@midscene/core 1.0.1-beta-20251208085324.0 → 1.0.1-beta-20251208112226.0

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

@@ -1 +1 @@
-{"version":3,"file":"agent/tasks.mjs","sources":["../../../src/agent/tasks.ts"],"sourcesContent":["import { ConversationHistory, plan, uiTarsPlanning } from '@/ai-model';\nimport type { TMultimodalPrompt, TUserPrompt } from '@/common';\nimport type { AbstractInterface } from '@/device';\nimport type Service from '@/service';\nimport type { TaskRunner } from '@/task-runner';\nimport { TaskExecutionError } from '@/task-runner';\nimport type {\n  ExecutionTaskApply,\n  ExecutionTaskInsightQueryApply,\n  ExecutionTaskPlanningApply,\n  ExecutionTaskProgressOptions,\n  InterfaceType,\n  MidsceneYamlFlowItem,\n  PlanningAIResponse,\n  PlanningAction,\n  PlanningActionParamSleep,\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 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    },\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.taskBuilder = new TaskBuilder({\n      interfaceInstance,\n      service,\n      taskCache: opts.taskCache,\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  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            more_actions_needed_by_instruction: 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    backgroundKnowledge?: string,\n    cacheable?: boolean,\n    replanningCycleLimitOverride?: number,\n    imagesIncludeCount?: number,\n  ): Promise<\n    ExecutionResult<\n      | {\n          yamlFlow?: MidsceneYamlFlowItem[]; // for cache use\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\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            aiActionContext: backgroundKnowledge,\n            imagesIncludeCount,\n          },\n          executor: async (param, executorContext) => {\n            const startTime = Date.now();\n            const { uiContext } = executorContext;\n            assert(uiContext, 'uiContext is required for Planning task');\n            const { vlMode } = modelConfigForPlanning;\n            const uiTarsModelVersion =\n              vlMode === 'vlm-ui-tars'\n                ? modelConfigForPlanning.uiTarsModelVersion\n                : undefined;\n\n            assert(\n              this.interface.actionSpace,\n              'actionSpace for device is not implemented',\n            );\n            const actionSpace = await this.interface.actionSpace();\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 planResult = await (uiTarsModelVersion\n              ? uiTarsPlanning\n              : plan)(param.userInstruction, {\n              context: uiContext,\n              actionContext: param.aiActionContext,\n              interfaceType: this.interface.interfaceType as InterfaceType,\n              actionSpace,\n              modelConfig: modelConfigForPlanning,\n              conversationHistory: this.conversationHistory,\n              includeBbox: includeBboxInPlanning,\n              imagesIncludeCount,\n            });\n            debug('planResult', JSON.stringify(planResult, null, 2));\n\n            const {\n              actions,\n              log,\n              more_actions_needed_by_instruction,\n              error,\n              usage,\n              rawResponse,\n              sleep,\n            } = planResult;\n\n            executorContext.task.log = {\n              ...(executorContext.task.log || {}),\n              rawResponse,\n            };\n            executorContext.task.usage = usage;\n            executorContext.task.output = {\n              actions: actions || [],\n              more_actions_needed_by_instruction,\n              log,\n              yamlFlow: planResult.yamlFlow,\n            };\n            executorContext.uiContext = uiContext;\n\n            const finalActions = [...(actions || [])];\n\n            if (sleep) {\n              const timeNow = Date.now();\n              const timeRemaining = sleep - (timeNow - startTime);\n              if (timeRemaining > 0) {\n                finalActions.push(this.sleepPlan(timeRemaining));\n              }\n            }\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      let errorFlag = false;\n      try {\n        await session.appendAndRun(executables.tasks);\n      } catch (error: any) {\n        errorFlag = true;\n        errorCountInOnePlanningLoop++;\n        this.conversationHistory.pendingFeedbackMessage = `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?.more_actions_needed_by_instruction) {\n        if (errorFlag) {\n          debug(\n            'more_actions_needed_by_instruction is false, but there are errors in one planning loop, continue to run',\n          );\n        } else {\n          break;\n        }\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 =\n          'I have finished the action previously planned.';\n      }\n    }\n\n    const finalResult = {\n      output: {\n        yamlFlow,\n      },\n      runner,\n    };\n    return finalResult;\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 } = extractResult;\n        applyDump(dump);\n\n        let outputResult = data;\n        if (ifTypeRestricted) {\n          // If AI returned a plain string instead of structured format, use it directly\n          if (typeof data === 'string') {\n            outputResult = data;\n          } else if (type === 'WaitFor') {\n            if (data === null || data === undefined) {\n              outputResult = false;\n            } else {\n              outputResult = (data as any)[keyOfResult];\n            }\n          } else if (data === null || data === undefined) {\n            outputResult = null;\n          } else {\n            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  private sleepPlan(timeMs: number): PlanningAction<PlanningActionParamSleep> {\n    return {\n      type: 'Sleep',\n      param: {\n        timeMs,\n      },\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 { timeoutMs, checkIntervalMs } = opt;\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        undefined,\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"],"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","backgroundKnowledge","cacheable","replanningCycleLimitOverride","imagesIncludeCount","replanCount","yamlFlow","replanningCycleLimit","undefined","errorCountInOnePlanningLoop","startTime","Date","vlMode","uiTarsModelVersion","actionSpace","action","Array","console","planResult","uiTarsPlanning","plan","JSON","actions","log","more_actions_needed_by_instruction","error","usage","rawResponse","sleep","finalActions","timeNow","timeRemaining","executables","errorFlag","String","Error","errorMsg","finalResult","type","demand","modelConfig","opt","multimodalPrompt","queryTask","taskContext","queryDump","applyDump","dump","ifTypeRestricted","demandInput","keyOfResult","booleanPrompt","extractResult","extraPageDescription","tree","descriptionOfTree","ServiceError","data","thought","outputResult","timeMs","_modelConfig","assertion","textPrompt","parsePrompt","description","timeoutMs","checkIntervalMs","overallStartTime","lastCheckStart","errorThought","currentCheckStart","now","sleepTask","interfaceInstance","service","opts","ConversationHistory","TaskBuilder"],"mappings":";;;;;;;;;;;;;;;;;;;;AA8CA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,wCAAwC;AAIvC,MAAMC;IAkBX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IA0BQ,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;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,oCAAoC;wBACpC,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,mBAA4B,EAC5BC,SAAmB,EACnBC,4BAAqC,EACrCC,kBAA2B,EAQ3B;QACA,IAAI,CAAC,mBAAmB,CAAC,KAAK;QAE9B,MAAMhB,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUU;QAEzB,MAAMJ,SAASP,QAAQ,SAAS;QAEhC,IAAIiB,cAAc;QAClB,MAAMC,WAAmC,EAAE;QAC3C,MAAMC,uBACJJ,gCAAgC,IAAI,CAAC,oBAAoB;QAC3DT,OACEa,AAAyBC,WAAzBD,sBACA;QAGF,IAAIE,8BAA8B;QAGlC,MAAO,KAAM;YACX,MAAMZ,SAAS,MAAMT,QAAQ,YAAY,CACvC;gBACE,MAAM;gBACN,SAAS;gBACT,OAAO;oBACL,iBAAiBW;oBACjB,iBAAiBE;oBACjBG;gBACF;gBACA,UAAU,OAAOb,OAAOC;oBACtB,MAAMkB,YAAYC,KAAK,GAAG;oBAC1B,MAAM,EAAElB,SAAS,EAAE,GAAGD;oBACtBE,OAAOD,WAAW;oBAClB,MAAM,EAAEmB,MAAM,EAAE,GAAG5B;oBACnB,MAAM6B,qBACJD,AAAW,kBAAXA,SACI5B,uBAAuB,kBAAkB,GACzCwB;oBAENd,OACE,IAAI,CAAC,SAAS,CAAC,WAAW,EAC1B;oBAEF,MAAMoB,cAAc,MAAM,IAAI,CAAC,SAAS,CAAC,WAAW;oBACpDvC,MACE,sCACAuC,YAAY,GAAG,CAAC,CAACC,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;oBAEhDrB,OAAOsB,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,aAAa,MAAOL,AAAAA,CAAAA,qBACtBM,iBACAC,IAAG,EAAG7B,MAAM,eAAe,EAAE;wBAC/B,SAASE;wBACT,eAAeF,MAAM,eAAe;wBACpC,eAAe,IAAI,CAAC,SAAS,CAAC,aAAa;wBAC3CuB;wBACA,aAAa9B;wBACb,qBAAqB,IAAI,CAAC,mBAAmB;wBAC7C,aAAagB;wBACbI;oBACF;oBACA7B,MAAM,cAAc8C,KAAK,SAAS,CAACH,YAAY,MAAM;oBAErD,MAAM,EACJI,OAAO,EACPC,GAAG,EACHC,kCAAkC,EAClCC,KAAK,EACLC,KAAK,EACLC,WAAW,EACXC,KAAK,EACN,GAAGV;oBAEJ1B,gBAAgB,IAAI,CAAC,GAAG,GAAG;wBACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;wBAClCmC;oBACF;oBACAnC,gBAAgB,IAAI,CAAC,KAAK,GAAGkC;oBAC7BlC,gBAAgB,IAAI,CAAC,MAAM,GAAG;wBAC5B,SAAS8B,WAAW,EAAE;wBACtBE;wBACAD;wBACA,UAAUL,WAAW,QAAQ;oBAC/B;oBACA1B,gBAAgB,SAAS,GAAGC;oBAE5B,MAAMoC,eAAe;2BAAKP,WAAW,EAAE;qBAAE;oBAEzC,IAAIM,OAAO;wBACT,MAAME,UAAUnB,KAAK,GAAG;wBACxB,MAAMoB,gBAAgBH,QAASE,CAAAA,UAAUpB,SAAQ;wBACjD,IAAIqB,gBAAgB,GAClBF,aAAa,IAAI,CAAC,IAAI,CAAC,SAAS,CAACE;oBAErC;oBAEArC,OAAO,CAAC+B,OAAO,CAAC,oBAAoB,EAAEA,MAAM,EAAE,EAAEF,OAAO,IAAI;oBAE3D,OAAO;wBACL,OAAO;4BACL,KAAK;wBACP;oBACF;gBACF;YACF,GACA;gBACE,gBAAgB;YAClB;YAGF,MAAML,aAAarB,QAAQ;YAG3B,MAAMd,QAAQmC,YAAY,WAAW,EAAE;YACvCZ,SAAS,IAAI,IAAKY,YAAY,YAAY,EAAE;YAE5C,IAAIc;YACJ,IAAI;gBACFA,cAAc,MAAM,IAAI,CAAC,uBAAuB,CAC9CjD,OACAC,wBACAC,6BACA;oBACEiB;oBACA,SAAS;gBACX;YAEJ,EAAE,OAAOuB,OAAO;gBACd,OAAOrC,QAAQ,eAAe,CAC5B,CAAC,4CAA4C,EAAEqC,MAAM,SAAS,EAAEJ,KAAK,SAAS,CAC5EtC,QACC;YAEP;YACA,IAAI,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,EACjDkC,QAAQ,IAAI,CACV,8FACA,IAAI,CAAC,mBAAmB,CAAC,sBAAsB;YAGnD,IAAIgB,YAAY;YAChB,IAAI;gBACF,MAAM7C,QAAQ,YAAY,CAAC4C,YAAY,KAAK;YAC9C,EAAE,OAAOP,OAAY;gBACnBQ,YAAY;gBACZxB;gBACA,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,GAAG,CAAC,+BAA+B,EAAEgB,OAAO,WAAWS,OAAOT,QAAQ;gBACrHlD,MACE,yFACAkD,iBAAiBU,QAAQV,MAAM,OAAO,GAAGS,OAAOT,QAChD,6CACAhB;YAEJ;YAEA,IAAIA,8BAA8BhC,uCAChC,OAAOW,QAAQ,eAAe,CAAC;YAIjC,IAAI,CAAC8B,YAAY,oCACf,IAAIe,WACF1D,MACE;iBAGF;YAKJ,EAAE8B;YAEF,IAAIA,cAAcE,sBAAsB;gBACtC,MAAM6B,WAAW,CAAC,UAAU,EAAE7B,qBAAqB,4JAA4J,CAAC;gBAChN,OAAOnB,QAAQ,eAAe,CAACgD;YACjC;YAEA,IAAI,CAAC,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,EAClD,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,GAC7C;QAEN;QAEA,MAAMC,cAAc;YAClB,QAAQ;gBACN/B;YACF;YACAX;QACF;QACA,OAAO0C;IACT;IAEQ,oBACNC,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,OAAOhD,OAAOqD;gBACtB,MAAM,EAAEtD,IAAI,EAAE,GAAGsD;gBACjB,IAAIC;gBACJ,MAAMC,YAAY,CAACC;oBACjBF,YAAYE;oBACZzD,KAAK,GAAG,GAAG;wBACTyD;oBACF;gBACF;gBAGA,MAAMtD,YAAYmD,YAAY,SAAS;gBACvClD,OAAOD,WAAW;gBAElB,MAAMuD,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;oBAC1DlE,MAAM;oBACN,MAAM+E,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,OAAOjB,OAAO;oBACd,IAAIA,iBAAiB+B,cACnBV,UAAUrB,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAEA,MAAM,EAAEgC,IAAI,EAAE/B,KAAK,EAAEgC,OAAO,EAAEX,IAAI,EAAE,GAAGK;gBACvCN,UAAUC;gBAEV,IAAIY,eAAeF;gBACnB,IAAIT,kBAEF,IAAI,AAAgB,YAAhB,OAAOS,MACTE,eAAeF;qBACV,IAAInB,AAAS,cAATA,MAEPqB,eADEF,QAAAA,OACa,QAECA,IAAY,CAACP,YAAY;qBAEtC,IAAIO,QAAAA,MACTE,eAAe;qBACV;oBACLjE,OACE+D,MAAM,CAACP,YAAY,KAAK1C,QACxB;oBAEFmD,eAAgBF,IAAY,CAACP,YAAY;gBAC3C;gBAGF,IAAIZ,AAAS,aAATA,QAAqB,CAACqB,cAAc;oBACtCrE,KAAK,KAAK,GAAGoC;oBACbpC,KAAK,OAAO,GAAGoE;oBACf,MAAM,IAAIvB,MAAM,CAAC,kBAAkB,EAAEuB,SAAS;gBAChD;gBAEA,OAAO;oBACL,QAAQC;oBACR,KAAKd;oBACLnB;oBACAgC;gBACF;YACF;QACF;QAEA,OAAOf;IACT;IACA,MAAM,yBACJL,IAA0D,EAC1DC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACP;QAC7B,MAAMtD,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aACEiD,MACA,AAAkB,YAAlB,OAAOC,SAAsBA,SAASlB,KAAK,SAAS,CAACkB;QAIzD,MAAMI,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9CL,MACAC,QACAC,aACAC,KACAC;QAGF,MAAM/C,SAASP,QAAQ,SAAS;QAChC,MAAMS,SAAS,MAAMT,QAAQ,YAAY,CAACuD;QAE1C,IAAI,CAAC9C,QACH,MAAM,IAAIsC,MACR;QAIJ,MAAM,EAAErC,MAAM,EAAE4D,OAAO,EAAE,GAAG7D;QAE5B,OAAO;YACLC;YACA4D;YACA/D;QACF;IACF;IAEQ,UAAUiE,MAAc,EAA4C;QAC1E,OAAO;YACL,MAAM;YACN,OAAO;gBACLA;YACF;QACF;IACF;IAEA,MAAM,aAAaA,MAAc,EAAEC,YAA0B,EAAE;QAC7D,OAAO,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;YACtCD;QACF;IACF;IAEA,MAAM,QACJE,SAAsB,EACtBrB,GAA+B,EAC/BD,WAAyB,EACO;QAChC,MAAM,EAAEuB,UAAU,EAAErB,gBAAgB,EAAE,GAAGsB,YAAYF;QAErD,MAAMG,cAAc,CAAC,SAAS,EAAEF,YAAY;QAC5C,MAAM3E,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,WAAW4E;QAE1B,MAAMtE,SAASP,QAAQ,SAAS;QAChC,MAAM,EAAE8E,SAAS,EAAEC,eAAe,EAAE,GAAG1B;QAEvC/C,OAAOoE,WAAW;QAClBpE,OAAOwE,WAAW;QAClBxE,OAAOyE,iBAAiB;QAExBzE,OACEyE,mBAAmBD,WACnB,CAAC,iGAAiG,EAAEC,gBAAgB,aAAa,EAAED,UAAU,CAAC,CAAC;QAGjJ,MAAME,mBAAmBzD,KAAK,GAAG;QACjC,IAAI0D,iBAAiBD;QACrB,IAAIE,eAAe;QAEnB,MAAOD,iBAAiBD,oBAAoBF,UAAW;YACrD,MAAMK,oBAAoB5D,KAAK,GAAG;YAClC0D,iBAAiBE;YACjB,MAAM5B,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9C,WACAoB,YACAvB,aACAhC,QACAkC;YAGF,MAAM7C,SAAU,MAAMT,QAAQ,YAAY,CAACuD;YAO3C,IAAI9C,QAAQ,QACV,OAAO;gBACL,QAAQW;gBACRb;YACF;YAGF2E,eACEzE,QAAQ,WACP,CAACA,UAAU,CAAC,0BAA0B,EAAEkE,YAAY,IACrD,CAAC,0CAA0C,EAAEA,YAAY;YAC3D,MAAMS,MAAM7D,KAAK,GAAG;YACpB,IAAI6D,MAAMD,oBAAoBJ,iBAAiB;gBAC7C,MAAMpC,gBAAgBoC,kBAAmBK,CAAAA,MAAMD,iBAAgB;gBAC/D,MAAME,YAAY,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;oBACjD,QAAQ1C;gBACV;gBACA,MAAM3C,QAAQ,MAAM,CAACqF;YACvB;QACF;QAEA,OAAOrF,QAAQ,eAAe,CAAC,CAAC,iBAAiB,EAAEkF,cAAc;IACnE;IAjkBA,YACEI,iBAAoC,EACpCC,OAAgB,EAChBC,IAKC,CACD;QA9BF;QAEA;QAEA;QAEA,uBAAiB,eAAjB;QAEA,uBAAQ,uBAAR;QAEA;QAEA,uBAAiB,SAAjB;QAEA;QAiBE,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,WAAW,GAAG,IAAIC,YAAY;YACjCJ;YACAC;YACA,WAAWC,KAAK,SAAS;QAC3B;IACF;AA4iBF"}
+{"version":3,"file":"agent/tasks.mjs","sources":["../../../src/agent/tasks.ts"],"sourcesContent":["import { ConversationHistory, plan, uiTarsPlanning } from '@/ai-model';\nimport type { TMultimodalPrompt, TUserPrompt } from '@/common';\nimport type { AbstractInterface } from '@/device';\nimport type Service from '@/service';\nimport type { TaskRunner } from '@/task-runner';\nimport { TaskExecutionError } from '@/task-runner';\nimport type {\n  ExecutionTaskApply,\n  ExecutionTaskInsightQueryApply,\n  ExecutionTaskPlanningApply,\n  ExecutionTaskProgressOptions,\n  InterfaceType,\n  MidsceneYamlFlowItem,\n  PlanningAIResponse,\n  PlanningAction,\n  PlanningActionParamSleep,\n  PlanningActionParamWaitFor,\n  ServiceDump,\n  ServiceExtractOption,\n  ServiceExtractParam,\n  DeviceAction,\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            more_actions_needed_by_instruction: 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    backgroundKnowledge?: string,\n    cacheable?: boolean,\n    replanningCycleLimitOverride?: number,\n    imagesIncludeCount?: number,\n  ): Promise<\n    ExecutionResult<\n      | {\n          yamlFlow?: MidsceneYamlFlowItem[]; // for cache use\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\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            aiActionContext: backgroundKnowledge,\n            imagesIncludeCount,\n          },\n          executor: async (param, executorContext) => {\n            const startTime = Date.now();\n            const { uiContext } = executorContext;\n            assert(uiContext, 'uiContext is required for Planning task');\n            const { vlMode } = modelConfigForPlanning;\n            const uiTarsModelVersion =\n              vlMode === 'vlm-ui-tars'\n                ? modelConfigForPlanning.uiTarsModelVersion\n                : undefined;\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 planResult = await (uiTarsModelVersion\n              ? uiTarsPlanning\n              : plan)(param.userInstruction, {\n              context: uiContext,\n              actionContext: param.aiActionContext,\n              interfaceType: this.interface.interfaceType as InterfaceType,\n              actionSpace,\n              modelConfig: modelConfigForPlanning,\n              conversationHistory: this.conversationHistory,\n              includeBbox: includeBboxInPlanning,\n              imagesIncludeCount,\n            });\n            debug('planResult', JSON.stringify(planResult, null, 2));\n\n            const {\n              actions,\n              log,\n              more_actions_needed_by_instruction,\n              error,\n              usage,\n              rawResponse,\n              sleep,\n            } = planResult;\n\n            executorContext.task.log = {\n              ...(executorContext.task.log || {}),\n              rawResponse,\n            };\n            executorContext.task.usage = usage;\n            executorContext.task.output = {\n              actions: actions || [],\n              more_actions_needed_by_instruction,\n              log,\n              yamlFlow: planResult.yamlFlow,\n            };\n            executorContext.uiContext = uiContext;\n\n            const finalActions = [...(actions || [])];\n\n            if (sleep) {\n              const timeNow = Date.now();\n              const timeRemaining = sleep - (timeNow - startTime);\n              if (timeRemaining > 0) {\n                finalActions.push(this.sleepPlan(timeRemaining));\n              }\n            }\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      let errorFlag = false;\n      try {\n        await session.appendAndRun(executables.tasks);\n      } catch (error: any) {\n        errorFlag = true;\n        errorCountInOnePlanningLoop++;\n        this.conversationHistory.pendingFeedbackMessage = `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?.more_actions_needed_by_instruction) {\n        if (errorFlag) {\n          debug(\n            'more_actions_needed_by_instruction is false, but there are errors in one planning loop, continue to run',\n          );\n        } else {\n          break;\n        }\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 =\n          'I have finished the action previously planned.';\n      }\n    }\n\n    const finalResult = {\n      output: {\n        yamlFlow,\n      },\n      runner,\n    };\n    return finalResult;\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 } = extractResult;\n        applyDump(dump);\n\n        let outputResult = data;\n        if (ifTypeRestricted) {\n          // If AI returned a plain string instead of structured format, use it directly\n          if (typeof data === 'string') {\n            outputResult = data;\n          } else if (type === 'WaitFor') {\n            if (data === null || data === undefined) {\n              outputResult = false;\n            } else {\n              outputResult = (data as any)[keyOfResult];\n            }\n          } else if (data === null || data === undefined) {\n            outputResult = null;\n          } else {\n            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  private sleepPlan(timeMs: number): PlanningAction<PlanningActionParamSleep> {\n    return {\n      type: 'Sleep',\n      param: {\n        timeMs,\n      },\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 { timeoutMs, checkIntervalMs } = opt;\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        undefined,\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"],"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","backgroundKnowledge","cacheable","replanningCycleLimitOverride","imagesIncludeCount","replanCount","yamlFlow","replanningCycleLimit","undefined","errorCountInOnePlanningLoop","startTime","Date","vlMode","uiTarsModelVersion","actionSpace","action","Array","console","planResult","uiTarsPlanning","plan","JSON","actions","log","more_actions_needed_by_instruction","error","usage","rawResponse","sleep","finalActions","timeNow","timeRemaining","executables","errorFlag","String","Error","errorMsg","finalResult","type","demand","modelConfig","opt","multimodalPrompt","queryTask","taskContext","queryDump","applyDump","dump","ifTypeRestricted","demandInput","keyOfResult","booleanPrompt","extractResult","extraPageDescription","tree","descriptionOfTree","ServiceError","data","thought","outputResult","timeMs","_modelConfig","assertion","textPrompt","parsePrompt","description","timeoutMs","checkIntervalMs","overallStartTime","lastCheckStart","errorThought","currentCheckStart","now","sleepTask","interfaceInstance","service","opts","ConversationHistory","TaskBuilder"],"mappings":";;;;;;;;;;;;;;;;;;;;AA+CA,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,oCAAoC;wBACpC,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,mBAA4B,EAC5BC,SAAmB,EACnBC,4BAAqC,EACrCC,kBAA2B,EAQ3B;QACA,IAAI,CAAC,mBAAmB,CAAC,KAAK;QAE9B,MAAMhB,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUU;QAEzB,MAAMJ,SAASP,QAAQ,SAAS;QAEhC,IAAIiB,cAAc;QAClB,MAAMC,WAAmC,EAAE;QAC3C,MAAMC,uBACJJ,gCAAgC,IAAI,CAAC,oBAAoB;QAC3DT,OACEa,AAAyBC,WAAzBD,sBACA;QAGF,IAAIE,8BAA8B;QAGlC,MAAO,KAAM;YACX,MAAMZ,SAAS,MAAMT,QAAQ,YAAY,CACvC;gBACE,MAAM;gBACN,SAAS;gBACT,OAAO;oBACL,iBAAiBW;oBACjB,iBAAiBE;oBACjBG;gBACF;gBACA,UAAU,OAAOb,OAAOC;oBACtB,MAAMkB,YAAYC,KAAK,GAAG;oBAC1B,MAAM,EAAElB,SAAS,EAAE,GAAGD;oBACtBE,OAAOD,WAAW;oBAClB,MAAM,EAAEmB,MAAM,EAAE,GAAG5B;oBACnB,MAAM6B,qBACJD,AAAW,kBAAXA,SACI5B,uBAAuB,kBAAkB,GACzCwB;oBAEN,MAAMM,cAAc,IAAI,CAAC,cAAc;oBACvCvC,MACE,sCACAuC,YAAY,GAAG,CAAC,CAACC,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;oBAEhDrB,OAAOsB,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,aAAa,MAAOL,AAAAA,CAAAA,qBACtBM,iBACAC,IAAG,EAAG7B,MAAM,eAAe,EAAE;wBAC/B,SAASE;wBACT,eAAeF,MAAM,eAAe;wBACpC,eAAe,IAAI,CAAC,SAAS,CAAC,aAAa;wBAC3CuB;wBACA,aAAa9B;wBACb,qBAAqB,IAAI,CAAC,mBAAmB;wBAC7C,aAAagB;wBACbI;oBACF;oBACA7B,MAAM,cAAc8C,KAAK,SAAS,CAACH,YAAY,MAAM;oBAErD,MAAM,EACJI,OAAO,EACPC,GAAG,EACHC,kCAAkC,EAClCC,KAAK,EACLC,KAAK,EACLC,WAAW,EACXC,KAAK,EACN,GAAGV;oBAEJ1B,gBAAgB,IAAI,CAAC,GAAG,GAAG;wBACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;wBAClCmC;oBACF;oBACAnC,gBAAgB,IAAI,CAAC,KAAK,GAAGkC;oBAC7BlC,gBAAgB,IAAI,CAAC,MAAM,GAAG;wBAC5B,SAAS8B,WAAW,EAAE;wBACtBE;wBACAD;wBACA,UAAUL,WAAW,QAAQ;oBAC/B;oBACA1B,gBAAgB,SAAS,GAAGC;oBAE5B,MAAMoC,eAAe;2BAAKP,WAAW,EAAE;qBAAE;oBAEzC,IAAIM,OAAO;wBACT,MAAME,UAAUnB,KAAK,GAAG;wBACxB,MAAMoB,gBAAgBH,QAASE,CAAAA,UAAUpB,SAAQ;wBACjD,IAAIqB,gBAAgB,GAClBF,aAAa,IAAI,CAAC,IAAI,CAAC,SAAS,CAACE;oBAErC;oBAEArC,OAAO,CAAC+B,OAAO,CAAC,oBAAoB,EAAEA,MAAM,EAAE,EAAEF,OAAO,IAAI;oBAE3D,OAAO;wBACL,OAAO;4BACL,KAAK;wBACP;oBACF;gBACF;YACF,GACA;gBACE,gBAAgB;YAClB;YAGF,MAAML,aAAarB,QAAQ;YAG3B,MAAMd,QAAQmC,YAAY,WAAW,EAAE;YACvCZ,SAAS,IAAI,IAAKY,YAAY,YAAY,EAAE;YAE5C,IAAIc;YACJ,IAAI;gBACFA,cAAc,MAAM,IAAI,CAAC,uBAAuB,CAC9CjD,OACAC,wBACAC,6BACA;oBACEiB;oBACA,SAAS;gBACX;YAEJ,EAAE,OAAOuB,OAAO;gBACd,OAAOrC,QAAQ,eAAe,CAC5B,CAAC,4CAA4C,EAAEqC,MAAM,SAAS,EAAEJ,KAAK,SAAS,CAC5EtC,QACC;YAEP;YACA,IAAI,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,EACjDkC,QAAQ,IAAI,CACV,8FACA,IAAI,CAAC,mBAAmB,CAAC,sBAAsB;YAGnD,IAAIgB,YAAY;YAChB,IAAI;gBACF,MAAM7C,QAAQ,YAAY,CAAC4C,YAAY,KAAK;YAC9C,EAAE,OAAOP,OAAY;gBACnBQ,YAAY;gBACZxB;gBACA,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,GAAG,CAAC,+BAA+B,EAAEgB,OAAO,WAAWS,OAAOT,QAAQ;gBACrHlD,MACE,yFACAkD,iBAAiBU,QAAQV,MAAM,OAAO,GAAGS,OAAOT,QAChD,6CACAhB;YAEJ;YAEA,IAAIA,8BAA8BhC,uCAChC,OAAOW,QAAQ,eAAe,CAAC;YAIjC,IAAI,CAAC8B,YAAY,oCACf,IAAIe,WACF1D,MACE;iBAGF;YAKJ,EAAE8B;YAEF,IAAIA,cAAcE,sBAAsB;gBACtC,MAAM6B,WAAW,CAAC,UAAU,EAAE7B,qBAAqB,4JAA4J,CAAC;gBAChN,OAAOnB,QAAQ,eAAe,CAACgD;YACjC;YAEA,IAAI,CAAC,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,EAClD,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,GAC7C;QAEN;QAEA,MAAMC,cAAc;YAClB,QAAQ;gBACN/B;YACF;YACAX;QACF;QACA,OAAO0C;IACT;IAEQ,oBACNC,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,OAAOhD,OAAOqD;gBACtB,MAAM,EAAEtD,IAAI,EAAE,GAAGsD;gBACjB,IAAIC;gBACJ,MAAMC,YAAY,CAACC;oBACjBF,YAAYE;oBACZzD,KAAK,GAAG,GAAG;wBACTyD;oBACF;gBACF;gBAGA,MAAMtD,YAAYmD,YAAY,SAAS;gBACvClD,OAAOD,WAAW;gBAElB,MAAMuD,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;oBAC1DlE,MAAM;oBACN,MAAM+E,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,OAAOjB,OAAO;oBACd,IAAIA,iBAAiB+B,cACnBV,UAAUrB,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAEA,MAAM,EAAEgC,IAAI,EAAE/B,KAAK,EAAEgC,OAAO,EAAEX,IAAI,EAAE,GAAGK;gBACvCN,UAAUC;gBAEV,IAAIY,eAAeF;gBACnB,IAAIT,kBAEF,IAAI,AAAgB,YAAhB,OAAOS,MACTE,eAAeF;qBACV,IAAInB,AAAS,cAATA,MAEPqB,eADEF,QAAAA,OACa,QAECA,IAAY,CAACP,YAAY;qBAEtC,IAAIO,QAAAA,MACTE,eAAe;qBACV;oBACLjE,OACE+D,MAAM,CAACP,YAAY,KAAK1C,QACxB;oBAEFmD,eAAgBF,IAAY,CAACP,YAAY;gBAC3C;gBAGF,IAAIZ,AAAS,aAATA,QAAqB,CAACqB,cAAc;oBACtCrE,KAAK,KAAK,GAAGoC;oBACbpC,KAAK,OAAO,GAAGoE;oBACf,MAAM,IAAIvB,MAAM,CAAC,kBAAkB,EAAEuB,SAAS;gBAChD;gBAEA,OAAO;oBACL,QAAQC;oBACR,KAAKd;oBACLnB;oBACAgC;gBACF;YACF;QACF;QAEA,OAAOf;IACT;IACA,MAAM,yBACJL,IAA0D,EAC1DC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACP;QAC7B,MAAMtD,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aACEiD,MACA,AAAkB,YAAlB,OAAOC,SAAsBA,SAASlB,KAAK,SAAS,CAACkB;QAIzD,MAAMI,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9CL,MACAC,QACAC,aACAC,KACAC;QAGF,MAAM/C,SAASP,QAAQ,SAAS;QAChC,MAAMS,SAAS,MAAMT,QAAQ,YAAY,CAACuD;QAE1C,IAAI,CAAC9C,QACH,MAAM,IAAIsC,MACR;QAIJ,MAAM,EAAErC,MAAM,EAAE4D,OAAO,EAAE,GAAG7D;QAE5B,OAAO;YACLC;YACA4D;YACA/D;QACF;IACF;IAEQ,UAAUiE,MAAc,EAA4C;QAC1E,OAAO;YACL,MAAM;YACN,OAAO;gBACLA;YACF;QACF;IACF;IAEA,MAAM,aAAaA,MAAc,EAAEC,YAA0B,EAAE;QAC7D,OAAO,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;YACtCD;QACF;IACF;IAEA,MAAM,QACJE,SAAsB,EACtBrB,GAA+B,EAC/BD,WAAyB,EACO;QAChC,MAAM,EAAEuB,UAAU,EAAErB,gBAAgB,EAAE,GAAGsB,YAAYF;QAErD,MAAMG,cAAc,CAAC,SAAS,EAAEF,YAAY;QAC5C,MAAM3E,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,WAAW4E;QAE1B,MAAMtE,SAASP,QAAQ,SAAS;QAChC,MAAM,EAAE8E,SAAS,EAAEC,eAAe,EAAE,GAAG1B;QAEvC/C,OAAOoE,WAAW;QAClBpE,OAAOwE,WAAW;QAClBxE,OAAOyE,iBAAiB;QAExBzE,OACEyE,mBAAmBD,WACnB,CAAC,iGAAiG,EAAEC,gBAAgB,aAAa,EAAED,UAAU,CAAC,CAAC;QAGjJ,MAAME,mBAAmBzD,KAAK,GAAG;QACjC,IAAI0D,iBAAiBD;QACrB,IAAIE,eAAe;QAEnB,MAAOD,iBAAiBD,oBAAoBF,UAAW;YACrD,MAAMK,oBAAoB5D,KAAK,GAAG;YAClC0D,iBAAiBE;YACjB,MAAM5B,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9C,WACAoB,YACAvB,aACAhC,QACAkC;YAGF,MAAM7C,SAAU,MAAMT,QAAQ,YAAY,CAACuD;YAO3C,IAAI9C,QAAQ,QACV,OAAO;gBACL,QAAQW;gBACRb;YACF;YAGF2E,eACEzE,QAAQ,WACP,CAACA,UAAU,CAAC,0BAA0B,EAAEkE,YAAY,IACrD,CAAC,0CAA0C,EAAEA,YAAY;YAC3D,MAAMS,MAAM7D,KAAK,GAAG;YACpB,IAAI6D,MAAMD,oBAAoBJ,iBAAiB;gBAC7C,MAAMpC,gBAAgBoC,kBAAmBK,CAAAA,MAAMD,iBAAgB;gBAC/D,MAAME,YAAY,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;oBACjD,QAAQ1C;gBACV;gBACA,MAAM3C,QAAQ,MAAM,CAACqF;YACvB;QACF;QAEA,OAAOrF,QAAQ,eAAe,CAAC,CAAC,iBAAiB,EAAEkF,cAAc;IACnE;IApkBA,YACEI,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;AA4iBF"}

package/dist/es/agent/utils.mjs

@@ -100,7 +100,7 @@ async function matchElementFromCache(context, cacheEntry, cachePrompt, cacheable
         return;
     }
 }
-const getMidsceneVersion = ()=>"1.0.1-beta-20251208085324.0";
+const getMidsceneVersion = ()=>"1.0.1-beta-20251208112226.0";
 const parsePrompt = (prompt)=>{
     if ('string' == typeof prompt) return {
         textPrompt: prompt,

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

@@ -134,7 +134,7 @@ Please tell what the next one action is (or null if no action should be done) to
 - Make sure the previous actions are completed successfully before performing the next step
 - If there are some error messages reported by the previous actions, don't give up, try parse a new action to recover. If the error persists for more than 5 times, you should think this is an error and set the "error" field to the error message.
 - If there is nothing to do but waiting, set the "sleep" field to the positive waiting time in milliseconds and null for the "action" field.
-- Assertions are also important steps. When getting the assertion instruction, a solid conclusion is required. You should explicitly state your conclusion in the "log" field like this: "Assert: <condition>. I can see <...>, and I think <...>, so the assertion is <truthy / falsy>". If the assertion is falsy, think this an fatal error and set the reason into the "error" field.
+- Assertions are also important steps. When getting the assertion instruction, a solid conclusion is required. You should explicitly state your conclusion by calling the "Print_Assert_Result" action.
 
 ## Supporting actions
 ${actionList}

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

@@ -1 +1 @@
-{"version":3,"file":"ai-model/prompt/llm-planning.mjs","sources":["../../../../src/ai-model/prompt/llm-planning.ts"],"sourcesContent":["import type { DeviceAction } from '@/types';\nimport type { TVlModeTypes } from '@midscene/shared/env';\nimport type { z } from 'zod';\nimport { ifMidsceneLocatorField } from '../../common';\nimport { bboxDescription } from './common';\n\n// Note: put the log field first to trigger the CoT\n\nconst commonOutputFields = `\"error\"?: string, // Error messages about unexpected situations, if any. Only think it is an error when the situation is not foreseeable according to the instruction. Use the same language as the user's instruction.\n  \"more_actions_needed_by_instruction\": boolean, // Consider if there is still more action(s) to do after the action in \"Log\" is done, according to the instruction. If so, set this field to true. Otherwise, set it to false.`;\n\nconst vlLocateParam = (vlMode: TVlModeTypes | undefined) => {\n  if (vlMode) {\n    return `{bbox: [number, number, number, number], prompt: string } // ${bboxDescription(vlMode)}`;\n  }\n  return '{ prompt: string /* description of the target element */ }';\n};\n\nexport const descriptionForAction = (\n  action: DeviceAction<any>,\n  locatorSchemaTypeDescription: string,\n) => {\n  const tab = '  ';\n  const fields: string[] = [];\n\n  // Add the action type field\n  fields.push(`- type: \"${action.name}\"`);\n\n  // Handle paramSchema if it exists\n  if (action.paramSchema) {\n    const paramLines: string[] = [];\n\n    // Check if paramSchema is a ZodObject with shape\n    const schema = action.paramSchema as any;\n    const isZodObject = schema._def?.typeName === 'ZodObject';\n\n    if (isZodObject && schema.shape) {\n      // Original logic for ZodObject schemas\n      const shape = schema.shape;\n\n      // Helper function to get type name from zod schema\n      const getTypeName = (field: any): string => {\n        // Recursively unwrap optional, nullable, and other wrapper types to get the actual inner type\n        const unwrapField = (f: any): any => {\n          if (!f._def) return f;\n\n          const typeName = f._def.typeName;\n\n          // Handle wrapper types that have innerType\n          if (\n            typeName === 'ZodOptional' ||\n            typeName === 'ZodNullable' ||\n            typeName === 'ZodDefault'\n          ) {\n            return unwrapField(f._def.innerType);\n          }\n\n          // Handle ZodEffects (transformations, refinements, preprocessors)\n          if (typeName === 'ZodEffects') {\n            // For ZodEffects, unwrap the schema field which contains the underlying type\n            if (f._def.schema) {\n              return unwrapField(f._def.schema);\n            }\n          }\n\n          return f;\n        };\n\n        const actualField = unwrapField(field);\n        const fieldTypeName = actualField._def?.typeName;\n\n        if (fieldTypeName === 'ZodString') return 'string';\n        if (fieldTypeName === 'ZodNumber') return 'number';\n        if (fieldTypeName === 'ZodBoolean') return 'boolean';\n        if (fieldTypeName === 'ZodArray') return 'array';\n        if (fieldTypeName === 'ZodObject') {\n          // Check if this is a passthrough object (like MidsceneLocation)\n          if (ifMidsceneLocatorField(actualField)) {\n            return locatorSchemaTypeDescription;\n          }\n          return 'object';\n        }\n        if (fieldTypeName === 'ZodEnum') {\n          const values =\n            (actualField._def?.values as unknown[] | undefined)\n              ?.map((option: unknown) => String(`'${option}'`))\n              .join(', ') ?? 'enum';\n\n          return `enum(${values})`;\n        }\n        // Handle ZodUnion by taking the first option (for display purposes)\n        if (fieldTypeName === 'ZodUnion') {\n          const options = actualField._def?.options as any[] | undefined;\n          if (options && options.length > 0) {\n            // For unions, list all types\n            const types = options.map((opt: any) => getTypeName(opt));\n            return types.join(' | ');\n          }\n          return 'union';\n        }\n\n        console.warn(\n          'failed to parse Zod type. This may lead to wrong params from the LLM.\\n',\n          actualField._def,\n        );\n        return actualField.toString();\n      };\n\n      // Helper function to get description from zod schema\n      const getDescription = (field: z.ZodTypeAny): string | null => {\n        // Recursively unwrap optional, nullable, and other wrapper types to get the actual inner type\n        const unwrapField = (f: any): any => {\n          if (!f._def) return f;\n\n          const typeName = f._def.typeName;\n\n          // Handle wrapper types that have innerType\n          if (\n            typeName === 'ZodOptional' ||\n            typeName === 'ZodNullable' ||\n            typeName === 'ZodDefault'\n          ) {\n            return unwrapField(f._def.innerType);\n          }\n\n          // Handle ZodEffects (transformations, refinements, preprocessors)\n          if (typeName === 'ZodEffects') {\n            // For ZodEffects, unwrap the schema field which contains the underlying type\n            if (f._def.schema) {\n              return unwrapField(f._def.schema);\n            }\n          }\n\n          return f;\n        };\n\n        // Check for direct description on the original field (wrapper may have description)\n        if ('description' in field) {\n          return field.description || null;\n        }\n\n        const actualField = unwrapField(field);\n\n        // Check for description on the unwrapped field\n        if ('description' in actualField) {\n          return actualField.description || null;\n        }\n\n        // Check for MidsceneLocation fields and add description\n        if (actualField._def?.typeName === 'ZodObject') {\n          if ('midscene_location_field_flag' in actualField._def.shape()) {\n            return 'Location information for the target element';\n          }\n        }\n\n        return null;\n      };\n\n      for (const [key, field] of Object.entries(shape)) {\n        if (field && typeof field === 'object') {\n          // Check if field is optional\n          const isOptional =\n            typeof (field as any).isOptional === 'function' &&\n            (field as any).isOptional();\n          const keyWithOptional = isOptional ? `${key}?` : key;\n\n          // Get the type name\n          const typeName = getTypeName(field);\n\n          // Get description\n          const description = getDescription(field as z.ZodTypeAny);\n\n          // Build param line for this field\n          let paramLine = `${keyWithOptional}: ${typeName}`;\n          if (description) {\n            paramLine += ` // ${description}`;\n          }\n\n          paramLines.push(paramLine);\n        }\n      }\n\n      // Add the param section to fields if there are paramLines\n      if (paramLines.length > 0) {\n        fields.push('- param:');\n        paramLines.forEach((line) => {\n          fields.push(`  - ${line}`);\n        });\n      }\n    } else {\n      // Handle non-object schemas (string, number, etc.)\n      // For simple primitive types, the param should be passed directly as the value\n      const schemaTypeName = schema._def?.typeName;\n      let typeName = 'unknown';\n\n      if (schemaTypeName === 'ZodString') typeName = 'string';\n      else if (schemaTypeName === 'ZodNumber') typeName = 'number';\n      else if (schemaTypeName === 'ZodBoolean') typeName = 'boolean';\n\n      // Get description if available\n      const description = 'description' in schema ? schema.description : null;\n\n      // For simple types, indicate that param should be the direct value, not an object\n      let paramDescription = `- param: ${typeName}`;\n      if (description) {\n        paramDescription += ` // ${description}`;\n      }\n      paramDescription += ' (pass the value directly, not as an object)';\n\n      fields.push(paramDescription);\n    }\n  }\n\n  return `- ${action.name}, ${action.description || 'No description provided'}\n${tab}${fields.join(`\\n${tab}`)}\n`.trim();\n};\n\nexport async function systemPromptToTaskPlanning({\n  actionSpace,\n  vlMode,\n  includeBbox,\n}: {\n  actionSpace: DeviceAction<any>[];\n  vlMode: TVlModeTypes | undefined;\n  includeBbox: boolean;\n}) {\n  // Validate parameters: if includeBbox is true, vlMode must be defined\n  if (includeBbox && !vlMode) {\n    throw new Error(\n      'vlMode cannot be undefined when includeBbox is true. A valid vlMode is required for bbox-based location.',\n    );\n  }\n\n  const actionDescriptionList = actionSpace.map((action) => {\n    return descriptionForAction(\n      action,\n      vlLocateParam(includeBbox ? vlMode : undefined),\n    );\n  });\n  const actionList = actionDescriptionList.join('\\n');\n\n  const logFieldInstruction = `\n## About the \\`log\\` field (preamble message)\n\nThe \\`log\\` field is a brief preamble message to the user explaining what you’re about to do. It should follow these principles and examples:\n\n- **Use the same language as the user's instruction**\n- **Keep it concise**: be no more than 1-2 sentences, focused on immediate, tangible next steps. (8–12 words or Chinese characters for quick updates).\n- **Build on prior context**: if this is not the first action to be done, use the preamble message to connect the dots with what’s been done so far and create a sense of momentum and clarity for the user to understand your next actions.\n- **Keep your tone light, friendly and curious**: add small touches of personality in preambles feel collaborative and engaging.\n\n**Examples:**\n- \"Click the login button\"\n- \"Scroll to find the 'Yes' button in popup\"\n- \"Previous actions failed to find the 'Yes' button, i will try again\"\n- \"Go back to find the login button\"\n`;\n\n  return `\nTarget: User will give you an instruction, some screenshots and previous logs indicating what have been done. Your task is to plan the next one action according to current situation to accomplish the instruction.\n\nPlease tell what the next one action is (or null if no action should be done) to do the tasks the instruction requires. \n\n## Rules\n\n- Don't give extra actions or plans beyond the instruction. For example, don't try to submit the form if the instruction is only to fill something.\n- Give just the next ONE action you should do\n- Consider the current screenshot and give the action that is most likely to accomplish the instruction. For example, if the next step is to click a button but it's not visible in the screenshot, you should try to find it first instead of give a click action.\n- Make sure the previous actions are completed successfully before performing the next step\n- If there are some error messages reported by the previous actions, don't give up, try parse a new action to recover. If the error persists for more than 5 times, you should think this is an error and set the \"error\" field to the error message.\n- If there is nothing to do but waiting, set the \"sleep\" field to the positive waiting time in milliseconds and null for the \"action\" field.\n- Assertions are also important steps. When getting the assertion instruction, a solid conclusion is required. You should explicitly state your conclusion in the \"log\" field like this: \"Assert: <condition>. I can see <...>, and I think <...>, so the assertion is <truthy / falsy>\". If the assertion is falsy, think this an fatal error and set the reason into the \"error\" field.\n\n## Supporting actions\n${actionList}\n\n${logFieldInstruction}\n\n## Return format\n\nReturn in JSON format:\n{\n  \"log\": string, // a brief preamble to the user explaining what you’re about to do\n  ${commonOutputFields}\n  \"action\": \n    {\n      \"type\": string, // the type of the action\n      \"param\"?: { // The parameter of the action, if any\n         // k-v style parameter fields\n      }, \n    } | null,\n  ,\n  \"sleep\"?: number, // The sleep time after the action, in milliseconds.\n}\n\nFor example, if the instruction is to login and the form has already been filled, this is a valid return value:\n\n{\n  \"log\": \"Click the login button\",\n  \"more_actions_needed_by_instruction\": false,\n  \"action\": {\n    \"type\": \"Tap\",\n    \"param\": {\n      \"locate\": { \n        \"prompt\": \"The login button\"${vlMode ? `, \"bbox\": [100, 200, 300, 400]` : ''}\n      }\n    }\n  }\n`;\n}\n"],"names":["commonOutputFields","vlLocateParam","vlMode","bboxDescription","descriptionForAction","action","locatorSchemaTypeDescription","tab","fields","paramLines","schema","isZodObject","shape","getTypeName","field","unwrapField","f","typeName","actualField","fieldTypeName","ifMidsceneLocatorField","values","option","String","options","types","opt","console","getDescription","key","Object","isOptional","keyWithOptional","description","paramLine","line","schemaTypeName","paramDescription","systemPromptToTaskPlanning","actionSpace","includeBbox","Error","actionDescriptionList","undefined","actionList","logFieldInstruction"],"mappings":";;AAQA,MAAMA,qBAAqB,CAAC;+NACmM,CAAC;AAEhO,MAAMC,gBAAgB,CAACC;IACrB,IAAIA,QACF,OAAO,CAAC,6DAA6D,EAAEC,gBAAgBD,SAAS;IAElG,OAAO;AACT;AAEO,MAAME,uBAAuB,CAClCC,QACAC;IAEA,MAAMC,MAAM;IACZ,MAAMC,SAAmB,EAAE;IAG3BA,OAAO,IAAI,CAAC,CAAC,SAAS,EAAEH,OAAO,IAAI,CAAC,CAAC,CAAC;IAGtC,IAAIA,OAAO,WAAW,EAAE;QACtB,MAAMI,aAAuB,EAAE;QAG/B,MAAMC,SAASL,OAAO,WAAW;QACjC,MAAMM,cAAcD,OAAO,IAAI,EAAE,aAAa;QAE9C,IAAIC,eAAeD,OAAO,KAAK,EAAE;YAE/B,MAAME,QAAQF,OAAO,KAAK;YAG1B,MAAMG,cAAc,CAACC;gBAEnB,MAAMC,cAAc,CAACC;oBACnB,IAAI,CAACA,EAAE,IAAI,EAAE,OAAOA;oBAEpB,MAAMC,WAAWD,EAAE,IAAI,CAAC,QAAQ;oBAGhC,IACEC,AAAa,kBAAbA,YACAA,AAAa,kBAAbA,YACAA,AAAa,iBAAbA,UAEA,OAAOF,YAAYC,EAAE,IAAI,CAAC,SAAS;oBAIrC,IAAIC,AAAa,iBAAbA,UAEF;wBAAA,IAAID,EAAE,IAAI,CAAC,MAAM,EACf,OAAOD,YAAYC,EAAE,IAAI,CAAC,MAAM;oBAClC;oBAGF,OAAOA;gBACT;gBAEA,MAAME,cAAcH,YAAYD;gBAChC,MAAMK,gBAAgBD,YAAY,IAAI,EAAE;gBAExC,IAAIC,AAAkB,gBAAlBA,eAA+B,OAAO;gBAC1C,IAAIA,AAAkB,gBAAlBA,eAA+B,OAAO;gBAC1C,IAAIA,AAAkB,iBAAlBA,eAAgC,OAAO;gBAC3C,IAAIA,AAAkB,eAAlBA,eAA8B,OAAO;gBACzC,IAAIA,AAAkB,gBAAlBA,eAA+B;oBAEjC,IAAIC,uBAAuBF,cACzB,OAAOZ;oBAET,OAAO;gBACT;gBACA,IAAIa,AAAkB,cAAlBA,eAA6B;oBAC/B,MAAME,SACHH,YAAY,IAAI,EAAE,QACf,IAAI,CAACI,SAAoBC,OAAO,CAAC,CAAC,EAAED,OAAO,CAAC,CAAC,GAC9C,KAAK,SAAS;oBAEnB,OAAO,CAAC,KAAK,EAAED,OAAO,CAAC,CAAC;gBAC1B;gBAEA,IAAIF,AAAkB,eAAlBA,eAA8B;oBAChC,MAAMK,UAAUN,YAAY,IAAI,EAAE;oBAClC,IAAIM,WAAWA,QAAQ,MAAM,GAAG,GAAG;wBAEjC,MAAMC,QAAQD,QAAQ,GAAG,CAAC,CAACE,MAAab,YAAYa;wBACpD,OAAOD,MAAM,IAAI,CAAC;oBACpB;oBACA,OAAO;gBACT;gBAEAE,QAAQ,IAAI,CACV,2EACAT,YAAY,IAAI;gBAElB,OAAOA,YAAY,QAAQ;YAC7B;YAGA,MAAMU,iBAAiB,CAACd;gBAEtB,MAAMC,cAAc,CAACC;oBACnB,IAAI,CAACA,EAAE,IAAI,EAAE,OAAOA;oBAEpB,MAAMC,WAAWD,EAAE,IAAI,CAAC,QAAQ;oBAGhC,IACEC,AAAa,kBAAbA,YACAA,AAAa,kBAAbA,YACAA,AAAa,iBAAbA,UAEA,OAAOF,YAAYC,EAAE,IAAI,CAAC,SAAS;oBAIrC,IAAIC,AAAa,iBAAbA,UAEF;wBAAA,IAAID,EAAE,IAAI,CAAC,MAAM,EACf,OAAOD,YAAYC,EAAE,IAAI,CAAC,MAAM;oBAClC;oBAGF,OAAOA;gBACT;gBAGA,IAAI,iBAAiBF,OACnB,OAAOA,MAAM,WAAW,IAAI;gBAG9B,MAAMI,cAAcH,YAAYD;gBAGhC,IAAI,iBAAiBI,aACnB,OAAOA,YAAY,WAAW,IAAI;gBAIpC,IAAIA,YAAY,IAAI,EAAE,aAAa,aACjC;oBAAA,IAAI,kCAAkCA,YAAY,IAAI,CAAC,KAAK,IAC1D,OAAO;gBACT;gBAGF,OAAO;YACT;YAEA,KAAK,MAAM,CAACW,KAAKf,MAAM,IAAIgB,OAAO,OAAO,CAAClB,OACxC,IAAIE,SAAS,AAAiB,YAAjB,OAAOA,OAAoB;gBAEtC,MAAMiB,aACJ,AAAqC,cAArC,OAAQjB,MAAc,UAAU,IAC/BA,MAAc,UAAU;gBAC3B,MAAMkB,kBAAkBD,aAAa,GAAGF,IAAI,CAAC,CAAC,GAAGA;gBAGjD,MAAMZ,WAAWJ,YAAYC;gBAG7B,MAAMmB,cAAcL,eAAed;gBAGnC,IAAIoB,YAAY,GAAGF,gBAAgB,EAAE,EAAEf,UAAU;gBACjD,IAAIgB,aACFC,aAAa,CAAC,IAAI,EAAED,aAAa;gBAGnCxB,WAAW,IAAI,CAACyB;YAClB;YAIF,IAAIzB,WAAW,MAAM,GAAG,GAAG;gBACzBD,OAAO,IAAI,CAAC;gBACZC,WAAW,OAAO,CAAC,CAAC0B;oBAClB3B,OAAO,IAAI,CAAC,CAAC,IAAI,EAAE2B,MAAM;gBAC3B;YACF;QACF,OAAO;YAGL,MAAMC,iBAAiB1B,OAAO,IAAI,EAAE;YACpC,IAAIO,WAAW;YAEf,IAAImB,AAAmB,gBAAnBA,gBAAgCnB,WAAW;iBAC1C,IAAImB,AAAmB,gBAAnBA,gBAAgCnB,WAAW;iBAC/C,IAAImB,AAAmB,iBAAnBA,gBAAiCnB,WAAW;YAGrD,MAAMgB,cAAc,iBAAiBvB,SAASA,OAAO,WAAW,GAAG;YAGnE,IAAI2B,mBAAmB,CAAC,SAAS,EAAEpB,UAAU;YAC7C,IAAIgB,aACFI,oBAAoB,CAAC,IAAI,EAAEJ,aAAa;YAE1CI,oBAAoB;YAEpB7B,OAAO,IAAI,CAAC6B;QACd;IACF;IAEA,OAAO,CAAC,EAAE,EAAEhC,OAAO,IAAI,CAAC,EAAE,EAAEA,OAAO,WAAW,IAAI,0BAA0B;AAC9E,EAAEE,MAAMC,OAAO,IAAI,CAAC,CAAC,EAAE,EAAED,KAAK,EAAE;AAChC,CAAC,CAAC,IAAI;AACN;AAEO,eAAe+B,2BAA2B,EAC/CC,WAAW,EACXrC,MAAM,EACNsC,WAAW,EAKZ;IAEC,IAAIA,eAAe,CAACtC,QAClB,MAAM,IAAIuC,MACR;IAIJ,MAAMC,wBAAwBH,YAAY,GAAG,CAAC,CAAClC,SACtCD,qBACLC,QACAJ,cAAcuC,cAActC,SAASyC;IAGzC,MAAMC,aAAaF,sBAAsB,IAAI,CAAC;IAE9C,MAAMG,sBAAsB,CAAC;;;;;;;;;;;;;;;AAe/B,CAAC;IAEC,OAAO,CAAC;;;;;;;;;;;;;;;;AAgBV,EAAED,WAAW;;AAEb,EAAEC,oBAAoB;;;;;;;EAOpB,EAAE7C,mBAAmB;;;;;;;;;;;;;;;;;;;;;oCAqBa,EAAEE,SAAS,mCAAmC,GAAG;;;;AAIrF,CAAC;AACD"}
+{"version":3,"file":"ai-model/prompt/llm-planning.mjs","sources":["../../../../src/ai-model/prompt/llm-planning.ts"],"sourcesContent":["import type { DeviceAction } from '@/types';\nimport type { TVlModeTypes } from '@midscene/shared/env';\nimport type { z } from 'zod';\nimport { ifMidsceneLocatorField } from '../../common';\nimport { bboxDescription } from './common';\n\n// Note: put the log field first to trigger the CoT\n\nconst commonOutputFields = `\"error\"?: string, // Error messages about unexpected situations, if any. Only think it is an error when the situation is not foreseeable according to the instruction. Use the same language as the user's instruction.\n  \"more_actions_needed_by_instruction\": boolean, // Consider if there is still more action(s) to do after the action in \"Log\" is done, according to the instruction. If so, set this field to true. Otherwise, set it to false.`;\n\nconst vlLocateParam = (vlMode: TVlModeTypes | undefined) => {\n  if (vlMode) {\n    return `{bbox: [number, number, number, number], prompt: string } // ${bboxDescription(vlMode)}`;\n  }\n  return '{ prompt: string /* description of the target element */ }';\n};\n\nexport const descriptionForAction = (\n  action: DeviceAction<any>,\n  locatorSchemaTypeDescription: string,\n) => {\n  const tab = '  ';\n  const fields: string[] = [];\n\n  // Add the action type field\n  fields.push(`- type: \"${action.name}\"`);\n\n  // Handle paramSchema if it exists\n  if (action.paramSchema) {\n    const paramLines: string[] = [];\n\n    // Check if paramSchema is a ZodObject with shape\n    const schema = action.paramSchema as any;\n    const isZodObject = schema._def?.typeName === 'ZodObject';\n\n    if (isZodObject && schema.shape) {\n      // Original logic for ZodObject schemas\n      const shape = schema.shape;\n\n      // Helper function to get type name from zod schema\n      const getTypeName = (field: any): string => {\n        // Recursively unwrap optional, nullable, and other wrapper types to get the actual inner type\n        const unwrapField = (f: any): any => {\n          if (!f._def) return f;\n\n          const typeName = f._def.typeName;\n\n          // Handle wrapper types that have innerType\n          if (\n            typeName === 'ZodOptional' ||\n            typeName === 'ZodNullable' ||\n            typeName === 'ZodDefault'\n          ) {\n            return unwrapField(f._def.innerType);\n          }\n\n          // Handle ZodEffects (transformations, refinements, preprocessors)\n          if (typeName === 'ZodEffects') {\n            // For ZodEffects, unwrap the schema field which contains the underlying type\n            if (f._def.schema) {\n              return unwrapField(f._def.schema);\n            }\n          }\n\n          return f;\n        };\n\n        const actualField = unwrapField(field);\n        const fieldTypeName = actualField._def?.typeName;\n\n        if (fieldTypeName === 'ZodString') return 'string';\n        if (fieldTypeName === 'ZodNumber') return 'number';\n        if (fieldTypeName === 'ZodBoolean') return 'boolean';\n        if (fieldTypeName === 'ZodArray') return 'array';\n        if (fieldTypeName === 'ZodObject') {\n          // Check if this is a passthrough object (like MidsceneLocation)\n          if (ifMidsceneLocatorField(actualField)) {\n            return locatorSchemaTypeDescription;\n          }\n          return 'object';\n        }\n        if (fieldTypeName === 'ZodEnum') {\n          const values =\n            (actualField._def?.values as unknown[] | undefined)\n              ?.map((option: unknown) => String(`'${option}'`))\n              .join(', ') ?? 'enum';\n\n          return `enum(${values})`;\n        }\n        // Handle ZodUnion by taking the first option (for display purposes)\n        if (fieldTypeName === 'ZodUnion') {\n          const options = actualField._def?.options as any[] | undefined;\n          if (options && options.length > 0) {\n            // For unions, list all types\n            const types = options.map((opt: any) => getTypeName(opt));\n            return types.join(' | ');\n          }\n          return 'union';\n        }\n\n        console.warn(\n          'failed to parse Zod type. This may lead to wrong params from the LLM.\\n',\n          actualField._def,\n        );\n        return actualField.toString();\n      };\n\n      // Helper function to get description from zod schema\n      const getDescription = (field: z.ZodTypeAny): string | null => {\n        // Recursively unwrap optional, nullable, and other wrapper types to get the actual inner type\n        const unwrapField = (f: any): any => {\n          if (!f._def) return f;\n\n          const typeName = f._def.typeName;\n\n          // Handle wrapper types that have innerType\n          if (\n            typeName === 'ZodOptional' ||\n            typeName === 'ZodNullable' ||\n            typeName === 'ZodDefault'\n          ) {\n            return unwrapField(f._def.innerType);\n          }\n\n          // Handle ZodEffects (transformations, refinements, preprocessors)\n          if (typeName === 'ZodEffects') {\n            // For ZodEffects, unwrap the schema field which contains the underlying type\n            if (f._def.schema) {\n              return unwrapField(f._def.schema);\n            }\n          }\n\n          return f;\n        };\n\n        // Check for direct description on the original field (wrapper may have description)\n        if ('description' in field) {\n          return field.description || null;\n        }\n\n        const actualField = unwrapField(field);\n\n        // Check for description on the unwrapped field\n        if ('description' in actualField) {\n          return actualField.description || null;\n        }\n\n        // Check for MidsceneLocation fields and add description\n        if (actualField._def?.typeName === 'ZodObject') {\n          if ('midscene_location_field_flag' in actualField._def.shape()) {\n            return 'Location information for the target element';\n          }\n        }\n\n        return null;\n      };\n\n      for (const [key, field] of Object.entries(shape)) {\n        if (field && typeof field === 'object') {\n          // Check if field is optional\n          const isOptional =\n            typeof (field as any).isOptional === 'function' &&\n            (field as any).isOptional();\n          const keyWithOptional = isOptional ? `${key}?` : key;\n\n          // Get the type name\n          const typeName = getTypeName(field);\n\n          // Get description\n          const description = getDescription(field as z.ZodTypeAny);\n\n          // Build param line for this field\n          let paramLine = `${keyWithOptional}: ${typeName}`;\n          if (description) {\n            paramLine += ` // ${description}`;\n          }\n\n          paramLines.push(paramLine);\n        }\n      }\n\n      // Add the param section to fields if there are paramLines\n      if (paramLines.length > 0) {\n        fields.push('- param:');\n        paramLines.forEach((line) => {\n          fields.push(`  - ${line}`);\n        });\n      }\n    } else {\n      // Handle non-object schemas (string, number, etc.)\n      // For simple primitive types, the param should be passed directly as the value\n      const schemaTypeName = schema._def?.typeName;\n      let typeName = 'unknown';\n\n      if (schemaTypeName === 'ZodString') typeName = 'string';\n      else if (schemaTypeName === 'ZodNumber') typeName = 'number';\n      else if (schemaTypeName === 'ZodBoolean') typeName = 'boolean';\n\n      // Get description if available\n      const description = 'description' in schema ? schema.description : null;\n\n      // For simple types, indicate that param should be the direct value, not an object\n      let paramDescription = `- param: ${typeName}`;\n      if (description) {\n        paramDescription += ` // ${description}`;\n      }\n      paramDescription += ' (pass the value directly, not as an object)';\n\n      fields.push(paramDescription);\n    }\n  }\n\n  return `- ${action.name}, ${action.description || 'No description provided'}\n${tab}${fields.join(`\\n${tab}`)}\n`.trim();\n};\n\nexport async function systemPromptToTaskPlanning({\n  actionSpace,\n  vlMode,\n  includeBbox,\n}: {\n  actionSpace: DeviceAction<any>[];\n  vlMode: TVlModeTypes | undefined;\n  includeBbox: boolean;\n}) {\n  // Validate parameters: if includeBbox is true, vlMode must be defined\n  if (includeBbox && !vlMode) {\n    throw new Error(\n      'vlMode cannot be undefined when includeBbox is true. A valid vlMode is required for bbox-based location.',\n    );\n  }\n\n  const actionDescriptionList = actionSpace.map((action) => {\n    return descriptionForAction(\n      action,\n      vlLocateParam(includeBbox ? vlMode : undefined),\n    );\n  });\n  const actionList = actionDescriptionList.join('\\n');\n\n  const logFieldInstruction = `\n## About the \\`log\\` field (preamble message)\n\nThe \\`log\\` field is a brief preamble message to the user explaining what you’re about to do. It should follow these principles and examples:\n\n- **Use the same language as the user's instruction**\n- **Keep it concise**: be no more than 1-2 sentences, focused on immediate, tangible next steps. (8–12 words or Chinese characters for quick updates).\n- **Build on prior context**: if this is not the first action to be done, use the preamble message to connect the dots with what’s been done so far and create a sense of momentum and clarity for the user to understand your next actions.\n- **Keep your tone light, friendly and curious**: add small touches of personality in preambles feel collaborative and engaging.\n\n**Examples:**\n- \"Click the login button\"\n- \"Scroll to find the 'Yes' button in popup\"\n- \"Previous actions failed to find the 'Yes' button, i will try again\"\n- \"Go back to find the login button\"\n`;\n\n  return `\nTarget: User will give you an instruction, some screenshots and previous logs indicating what have been done. Your task is to plan the next one action according to current situation to accomplish the instruction.\n\nPlease tell what the next one action is (or null if no action should be done) to do the tasks the instruction requires. \n\n## Rules\n\n- Don't give extra actions or plans beyond the instruction. For example, don't try to submit the form if the instruction is only to fill something.\n- Give just the next ONE action you should do\n- Consider the current screenshot and give the action that is most likely to accomplish the instruction. For example, if the next step is to click a button but it's not visible in the screenshot, you should try to find it first instead of give a click action.\n- Make sure the previous actions are completed successfully before performing the next step\n- If there are some error messages reported by the previous actions, don't give up, try parse a new action to recover. If the error persists for more than 5 times, you should think this is an error and set the \"error\" field to the error message.\n- If there is nothing to do but waiting, set the \"sleep\" field to the positive waiting time in milliseconds and null for the \"action\" field.\n- Assertions are also important steps. When getting the assertion instruction, a solid conclusion is required. You should explicitly state your conclusion by calling the \"Print_Assert_Result\" action.\n\n## Supporting actions\n${actionList}\n\n${logFieldInstruction}\n\n## Return format\n\nReturn in JSON format:\n{\n  \"log\": string, // a brief preamble to the user explaining what you’re about to do\n  ${commonOutputFields}\n  \"action\": \n    {\n      \"type\": string, // the type of the action\n      \"param\"?: { // The parameter of the action, if any\n         // k-v style parameter fields\n      }, \n    } | null,\n  ,\n  \"sleep\"?: number, // The sleep time after the action, in milliseconds.\n}\n\nFor example, if the instruction is to login and the form has already been filled, this is a valid return value:\n\n{\n  \"log\": \"Click the login button\",\n  \"more_actions_needed_by_instruction\": false,\n  \"action\": {\n    \"type\": \"Tap\",\n    \"param\": {\n      \"locate\": { \n        \"prompt\": \"The login button\"${vlMode ? `, \"bbox\": [100, 200, 300, 400]` : ''}\n      }\n    }\n  }\n`;\n}\n"],"names":["commonOutputFields","vlLocateParam","vlMode","bboxDescription","descriptionForAction","action","locatorSchemaTypeDescription","tab","fields","paramLines","schema","isZodObject","shape","getTypeName","field","unwrapField","f","typeName","actualField","fieldTypeName","ifMidsceneLocatorField","values","option","String","options","types","opt","console","getDescription","key","Object","isOptional","keyWithOptional","description","paramLine","line","schemaTypeName","paramDescription","systemPromptToTaskPlanning","actionSpace","includeBbox","Error","actionDescriptionList","undefined","actionList","logFieldInstruction"],"mappings":";;AAQA,MAAMA,qBAAqB,CAAC;+NACmM,CAAC;AAEhO,MAAMC,gBAAgB,CAACC;IACrB,IAAIA,QACF,OAAO,CAAC,6DAA6D,EAAEC,gBAAgBD,SAAS;IAElG,OAAO;AACT;AAEO,MAAME,uBAAuB,CAClCC,QACAC;IAEA,MAAMC,MAAM;IACZ,MAAMC,SAAmB,EAAE;IAG3BA,OAAO,IAAI,CAAC,CAAC,SAAS,EAAEH,OAAO,IAAI,CAAC,CAAC,CAAC;IAGtC,IAAIA,OAAO,WAAW,EAAE;QACtB,MAAMI,aAAuB,EAAE;QAG/B,MAAMC,SAASL,OAAO,WAAW;QACjC,MAAMM,cAAcD,OAAO,IAAI,EAAE,aAAa;QAE9C,IAAIC,eAAeD,OAAO,KAAK,EAAE;YAE/B,MAAME,QAAQF,OAAO,KAAK;YAG1B,MAAMG,cAAc,CAACC;gBAEnB,MAAMC,cAAc,CAACC;oBACnB,IAAI,CAACA,EAAE,IAAI,EAAE,OAAOA;oBAEpB,MAAMC,WAAWD,EAAE,IAAI,CAAC,QAAQ;oBAGhC,IACEC,AAAa,kBAAbA,YACAA,AAAa,kBAAbA,YACAA,AAAa,iBAAbA,UAEA,OAAOF,YAAYC,EAAE,IAAI,CAAC,SAAS;oBAIrC,IAAIC,AAAa,iBAAbA,UAEF;wBAAA,IAAID,EAAE,IAAI,CAAC,MAAM,EACf,OAAOD,YAAYC,EAAE,IAAI,CAAC,MAAM;oBAClC;oBAGF,OAAOA;gBACT;gBAEA,MAAME,cAAcH,YAAYD;gBAChC,MAAMK,gBAAgBD,YAAY,IAAI,EAAE;gBAExC,IAAIC,AAAkB,gBAAlBA,eAA+B,OAAO;gBAC1C,IAAIA,AAAkB,gBAAlBA,eAA+B,OAAO;gBAC1C,IAAIA,AAAkB,iBAAlBA,eAAgC,OAAO;gBAC3C,IAAIA,AAAkB,eAAlBA,eAA8B,OAAO;gBACzC,IAAIA,AAAkB,gBAAlBA,eAA+B;oBAEjC,IAAIC,uBAAuBF,cACzB,OAAOZ;oBAET,OAAO;gBACT;gBACA,IAAIa,AAAkB,cAAlBA,eAA6B;oBAC/B,MAAME,SACHH,YAAY,IAAI,EAAE,QACf,IAAI,CAACI,SAAoBC,OAAO,CAAC,CAAC,EAAED,OAAO,CAAC,CAAC,GAC9C,KAAK,SAAS;oBAEnB,OAAO,CAAC,KAAK,EAAED,OAAO,CAAC,CAAC;gBAC1B;gBAEA,IAAIF,AAAkB,eAAlBA,eAA8B;oBAChC,MAAMK,UAAUN,YAAY,IAAI,EAAE;oBAClC,IAAIM,WAAWA,QAAQ,MAAM,GAAG,GAAG;wBAEjC,MAAMC,QAAQD,QAAQ,GAAG,CAAC,CAACE,MAAab,YAAYa;wBACpD,OAAOD,MAAM,IAAI,CAAC;oBACpB;oBACA,OAAO;gBACT;gBAEAE,QAAQ,IAAI,CACV,2EACAT,YAAY,IAAI;gBAElB,OAAOA,YAAY,QAAQ;YAC7B;YAGA,MAAMU,iBAAiB,CAACd;gBAEtB,MAAMC,cAAc,CAACC;oBACnB,IAAI,CAACA,EAAE,IAAI,EAAE,OAAOA;oBAEpB,MAAMC,WAAWD,EAAE,IAAI,CAAC,QAAQ;oBAGhC,IACEC,AAAa,kBAAbA,YACAA,AAAa,kBAAbA,YACAA,AAAa,iBAAbA,UAEA,OAAOF,YAAYC,EAAE,IAAI,CAAC,SAAS;oBAIrC,IAAIC,AAAa,iBAAbA,UAEF;wBAAA,IAAID,EAAE,IAAI,CAAC,MAAM,EACf,OAAOD,YAAYC,EAAE,IAAI,CAAC,MAAM;oBAClC;oBAGF,OAAOA;gBACT;gBAGA,IAAI,iBAAiBF,OACnB,OAAOA,MAAM,WAAW,IAAI;gBAG9B,MAAMI,cAAcH,YAAYD;gBAGhC,IAAI,iBAAiBI,aACnB,OAAOA,YAAY,WAAW,IAAI;gBAIpC,IAAIA,YAAY,IAAI,EAAE,aAAa,aACjC;oBAAA,IAAI,kCAAkCA,YAAY,IAAI,CAAC,KAAK,IAC1D,OAAO;gBACT;gBAGF,OAAO;YACT;YAEA,KAAK,MAAM,CAACW,KAAKf,MAAM,IAAIgB,OAAO,OAAO,CAAClB,OACxC,IAAIE,SAAS,AAAiB,YAAjB,OAAOA,OAAoB;gBAEtC,MAAMiB,aACJ,AAAqC,cAArC,OAAQjB,MAAc,UAAU,IAC/BA,MAAc,UAAU;gBAC3B,MAAMkB,kBAAkBD,aAAa,GAAGF,IAAI,CAAC,CAAC,GAAGA;gBAGjD,MAAMZ,WAAWJ,YAAYC;gBAG7B,MAAMmB,cAAcL,eAAed;gBAGnC,IAAIoB,YAAY,GAAGF,gBAAgB,EAAE,EAAEf,UAAU;gBACjD,IAAIgB,aACFC,aAAa,CAAC,IAAI,EAAED,aAAa;gBAGnCxB,WAAW,IAAI,CAACyB;YAClB;YAIF,IAAIzB,WAAW,MAAM,GAAG,GAAG;gBACzBD,OAAO,IAAI,CAAC;gBACZC,WAAW,OAAO,CAAC,CAAC0B;oBAClB3B,OAAO,IAAI,CAAC,CAAC,IAAI,EAAE2B,MAAM;gBAC3B;YACF;QACF,OAAO;YAGL,MAAMC,iBAAiB1B,OAAO,IAAI,EAAE;YACpC,IAAIO,WAAW;YAEf,IAAImB,AAAmB,gBAAnBA,gBAAgCnB,WAAW;iBAC1C,IAAImB,AAAmB,gBAAnBA,gBAAgCnB,WAAW;iBAC/C,IAAImB,AAAmB,iBAAnBA,gBAAiCnB,WAAW;YAGrD,MAAMgB,cAAc,iBAAiBvB,SAASA,OAAO,WAAW,GAAG;YAGnE,IAAI2B,mBAAmB,CAAC,SAAS,EAAEpB,UAAU;YAC7C,IAAIgB,aACFI,oBAAoB,CAAC,IAAI,EAAEJ,aAAa;YAE1CI,oBAAoB;YAEpB7B,OAAO,IAAI,CAAC6B;QACd;IACF;IAEA,OAAO,CAAC,EAAE,EAAEhC,OAAO,IAAI,CAAC,EAAE,EAAEA,OAAO,WAAW,IAAI,0BAA0B;AAC9E,EAAEE,MAAMC,OAAO,IAAI,CAAC,CAAC,EAAE,EAAED,KAAK,EAAE;AAChC,CAAC,CAAC,IAAI;AACN;AAEO,eAAe+B,2BAA2B,EAC/CC,WAAW,EACXrC,MAAM,EACNsC,WAAW,EAKZ;IAEC,IAAIA,eAAe,CAACtC,QAClB,MAAM,IAAIuC,MACR;IAIJ,MAAMC,wBAAwBH,YAAY,GAAG,CAAC,CAAClC,SACtCD,qBACLC,QACAJ,cAAcuC,cAActC,SAASyC;IAGzC,MAAMC,aAAaF,sBAAsB,IAAI,CAAC;IAE9C,MAAMG,sBAAsB,CAAC;;;;;;;;;;;;;;;AAe/B,CAAC;IAEC,OAAO,CAAC;;;;;;;;;;;;;;;;AAgBV,EAAED,WAAW;;AAEb,EAAEC,oBAAoB;;;;;;;EAOpB,EAAE7C,mBAAmB;;;;;;;;;;;;;;;;;;;;;oCAqBa,EAAEE,SAAS,mCAAmC,GAAG;;;;AAIrF,CAAC;AACD"}

package/dist/es/device/index.mjs

@@ -1,4 +1,5 @@
 import { getMidsceneLocationSchema } from "../common.mjs";
+import { getDebug } from "@midscene/shared/logger";
 import { z } from "zod";
 class AbstractInterface {
 }
@@ -149,14 +150,19 @@ const defineActionClearInput = (call)=>defineAction({
         call
     });
 const actionAssertParamSchema = z.object({
-    condition: z.string().describe('The condition of the assertion')
+    condition: z.string().describe('The condition of the assertion'),
+    thought: z.string().describe('The thought of the assertion, like "I can see there are A, B, C elements on the page, which means ... , so the assertion is true"'),
+    result: z.boolean().describe('The result of the assertion, true or false')
 });
-const defineActionAssert = (call)=>defineAction({
-        name: 'Assert',
-        description: 'If the user explicitly requires making an assertion (like "there should be a button with text "Yes" in the popup"), this tool will think about it, and then provide a solid result',
-        interfaceAlias: 'aiAssert',
+const defineActionAssert = ()=>defineAction({
+        name: 'Print_Assert_Result',
+        description: 'Print the result of the assertion',
         paramSchema: actionAssertParamSchema,
-        call
+        call: async (param)=>{
+            if ('boolean' != typeof param?.result) throw new Error(`The result of the assertion must be a boolean, but got: ${typeof param?.result}. ${param.thought || '(no thought)'}`);
+            getDebug('device:common-action')(`Assert: ${param.condition}, Thought: ${param.thought}, Result: ${param.result}`);
+            if (!param.result) throw new Error(`Assertion failed: ${param.thought || '(no thought)'}. (Assertion = ${param.condition})`);
+        }
     });
 export { AbstractInterface, ActionLongPressParamSchema, ActionSwipeParamSchema, actionAssertParamSchema, actionClearInputParamSchema, actionDoubleClickParamSchema, actionDragAndDropParamSchema, actionHoverParamSchema, actionInputParamSchema, actionKeyboardPressParamSchema, actionRightClickParamSchema, actionScrollParamSchema, actionTapParamSchema, defineAction, defineActionAssert, defineActionClearInput, defineActionDoubleClick, defineActionDragAndDrop, defineActionHover, defineActionInput, defineActionKeyboardPress, defineActionLongPress, defineActionRightClick, defineActionScroll, defineActionSwipe, defineActionTap };
 

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

@@ -1 +1 @@
-{"version":3,"file":"device/index.mjs","sources":["../../../src/device/index.ts"],"sourcesContent":["import { getMidsceneLocationSchema } from '@/common';\nimport type {\n  ActionScrollParam,\n  DeviceAction,\n  LocateResultElement,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport type { ElementNode } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { z } from 'zod';\nimport type { ElementCacheFeature, Rect, Size, UIContext } from '../types';\n\nexport abstract class AbstractInterface {\n  abstract interfaceType: string;\n\n  abstract screenshotBase64(): Promise<string>;\n  abstract size(): Promise<Size>;\n  abstract actionSpace(): DeviceAction[] | Promise<DeviceAction[]>;\n\n  abstract cacheFeatureForRect?(\n    rect: Rect,\n    options?: {\n      targetDescription?: string;\n      modelConfig?: IModelConfig;\n    },\n  ): Promise<ElementCacheFeature>;\n  abstract rectMatchesCacheFeature?(\n    feature: ElementCacheFeature,\n  ): Promise<Rect>;\n\n  abstract destroy?(): Promise<void>;\n\n  abstract describe?(): string;\n  abstract beforeInvokeAction?(actionName: string, param: any): Promise<void>;\n  abstract afterInvokeAction?(actionName: string, param: any): Promise<void>;\n\n  // @deprecated do NOT extend this method\n  abstract getElementsNodeTree?: () => Promise<ElementNode>;\n\n  // @deprecated do NOT extend this method\n  abstract url?: () => string | Promise<string>;\n\n  // @deprecated do NOT extend this method\n  abstract evaluateJavaScript?<T = any>(script: string): Promise<T>;\n\n  // @deprecated do NOT extend this method\n  abstract getContext?(): Promise<UIContext>;\n}\n\n// Generic function to define actions with proper type inference\n// TRuntime allows specifying a different type for the runtime parameter (after location resolution)\n// TReturn allows specifying the return type of the action\nexport const defineAction = <\n  TSchema extends z.ZodType | undefined = undefined,\n  TRuntime = TSchema extends z.ZodType ? z.infer<TSchema> : undefined,\n  TReturn = any,\n>(\n  config: {\n    name: string;\n    description: string;\n    interfaceAlias?: string;\n    paramSchema?: TSchema;\n    call: (param: TRuntime) => Promise<TReturn> | TReturn;\n  } & Partial<\n    Omit<\n      DeviceAction<TRuntime, TReturn>,\n      'name' | 'description' | 'interfaceAlias' | 'paramSchema' | 'call'\n    >\n  >,\n): DeviceAction<TRuntime, TReturn> => {\n  return config as any; // Type assertion needed because schema validation type differs from runtime type\n};\n\n// Tap\nexport const actionTapParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The element to be tapped'),\n});\n// Override the inferred type to use LocateResultElement for the runtime locate field\nexport type ActionTapParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionTap = (\n  call: (param: ActionTapParam) => Promise<void>,\n): DeviceAction<ActionTapParam> => {\n  return defineAction<typeof actionTapParamSchema, ActionTapParam>({\n    name: 'Tap',\n    description: 'Tap the element',\n    interfaceAlias: 'aiTap',\n    paramSchema: actionTapParamSchema,\n    call,\n  });\n};\n\n// RightClick\nexport const actionRightClickParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be right clicked',\n  ),\n});\nexport type ActionRightClickParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionRightClick = (\n  call: (param: ActionRightClickParam) => Promise<void>,\n): DeviceAction<ActionRightClickParam> => {\n  return defineAction<\n    typeof actionRightClickParamSchema,\n    ActionRightClickParam\n  >({\n    name: 'RightClick',\n    description: 'Right click the element',\n    interfaceAlias: 'aiRightClick',\n    paramSchema: actionRightClickParamSchema,\n    call,\n  });\n};\n\n// DoubleClick\nexport const actionDoubleClickParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be double clicked',\n  ),\n});\nexport type ActionDoubleClickParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionDoubleClick = (\n  call: (param: ActionDoubleClickParam) => Promise<void>,\n): DeviceAction<ActionDoubleClickParam> => {\n  return defineAction<\n    typeof actionDoubleClickParamSchema,\n    ActionDoubleClickParam\n  >({\n    name: 'DoubleClick',\n    description: 'Double click the element',\n    interfaceAlias: 'aiDoubleClick',\n    paramSchema: actionDoubleClickParamSchema,\n    call,\n  });\n};\n\n// Hover\nexport const actionHoverParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The element to be hovered'),\n});\nexport type ActionHoverParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionHover = (\n  call: (param: ActionHoverParam) => Promise<void>,\n): DeviceAction<ActionHoverParam> => {\n  return defineAction<typeof actionHoverParamSchema, ActionHoverParam>({\n    name: 'Hover',\n    description: 'Move the mouse to the element',\n    interfaceAlias: 'aiHover',\n    paramSchema: actionHoverParamSchema,\n    call,\n  });\n};\n\n// Input\nconst inputLocateDescription =\n  'the position of the placeholder or text content in the target input field. If there is no content, locate the center of the input field.';\nexport const actionInputParamSchema = z.object({\n  value: z\n    .union([z.string(), z.number()])\n    .transform((val) => String(val))\n    .describe(\n      'The text to input. Provide the final content for replace/append modes, or an empty string when using clear mode to remove existing text.',\n    ),\n  locate: getMidsceneLocationSchema()\n    .describe(inputLocateDescription)\n    .optional(),\n  mode: z\n    .enum(['replace', 'clear', 'append'])\n    .default('replace')\n    .optional()\n    .describe(\n      'Input mode: \"replace\" (default) - clear the field and input the value; \"append\" - append the value to existing content; \"clear\" - clear the field without inputting new text.',\n    ),\n});\nexport type ActionInputParam = {\n  value: string;\n  locate?: LocateResultElement;\n  mode?: 'replace' | 'clear' | 'append';\n};\n\nexport const defineActionInput = (\n  call: (param: ActionInputParam) => Promise<void>,\n): DeviceAction<ActionInputParam> => {\n  return defineAction<typeof actionInputParamSchema, ActionInputParam>({\n    name: 'Input',\n    description: 'Input the value into the element',\n    interfaceAlias: 'aiInput',\n    paramSchema: actionInputParamSchema,\n    call,\n  });\n};\n\n// KeyboardPress\nexport const actionKeyboardPressParamSchema = z.object({\n  locate: getMidsceneLocationSchema()\n    .describe('The element to be clicked before pressing the key')\n    .optional(),\n  keyName: z\n    .string()\n    .describe(\n      \"The key to be pressed. Use '+' for key combinations, e.g., 'Control+A', 'Shift+Enter'\",\n    ),\n});\nexport type ActionKeyboardPressParam = {\n  locate?: LocateResultElement;\n  keyName: string;\n};\n\nexport const defineActionKeyboardPress = (\n  call: (param: ActionKeyboardPressParam) => Promise<void>,\n): DeviceAction<ActionKeyboardPressParam> => {\n  return defineAction<\n    typeof actionKeyboardPressParamSchema,\n    ActionKeyboardPressParam\n  >({\n    name: 'KeyboardPress',\n    description:\n      'Press a key or key combination, like \"Enter\", \"Tab\", \"Escape\", or \"Control+A\", \"Shift+Enter\". Do not use this to type text.',\n    interfaceAlias: 'aiKeyboardPress',\n    paramSchema: actionKeyboardPressParamSchema,\n    call,\n  });\n};\n\n// Scroll\nexport const actionScrollParamSchema = z.object({\n  direction: z\n    .enum(['down', 'up', 'right', 'left'])\n    .default('down')\n    .describe('The direction to scroll'),\n  scrollType: z\n    .enum([\n      'singleAction',\n      'scrollToBottom',\n      'scrollToTop',\n      'scrollToRight',\n      'scrollToLeft',\n    ])\n    .default('singleAction')\n    .describe(\n      'The scroll behavior: \"singleAction\" for a single scroll action, \"scrollToBottom\" for scrolling to the bottom, \"scrollToTop\" for scrolling to the top, \"scrollToRight\" for scrolling to the right, \"scrollToLeft\" for scrolling to the left',\n    ),\n  distance: z\n    .number()\n    .nullable()\n    .optional()\n    .describe('The distance in pixels to scroll'),\n  locate: getMidsceneLocationSchema()\n    .optional()\n    .describe('The target element to be scrolled'),\n});\n\nexport const defineActionScroll = (\n  call: (param: ActionScrollParam) => Promise<void>,\n): DeviceAction<ActionScrollParam> => {\n  return defineAction<typeof actionScrollParamSchema, ActionScrollParam>({\n    name: 'Scroll',\n    description:\n      'Scroll the page or an element. The direction to scroll, the scroll type, and the distance to scroll. The distance is the number of pixels to scroll. If not specified, use `down` direction, `once` scroll type, and `null` distance.',\n    interfaceAlias: 'aiScroll',\n    paramSchema: actionScrollParamSchema,\n    call,\n  });\n};\n\n// DragAndDrop\nexport const actionDragAndDropParamSchema = z.object({\n  from: getMidsceneLocationSchema().describe('The position to be dragged'),\n  to: getMidsceneLocationSchema().describe('The position to be dropped'),\n});\nexport type ActionDragAndDropParam = {\n  from: LocateResultElement;\n  to: LocateResultElement;\n};\n\nexport const defineActionDragAndDrop = (\n  call: (param: ActionDragAndDropParam) => Promise<void>,\n): DeviceAction<ActionDragAndDropParam> => {\n  return defineAction<\n    typeof actionDragAndDropParamSchema,\n    ActionDragAndDropParam\n  >({\n    name: 'DragAndDrop',\n    description:\n      'Drag and drop (hold the mouse or finger down and move the mouse) ',\n    interfaceAlias: 'aiDragAndDrop',\n    paramSchema: actionDragAndDropParamSchema,\n    call,\n  });\n};\n\nexport const ActionLongPressParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be long pressed',\n  ),\n  duration: z\n    .number()\n    .default(500)\n    .optional()\n    .describe('Long press duration in milliseconds'),\n});\n\nexport type ActionLongPressParam = {\n  locate: LocateResultElement;\n  duration?: number;\n};\nexport const defineActionLongPress = (\n  call: (param: ActionLongPressParam) => Promise<void>,\n): DeviceAction<ActionLongPressParam> => {\n  return defineAction<typeof ActionLongPressParamSchema, ActionLongPressParam>({\n    name: 'LongPress',\n    description: 'Long press the element',\n    paramSchema: ActionLongPressParamSchema,\n    call,\n  });\n};\n\nexport const ActionSwipeParamSchema = z.object({\n  start: getMidsceneLocationSchema()\n    .optional()\n    .describe(\n      'Starting point of the swipe gesture, if not specified, the center of the page will be used',\n    ),\n  direction: z\n    .enum(['up', 'down', 'left', 'right'])\n    .optional()\n    .describe(\n      'The direction to swipe (required when using distance). The direction means the direction of the finger swipe.',\n    ),\n  distance: z\n    .number()\n    .optional()\n    .describe('The distance in pixels to swipe (mutually exclusive with end)'),\n  end: getMidsceneLocationSchema()\n    .optional()\n    .describe(\n      'Ending point of the swipe gesture (mutually exclusive with distance)',\n    ),\n  duration: z\n    .number()\n    .default(300)\n    .describe('Duration of the swipe gesture in milliseconds'),\n  repeat: z\n    .number()\n    .optional()\n    .describe(\n      'The number of times to repeat the swipe gesture. 1 for default, 0 for infinite (e.g. endless swipe until the end of the page)',\n    ),\n});\n\nexport type ActionSwipeParam = {\n  start?: LocateResultElement;\n  direction?: 'up' | 'down' | 'left' | 'right';\n  distance?: number;\n  end?: LocateResultElement;\n  duration?: number;\n  repeat?: number;\n};\n\nexport const defineActionSwipe = (\n  call: (param: ActionSwipeParam) => Promise<void>,\n): DeviceAction<ActionSwipeParam> => {\n  return defineAction<typeof ActionSwipeParamSchema, ActionSwipeParam>({\n    name: 'Swipe',\n    description:\n      'Perform a swipe gesture. You must specify either \"end\" (target location) or \"distance\" + \"direction\" - they are mutually exclusive. Use \"end\" for precise location-based swipes, or \"distance\" + \"direction\" for relative movement.',\n    paramSchema: ActionSwipeParamSchema,\n    call,\n  });\n};\n\n// ClearInput\nexport const actionClearInputParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The input field to be cleared'),\n});\nexport type ActionClearInputParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionClearInput = (\n  call: (param: ActionClearInputParam) => Promise<void>,\n): DeviceAction<ActionClearInputParam> => {\n  return defineAction<\n    typeof actionClearInputParamSchema,\n    ActionClearInputParam\n  >({\n    name: 'ClearInput',\n    description: inputLocateDescription,\n    interfaceAlias: 'aiClearInput',\n    paramSchema: actionClearInputParamSchema,\n    call,\n  });\n};\n\n// Assert\nexport const actionAssertParamSchema = z.object({\n  condition: z.string().describe('The condition of the assertion'),\n});\nexport type ActionAssertParam = {\n  condition: string;\n};\n\nexport const defineActionAssert = (\n  call: (param: ActionAssertParam) => Promise<void>,\n): DeviceAction<ActionAssertParam> => {\n  return defineAction<typeof actionAssertParamSchema, ActionAssertParam>({\n    name: 'Assert',\n    description:\n      'If the user explicitly requires making an assertion (like \"there should be a button with text \"Yes\" in the popup\"), this tool will think about it, and then provide a solid result',\n    interfaceAlias: 'aiAssert',\n    paramSchema: actionAssertParamSchema,\n    call,\n  });\n};\n\nexport type { DeviceAction } from '../types';\nexport type {\n  AndroidDeviceOpt,\n  AndroidDeviceInputOpt,\n  IOSDeviceOpt,\n  IOSDeviceInputOpt,\n} from './device-options';\n"],"names":["AbstractInterface","defineAction","config","actionTapParamSchema","z","getMidsceneLocationSchema","defineActionTap","call","actionRightClickParamSchema","defineActionRightClick","actionDoubleClickParamSchema","defineActionDoubleClick","actionHoverParamSchema","defineActionHover","inputLocateDescription","actionInputParamSchema","val","String","defineActionInput","actionKeyboardPressParamSchema","defineActionKeyboardPress","actionScrollParamSchema","defineActionScroll","actionDragAndDropParamSchema","defineActionDragAndDrop","ActionLongPressParamSchema","defineActionLongPress","ActionSwipeParamSchema","defineActionSwipe","actionClearInputParamSchema","defineActionClearInput","actionAssertParamSchema","defineActionAssert"],"mappings":";;AAaO,MAAeA;AAmCtB;AAKO,MAAMC,eAAe,CAK1BC,SAaOA;AAIF,MAAMC,uBAAuBC,EAAE,MAAM,CAAC;IAC3C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAMO,MAAMC,kBAAkB,CAC7BC,OAEON,aAA0D;QAC/D,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaE;QACbI;IACF;AAIK,MAAMC,8BAA8BJ,EAAE,MAAM,CAAC;IAClD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMI,yBAAyB,CACpCF,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaO;QACbD;IACF;AAIK,MAAMG,+BAA+BN,EAAE,MAAM,CAAC;IACnD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMM,0BAA0B,CACrCJ,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaS;QACbH;IACF;AAIK,MAAMK,yBAAyBR,EAAE,MAAM,CAAC;IAC7C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMQ,oBAAoB,CAC/BN,OAEON,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaW;QACbL;IACF;AAIF,MAAMO,yBACJ;AACK,MAAMC,yBAAyBX,EAAE,MAAM,CAAC;IAC7C,OAAOA,EAAAA,KACC,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,MAAM;KAAG,EAC9B,SAAS,CAAC,CAACY,MAAQC,OAAOD,MAC1B,QAAQ,CACP;IAEJ,QAAQX,4BACL,QAAQ,CAACS,wBACT,QAAQ;IACX,MAAMV,CAAC,CAADA,OACC,CAAC;QAAC;QAAW;QAAS;KAAS,EACnC,OAAO,CAAC,WACR,QAAQ,GACR,QAAQ,CACP;AAEN;AAOO,MAAMc,oBAAoB,CAC/BX,OAEON,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAac;QACbR;IACF;AAIK,MAAMY,iCAAiCf,EAAE,MAAM,CAAC;IACrD,QAAQC,4BACL,QAAQ,CAAC,qDACT,QAAQ;IACX,SAASD,EAAAA,MACA,GACN,QAAQ,CACP;AAEN;AAMO,MAAMgB,4BAA4B,CACvCb,OAEON,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAakB;QACbZ;IACF;AAIK,MAAMc,0BAA0BjB,EAAE,MAAM,CAAC;IAC9C,WAAWA,CAAC,CAADA,OACJ,CAAC;QAAC;QAAQ;QAAM;QAAS;KAAO,EACpC,OAAO,CAAC,QACR,QAAQ,CAAC;IACZ,YAAYA,CAAC,CAADA,OACL,CAAC;QACJ;QACA;QACA;QACA;QACA;KACD,EACA,OAAO,CAAC,gBACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,GACR,QAAQ,CAAC;IACZ,QAAQC,4BACL,QAAQ,GACR,QAAQ,CAAC;AACd;AAEO,MAAMiB,qBAAqB,CAChCf,OAEON,aAAgE;QACrE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAaoB;QACbd;IACF;AAIK,MAAMgB,+BAA+BnB,EAAE,MAAM,CAAC;IACnD,MAAMC,4BAA4B,QAAQ,CAAC;IAC3C,IAAIA,4BAA4B,QAAQ,CAAC;AAC3C;AAMO,MAAMmB,0BAA0B,CACrCjB,OAEON,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAasB;QACbhB;IACF;AAGK,MAAMkB,6BAA6BrB,EAAE,MAAM,CAAC;IACjD,QAAQC,4BAA4B,QAAQ,CAC1C;IAEF,UAAUD,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,GACR,QAAQ,CAAC;AACd;AAMO,MAAMsB,wBAAwB,CACnCnB,OAEON,aAAsE;QAC3E,MAAM;QACN,aAAa;QACb,aAAawB;QACblB;IACF;AAGK,MAAMoB,yBAAyBvB,EAAE,MAAM,CAAC;IAC7C,OAAOC,4BACJ,QAAQ,GACR,QAAQ,CACP;IAEJ,WAAWD,CAAC,CAADA,OACJ,CAAC;QAAC;QAAM;QAAQ;QAAQ;KAAQ,EACpC,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,CAAC;IACZ,KAAKC,4BACF,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUD,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,CAAC;IACZ,QAAQA,EAAAA,MACC,GACN,QAAQ,GACR,QAAQ,CACP;AAEN;AAWO,MAAMwB,oBAAoB,CAC/BrB,OAEON,aAA8D;QACnE,MAAM;QACN,aACE;QACF,aAAa0B;QACbpB;IACF;AAIK,MAAMsB,8BAA8BzB,EAAE,MAAM,CAAC;IAClD,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMyB,yBAAyB,CACpCvB,OAEON,aAGL;QACA,MAAM;QACN,aAAaa;QACb,gBAAgB;QAChB,aAAae;QACbtB;IACF;AAIK,MAAMwB,0BAA0B3B,EAAE,MAAM,CAAC;IAC9C,WAAWA,EAAE,MAAM,GAAG,QAAQ,CAAC;AACjC;AAKO,MAAM4B,qBAAqB,CAChCzB,OAEON,aAAgE;QACrE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAa8B;QACbxB;IACF"}
+{"version":3,"file":"device/index.mjs","sources":["../../../src/device/index.ts"],"sourcesContent":["import { getMidsceneLocationSchema } from '@/common';\nimport type {\n  ActionScrollParam,\n  DeviceAction,\n  LocateResultElement,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport type { ElementNode } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { z } from 'zod';\nimport type { ElementCacheFeature, Rect, Size, UIContext } from '../types';\n\nexport abstract class AbstractInterface {\n  abstract interfaceType: string;\n\n  abstract screenshotBase64(): Promise<string>;\n  abstract size(): Promise<Size>;\n  abstract actionSpace(): DeviceAction[];\n\n  abstract cacheFeatureForRect?(\n    rect: Rect,\n    options?: {\n      targetDescription?: string;\n      modelConfig?: IModelConfig;\n    },\n  ): Promise<ElementCacheFeature>;\n  abstract rectMatchesCacheFeature?(\n    feature: ElementCacheFeature,\n  ): Promise<Rect>;\n\n  abstract destroy?(): Promise<void>;\n\n  abstract describe?(): string;\n  abstract beforeInvokeAction?(actionName: string, param: any): Promise<void>;\n  abstract afterInvokeAction?(actionName: string, param: any): Promise<void>;\n\n  // @deprecated do NOT extend this method\n  abstract getElementsNodeTree?: () => Promise<ElementNode>;\n\n  // @deprecated do NOT extend this method\n  abstract url?: () => string | Promise<string>;\n\n  // @deprecated do NOT extend this method\n  abstract evaluateJavaScript?<T = any>(script: string): Promise<T>;\n\n  // @deprecated do NOT extend this method\n  abstract getContext?(): Promise<UIContext>;\n}\n\n// Generic function to define actions with proper type inference\n// TRuntime allows specifying a different type for the runtime parameter (after location resolution)\n// TReturn allows specifying the return type of the action\nexport const defineAction = <\n  TSchema extends z.ZodType | undefined = undefined,\n  TRuntime = TSchema extends z.ZodType ? z.infer<TSchema> : undefined,\n  TReturn = any,\n>(\n  config: {\n    name: string;\n    description: string;\n    interfaceAlias?: string;\n    paramSchema?: TSchema;\n    call: (param: TRuntime) => Promise<TReturn> | TReturn;\n  } & Partial<\n    Omit<\n      DeviceAction<TRuntime, TReturn>,\n      'name' | 'description' | 'interfaceAlias' | 'paramSchema' | 'call'\n    >\n  >,\n): DeviceAction<TRuntime, TReturn> => {\n  return config as any; // Type assertion needed because schema validation type differs from runtime type\n};\n\n// Tap\nexport const actionTapParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The element to be tapped'),\n});\n// Override the inferred type to use LocateResultElement for the runtime locate field\nexport type ActionTapParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionTap = (\n  call: (param: ActionTapParam) => Promise<void>,\n): DeviceAction<ActionTapParam> => {\n  return defineAction<typeof actionTapParamSchema, ActionTapParam>({\n    name: 'Tap',\n    description: 'Tap the element',\n    interfaceAlias: 'aiTap',\n    paramSchema: actionTapParamSchema,\n    call,\n  });\n};\n\n// RightClick\nexport const actionRightClickParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be right clicked',\n  ),\n});\nexport type ActionRightClickParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionRightClick = (\n  call: (param: ActionRightClickParam) => Promise<void>,\n): DeviceAction<ActionRightClickParam> => {\n  return defineAction<\n    typeof actionRightClickParamSchema,\n    ActionRightClickParam\n  >({\n    name: 'RightClick',\n    description: 'Right click the element',\n    interfaceAlias: 'aiRightClick',\n    paramSchema: actionRightClickParamSchema,\n    call,\n  });\n};\n\n// DoubleClick\nexport const actionDoubleClickParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be double clicked',\n  ),\n});\nexport type ActionDoubleClickParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionDoubleClick = (\n  call: (param: ActionDoubleClickParam) => Promise<void>,\n): DeviceAction<ActionDoubleClickParam> => {\n  return defineAction<\n    typeof actionDoubleClickParamSchema,\n    ActionDoubleClickParam\n  >({\n    name: 'DoubleClick',\n    description: 'Double click the element',\n    interfaceAlias: 'aiDoubleClick',\n    paramSchema: actionDoubleClickParamSchema,\n    call,\n  });\n};\n\n// Hover\nexport const actionHoverParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The element to be hovered'),\n});\nexport type ActionHoverParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionHover = (\n  call: (param: ActionHoverParam) => Promise<void>,\n): DeviceAction<ActionHoverParam> => {\n  return defineAction<typeof actionHoverParamSchema, ActionHoverParam>({\n    name: 'Hover',\n    description: 'Move the mouse to the element',\n    interfaceAlias: 'aiHover',\n    paramSchema: actionHoverParamSchema,\n    call,\n  });\n};\n\n// Input\nconst inputLocateDescription =\n  'the position of the placeholder or text content in the target input field. If there is no content, locate the center of the input field.';\nexport const actionInputParamSchema = z.object({\n  value: z\n    .union([z.string(), z.number()])\n    .transform((val) => String(val))\n    .describe(\n      'The text to input. Provide the final content for replace/append modes, or an empty string when using clear mode to remove existing text.',\n    ),\n  locate: getMidsceneLocationSchema()\n    .describe(inputLocateDescription)\n    .optional(),\n  mode: z\n    .enum(['replace', 'clear', 'append'])\n    .default('replace')\n    .optional()\n    .describe(\n      'Input mode: \"replace\" (default) - clear the field and input the value; \"append\" - append the value to existing content; \"clear\" - clear the field without inputting new text.',\n    ),\n});\nexport type ActionInputParam = {\n  value: string;\n  locate?: LocateResultElement;\n  mode?: 'replace' | 'clear' | 'append';\n};\n\nexport const defineActionInput = (\n  call: (param: ActionInputParam) => Promise<void>,\n): DeviceAction<ActionInputParam> => {\n  return defineAction<typeof actionInputParamSchema, ActionInputParam>({\n    name: 'Input',\n    description: 'Input the value into the element',\n    interfaceAlias: 'aiInput',\n    paramSchema: actionInputParamSchema,\n    call,\n  });\n};\n\n// KeyboardPress\nexport const actionKeyboardPressParamSchema = z.object({\n  locate: getMidsceneLocationSchema()\n    .describe('The element to be clicked before pressing the key')\n    .optional(),\n  keyName: z\n    .string()\n    .describe(\n      \"The key to be pressed. Use '+' for key combinations, e.g., 'Control+A', 'Shift+Enter'\",\n    ),\n});\nexport type ActionKeyboardPressParam = {\n  locate?: LocateResultElement;\n  keyName: string;\n};\n\nexport const defineActionKeyboardPress = (\n  call: (param: ActionKeyboardPressParam) => Promise<void>,\n): DeviceAction<ActionKeyboardPressParam> => {\n  return defineAction<\n    typeof actionKeyboardPressParamSchema,\n    ActionKeyboardPressParam\n  >({\n    name: 'KeyboardPress',\n    description:\n      'Press a key or key combination, like \"Enter\", \"Tab\", \"Escape\", or \"Control+A\", \"Shift+Enter\". Do not use this to type text.',\n    interfaceAlias: 'aiKeyboardPress',\n    paramSchema: actionKeyboardPressParamSchema,\n    call,\n  });\n};\n\n// Scroll\nexport const actionScrollParamSchema = z.object({\n  direction: z\n    .enum(['down', 'up', 'right', 'left'])\n    .default('down')\n    .describe('The direction to scroll'),\n  scrollType: z\n    .enum([\n      'singleAction',\n      'scrollToBottom',\n      'scrollToTop',\n      'scrollToRight',\n      'scrollToLeft',\n    ])\n    .default('singleAction')\n    .describe(\n      'The scroll behavior: \"singleAction\" for a single scroll action, \"scrollToBottom\" for scrolling to the bottom, \"scrollToTop\" for scrolling to the top, \"scrollToRight\" for scrolling to the right, \"scrollToLeft\" for scrolling to the left',\n    ),\n  distance: z\n    .number()\n    .nullable()\n    .optional()\n    .describe('The distance in pixels to scroll'),\n  locate: getMidsceneLocationSchema()\n    .optional()\n    .describe('The target element to be scrolled'),\n});\n\nexport const defineActionScroll = (\n  call: (param: ActionScrollParam) => Promise<void>,\n): DeviceAction<ActionScrollParam> => {\n  return defineAction<typeof actionScrollParamSchema, ActionScrollParam>({\n    name: 'Scroll',\n    description:\n      'Scroll the page or an element. The direction to scroll, the scroll type, and the distance to scroll. The distance is the number of pixels to scroll. If not specified, use `down` direction, `once` scroll type, and `null` distance.',\n    interfaceAlias: 'aiScroll',\n    paramSchema: actionScrollParamSchema,\n    call,\n  });\n};\n\n// DragAndDrop\nexport const actionDragAndDropParamSchema = z.object({\n  from: getMidsceneLocationSchema().describe('The position to be dragged'),\n  to: getMidsceneLocationSchema().describe('The position to be dropped'),\n});\nexport type ActionDragAndDropParam = {\n  from: LocateResultElement;\n  to: LocateResultElement;\n};\n\nexport const defineActionDragAndDrop = (\n  call: (param: ActionDragAndDropParam) => Promise<void>,\n): DeviceAction<ActionDragAndDropParam> => {\n  return defineAction<\n    typeof actionDragAndDropParamSchema,\n    ActionDragAndDropParam\n  >({\n    name: 'DragAndDrop',\n    description:\n      'Drag and drop (hold the mouse or finger down and move the mouse) ',\n    interfaceAlias: 'aiDragAndDrop',\n    paramSchema: actionDragAndDropParamSchema,\n    call,\n  });\n};\n\nexport const ActionLongPressParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be long pressed',\n  ),\n  duration: z\n    .number()\n    .default(500)\n    .optional()\n    .describe('Long press duration in milliseconds'),\n});\n\nexport type ActionLongPressParam = {\n  locate: LocateResultElement;\n  duration?: number;\n};\nexport const defineActionLongPress = (\n  call: (param: ActionLongPressParam) => Promise<void>,\n): DeviceAction<ActionLongPressParam> => {\n  return defineAction<typeof ActionLongPressParamSchema, ActionLongPressParam>({\n    name: 'LongPress',\n    description: 'Long press the element',\n    paramSchema: ActionLongPressParamSchema,\n    call,\n  });\n};\n\nexport const ActionSwipeParamSchema = z.object({\n  start: getMidsceneLocationSchema()\n    .optional()\n    .describe(\n      'Starting point of the swipe gesture, if not specified, the center of the page will be used',\n    ),\n  direction: z\n    .enum(['up', 'down', 'left', 'right'])\n    .optional()\n    .describe(\n      'The direction to swipe (required when using distance). The direction means the direction of the finger swipe.',\n    ),\n  distance: z\n    .number()\n    .optional()\n    .describe('The distance in pixels to swipe (mutually exclusive with end)'),\n  end: getMidsceneLocationSchema()\n    .optional()\n    .describe(\n      'Ending point of the swipe gesture (mutually exclusive with distance)',\n    ),\n  duration: z\n    .number()\n    .default(300)\n    .describe('Duration of the swipe gesture in milliseconds'),\n  repeat: z\n    .number()\n    .optional()\n    .describe(\n      'The number of times to repeat the swipe gesture. 1 for default, 0 for infinite (e.g. endless swipe until the end of the page)',\n    ),\n});\n\nexport type ActionSwipeParam = {\n  start?: LocateResultElement;\n  direction?: 'up' | 'down' | 'left' | 'right';\n  distance?: number;\n  end?: LocateResultElement;\n  duration?: number;\n  repeat?: number;\n};\n\nexport const defineActionSwipe = (\n  call: (param: ActionSwipeParam) => Promise<void>,\n): DeviceAction<ActionSwipeParam> => {\n  return defineAction<typeof ActionSwipeParamSchema, ActionSwipeParam>({\n    name: 'Swipe',\n    description:\n      'Perform a swipe gesture. You must specify either \"end\" (target location) or \"distance\" + \"direction\" - they are mutually exclusive. Use \"end\" for precise location-based swipes, or \"distance\" + \"direction\" for relative movement.',\n    paramSchema: ActionSwipeParamSchema,\n    call,\n  });\n};\n\n// ClearInput\nexport const actionClearInputParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The input field to be cleared'),\n});\nexport type ActionClearInputParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionClearInput = (\n  call: (param: ActionClearInputParam) => Promise<void>,\n): DeviceAction<ActionClearInputParam> => {\n  return defineAction<\n    typeof actionClearInputParamSchema,\n    ActionClearInputParam\n  >({\n    name: 'ClearInput',\n    description: inputLocateDescription,\n    interfaceAlias: 'aiClearInput',\n    paramSchema: actionClearInputParamSchema,\n    call,\n  });\n};\n\n// Assert\nexport const actionAssertParamSchema = z.object({\n  condition: z.string().describe('The condition of the assertion'),\n  thought: z\n    .string()\n    .describe(\n      'The thought of the assertion, like \"I can see there are A, B, C elements on the page, which means ... , so the assertion is true\"',\n    ),\n  result: z.boolean().describe('The result of the assertion, true or false'),\n});\nexport type ActionAssertParam = {\n  condition: string;\n  thought: string;\n  result: boolean;\n};\n\nexport const defineActionAssert = (): DeviceAction<ActionAssertParam> => {\n  return defineAction<typeof actionAssertParamSchema, ActionAssertParam>({\n    name: 'Print_Assert_Result',\n    description: 'Print the result of the assertion',\n    paramSchema: actionAssertParamSchema,\n    call: async (param) => {\n      if (typeof param?.result !== 'boolean') {\n        throw new Error(\n          `The result of the assertion must be a boolean, but got: ${typeof param?.result}. ${param.thought || '(no thought)'}`,\n        );\n      }\n\n      getDebug('device:common-action')(\n        `Assert: ${param.condition}, Thought: ${param.thought}, Result: ${param.result}`,\n      );\n\n      if (!param.result) {\n        throw new Error(\n          `Assertion failed: ${param.thought || '(no thought)'}. (Assertion = ${param.condition})`,\n        );\n      }\n    },\n  });\n};\n\nexport type { DeviceAction } from '../types';\nexport type {\n  AndroidDeviceOpt,\n  AndroidDeviceInputOpt,\n  IOSDeviceOpt,\n  IOSDeviceInputOpt,\n} from './device-options';\n"],"names":["AbstractInterface","defineAction","config","actionTapParamSchema","z","getMidsceneLocationSchema","defineActionTap","call","actionRightClickParamSchema","defineActionRightClick","actionDoubleClickParamSchema","defineActionDoubleClick","actionHoverParamSchema","defineActionHover","inputLocateDescription","actionInputParamSchema","val","String","defineActionInput","actionKeyboardPressParamSchema","defineActionKeyboardPress","actionScrollParamSchema","defineActionScroll","actionDragAndDropParamSchema","defineActionDragAndDrop","ActionLongPressParamSchema","defineActionLongPress","ActionSwipeParamSchema","defineActionSwipe","actionClearInputParamSchema","defineActionClearInput","actionAssertParamSchema","defineActionAssert","param","Error","getDebug"],"mappings":";;;AAaO,MAAeA;AAmCtB;AAKO,MAAMC,eAAe,CAK1BC,SAaOA;AAIF,MAAMC,uBAAuBC,EAAE,MAAM,CAAC;IAC3C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAMO,MAAMC,kBAAkB,CAC7BC,OAEON,aAA0D;QAC/D,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaE;QACbI;IACF;AAIK,MAAMC,8BAA8BJ,EAAE,MAAM,CAAC;IAClD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMI,yBAAyB,CACpCF,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaO;QACbD;IACF;AAIK,MAAMG,+BAA+BN,EAAE,MAAM,CAAC;IACnD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMM,0BAA0B,CACrCJ,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaS;QACbH;IACF;AAIK,MAAMK,yBAAyBR,EAAE,MAAM,CAAC;IAC7C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMQ,oBAAoB,CAC/BN,OAEON,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaW;QACbL;IACF;AAIF,MAAMO,yBACJ;AACK,MAAMC,yBAAyBX,EAAE,MAAM,CAAC;IAC7C,OAAOA,EAAAA,KACC,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,MAAM;KAAG,EAC9B,SAAS,CAAC,CAACY,MAAQC,OAAOD,MAC1B,QAAQ,CACP;IAEJ,QAAQX,4BACL,QAAQ,CAACS,wBACT,QAAQ;IACX,MAAMV,CAAC,CAADA,OACC,CAAC;QAAC;QAAW;QAAS;KAAS,EACnC,OAAO,CAAC,WACR,QAAQ,GACR,QAAQ,CACP;AAEN;AAOO,MAAMc,oBAAoB,CAC/BX,OAEON,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAac;QACbR;IACF;AAIK,MAAMY,iCAAiCf,EAAE,MAAM,CAAC;IACrD,QAAQC,4BACL,QAAQ,CAAC,qDACT,QAAQ;IACX,SAASD,EAAAA,MACA,GACN,QAAQ,CACP;AAEN;AAMO,MAAMgB,4BAA4B,CACvCb,OAEON,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAakB;QACbZ;IACF;AAIK,MAAMc,0BAA0BjB,EAAE,MAAM,CAAC;IAC9C,WAAWA,CAAC,CAADA,OACJ,CAAC;QAAC;QAAQ;QAAM;QAAS;KAAO,EACpC,OAAO,CAAC,QACR,QAAQ,CAAC;IACZ,YAAYA,CAAC,CAADA,OACL,CAAC;QACJ;QACA;QACA;QACA;QACA;KACD,EACA,OAAO,CAAC,gBACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,GACR,QAAQ,CAAC;IACZ,QAAQC,4BACL,QAAQ,GACR,QAAQ,CAAC;AACd;AAEO,MAAMiB,qBAAqB,CAChCf,OAEON,aAAgE;QACrE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAaoB;QACbd;IACF;AAIK,MAAMgB,+BAA+BnB,EAAE,MAAM,CAAC;IACnD,MAAMC,4BAA4B,QAAQ,CAAC;IAC3C,IAAIA,4BAA4B,QAAQ,CAAC;AAC3C;AAMO,MAAMmB,0BAA0B,CACrCjB,OAEON,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAasB;QACbhB;IACF;AAGK,MAAMkB,6BAA6BrB,EAAE,MAAM,CAAC;IACjD,QAAQC,4BAA4B,QAAQ,CAC1C;IAEF,UAAUD,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,GACR,QAAQ,CAAC;AACd;AAMO,MAAMsB,wBAAwB,CACnCnB,OAEON,aAAsE;QAC3E,MAAM;QACN,aAAa;QACb,aAAawB;QACblB;IACF;AAGK,MAAMoB,yBAAyBvB,EAAE,MAAM,CAAC;IAC7C,OAAOC,4BACJ,QAAQ,GACR,QAAQ,CACP;IAEJ,WAAWD,CAAC,CAADA,OACJ,CAAC;QAAC;QAAM;QAAQ;QAAQ;KAAQ,EACpC,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,CAAC;IACZ,KAAKC,4BACF,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUD,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,CAAC;IACZ,QAAQA,EAAAA,MACC,GACN,QAAQ,GACR,QAAQ,CACP;AAEN;AAWO,MAAMwB,oBAAoB,CAC/BrB,OAEON,aAA8D;QACnE,MAAM;QACN,aACE;QACF,aAAa0B;QACbpB;IACF;AAIK,MAAMsB,8BAA8BzB,EAAE,MAAM,CAAC;IAClD,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMyB,yBAAyB,CACpCvB,OAEON,aAGL;QACA,MAAM;QACN,aAAaa;QACb,gBAAgB;QAChB,aAAae;QACbtB;IACF;AAIK,MAAMwB,0BAA0B3B,EAAE,MAAM,CAAC;IAC9C,WAAWA,EAAE,MAAM,GAAG,QAAQ,CAAC;IAC/B,SAASA,EAAAA,MACA,GACN,QAAQ,CACP;IAEJ,QAAQA,EAAE,OAAO,GAAG,QAAQ,CAAC;AAC/B;AAOO,MAAM4B,qBAAqB,IACzB/B,aAAgE;QACrE,MAAM;QACN,aAAa;QACb,aAAa8B;QACb,MAAM,OAAOE;YACX,IAAI,AAAyB,aAAzB,OAAOA,OAAO,QAChB,MAAM,IAAIC,MACR,CAAC,wDAAwD,EAAE,OAAOD,OAAO,OAAO,EAAE,EAAEA,MAAM,OAAO,IAAI,gBAAgB;YAIzHE,SAAS,wBACP,CAAC,QAAQ,EAAEF,MAAM,SAAS,CAAC,WAAW,EAAEA,MAAM,OAAO,CAAC,UAAU,EAAEA,MAAM,MAAM,EAAE;YAGlF,IAAI,CAACA,MAAM,MAAM,EACf,MAAM,IAAIC,MACR,CAAC,kBAAkB,EAAED,MAAM,OAAO,IAAI,eAAe,eAAe,EAAEA,MAAM,SAAS,CAAC,CAAC,CAAC;QAG9F;IACF"}