,KACAC;QAGF,MAAMjD,SAASP,QAAQ,SAAS;QAChC,MAAMS,SAAS,MAAMT,QAAQ,YAAY,CAACyD;QAE1C,IAAI,CAAChD,QACH,MAAM,IAAIyC,MACR;QAIJ,MAAM,EAAExC,MAAM,EAAE8B,OAAO,EAAE,GAAG/B;QAE5B,OAAO;YACLC;YACA8B;YACAjC;QACF;IACF;IAEA,MAAM,aAAakE,MAAc,EAAEC,YAA0B,EAAE;QAC7D,OAAO,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;YACtCD;QACF;IACF;IAEA,MAAM,QACJE,SAAsB,EACtBpB,GAA+B,EAC/BD,WAAyB,EACO;QAChC,MAAM,EAAEsB,UAAU,EAAEpB,gBAAgB,EAAE,GAAGqB,YAAYF;QAErD,MAAMG,cAAc,CAAC,SAAS,EAAEF,YAAY;QAC5C,MAAM5E,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,WAAW6E;QAE1B,MAAMvE,SAASP,QAAQ,SAAS;QAChC,MAAM,EACJ+E,SAAS,EACTC,eAAe,EACfC,WAAW,EACXC,kBAAkB,EAClB,GAAGC,SACJ,GAAG5B;QACJ,MAAM6B,oBAA0C;YAC9CH;YACAC;YACA,GAAGC,OAAO;QACZ;QAEA7E,OAAOqE,WAAW;QAClBrE,OAAOyE,WAAW;QAClBzE,OAAO0E,iBAAiB;QAExB1E,OACE0E,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,MAAMhC,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9C,WACAmB,YACAtB,aACA8B,mBACA5B;YAGF,MAAM/C,SAAU,MAAMT,QAAQ,YAAY,CAACyD;YAO3C,IAAIhD,QAAQ,QACV,OAAO;gBACL,QAAQc;gBACRhB;YACF;YAGFiF,eACE/E,QAAQ,WACP,CAACA,UAAU,CAAC,0BAA0B,EAAEmE,YAAY,IACrD,CAAC,0CAA0C,EAAEA,YAAY;YAC3D,MAAMc,MAAMJ,KAAK,GAAG;YACpB,IAAII,MAAMD,oBAAoBT,iBAAiB;gBAC7C,MAAMW,gBAAgBX,kBAAmBU,CAAAA,MAAMD,iBAAgB;gBAC/D,MAAMG,YAAY,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;oBACjD,QAAQD;gBACV;gBACA,MAAM3F,QAAQ,MAAM,CAAC4F;YACvB;QACF;QAEA,OAAO5F,QAAQ,eAAe,CAAC,CAAC,iBAAiB,EAAEwF,cAAc;IACnE;IApmBA,YACEK,iBAAoC,EACpCC,OAAgB,EAChBC,IAMC,CACD;QAjCF;QAEA;QAEA;QAEA,uBAAiB,uBAAjB;QAEA,uBAAiB,eAAjB;QAEA,uBAAQ,uBAAR;QAEA;QAEA,uBAAiB,SAAjB;QAEA;QAkBE,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,KAAK,GAAGA,KAAK,KAAK;QACvB,IAAI,CAAC,mBAAmB,GAAG,IAAIC;QAC/B,IAAI,CAAC,mBAAmB,GAAGD,KAAK,WAAW;QAC3C,IAAI,CAAC,WAAW,GAAG,IAAIE,YAAY;YACjCJ;YACAC;YACA,WAAWC,KAAK,SAAS;YACzB,aAAa,IAAI,CAAC,cAAc;QAClC;IACF;AA4kBF;AAEO,eAAe5E,gBACpB0E,iBAAoC,EACpC3E,iBAAuC,EACvCU,MAAwB;IAExB,IAAI,CAACV,mBAAmB,QACtB,OAAOU;IAGT,IAAI,CAACiE,kBAAkB,2BAA2B,EAChD,MAAM,IAAI3C,MACR,CAAC,gCAAgC,EAAE2C,kBAAkB,aAAa,EAAE;IAIxE,MAAMK,UAAU,OAAOC;QACrB,MAAMA,QAAQ,MAAM,CAACjF;IACvB;IAEA,MAAM,EAAEkF,OAAO,EAAEC,QAAQ,EAAE,GACzB,MAAMR,kBAAkB,2BAA2B,CAACK;IACtD,IAAI;QACF,MAAMzF,SAAS,MAAMmB;QAErB,MAAMe,QAAQ0D;QACd,IAAI1D,OACF,MAAMA;QAER,OAAOlC;IACT,SAAU;QACR2F;IACF;AACF"}
+{"version":3,"file":"agent/tasks.mjs","sources":["../../../src/agent/tasks.ts"],"sourcesContent":["import {\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 } 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 { 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  private conversationHistory: ConversationHistory;\n\n  onTaskStartCallback?: ExecutionTaskProgressOptions['onTaskStart'];\n\n  private readonly hooks?: TaskExecutorHooks;\n\n  replanningCycleLimit?: number;\n\n  waitAfterAction?: number;\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      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.hooks = opts.hooks;\n    this.conversationHistory = new ConversationHistory();\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  public async convertPlanToExecutable(\n    plans: PlanningAction[],\n    modelConfigForPlanning: IModelConfig,\n    modelConfigForDefaultIntent: IModelConfig,\n    options?: {\n      cacheable?: boolean;\n      subTask?: boolean;\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('Action', 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  ): 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      );\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  ): Promise<\n    ExecutionResult<\n      | {\n          yamlFlow?: MidsceneYamlFlowItem[]; // for cache use\n          output?: string;\n        }\n      | undefined\n    >\n  > {\n    this.conversationHistory.reset();\n\n    const session = this.createExecutionSession(\n      taskTitleStr('Action', 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      const result = await session.appendAndRun(\n        {\n          type: 'Planning',\n          subType: 'Plan',\n          param: {\n            userInstruction: userPrompt,\n            aiActContext,\n            imagesIncludeCount,\n            deepThink,\n          },\n          executor: async (param, executorContext) => {\n            const { uiContext } = executorContext;\n            assert(uiContext, 'uiContext is required for Planning task');\n            const { modelFamily } = modelConfigForPlanning;\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            const 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: this.conversationHistory,\n              includeBbox: includeBboxInPlanning,\n              imagesIncludeCount,\n              deepThink,\n            });\n            debug('planResult', JSON.stringify(planResult, null, 2));\n\n            const {\n              actions,\n              thought,\n              log,\n              note,\n              error,\n              usage,\n              rawResponse,\n              reasoning_content,\n              finalizeSuccess,\n              finalizeMessage,\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              note,\n              yamlFlow: planResult.yamlFlow,\n              output: finalizeMessage,\n              shouldContinuePlanning: planResult.shouldContinuePlanning,\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            subTask: true,\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 (this.conversationHistory.pendingFeedbackMessage) {\n        console.warn(\n          'unconsumed pending feedback message detected, this may lead to unexpected planning result:',\n          this.conversationHistory.pendingFeedbackMessage,\n        );\n      }\n      // todo: set time string\n\n      try {\n        await session.appendAndRun(executables.tasks);\n      } catch (error: any) {\n        // errorFlag = true;\n        errorCountInOnePlanningLoop++;\n        this.conversationHistory.pendingFeedbackMessage = `Time: ${getReadableTimeString()}, 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 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 (!this.conversationHistory.pendingFeedbackMessage) {\n        this.conversationHistory.pendingFeedbackMessage = `Time: ${getReadableTimeString()}, 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        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          };\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: ${demand}`;\n          demandInput = {\n            [keyOfResult]: booleanPrompt,\n          };\n        } else if (ifTypeRestricted) {\n          demandInput = {\n            [keyOfResult]: `${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          );\n        } catch (error) {\n          if (error instanceof ServiceError) {\n            applyDump(error.dump);\n          }\n          throw error;\n        }\n\n        const { data, usage, thought, dump, reasoning_content } = extractResult;\n        applyDump(dump);\n        task.reasoning_content = reasoning_content;\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            assert(\n              data?.[keyOfResult] !== undefined,\n              'No result in query data',\n            );\n            outputResult = (data as any)[keyOfResult];\n          }\n        }\n\n        if (type === 'Assert' && !outputResult) {\n          task.usage = usage;\n          task.thought = thought;\n          throw new Error(`Assertion failed: ${thought}`);\n        }\n\n        return {\n          output: outputResult,\n          log: queryDump,\n          usage,\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 taskForSleep(timeMs: number, _modelConfig: IModelConfig) {\n    return this.taskBuilder.createSleepTask({\n      timeMs,\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 timeRemaining = checkIntervalMs - (now - currentCheckStart);\n        const sleepTask = this.taskBuilder.createSleepTask({\n          timeMs: timeRemaining,\n        });\n        await session.append(sleepTask);\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","plans","modelConfigForPlanning","modelConfigForDefaultIntent","userInstruction","yamlString","session","taskTitleStr","task","param","executorContext","uiContext","assert","runner","tasks","result","output","userPrompt","includeBboxInPlanning","aiActContext","cacheable","replanningCycleLimitOverride","imagesIncludeCount","deepThink","fileChooserAccept","withFileChooser","replanCount","yamlFlow","replanningCycleLimit","undefined","errorCountInOnePlanningLoop","outputString","modelFamily","actionSpace","action","Array","console","planImpl","isUITars","uiTarsPlanning","isAutoGLM","autoGLMPlanning","plan","planResult","JSON","actions","thought","log","note","error","usage","rawResponse","reasoning_content","finalizeSuccess","finalizeMessage","executables","getReadableTimeString","String","Error","errorMsg","type","demand","modelConfig","opt","multimodalPrompt","queryTask","taskContext","queryDump","applyDump","dump","ifTypeRestricted","demandInput","keyOfResult","booleanPrompt","extractResult","extraPageDescription","tree","descriptionOfTree","ServiceError","data","outputResult","timeMs","_modelConfig","assertion","textPrompt","parsePrompt","description","timeoutMs","checkIntervalMs","domIncluded","screenshotIncluded","restOpt","serviceExtractOpt","overallStartTime","Date","lastCheckStart","errorThought","currentCheckStart","now","timeRemaining","sleepTask","interfaceInstance","service","opts","ConversationHistory","TaskBuilder","handler","chooser","dispose","getError"],"mappings":";;;;;;;;;;;;;;;;;;;;;;AAyDA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,wCAAwC;AAIvC,MAAMC;IAsBX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IAgCQ,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;IAEA,MAAa,wBACXG,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACzCL,OAGC,EACD;QACA,OAAO,IAAI,CAAC,WAAW,CAAC,KAAK,CAC3BG,OACAC,wBACAC,6BACAL;IAEJ;IAEA,MAAM,uBAAuBM,eAAuB,EAAEC,UAAkB,EAAE;QACxE,MAAMC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUH;QAGzB,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,SACJhB,KAAa,EACbI,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACf;QAC1B,MAAMG,UAAU,IAAI,CAAC,sBAAsB,CAACT;QAC5C,MAAM,EAAEiB,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,EAS5B;QACA,OAAOC,gBAAgB,IAAI,CAAC,SAAS,EAAED,mBAAmB,UACjD,IAAI,CAAC,SAAS,CACnBP,YACAf,wBACAC,6BACAe,uBACAC,cACAC,WACAC,8BACAC,oBACAC;IAGN;IAEA,MAAc,UACZN,UAAkB,EAClBf,sBAAoC,EACpCC,2BAAyC,EACzCe,qBAA8B,EAC9BC,YAAqB,EACrBC,SAAmB,EACnBC,4BAAqC,EACrCC,kBAA2B,EAC3BC,SAA2B,EAS3B;QACA,IAAI,CAAC,mBAAmB,CAAC,KAAK;QAE9B,MAAMjB,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUU;QAEzB,MAAMJ,SAASP,QAAQ,SAAS;QAEhC,IAAIoB,cAAc;QAClB,MAAMC,WAAmC,EAAE;QAC3C,MAAMC,uBACJP,gCAAgC,IAAI,CAAC,oBAAoB;QAC3DT,OACEgB,AAAyBC,WAAzBD,sBACA;QAGF,IAAIE,8BAA8B;QAClC,IAAIC;QAGJ,MAAO,KAAM;YACX,MAAMhB,SAAS,MAAMT,QAAQ,YAAY,CACvC;gBACE,MAAM;gBACN,SAAS;gBACT,OAAO;oBACL,iBAAiBW;oBACjBE;oBACAG;oBACAC;gBACF;gBACA,UAAU,OAAOd,OAAOC;oBACtB,MAAM,EAAEC,SAAS,EAAE,GAAGD;oBACtBE,OAAOD,WAAW;oBAClB,MAAM,EAAEqB,WAAW,EAAE,GAAG9B;oBAExB,MAAM+B,cAAc,IAAI,CAAC,cAAc;oBACvCxC,MACE,sCACAwC,YAAY,GAAG,CAAC,CAACC,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;oBAEhDtB,OAAOuB,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,SAASN,eACtBO,iBACAC,UAAUR,eACRS,kBACAC;oBAEN,MAAMC,aAAa,MAAMN,SAAS5B,MAAM,eAAe,EAAE;wBACvD,SAASE;wBACT,eAAeF,MAAM,YAAY;wBACjC,eAAe,IAAI,CAAC,SAAS,CAAC,aAAa;wBAC3CwB;wBACA,aAAa/B;wBACb,qBAAqB,IAAI,CAAC,mBAAmB;wBAC7C,aAAagB;wBACbI;wBACAC;oBACF;oBACA9B,MAAM,cAAcmD,KAAK,SAAS,CAACD,YAAY,MAAM;oBAErD,MAAM,EACJE,OAAO,EACPC,OAAO,EACPC,GAAG,EACHC,IAAI,EACJC,KAAK,EACLC,KAAK,EACLC,WAAW,EACXC,iBAAiB,EACjBC,eAAe,EACfC,eAAe,EAChB,GAAGX;oBACJZ,eAAeuB;oBAEf5C,gBAAgB,IAAI,CAAC,GAAG,GAAG;wBACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;wBAClCyC;oBACF;oBACAzC,gBAAgB,IAAI,CAAC,KAAK,GAAGwC;oBAC7BxC,gBAAgB,IAAI,CAAC,iBAAiB,GAAG0C;oBACzC1C,gBAAgB,IAAI,CAAC,MAAM,GAAG;wBAC5B,SAASmC,WAAW,EAAE;wBACtBE;wBACAD;wBACAE;wBACA,UAAUL,WAAW,QAAQ;wBAC7B,QAAQW;wBACR,wBAAwBX,WAAW,sBAAsB;oBAC3D;oBACAjC,gBAAgB,SAAS,GAAGC;oBAE5BC,OAAO,CAACqC,OAAO,CAAC,oBAAoB,EAAEA,MAAM,EAAE,EAAEF,OAAO,IAAI;oBAG3D,IAAIM,AAAoB,UAApBA,iBACFzC,OACE,OACA,CAAC,aAAa,EAAE0C,mBAAmB,4BAA4B,EAAE,EAAEP,OAAO,IAAI;oBAIlF,OAAO;wBACL,OAAO;4BACL,KAAK;wBACP;oBACF;gBACF;YACF,GACA;gBACE,gBAAgB;YAClB;YAGF,MAAMJ,aAAa5B,QAAQ;YAG3B,MAAMd,QAAQ0C,YAAY,WAAW,EAAE;YACvChB,SAAS,IAAI,IAAKgB,YAAY,YAAY,EAAE;YAE5C,IAAIY;YACJ,IAAI;gBACFA,cAAc,MAAM,IAAI,CAAC,uBAAuB,CAC9CtD,OACAC,wBACAC,6BACA;oBACEiB;oBACA,SAAS;gBACX;YAEJ,EAAE,OAAO6B,OAAO;gBACd,OAAO3C,QAAQ,eAAe,CAC5B,CAAC,4CAA4C,EAAE2C,MAAM,SAAS,EAAEL,KAAK,SAAS,CAC5E3C,QACC;YAEP;YACA,IAAI,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,EACjDmC,QAAQ,IAAI,CACV,8FACA,IAAI,CAAC,mBAAmB,CAAC,sBAAsB;YAKnD,IAAI;gBACF,MAAM9B,QAAQ,YAAY,CAACiD,YAAY,KAAK;YAC9C,EAAE,OAAON,OAAY;gBAEnBnB;gBACA,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,GAAG,CAAC,MAAM,EAAE0B,wBAAwB,iCAAiC,EAAEP,OAAO,WAAWQ,OAAOR,QAAQ;gBACvJxD,MACE,yFACAwD,iBAAiBS,QAAQT,MAAM,OAAO,GAAGQ,OAAOR,QAChD,6CACAnB;YAEJ;YAEA,IAAIA,8BAA8BnC,uCAChC,OAAOW,QAAQ,eAAe,CAAC;YAIjC,IAAI,CAACqC,YAAY,wBACf;YAIF,EAAEjB;YAEF,IAAIA,cAAcE,sBAAsB;gBACtC,MAAM+B,WAAW,CAAC,UAAU,EAAE/B,qBAAqB,4JAA4J,CAAC;gBAChN,OAAOtB,QAAQ,eAAe,CAACqD;YACjC;YAEA,IAAI,CAAC,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,EAClD,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,GAAG,CAAC,MAAM,EAAEH,wBAAwB,gDAAgD,CAAC;QAExI;QAEA,OAAO;YACL,QAAQ;gBACN7B;gBACA,QAAQI;YACV;YACAlB;QACF;IACF;IAEQ,oBACN+C,IAAsE,EACtEC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACpC;QACA,MAAMC,YAA4C;YAChD,MAAM;YACN,SAASL;YACT,OAAO;gBACL,YAAYI,mBACP;oBACCH;oBACAG;gBACF,IACAH;YACN;YACA,UAAU,OAAOpD,OAAOyD;gBACtB,MAAM,EAAE1D,IAAI,EAAE,GAAG0D;gBACjB,IAAIC;gBACJ,MAAMC,YAAY,CAACC;oBACjBF,YAAYE;oBACZ7D,KAAK,GAAG,GAAG;wBACT6D;oBACF;gBACF;gBAGA,MAAM1D,YAAYuD,YAAY,SAAS;gBACvCtD,OAAOD,WAAW;gBAElB,MAAM2D,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,kBACTC,cAAc;oBACZ,CAACC,YAAY,EAAE,GAAGZ,KAAK,EAAE,EAAEC,QAAQ;gBACrC;gBAGF,IAAIa;gBAEJ,IAAIC,uBAAuB;gBAC3B,IAAIZ,KAAK,eAAe,IAAI,CAAC,SAAS,CAAC,mBAAmB,EAAE;oBAC1DtE,MAAM;oBACN,MAAMmF,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;gBAEJ,EAAE,OAAOf,OAAO;oBACd,IAAIA,iBAAiB6B,cACnBV,UAAUnB,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAEA,MAAM,EAAE8B,IAAI,EAAE7B,KAAK,EAAEJ,OAAO,EAAEuB,IAAI,EAAEjB,iBAAiB,EAAE,GAAGsB;gBAC1DN,UAAUC;gBACV7D,KAAK,iBAAiB,GAAG4C;gBAEzB,IAAI4B,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;qBACV;oBACLpE,OACEmE,MAAM,CAACP,YAAY,KAAK3C,QACxB;oBAEFmD,eAAgBD,IAAY,CAACP,YAAY;gBAC3C;gBAGF,IAAIZ,AAAS,aAATA,QAAqB,CAACoB,cAAc;oBACtCxE,KAAK,KAAK,GAAG0C;oBACb1C,KAAK,OAAO,GAAGsC;oBACf,MAAM,IAAIY,MAAM,CAAC,kBAAkB,EAAEZ,SAAS;gBAChD;gBAEA,OAAO;oBACL,QAAQkC;oBACR,KAAKb;oBACLjB;oBACAJ;gBACF;YACF;QACF;QAEA,OAAOmB;IACT;IACA,MAAM,yBACJL,IAA0D,EAC1DC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACP;QAC7B,MAAM1D,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aACEqD,MACA,AAAkB,YAAlB,OAAOC,SAAsBA,SAASjB,KAAK,SAAS,CAACiB;QAIzD,MAAMI,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9CL,MACAC,QACAC,aACAC,KACAC;QAGF,MAAMnD,SAASP,QAAQ,SAAS;QAChC,MAAMS,SAAS,MAAMT,QAAQ,YAAY,CAAC2D;QAE1C,IAAI,CAAClD,QACH,MAAM,IAAI2C,MACR;QAIJ,MAAM,EAAE1C,MAAM,EAAE8B,OAAO,EAAE,GAAG/B;QAE5B,OAAO;YACLC;YACA8B;YACAjC;QACF;IACF;IAEA,MAAM,aAAaoE,MAAc,EAAEC,YAA0B,EAAE;QAC7D,OAAO,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;YACtCD;QACF;IACF;IAEA,MAAM,QACJE,SAAsB,EACtBpB,GAA+B,EAC/BD,WAAyB,EACO;QAChC,MAAM,EAAEsB,UAAU,EAAEpB,gBAAgB,EAAE,GAAGqB,YAAYF;QAErD,MAAMG,cAAc,CAAC,SAAS,EAAEF,YAAY;QAC5C,MAAM9E,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,WAAW+E;QAE1B,MAAMzE,SAASP,QAAQ,SAAS;QAChC,MAAM,EACJiF,SAAS,EACTC,eAAe,EACfC,WAAW,EACXC,kBAAkB,EAClB,GAAGC,SACJ,GAAG5B;QACJ,MAAM6B,oBAA0C;YAC9CH;YACAC;YACA,GAAGC,OAAO;QACZ;QAEA/E,OAAOuE,WAAW;QAClBvE,OAAO2E,WAAW;QAClB3E,OAAO4E,iBAAiB;QAExB5E,OACE4E,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,MAAMhC,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9C,WACAmB,YACAtB,aACA8B,mBACA5B;YAGF,MAAMjD,SAAU,MAAMT,QAAQ,YAAY,CAAC2D;YAO3C,IAAIlD,QAAQ,QACV,OAAO;gBACL,QAAQc;gBACRhB;YACF;YAGFmF,eACEjF,QAAQ,WACP,CAACA,UAAU,CAAC,0BAA0B,EAAEqE,YAAY,IACrD,CAAC,0CAA0C,EAAEA,YAAY;YAC3D,MAAMc,MAAMJ,KAAK,GAAG;YACpB,IAAII,MAAMD,oBAAoBT,iBAAiB;gBAC7C,MAAMW,gBAAgBX,kBAAmBU,CAAAA,MAAMD,iBAAgB;gBAC/D,MAAMG,YAAY,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;oBACjD,QAAQD;gBACV;gBACA,MAAM7F,QAAQ,MAAM,CAAC8F;YACvB;QACF;QAEA,OAAO9F,QAAQ,eAAe,CAAC,CAAC,iBAAiB,EAAE0F,cAAc;IACnE;IAjnBA,YACEK,iBAAoC,EACpCC,OAAgB,EAChBC,IAOC,CACD;QApCF;QAEA;QAEA;QAEA,uBAAiB,uBAAjB;QAEA,uBAAiB,eAAjB;QAEA,uBAAQ,uBAAR;QAEA;QAEA,uBAAiB,SAAjB;QAEA;QAEA;QAmBE,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,KAAK,GAAGA,KAAK,KAAK;QACvB,IAAI,CAAC,mBAAmB,GAAG,IAAIC;QAC/B,IAAI,CAAC,mBAAmB,GAAGD,KAAK,WAAW;QAC3C,IAAI,CAAC,WAAW,GAAG,IAAIE,YAAY;YACjCJ;YACAC;YACA,WAAWC,KAAK,SAAS;YACzB,aAAa,IAAI,CAAC,cAAc;YAChC,iBAAiBA,KAAK,eAAe;QACvC;IACF;AAslBF;AAEO,eAAe9E,gBACpB4E,iBAAoC,EACpC7E,iBAAuC,EACvCU,MAAwB;IAExB,IAAI,CAACV,mBAAmB,QACtB,OAAOU;IAGT,IAAI,CAACmE,kBAAkB,2BAA2B,EAChD,MAAM,IAAI3C,MACR,CAAC,gCAAgC,EAAE2C,kBAAkB,aAAa,EAAE;IAIxE,MAAMK,UAAU,OAAOC;QACrB,MAAMA,QAAQ,MAAM,CAACnF;IACvB;IAEA,MAAM,EAAEoF,OAAO,EAAEC,QAAQ,EAAE,GACzB,MAAMR,kBAAkB,2BAA2B,CAACK;IACtD,IAAI;QACF,MAAM3F,SAAS,MAAMmB;QAErB,MAAMe,QAAQ4D;QACd,IAAI5D,OACF,MAAMA;QAER,OAAOlC;IACT,SAAU;QACR6F;IACF;AACF"}

package/dist/es/agent/utils.mjs

@@ -103,7 +103,7 @@ async function matchElementFromCache(context, cacheEntry, cachePrompt, cacheable
         return;
     }
 }
-const getMidsceneVersion = ()=>"1.2.2-beta-20260120033218.0";
+const getMidsceneVersion = ()=>"1.2.3-beta-20260120082504.0";
 const parsePrompt = (prompt)=>{
     if ('string' == typeof prompt) return {
         textPrompt: prompt,

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

@@ -6,11 +6,11 @@ import { adaptBboxToRect, expandSearchArea, mergeRects } from "../common.mjs";
 import { parseAutoGLMLocateResponse } from "./auto-glm/parser.mjs";
 import { getAutoGLMLocatePrompt } from "./auto-glm/prompt.mjs";
 import { isAutoGLM } from "./auto-glm/util.mjs";
-import { extractDataQueryPrompt, systemPromptToExtract } from "./prompt/extraction.mjs";
+import { extractDataQueryPrompt, parseXMLExtractionResponse, systemPromptToExtract } from "./prompt/extraction.mjs";
 import { findElementPrompt, systemPromptToLocateElement } from "./prompt/llm-locator.mjs";
 import { sectionLocatorInstruction, systemPromptToLocateSection } from "./prompt/llm-section-locator.mjs";
 import { orderSensitiveJudgePrompt, systemPromptToJudgeOrderSensitive } from "./prompt/order-sensitive-judge.mjs";
-import { callAIWithObjectResponse, callAIWithStringResponse } from "./service-caller/index.mjs";
+import { callAI, callAIWithObjectResponse, callAIWithStringResponse } from "./service-caller/index.mjs";
 const debugInspect = getDebug('ai:inspect');
 const debugSection = getDebug('ai:section');
 const extraTextFromUserPrompt = (prompt)=>{
@@ -313,11 +313,12 @@ async function AiExtractElementInfo(options) {
         });
         msgs.push(...addOns);
     }
-    const result = await callAIWithObjectResponse(msgs, modelConfig);
+    const { content: rawResponse, usage, reasoning_content } = await callAI(msgs, modelConfig);
+    const parseResult = parseXMLExtractionResponse(rawResponse);
     return {
-        parseResult: result.content,
-        usage: result.usage,
-        reasoning_content: result.reasoning_content
+        parseResult,
+        usage,
+        reasoning_content
     };
 }
 async function AiJudgeOrderSensitive(description, callAIFn, modelConfig) {

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  ReferenceImage,\n  ServiceExtractOption,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { generateElementByPosition } from '@midscene/shared/extractor/dom-util';\nimport {\n  cropByRect,\n  paddingToMatchBlockByBase64,\n  preProcessImageUrl,\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  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  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\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  callAIFn: typeof callAIWithObjectResponse<\n    AIElementResponse | [number, number]\n  >;\n  searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;\n  modelConfig: IModelConfig;\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, callAIFn, 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.size.width;\n  let imageHeight = context.size.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\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      // Create a small bbox around the point\n      const bboxSize = 10;\n      const x1 = Math.max(pixelX - bboxSize / 2, 0);\n      const y1 = Math.max(pixelY - bboxSize / 2, 0);\n      const x2 = Math.min(pixelX + bboxSize / 2, imageWidth);\n      const y2 = Math.min(pixelY + bboxSize / 2, imageHeight);\n\n      // Convert to Rect format\n      resRect = {\n        left: x1,\n        top: y1,\n        width: x2 - x1,\n        height: y2 - y1,\n      };\n\n      // Apply offset if searching in a cropped area\n      if (options.searchConfig?.rect) {\n        resRect.left += options.searchConfig.rect.left;\n        resRect.top += options.searchConfig.rect.top;\n      }\n\n      debugInspect('auto-glm resRect:', resRect);\n\n      const rectCenter = {\n        x: resRect.left + resRect.width / 2,\n        y: resRect.top + resRect.height / 2,\n      };\n\n      const element: LocateResultElement = generateElementByPosition(\n        rectCenter,\n        targetElementDescriptionText as string,\n      );\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  const res = await callAIFn(msgs, modelConfig);\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      );\n\n      debugInspect('resRect', resRect);\n\n      const rectCenter = {\n        x: resRect.left + resRect.width / 2,\n        y: resRect.top + resRect.height / 2,\n      };\n\n      const element: LocateResultElement = generateElementByPosition(\n        rectCenter,\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}): Promise<{\n  rect?: Rect;\n  imageBase64?: string;\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  const result = await callAIWithObjectResponse<AISectionLocatorResponse>(\n    msgs,\n    modelConfig,\n  );\n\n  let sectionRect: Rect | undefined;\n  const sectionBbox = result.content.bbox;\n  if (sectionBbox) {\n    const targetRect = adaptBboxToRect(\n      sectionBbox,\n      context.size.width,\n      context.size.height,\n      0,\n      0,\n      context.size.width,\n      context.size.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.size.width,\n          context.size.height,\n          0,\n          0,\n          context.size.width,\n          context.size.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    // expand search area to at least 200 x 200\n    sectionRect = expandSearchArea(mergedRect, context.size, modelFamily);\n    debugSection('expanded sectionRect %j', sectionRect);\n  }\n\n  let imageBase64 = screenshotBase64;\n  if (sectionRect) {\n    const croppedResult = await cropByRect(\n      screenshotBase64,\n      sectionRect,\n      modelFamily === 'qwen2.5-vl',\n    );\n    imageBase64 = croppedResult.imageBase64;\n    sectionRect.width = croppedResult.width;\n    sectionRect.height = croppedResult.height;\n  }\n\n  return {\n    rect: sectionRect,\n    imageBase64,\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 result = await callAIWithObjectResponse<AIDataExtractionResponse<T>>(\n    msgs,\n    modelConfig,\n  );\n  return {\n    parseResult: result.content,\n    usage: result.usage,\n    reasoning_content: result.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  const result = await callAIFn(msgs, modelConfig);\n\n  return {\n    isOrderSensitive: result.content.isOrderSensitive ?? false,\n    usage: result.usage,\n  };\n}\n"],"names":["debugInspect","getDebug","debugSection","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","msgs","item","base64","preProcessImageUrl","AiLocateElement","options","context","targetElementDescription","callAIFn","modelConfig","modelFamily","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","bboxSize","x1","y1","x2","y2","rectCenter","element","generateElementByPosition","res","rawResponse","JSON","Array","adaptBboxToRect","e","msg","Error","AiLocateSection","sectionDescription","systemPromptToLocateSection","sectionLocatorInstructionText","sectionLocatorInstruction","result","callAIWithObjectResponse","sectionRect","sectionBbox","targetRect","referenceBboxList","referenceRects","bbox","mergedRect","mergeRects","expandSearchArea","imageBase64","croppedResult","cropByRect","AiExtractElementInfo","dataQuery","extractOption","systemPromptToExtract","extractDataPromptText","extractDataQueryPrompt","userContent","AiJudgeOrderSensitive","description","systemPromptToJudgeOrderSensitive","userPrompt","orderSensitiveJudgePrompt"],"mappings":";;;;;;;;;;;;;AAuDA,MAAMA,eAAeC,SAAS;AAC9B,MAAMC,eAAeD,SAAS;AAE9B,MAAME,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,gBAAgBC,OAQrC;IAUC,MAAM,EAAEC,OAAO,EAAEC,wBAAwB,EAAEC,QAAQ,EAAEC,WAAW,EAAE,GAAGJ;IACrE,MAAM,EAAEK,WAAW,EAAE,GAAGD;IACxB,MAAME,mBAAmBL,QAAQ,UAAU,CAAC,MAAM;IAElDM,OACEL,0BACA;IAEF,MAAMM,+BAA+BjB,wBACnCW;IAEF,MAAMO,wBAAwBC,kBAAkBF;IAChD,MAAMG,eAAeC,UAAUP,eAC3BQ,uBAAuBR,eACvBS,4BAA4BT;IAEhC,IAAIU,eAAeT;IACnB,IAAIU,aAAaf,QAAQ,IAAI,CAAC,KAAK;IACnC,IAAIgB,cAAchB,QAAQ,IAAI,CAAC,MAAM;IACrC,IAAIiB,qBAAqBF;IACzB,IAAIG,sBAAsBF;IAE1B,IAAIjB,QAAQ,YAAY,EAAE;QACxBO,OACEP,QAAQ,YAAY,CAAC,IAAI,EACzB;QAEFO,OACEP,QAAQ,YAAY,CAAC,WAAW,EAChC;QAGFe,eAAef,QAAQ,YAAY,CAAC,WAAW;QAC/CgB,aAAahB,QAAQ,YAAY,CAAC,IAAI,EAAE;QACxCiB,cAAcjB,QAAQ,YAAY,CAAC,IAAI,EAAE;QACzCkB,qBAAqBF;QACrBG,sBAAsBF;IACxB,OAAO,IAAIZ,AAAgB,iBAAhBA,aAA8B;QACvC,MAAMe,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAMzB,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKI;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMH,UAAUP,eACZ,CAAC,KAAK,EAAEI,uBAAuB,GAC/BA;gBACN;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAOP,0BAAuC;QAChD,MAAMoB,SAAS,MAAM7B,mBAAmB;YACtC,QAAQS,yBAAyB,MAAM;YACvC,yBAAyBA,yBAAyB,uBAAuB;QAC3E;QACAP,KAAK,IAAI,IAAI2B;IACf;IAEA,IAAIV,UAAUP,cAAc;QAC1B,MAAM,EAAE,SAASkB,kBAAkB,EAAEC,KAAK,EAAE,GAC1C,MAAMC,yBAAyB9B,MAAMS;QAEvChB,aAAa,yBAAyBmC;QAEtC,MAAMG,SAASC,2BAA2BJ;QAE1CnC,aAAa,sBAAsBsC,OAAO,KAAK;QAC/CtC,aAAa,yBAAyBsC,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;YAC9DtC,aAAa,yBAAyB0C,MAAM,CAAC,EAAE;QACjD,OAAO;YACL,MAAM,EAAEC,CAAC,EAAEC,CAAC,EAAE,GAAGN,OAAO,WAAW;YAEnCtC,aAAa,iCAAiC;gBAAE2C;gBAAGC;YAAE;YAIrD,MAAMC,SAASC,KAAK,KAAK,CAAEH,IAAIf,aAAc;YAC7C,MAAMmB,SAASD,KAAK,KAAK,CAAEF,IAAIf,cAAe;YAE9C7B,aAAa,+BAA+B;gBAAE6C;gBAAQE;YAAO;YAG7D,MAAMC,WAAW;YACjB,MAAMC,KAAKH,KAAK,GAAG,CAACD,SAASG,WAAW,GAAG;YAC3C,MAAME,KAAKJ,KAAK,GAAG,CAACC,SAASC,WAAW,GAAG;YAC3C,MAAMG,KAAKL,KAAK,GAAG,CAACD,SAASG,WAAW,GAAGpB;YAC3C,MAAMwB,KAAKN,KAAK,GAAG,CAACC,SAASC,WAAW,GAAGnB;YAG3CW,UAAU;gBACR,MAAMS;gBACN,KAAKC;gBACL,OAAOC,KAAKF;gBACZ,QAAQG,KAAKF;YACf;YAGA,IAAItC,QAAQ,YAAY,EAAE,MAAM;gBAC9B4B,QAAQ,IAAI,IAAI5B,QAAQ,YAAY,CAAC,IAAI,CAAC,IAAI;gBAC9C4B,QAAQ,GAAG,IAAI5B,QAAQ,YAAY,CAAC,IAAI,CAAC,GAAG;YAC9C;YAEAZ,aAAa,qBAAqBwC;YAElC,MAAMa,aAAa;gBACjB,GAAGb,QAAQ,IAAI,GAAGA,QAAQ,KAAK,GAAG;gBAClC,GAAGA,QAAQ,GAAG,GAAGA,QAAQ,MAAM,GAAG;YACpC;YAEA,MAAMc,UAA+BC,0BACnCF,YACAjC;YAGF,IAAIkC,SACFb,kBAAkB;gBAACa;aAAQ;QAE/B;QAEA,OAAO;YACL,MAAMd;YACN,aAAa;gBACX,UAAUC;gBACVC;YACF;YACA,aAAaP;YACbC;YACA,mBAAmBE,OAAO,KAAK;QACjC;IACF;IAEA,MAAMkB,MAAM,MAAMzC,SAASR,MAAMS;IAEjC,MAAMyC,cAAcC,KAAK,SAAS,CAACF,IAAI,OAAO;IAE9C,IAAIhB;IACJ,IAAIC,kBAAyC,EAAE;IAC/C,IAAIC,SACF,YAAYc,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,MAAM,GAAG,EAAE;IACnD,IAAI;QACF,IACE,UAAUA,IAAI,OAAO,IACrBG,MAAM,OAAO,CAACH,IAAI,OAAO,CAAC,IAAI,KAC9BA,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,IAAI,GAC3B;YACAhB,UAAUoB,gBACRJ,IAAI,OAAO,CAAC,IAAI,EAChB5B,YACAC,aACAjB,QAAQ,YAAY,EAAE,MAAM,MAC5BA,QAAQ,YAAY,EAAE,MAAM,KAC5BkB,oBACAC,qBACAd;YAGFjB,aAAa,WAAWwC;YAExB,MAAMa,aAAa;gBACjB,GAAGb,QAAQ,IAAI,GAAGA,QAAQ,KAAK,GAAG;gBAClC,GAAGA,QAAQ,GAAG,GAAGA,QAAQ,MAAM,GAAG;YACpC;YAEA,MAAMc,UAA+BC,0BACnCF,YACAjC;YAEFsB,SAAS,EAAE;YAEX,IAAIY,SACFb,kBAAkB;gBAACa;aAAQ;QAE/B;IACF,EAAE,OAAOO,GAAG;QACV,MAAMC,MACJD,aAAaE,QACT,CAAC,sBAAsB,EAAEF,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACnB,UAAUA,QAAQ,WAAW,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAEoB,IAAI,CAAC,CAAC;aAFtBpB,SAAS;YAACoB;SAAI;IAIlB;IAEA,OAAO;QACL,MAAMtB;QACN,aAAa;YACX,UAAUC;YACV,QAAQC;QACV;QACAe;QACA,OAAOD,IAAI,KAAK;QAChB,mBAAmBA,IAAI,iBAAiB;IAC1C;AACF;AAEO,eAAeQ,gBAAgBpD,OAIrC;IAOC,MAAM,EAAEC,OAAO,EAAEoD,kBAAkB,EAAEjD,WAAW,EAAE,GAAGJ;IACrD,MAAM,EAAEK,WAAW,EAAE,GAAGD;IACxB,MAAME,mBAAmBL,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAMU,eAAe2C,4BAA4BjD;IACjD,MAAMkD,gCAAgCC,0BACpCjE,wBAAwB8D;IAE1B,MAAM1D,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKL;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMiD;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAM/B,SAAS,MAAM7B,mBAAmB;YACtC,QAAQ4D,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACA1D,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAMmC,SAAS,MAAMC,yBACnB/D,MACAS;IAGF,IAAIuD;IACJ,MAAMC,cAAcH,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIG,aAAa;QACf,MAAMC,aAAab,gBACjBY,aACA3D,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAEFf,aAAa,0BAA0BuE;QAEvC,MAAMC,oBAAoBL,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9DnE,aAAa,wBAAwBwE;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAASjB,MAAM,OAAO,CAACiB,OAC/B,GAAG,CAAC,CAACA,OACGhB,gBACLgB,MACA/D,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAGNf,aAAa,qBAAqByE;QAGlC,MAAME,aAAaC,WAAW;YAACL;eAAeE;SAAe;QAC7DzE,aAAa,iBAAiB2E;QAG9BN,cAAcQ,iBAAiBF,YAAYhE,QAAQ,IAAI,EAAEI;QACzDf,aAAa,2BAA2BqE;IAC1C;IAEA,IAAIS,cAAc9D;IAClB,IAAIqD,aAAa;QACf,MAAMU,gBAAgB,MAAMC,WAC1BhE,kBACAqD,aACAtD,AAAgB,iBAAhBA;QAEF+D,cAAcC,cAAc,WAAW;QACvCV,YAAY,KAAK,GAAGU,cAAc,KAAK;QACvCV,YAAY,MAAM,GAAGU,cAAc,MAAM;IAC3C;IAEA,OAAO;QACL,MAAMV;QACNS;QACA,OAAOX,OAAO,OAAO,CAAC,KAAK;QAC3B,aAAaX,KAAK,SAAS,CAACW,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAec,qBAAwBvE,OAO7C;IACC,MAAM,EAAEwE,SAAS,EAAEvE,OAAO,EAAEwE,aAAa,EAAE/E,gBAAgB,EAAEU,WAAW,EAAE,GACxEJ;IACF,MAAMW,eAAe+D;IACrB,MAAMpE,mBAAmBL,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM0E,wBAAwBC,uBAC5B5E,QAAQ,eAAe,IAAI,IAC3BwE;IAGF,MAAMK,cAAyD,EAAE;IAEjE,IAAIJ,eAAe,uBAAuB,OACxCI,YAAY,IAAI,CAAC;QACf,MAAM;QACN,WAAW;YACT,KAAKvE;YACL,QAAQ;QACV;IACF;IAGFuE,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAMhF,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAASkE;QACX;KACD;IAED,IAAInF,kBAAkB;QACpB,MAAM4B,SAAS,MAAM7B,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAC,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAMmC,SAAS,MAAMC,yBACnB/D,MACAS;IAEF,OAAO;QACL,aAAaqD,OAAO,OAAO;QAC3B,OAAOA,OAAO,KAAK;QACnB,mBAAmBA,OAAO,iBAAiB;IAC7C;AACF;AAEO,eAAeqB,sBACpBC,WAAmB,EACnB5E,QAAwE,EACxEC,WAAyB;IAKzB,MAAMO,eAAeqE;IACrB,MAAMC,aAAaC,0BAA0BH;IAE7C,MAAMpF,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAASsE;QACX;KACD;IAED,MAAMxB,SAAS,MAAMtD,SAASR,MAAMS;IAEpC,OAAO;QACL,kBAAkBqD,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  AIElementResponse,\n  AISectionLocatorResponse,\n  AIUsageInfo,\n  Rect,\n  ServiceExtractOption,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { generateElementByPosition } from '@midscene/shared/extractor/dom-util';\nimport {\n  cropByRect,\n  paddingToMatchBlockByBase64,\n  preProcessImageUrl,\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  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\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  callAIFn: typeof callAIWithObjectResponse<\n    AIElementResponse | [number, number]\n  >;\n  searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;\n  modelConfig: IModelConfig;\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, callAIFn, 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.size.width;\n  let imageHeight = context.size.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\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      // Create a small bbox around the point\n      const bboxSize = 10;\n      const x1 = Math.max(pixelX - bboxSize / 2, 0);\n      const y1 = Math.max(pixelY - bboxSize / 2, 0);\n      const x2 = Math.min(pixelX + bboxSize / 2, imageWidth);\n      const y2 = Math.min(pixelY + bboxSize / 2, imageHeight);\n\n      // Convert to Rect format\n      resRect = {\n        left: x1,\n        top: y1,\n        width: x2 - x1,\n        height: y2 - y1,\n      };\n\n      // Apply offset if searching in a cropped area\n      if (options.searchConfig?.rect) {\n        resRect.left += options.searchConfig.rect.left;\n        resRect.top += options.searchConfig.rect.top;\n      }\n\n      debugInspect('auto-glm resRect:', resRect);\n\n      const rectCenter = {\n        x: resRect.left + resRect.width / 2,\n        y: resRect.top + resRect.height / 2,\n      };\n\n      const element: LocateResultElement = generateElementByPosition(\n        rectCenter,\n        targetElementDescriptionText as string,\n      );\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  const res = await callAIFn(msgs, modelConfig);\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      );\n\n      debugInspect('resRect', resRect);\n\n      const rectCenter = {\n        x: resRect.left + resRect.width / 2,\n        y: resRect.top + resRect.height / 2,\n      };\n\n      const element: LocateResultElement = generateElementByPosition(\n        rectCenter,\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}): Promise<{\n  rect?: Rect;\n  imageBase64?: string;\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  const result = await callAIWithObjectResponse<AISectionLocatorResponse>(\n    msgs,\n    modelConfig,\n  );\n\n  let sectionRect: Rect | undefined;\n  const sectionBbox = result.content.bbox;\n  if (sectionBbox) {\n    const targetRect = adaptBboxToRect(\n      sectionBbox,\n      context.size.width,\n      context.size.height,\n      0,\n      0,\n      context.size.width,\n      context.size.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.size.width,\n          context.size.height,\n          0,\n          0,\n          context.size.width,\n          context.size.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    // expand search area to at least 200 x 200\n    sectionRect = expandSearchArea(mergedRect, context.size, modelFamily);\n    debugSection('expanded sectionRect %j', sectionRect);\n  }\n\n  let imageBase64 = screenshotBase64;\n  if (sectionRect) {\n    const croppedResult = await cropByRect(\n      screenshotBase64,\n      sectionRect,\n      modelFamily === 'qwen2.5-vl',\n    );\n    imageBase64 = croppedResult.imageBase64;\n    sectionRect.width = croppedResult.width;\n    sectionRect.height = croppedResult.height;\n  }\n\n  return {\n    rect: sectionRect,\n    imageBase64,\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  const parseResult = parseXMLExtractionResponse<T>(rawResponse);\n\n  return {\n    parseResult,\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  const result = await callAIFn(msgs, modelConfig);\n\n  return {\n    isOrderSensitive: result.content.isOrderSensitive ?? false,\n    usage: result.usage,\n  };\n}\n"],"names":["debugInspect","getDebug","debugSection","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","msgs","item","base64","preProcessImageUrl","AiLocateElement","options","context","targetElementDescription","callAIFn","modelConfig","modelFamily","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","bboxSize","x1","y1","x2","y2","rectCenter","element","generateElementByPosition","res","rawResponse","JSON","Array","adaptBboxToRect","e","msg","Error","AiLocateSection","sectionDescription","systemPromptToLocateSection","sectionLocatorInstructionText","sectionLocatorInstruction","result","callAIWithObjectResponse","sectionRect","sectionBbox","targetRect","referenceBboxList","referenceRects","bbox","mergedRect","mergeRects","expandSearchArea","imageBase64","croppedResult","cropByRect","AiExtractElementInfo","dataQuery","extractOption","systemPromptToExtract","extractDataPromptText","extractDataQueryPrompt","userContent","reasoning_content","callAI","parseResult","parseXMLExtractionResponse","AiJudgeOrderSensitive","description","systemPromptToJudgeOrderSensitive","userPrompt","orderSensitiveJudgePrompt"],"mappings":";;;;;;;;;;;;;AAuDA,MAAMA,eAAeC,SAAS;AAC9B,MAAMC,eAAeD,SAAS;AAE9B,MAAME,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,gBAAgBC,OAQrC;IAUC,MAAM,EAAEC,OAAO,EAAEC,wBAAwB,EAAEC,QAAQ,EAAEC,WAAW,EAAE,GAAGJ;IACrE,MAAM,EAAEK,WAAW,EAAE,GAAGD;IACxB,MAAME,mBAAmBL,QAAQ,UAAU,CAAC,MAAM;IAElDM,OACEL,0BACA;IAEF,MAAMM,+BAA+BjB,wBACnCW;IAEF,MAAMO,wBAAwBC,kBAAkBF;IAChD,MAAMG,eAAeC,UAAUP,eAC3BQ,uBAAuBR,eACvBS,4BAA4BT;IAEhC,IAAIU,eAAeT;IACnB,IAAIU,aAAaf,QAAQ,IAAI,CAAC,KAAK;IACnC,IAAIgB,cAAchB,QAAQ,IAAI,CAAC,MAAM;IACrC,IAAIiB,qBAAqBF;IACzB,IAAIG,sBAAsBF;IAE1B,IAAIjB,QAAQ,YAAY,EAAE;QACxBO,OACEP,QAAQ,YAAY,CAAC,IAAI,EACzB;QAEFO,OACEP,QAAQ,YAAY,CAAC,WAAW,EAChC;QAGFe,eAAef,QAAQ,YAAY,CAAC,WAAW;QAC/CgB,aAAahB,QAAQ,YAAY,CAAC,IAAI,EAAE;QACxCiB,cAAcjB,QAAQ,YAAY,CAAC,IAAI,EAAE;QACzCkB,qBAAqBF;QACrBG,sBAAsBF;IACxB,OAAO,IAAIZ,AAAgB,iBAAhBA,aAA8B;QACvC,MAAMe,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAMzB,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKI;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMH,UAAUP,eACZ,CAAC,KAAK,EAAEI,uBAAuB,GAC/BA;gBACN;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAOP,0BAAuC;QAChD,MAAMoB,SAAS,MAAM7B,mBAAmB;YACtC,QAAQS,yBAAyB,MAAM;YACvC,yBAAyBA,yBAAyB,uBAAuB;QAC3E;QACAP,KAAK,IAAI,IAAI2B;IACf;IAEA,IAAIV,UAAUP,cAAc;QAC1B,MAAM,EAAE,SAASkB,kBAAkB,EAAEC,KAAK,EAAE,GAC1C,MAAMC,yBAAyB9B,MAAMS;QAEvChB,aAAa,yBAAyBmC;QAEtC,MAAMG,SAASC,2BAA2BJ;QAE1CnC,aAAa,sBAAsBsC,OAAO,KAAK;QAC/CtC,aAAa,yBAAyBsC,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;YAC9DtC,aAAa,yBAAyB0C,MAAM,CAAC,EAAE;QACjD,OAAO;YACL,MAAM,EAAEC,CAAC,EAAEC,CAAC,EAAE,GAAGN,OAAO,WAAW;YAEnCtC,aAAa,iCAAiC;gBAAE2C;gBAAGC;YAAE;YAIrD,MAAMC,SAASC,KAAK,KAAK,CAAEH,IAAIf,aAAc;YAC7C,MAAMmB,SAASD,KAAK,KAAK,CAAEF,IAAIf,cAAe;YAE9C7B,aAAa,+BAA+B;gBAAE6C;gBAAQE;YAAO;YAG7D,MAAMC,WAAW;YACjB,MAAMC,KAAKH,KAAK,GAAG,CAACD,SAASG,WAAW,GAAG;YAC3C,MAAME,KAAKJ,KAAK,GAAG,CAACC,SAASC,WAAW,GAAG;YAC3C,MAAMG,KAAKL,KAAK,GAAG,CAACD,SAASG,WAAW,GAAGpB;YAC3C,MAAMwB,KAAKN,KAAK,GAAG,CAACC,SAASC,WAAW,GAAGnB;YAG3CW,UAAU;gBACR,MAAMS;gBACN,KAAKC;gBACL,OAAOC,KAAKF;gBACZ,QAAQG,KAAKF;YACf;YAGA,IAAItC,QAAQ,YAAY,EAAE,MAAM;gBAC9B4B,QAAQ,IAAI,IAAI5B,QAAQ,YAAY,CAAC,IAAI,CAAC,IAAI;gBAC9C4B,QAAQ,GAAG,IAAI5B,QAAQ,YAAY,CAAC,IAAI,CAAC,GAAG;YAC9C;YAEAZ,aAAa,qBAAqBwC;YAElC,MAAMa,aAAa;gBACjB,GAAGb,QAAQ,IAAI,GAAGA,QAAQ,KAAK,GAAG;gBAClC,GAAGA,QAAQ,GAAG,GAAGA,QAAQ,MAAM,GAAG;YACpC;YAEA,MAAMc,UAA+BC,0BACnCF,YACAjC;YAGF,IAAIkC,SACFb,kBAAkB;gBAACa;aAAQ;QAE/B;QAEA,OAAO;YACL,MAAMd;YACN,aAAa;gBACX,UAAUC;gBACVC;YACF;YACA,aAAaP;YACbC;YACA,mBAAmBE,OAAO,KAAK;QACjC;IACF;IAEA,MAAMkB,MAAM,MAAMzC,SAASR,MAAMS;IAEjC,MAAMyC,cAAcC,KAAK,SAAS,CAACF,IAAI,OAAO;IAE9C,IAAIhB;IACJ,IAAIC,kBAAyC,EAAE;IAC/C,IAAIC,SACF,YAAYc,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,MAAM,GAAG,EAAE;IACnD,IAAI;QACF,IACE,UAAUA,IAAI,OAAO,IACrBG,MAAM,OAAO,CAACH,IAAI,OAAO,CAAC,IAAI,KAC9BA,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,IAAI,GAC3B;YACAhB,UAAUoB,gBACRJ,IAAI,OAAO,CAAC,IAAI,EAChB5B,YACAC,aACAjB,QAAQ,YAAY,EAAE,MAAM,MAC5BA,QAAQ,YAAY,EAAE,MAAM,KAC5BkB,oBACAC,qBACAd;YAGFjB,aAAa,WAAWwC;YAExB,MAAMa,aAAa;gBACjB,GAAGb,QAAQ,IAAI,GAAGA,QAAQ,KAAK,GAAG;gBAClC,GAAGA,QAAQ,GAAG,GAAGA,QAAQ,MAAM,GAAG;YACpC;YAEA,MAAMc,UAA+BC,0BACnCF,YACAjC;YAEFsB,SAAS,EAAE;YAEX,IAAIY,SACFb,kBAAkB;gBAACa;aAAQ;QAE/B;IACF,EAAE,OAAOO,GAAG;QACV,MAAMC,MACJD,aAAaE,QACT,CAAC,sBAAsB,EAAEF,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACnB,UAAUA,QAAQ,WAAW,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAEoB,IAAI,CAAC,CAAC;aAFtBpB,SAAS;YAACoB;SAAI;IAIlB;IAEA,OAAO;QACL,MAAMtB;QACN,aAAa;YACX,UAAUC;YACV,QAAQC;QACV;QACAe;QACA,OAAOD,IAAI,KAAK;QAChB,mBAAmBA,IAAI,iBAAiB;IAC1C;AACF;AAEO,eAAeQ,gBAAgBpD,OAIrC;IAOC,MAAM,EAAEC,OAAO,EAAEoD,kBAAkB,EAAEjD,WAAW,EAAE,GAAGJ;IACrD,MAAM,EAAEK,WAAW,EAAE,GAAGD;IACxB,MAAME,mBAAmBL,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAMU,eAAe2C,4BAA4BjD;IACjD,MAAMkD,gCAAgCC,0BACpCjE,wBAAwB8D;IAE1B,MAAM1D,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKL;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMiD;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAM/B,SAAS,MAAM7B,mBAAmB;YACtC,QAAQ4D,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACA1D,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAMmC,SAAS,MAAMC,yBACnB/D,MACAS;IAGF,IAAIuD;IACJ,MAAMC,cAAcH,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIG,aAAa;QACf,MAAMC,aAAab,gBACjBY,aACA3D,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAEFf,aAAa,0BAA0BuE;QAEvC,MAAMC,oBAAoBL,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9DnE,aAAa,wBAAwBwE;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAASjB,MAAM,OAAO,CAACiB,OAC/B,GAAG,CAAC,CAACA,OACGhB,gBACLgB,MACA/D,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAGNf,aAAa,qBAAqByE;QAGlC,MAAME,aAAaC,WAAW;YAACL;eAAeE;SAAe;QAC7DzE,aAAa,iBAAiB2E;QAG9BN,cAAcQ,iBAAiBF,YAAYhE,QAAQ,IAAI,EAAEI;QACzDf,aAAa,2BAA2BqE;IAC1C;IAEA,IAAIS,cAAc9D;IAClB,IAAIqD,aAAa;QACf,MAAMU,gBAAgB,MAAMC,WAC1BhE,kBACAqD,aACAtD,AAAgB,iBAAhBA;QAEF+D,cAAcC,cAAc,WAAW;QACvCV,YAAY,KAAK,GAAGU,cAAc,KAAK;QACvCV,YAAY,MAAM,GAAGU,cAAc,MAAM;IAC3C;IAEA,OAAO;QACL,MAAMV;QACNS;QACA,OAAOX,OAAO,OAAO,CAAC,KAAK;QAC3B,aAAaX,KAAK,SAAS,CAACW,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAec,qBAAwBvE,OAO7C;IACC,MAAM,EAAEwE,SAAS,EAAEvE,OAAO,EAAEwE,aAAa,EAAE/E,gBAAgB,EAAEU,WAAW,EAAE,GACxEJ;IACF,MAAMW,eAAe+D;IACrB,MAAMpE,mBAAmBL,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM0E,wBAAwBC,uBAC5B5E,QAAQ,eAAe,IAAI,IAC3BwE;IAGF,MAAMK,cAAyD,EAAE;IAEjE,IAAIJ,eAAe,uBAAuB,OACxCI,YAAY,IAAI,CAAC;QACf,MAAM;QACN,WAAW;YACT,KAAKvE;YACL,QAAQ;QACV;IACF;IAGFuE,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAMhF,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAASkE;QACX;KACD;IAED,IAAInF,kBAAkB;QACpB,MAAM4B,SAAS,MAAM7B,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAC,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAM,EACJ,SAASuB,WAAW,EACpBrB,KAAK,EACLsD,iBAAiB,EAClB,GAAG,MAAMC,OAAOpF,MAAMS;IAGvB,MAAM4E,cAAcC,2BAA8BpC;IAElD,OAAO;QACLmC;QACAxD;QACAsD;IACF;AACF;AAEO,eAAeI,sBACpBC,WAAmB,EACnBhF,QAAwE,EACxEC,WAAyB;IAKzB,MAAMO,eAAeyE;IACrB,MAAMC,aAAaC,0BAA0BH;IAE7C,MAAMxF,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS0E;QACX;KACD;IAED,MAAM5B,SAAS,MAAMtD,SAASR,MAAMS;IAEpC,OAAO;QACL,kBAAkBqD,OAAO,OAAO,CAAC,gBAAgB,IAAI;QACrD,OAAOA,OAAO,KAAK;IACrB;AACF"}

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

@@ -1,10 +1,63 @@
 import { paddingToMatchBlockByBase64 } from "@midscene/shared/img";
 import { getDebug } from "@midscene/shared/logger";
 import { assert } from "@midscene/shared/utils";
-import { buildYamlFlowFromPlans, fillBboxParam, finalizeActionName, findAllMidsceneLocatorField } from "../common.mjs";
+import { buildYamlFlowFromPlans, fillBboxParam, findAllMidsceneLocatorField } from "../common.mjs";
 import { systemPromptToTaskPlanning } from "./prompt/llm-planning.mjs";
-import { callAIWithObjectResponse } from "./service-caller/index.mjs";
+import { extractXMLTag } from "./prompt/util.mjs";
+import { callAI, safeParseJson } from "./service-caller/index.mjs";
 const debug = getDebug('planning');
+function parseXMLPlanningResponse(xmlString, modelFamily) {
+    const thought = extractXMLTag(xmlString, 'thought');
+    const note = extractXMLTag(xmlString, 'note');
+    const log = extractXMLTag(xmlString, 'log');
+    const error = extractXMLTag(xmlString, 'error');
+    const actionType = extractXMLTag(xmlString, 'action-type');
+    const actionParamStr = extractXMLTag(xmlString, 'action-param-json');
+    const completeTaskRegex = /<complete-task\s+success="(true|false)">([\s\S]*?)<\/complete-task>/i;
+    const completeTaskMatch = xmlString.match(completeTaskRegex);
+    let finalizeMessage;
+    let finalizeSuccess;
+    if (completeTaskMatch) {
+        finalizeSuccess = 'true' === completeTaskMatch[1];
+        finalizeMessage = completeTaskMatch[2]?.trim() || void 0;
+    }
+    if (!log) throw new Error('Missing required field: log');
+    let action = null;
+    if (actionType && 'null' !== actionType.toLowerCase()) {
+        const type = actionType.trim();
+        let param;
+        if (actionParamStr) try {
+            param = safeParseJson(actionParamStr, modelFamily);
+        } catch (e) {
+            throw new Error(`Failed to parse action-param-json: ${e}`);
+        }
+        action = {
+            type,
+            ...void 0 !== param ? {
+                param
+            } : {}
+        };
+    }
+    return {
+        ...thought ? {
+            thought
+        } : {},
+        ...note ? {
+            note
+        } : {},
+        log,
+        ...error ? {
+            error
+        } : {},
+        action,
+        ...void 0 !== finalizeMessage ? {
+            finalizeMessage
+        } : {},
+        ...void 0 !== finalizeSuccess ? {
+            finalizeSuccess
+        } : {}
+    };
+}
 async function plan(userInstruction, opts) {
     const { context, modelConfig, conversationHistory } = opts;
     const { size } = context;
@@ -84,15 +137,16 @@ async function plan(userInstruction, opts) {
         ...instruction,
         ...historyLog
     ];
-    const { content: planFromAI, contentString: rawResponse, usage, reasoning_content } = await callAIWithObjectResponse(msgs, modelConfig, {
+    const { content: rawResponse, usage, reasoning_content } = await callAI(msgs, modelConfig, {
         deepThink: 'unset' === opts.deepThink ? void 0 : opts.deepThink
     });
+    const planFromAI = parseXMLPlanningResponse(rawResponse, modelFamily);
     const actions = planFromAI.action ? [
         planFromAI.action
     ] : [];
     let shouldContinuePlanning = true;
-    if (actions[0]?.type === finalizeActionName) {
-        debug('finalize action planned, stop planning');
+    if (void 0 !== planFromAI.finalizeSuccess) {
+        debug('task finalized via complete-task tag, stop planning');
         shouldContinuePlanning = false;
     }
     const returnValue = {
@@ -127,6 +181,6 @@ async function plan(userInstruction, opts) {
     });
     return returnValue;
 }
-export { plan };
+export { parseXMLPlanningResponse, plan };
 
 //# sourceMappingURL=llm-planning.mjs.map

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

@@ -1 +1 @@
-{"version":3,"file":"ai-model/llm-planning.mjs","sources":["../../../src/ai-model/llm-planning.ts"],"sourcesContent":["import type {\n  DeepThinkOption,\n  DeviceAction,\n  InterfaceType,\n  PlanningAIResponse,\n  RawResponsePlanningAIResponse,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { paddingToMatchBlockByBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport {\n  buildYamlFlowFromPlans,\n  fillBboxParam,\n  finalizeActionName,\n  findAllMidsceneLocatorField,\n} from '../common';\nimport type { ConversationHistory } from './conversation-history';\nimport { systemPromptToTaskPlanning } from './prompt/llm-planning';\nimport { callAIWithObjectResponse } from './service-caller/index';\n\nconst debug = getDebug('planning');\n\nexport async function plan(\n  userInstruction: string,\n  opts: {\n    context: UIContext;\n    interfaceType: InterfaceType;\n    actionSpace: DeviceAction<any>[];\n    actionContext?: string;\n    modelConfig: IModelConfig;\n    conversationHistory: ConversationHistory;\n    includeBbox: boolean;\n    imagesIncludeCount?: number;\n    deepThink?: DeepThinkOption;\n  },\n): Promise<PlanningAIResponse> {\n  const { context, modelConfig, conversationHistory } = opts;\n  const { size } = context;\n  const screenshotBase64 = context.screenshot.base64;\n\n  const { modelFamily } = modelConfig;\n\n  const systemPrompt = await systemPromptToTaskPlanning({\n    actionSpace: opts.actionSpace,\n    modelFamily,\n    includeBbox: opts.includeBbox,\n    includeThought: opts.deepThink !== true,\n  });\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = size.width;\n  let imageHeight = size.height;\n  const rightLimit = imageWidth;\n  const bottomLimit = imageHeight;\n\n  // Process image based on VL mode requirements\n  if (modelFamily === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  }\n\n  const actionContext = opts.actionContext\n    ? `<high_priority_knowledge>${opts.actionContext}</high_priority_knowledge>\\n`\n    : '';\n\n  const instruction: ChatCompletionMessageParam[] = [\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${actionContext}<user_instruction>${userInstruction}</user_instruction>`,\n        },\n      ],\n    },\n  ];\n\n  let latestFeedbackMessage: ChatCompletionMessageParam;\n\n  if (conversationHistory.pendingFeedbackMessage) {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${conversationHistory.pendingFeedbackMessage}. The last screenshot is attached. Please going on according to the instruction.`,\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n\n    conversationHistory.resetPendingFeedbackMessageIfExists();\n  } else {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: 'this is the latest screenshot',\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n  }\n  conversationHistory.append(latestFeedbackMessage);\n  const historyLog = conversationHistory.snapshot(opts.imagesIncludeCount);\n\n  const msgs: ChatCompletionMessageParam[] = [\n    { role: 'system', content: systemPrompt },\n    ...instruction,\n    ...historyLog,\n  ];\n\n  const {\n    content: planFromAI,\n    contentString: rawResponse,\n    usage,\n    reasoning_content,\n  } = await callAIWithObjectResponse<RawResponsePlanningAIResponse>(\n    msgs,\n    modelConfig,\n    {\n      deepThink: opts.deepThink === 'unset' ? undefined : opts.deepThink,\n    },\n  );\n\n  const actions = planFromAI.action ? [planFromAI.action] : [];\n  let shouldContinuePlanning = true;\n  if (actions[0]?.type === finalizeActionName) {\n    debug('finalize action planned, stop planning');\n    shouldContinuePlanning = false;\n  }\n  const returnValue: PlanningAIResponse = {\n    ...planFromAI,\n    actions,\n    rawResponse,\n    usage,\n    reasoning_content,\n    yamlFlow: buildYamlFlowFromPlans(actions, opts.actionSpace),\n    shouldContinuePlanning,\n  };\n\n  assert(planFromAI, \"can't get plans from AI\");\n\n  actions.forEach((action) => {\n    const type = action.type;\n    const actionInActionSpace = opts.actionSpace.find(\n      (action) => action.name === type,\n    );\n\n    debug('actionInActionSpace matched', actionInActionSpace);\n    const locateFields = actionInActionSpace\n      ? findAllMidsceneLocatorField(actionInActionSpace.paramSchema)\n      : [];\n\n    debug('locateFields', locateFields);\n\n    locateFields.forEach((field) => {\n      const locateResult = action.param[field];\n      if (locateResult && modelFamily !== undefined) {\n        // Always use model family to fill bbox parameters\n        action.param[field] = fillBboxParam(\n          locateResult,\n          imageWidth,\n          imageHeight,\n          rightLimit,\n          bottomLimit,\n          modelFamily,\n        );\n      }\n    });\n  });\n\n  conversationHistory.append({\n    role: 'assistant',\n    content: [\n      {\n        type: 'text',\n        text: rawResponse,\n      },\n    ],\n  });\n\n  return returnValue;\n}\n"],"names":["debug","getDebug","plan","userInstruction","opts","context","modelConfig","conversationHistory","size","screenshotBase64","modelFamily","systemPrompt","systemPromptToTaskPlanning","imagePayload","imageWidth","imageHeight","rightLimit","bottomLimit","paddedResult","paddingToMatchBlockByBase64","actionContext","instruction","latestFeedbackMessage","historyLog","msgs","planFromAI","rawResponse","usage","reasoning_content","callAIWithObjectResponse","undefined","actions","shouldContinuePlanning","finalizeActionName","returnValue","buildYamlFlowFromPlans","assert","action","type","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam"],"mappings":";;;;;;AAuBA,MAAMA,QAAQC,SAAS;AAEhB,eAAeC,KACpBC,eAAuB,EACvBC,IAUC;IAED,MAAM,EAAEC,OAAO,EAAEC,WAAW,EAAEC,mBAAmB,EAAE,GAAGH;IACtD,MAAM,EAAEI,IAAI,EAAE,GAAGH;IACjB,MAAMI,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM,EAAEK,WAAW,EAAE,GAAGJ;IAExB,MAAMK,eAAe,MAAMC,2BAA2B;QACpD,aAAaR,KAAK,WAAW;QAC7BM;QACA,aAAaN,KAAK,WAAW;QAC7B,gBAAgBA,AAAmB,SAAnBA,KAAK,SAAS;IAChC;IAEA,IAAIS,eAAeJ;IACnB,IAAIK,aAAaN,KAAK,KAAK;IAC3B,IAAIO,cAAcP,KAAK,MAAM;IAC7B,MAAMQ,aAAaF;IACnB,MAAMG,cAAcF;IAGpB,IAAIL,AAAgB,iBAAhBA,aAA8B;QAChC,MAAMQ,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAME,gBAAgBhB,KAAK,aAAa,GACpC,CAAC,yBAAyB,EAAEA,KAAK,aAAa,CAAC,4BAA4B,CAAC,GAC5E;IAEJ,MAAMiB,cAA4C;QAChD;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGD,cAAc,kBAAkB,EAAEjB,gBAAgB,mBAAmB,CAAC;gBACjF;aACD;QACH;KACD;IAED,IAAImB;IAEJ,IAAIf,oBAAoB,sBAAsB,EAAE;QAC9Ce,wBAAwB;YACtB,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGf,oBAAoB,sBAAsB,CAAC,gFAAgF,CAAC;gBACvI;gBACA;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKM;wBACL,QAAQ;oBACV;gBACF;aACD;QACH;QAEAN,oBAAoB,mCAAmC;IACzD,OACEe,wBAAwB;QACtB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAM;YACR;YACA;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKT;oBACL,QAAQ;gBACV;YACF;SACD;IACH;IAEFN,oBAAoB,MAAM,CAACe;IAC3B,MAAMC,aAAahB,oBAAoB,QAAQ,CAACH,KAAK,kBAAkB;IAEvE,MAAMoB,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASb;QAAa;WACrCU;WACAE;KACJ;IAED,MAAM,EACJ,SAASE,UAAU,EACnB,eAAeC,WAAW,EAC1BC,KAAK,EACLC,iBAAiB,EAClB,GAAG,MAAMC,yBACRL,MACAlB,aACA;QACE,WAAWF,AAAmB,YAAnBA,KAAK,SAAS,GAAe0B,SAAY1B,KAAK,SAAS;IACpE;IAGF,MAAM2B,UAAUN,WAAW,MAAM,GAAG;QAACA,WAAW,MAAM;KAAC,GAAG,EAAE;IAC5D,IAAIO,yBAAyB;IAC7B,IAAID,OAAO,CAAC,EAAE,EAAE,SAASE,oBAAoB;QAC3CjC,MAAM;QACNgC,yBAAyB;IAC3B;IACA,MAAME,cAAkC;QACtC,GAAGT,UAAU;QACbM;QACAL;QACAC;QACAC;QACA,UAAUO,uBAAuBJ,SAAS3B,KAAK,WAAW;QAC1D4B;IACF;IAEAI,OAAOX,YAAY;IAEnBM,QAAQ,OAAO,CAAC,CAACM;QACf,MAAMC,OAAOD,OAAO,IAAI;QACxB,MAAME,sBAAsBnC,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACiC,SAAWA,OAAO,IAAI,KAAKC;QAG9BtC,MAAM,+BAA+BuC;QACrC,MAAMC,eAAeD,sBACjBE,4BAA4BF,oBAAoB,WAAW,IAC3D,EAAE;QAENvC,MAAM,gBAAgBwC;QAEtBA,aAAa,OAAO,CAAC,CAACE;YACpB,MAAMC,eAAeN,OAAO,KAAK,CAACK,MAAM;YACxC,IAAIC,gBAAgBjC,AAAgBoB,WAAhBpB,aAElB2B,OAAO,KAAK,CAACK,MAAM,GAAGE,cACpBD,cACA7B,YACAC,aACAC,YACAC,aACAP;QAGN;IACF;IAEAH,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAMmB;YACR;SACD;IACH;IAEA,OAAOQ;AACT"}
+{"version":3,"file":"ai-model/llm-planning.mjs","sources":["../../../src/ai-model/llm-planning.ts"],"sourcesContent":["import type {\n  DeepThinkOption,\n  DeviceAction,\n  InterfaceType,\n  PlanningAIResponse,\n  RawResponsePlanningAIResponse,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig, TModelFamily } from '@midscene/shared/env';\nimport { paddingToMatchBlockByBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport {\n  buildYamlFlowFromPlans,\n  fillBboxParam,\n  findAllMidsceneLocatorField,\n} from '../common';\nimport type { ConversationHistory } from './conversation-history';\nimport { systemPromptToTaskPlanning } from './prompt/llm-planning';\nimport { extractXMLTag } from './prompt/util';\nimport { callAI } from './service-caller/index';\nimport { safeParseJson } from './service-caller/index';\n\nconst debug = getDebug('planning');\n\n/**\n * Parse XML response from LLM and convert to RawResponsePlanningAIResponse\n */\nexport function parseXMLPlanningResponse(\n  xmlString: string,\n  modelFamily: TModelFamily | undefined,\n): RawResponsePlanningAIResponse {\n  const thought = extractXMLTag(xmlString, 'thought');\n  const note = extractXMLTag(xmlString, 'note');\n  const log = extractXMLTag(xmlString, 'log');\n  const error = extractXMLTag(xmlString, 'error');\n  const actionType = extractXMLTag(xmlString, 'action-type');\n  const actionParamStr = extractXMLTag(xmlString, 'action-param-json');\n\n  // Parse complete-task tag with success attribute\n  const completeTaskRegex =\n    /<complete-task\\s+success=\"(true|false)\">([\\s\\S]*?)<\\/complete-task>/i;\n  const completeTaskMatch = xmlString.match(completeTaskRegex);\n  let finalizeMessage: string | undefined;\n  let finalizeSuccess: boolean | undefined;\n\n  if (completeTaskMatch) {\n    finalizeSuccess = completeTaskMatch[1] === 'true';\n    finalizeMessage = completeTaskMatch[2]?.trim() || undefined;\n  }\n\n  // Validate required fields\n  if (!log) {\n    throw new Error('Missing required field: log');\n  }\n\n  // Parse action\n  let action: any = null;\n  if (actionType && actionType.toLowerCase() !== 'null') {\n    const type = actionType.trim();\n    let param: any = undefined;\n\n    if (actionParamStr) {\n      try {\n        // Parse the JSON string in action-param-json\n        param = safeParseJson(actionParamStr, modelFamily);\n      } catch (e) {\n        throw new Error(`Failed to parse action-param-json: ${e}`);\n      }\n    }\n\n    action = {\n      type,\n      ...(param !== undefined ? { param } : {}),\n    };\n  }\n\n  return {\n    ...(thought ? { thought } : {}),\n    ...(note ? { note } : {}),\n    log,\n    ...(error ? { error } : {}),\n    action,\n    ...(finalizeMessage !== undefined ? { finalizeMessage } : {}),\n    ...(finalizeSuccess !== undefined ? { finalizeSuccess } : {}),\n  };\n}\n\nexport async function plan(\n  userInstruction: string,\n  opts: {\n    context: UIContext;\n    interfaceType: InterfaceType;\n    actionSpace: DeviceAction<any>[];\n    actionContext?: string;\n    modelConfig: IModelConfig;\n    conversationHistory: ConversationHistory;\n    includeBbox: boolean;\n    imagesIncludeCount?: number;\n    deepThink?: DeepThinkOption;\n  },\n): Promise<PlanningAIResponse> {\n  const { context, modelConfig, conversationHistory } = opts;\n  const { size } = context;\n  const screenshotBase64 = context.screenshot.base64;\n\n  const { modelFamily } = modelConfig;\n\n  const systemPrompt = await systemPromptToTaskPlanning({\n    actionSpace: opts.actionSpace,\n    modelFamily,\n    includeBbox: opts.includeBbox,\n    includeThought: opts.deepThink !== true,\n  });\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = size.width;\n  let imageHeight = size.height;\n  const rightLimit = imageWidth;\n  const bottomLimit = imageHeight;\n\n  // Process image based on VL mode requirements\n  if (modelFamily === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  }\n\n  const actionContext = opts.actionContext\n    ? `<high_priority_knowledge>${opts.actionContext}</high_priority_knowledge>\\n`\n    : '';\n\n  const instruction: ChatCompletionMessageParam[] = [\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${actionContext}<user_instruction>${userInstruction}</user_instruction>`,\n        },\n      ],\n    },\n  ];\n\n  let latestFeedbackMessage: ChatCompletionMessageParam;\n\n  if (conversationHistory.pendingFeedbackMessage) {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${conversationHistory.pendingFeedbackMessage}. The last screenshot is attached. Please going on according to the instruction.`,\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n\n    conversationHistory.resetPendingFeedbackMessageIfExists();\n  } else {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: 'this is the latest screenshot',\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n  }\n  conversationHistory.append(latestFeedbackMessage);\n  const historyLog = conversationHistory.snapshot(opts.imagesIncludeCount);\n\n  const msgs: ChatCompletionMessageParam[] = [\n    { role: 'system', content: systemPrompt },\n    ...instruction,\n    ...historyLog,\n  ];\n\n  const {\n    content: rawResponse,\n    usage,\n    reasoning_content,\n  } = await callAI(msgs, modelConfig, {\n    deepThink: opts.deepThink === 'unset' ? undefined : opts.deepThink,\n  });\n\n  // Parse XML response to JSON object\n  const planFromAI = parseXMLPlanningResponse(rawResponse, modelFamily);\n\n  const actions = planFromAI.action ? [planFromAI.action] : [];\n  let shouldContinuePlanning = true;\n\n  // Check if task is finalized via complete-task tag\n  if (planFromAI.finalizeSuccess !== undefined) {\n    debug('task finalized via complete-task tag, stop planning');\n    shouldContinuePlanning = false;\n  }\n\n  const returnValue: PlanningAIResponse = {\n    ...planFromAI,\n    actions,\n    rawResponse,\n    usage,\n    reasoning_content,\n    yamlFlow: buildYamlFlowFromPlans(actions, opts.actionSpace),\n    shouldContinuePlanning,\n  };\n\n  assert(planFromAI, \"can't get plans from AI\");\n\n  actions.forEach((action) => {\n    const type = action.type;\n    const actionInActionSpace = opts.actionSpace.find(\n      (action) => action.name === type,\n    );\n\n    debug('actionInActionSpace matched', actionInActionSpace);\n    const locateFields = actionInActionSpace\n      ? findAllMidsceneLocatorField(actionInActionSpace.paramSchema)\n      : [];\n\n    debug('locateFields', locateFields);\n\n    locateFields.forEach((field) => {\n      const locateResult = action.param[field];\n      if (locateResult && modelFamily !== undefined) {\n        // Always use model family to fill bbox parameters\n        action.param[field] = fillBboxParam(\n          locateResult,\n          imageWidth,\n          imageHeight,\n          rightLimit,\n          bottomLimit,\n          modelFamily,\n        );\n      }\n    });\n  });\n\n  conversationHistory.append({\n    role: 'assistant',\n    content: [\n      {\n        type: 'text',\n        text: rawResponse,\n      },\n    ],\n  });\n\n  return returnValue;\n}\n"],"names":["debug","getDebug","parseXMLPlanningResponse","xmlString","modelFamily","thought","extractXMLTag","note","log","error","actionType","actionParamStr","completeTaskRegex","completeTaskMatch","finalizeMessage","finalizeSuccess","undefined","Error","action","type","param","safeParseJson","e","plan","userInstruction","opts","context","modelConfig","conversationHistory","size","screenshotBase64","systemPrompt","systemPromptToTaskPlanning","imagePayload","imageWidth","imageHeight","rightLimit","bottomLimit","paddedResult","paddingToMatchBlockByBase64","actionContext","instruction","latestFeedbackMessage","historyLog","msgs","rawResponse","usage","reasoning_content","callAI","planFromAI","actions","shouldContinuePlanning","returnValue","buildYamlFlowFromPlans","assert","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam"],"mappings":";;;;;;;AAwBA,MAAMA,QAAQC,SAAS;AAKhB,SAASC,yBACdC,SAAiB,EACjBC,WAAqC;IAErC,MAAMC,UAAUC,cAAcH,WAAW;IACzC,MAAMI,OAAOD,cAAcH,WAAW;IACtC,MAAMK,MAAMF,cAAcH,WAAW;IACrC,MAAMM,QAAQH,cAAcH,WAAW;IACvC,MAAMO,aAAaJ,cAAcH,WAAW;IAC5C,MAAMQ,iBAAiBL,cAAcH,WAAW;IAGhD,MAAMS,oBACJ;IACF,MAAMC,oBAAoBV,UAAU,KAAK,CAACS;IAC1C,IAAIE;IACJ,IAAIC;IAEJ,IAAIF,mBAAmB;QACrBE,kBAAkBF,AAAyB,WAAzBA,iBAAiB,CAAC,EAAE;QACtCC,kBAAkBD,iBAAiB,CAAC,EAAE,EAAE,UAAUG;IACpD;IAGA,IAAI,CAACR,KACH,MAAM,IAAIS,MAAM;IAIlB,IAAIC,SAAc;IAClB,IAAIR,cAAcA,AAA6B,WAA7BA,WAAW,WAAW,IAAe;QACrD,MAAMS,OAAOT,WAAW,IAAI;QAC5B,IAAIU;QAEJ,IAAIT,gBACF,IAAI;YAEFS,QAAQC,cAAcV,gBAAgBP;QACxC,EAAE,OAAOkB,GAAG;YACV,MAAM,IAAIL,MAAM,CAAC,mCAAmC,EAAEK,GAAG;QAC3D;QAGFJ,SAAS;YACPC;YACA,GAAIC,AAAUJ,WAAVI,QAAsB;gBAAEA;YAAM,IAAI,CAAC,CAAC;QAC1C;IACF;IAEA,OAAO;QACL,GAAIf,UAAU;YAAEA;QAAQ,IAAI,CAAC,CAAC;QAC9B,GAAIE,OAAO;YAAEA;QAAK,IAAI,CAAC,CAAC;QACxBC;QACA,GAAIC,QAAQ;YAAEA;QAAM,IAAI,CAAC,CAAC;QAC1BS;QACA,GAAIJ,AAAoBE,WAApBF,kBAAgC;YAAEA;QAAgB,IAAI,CAAC,CAAC;QAC5D,GAAIC,AAAoBC,WAApBD,kBAAgC;YAAEA;QAAgB,IAAI,CAAC,CAAC;IAC9D;AACF;AAEO,eAAeQ,KACpBC,eAAuB,EACvBC,IAUC;IAED,MAAM,EAAEC,OAAO,EAAEC,WAAW,EAAEC,mBAAmB,EAAE,GAAGH;IACtD,MAAM,EAAEI,IAAI,EAAE,GAAGH;IACjB,MAAMI,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM,EAAEtB,WAAW,EAAE,GAAGuB;IAExB,MAAMI,eAAe,MAAMC,2BAA2B;QACpD,aAAaP,KAAK,WAAW;QAC7BrB;QACA,aAAaqB,KAAK,WAAW;QAC7B,gBAAgBA,AAAmB,SAAnBA,KAAK,SAAS;IAChC;IAEA,IAAIQ,eAAeH;IACnB,IAAII,aAAaL,KAAK,KAAK;IAC3B,IAAIM,cAAcN,KAAK,MAAM;IAC7B,MAAMO,aAAaF;IACnB,MAAMG,cAAcF;IAGpB,IAAI/B,AAAgB,iBAAhBA,aAA8B;QAChC,MAAMkC,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAME,gBAAgBf,KAAK,aAAa,GACpC,CAAC,yBAAyB,EAAEA,KAAK,aAAa,CAAC,4BAA4B,CAAC,GAC5E;IAEJ,MAAMgB,cAA4C;QAChD;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGD,cAAc,kBAAkB,EAAEhB,gBAAgB,mBAAmB,CAAC;gBACjF;aACD;QACH;KACD;IAED,IAAIkB;IAEJ,IAAId,oBAAoB,sBAAsB,EAAE;QAC9Cc,wBAAwB;YACtB,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGd,oBAAoB,sBAAsB,CAAC,gFAAgF,CAAC;gBACvI;gBACA;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKK;wBACL,QAAQ;oBACV;gBACF;aACD;QACH;QAEAL,oBAAoB,mCAAmC;IACzD,OACEc,wBAAwB;QACtB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAM;YACR;YACA;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKT;oBACL,QAAQ;gBACV;YACF;SACD;IACH;IAEFL,oBAAoB,MAAM,CAACc;IAC3B,MAAMC,aAAaf,oBAAoB,QAAQ,CAACH,KAAK,kBAAkB;IAEvE,MAAMmB,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASb;QAAa;WACrCU;WACAE;KACJ;IAED,MAAM,EACJ,SAASE,WAAW,EACpBC,KAAK,EACLC,iBAAiB,EAClB,GAAG,MAAMC,OAAOJ,MAAMjB,aAAa;QAClC,WAAWF,AAAmB,YAAnBA,KAAK,SAAS,GAAeT,SAAYS,KAAK,SAAS;IACpE;IAGA,MAAMwB,aAAa/C,yBAAyB2C,aAAazC;IAEzD,MAAM8C,UAAUD,WAAW,MAAM,GAAG;QAACA,WAAW,MAAM;KAAC,GAAG,EAAE;IAC5D,IAAIE,yBAAyB;IAG7B,IAAIF,AAA+BjC,WAA/BiC,WAAW,eAAe,EAAgB;QAC5CjD,MAAM;QACNmD,yBAAyB;IAC3B;IAEA,MAAMC,cAAkC;QACtC,GAAGH,UAAU;QACbC;QACAL;QACAC;QACAC;QACA,UAAUM,uBAAuBH,SAASzB,KAAK,WAAW;QAC1D0B;IACF;IAEAG,OAAOL,YAAY;IAEnBC,QAAQ,OAAO,CAAC,CAAChC;QACf,MAAMC,OAAOD,OAAO,IAAI;QACxB,MAAMqC,sBAAsB9B,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACP,SAAWA,OAAO,IAAI,KAAKC;QAG9BnB,MAAM,+BAA+BuD;QACrC,MAAMC,eAAeD,sBACjBE,4BAA4BF,oBAAoB,WAAW,IAC3D,EAAE;QAENvD,MAAM,gBAAgBwD;QAEtBA,aAAa,OAAO,CAAC,CAACE;YACpB,MAAMC,eAAezC,OAAO,KAAK,CAACwC,MAAM;YACxC,IAAIC,gBAAgBvD,AAAgBY,WAAhBZ,aAElBc,OAAO,KAAK,CAACwC,MAAM,GAAGE,cACpBD,cACAzB,YACAC,aACAC,YACAC,aACAjC;QAGN;IACF;IAEAwB,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAMiB;YACR;SACD;IACH;IAEA,OAAOO;AACT"}