@midscene/core 1.0.1-beta-20251022061922.0 → 1.0.1-beta-20251024064637.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/es/agent/agent.mjs +2 -3
- package/dist/es/agent/agent.mjs.map +1 -1
- package/dist/es/agent/index.mjs +2 -2
- package/dist/es/agent/task-builder.mjs +11 -7
- package/dist/es/agent/task-builder.mjs.map +1 -1
- package/dist/es/agent/tasks.mjs +11 -2
- package/dist/es/agent/tasks.mjs.map +1 -1
- package/dist/es/agent/ui-utils.mjs +10 -10
- package/dist/es/agent/ui-utils.mjs.map +1 -1
- package/dist/es/agent/utils.mjs +6 -50
- package/dist/es/agent/utils.mjs.map +1 -1
- package/dist/es/ai-model/common.mjs.map +1 -1
- package/dist/es/ai-model/index.mjs +2 -2
- package/dist/es/ai-model/inspect.mjs +12 -31
- package/dist/es/ai-model/inspect.mjs.map +1 -1
- package/dist/es/ai-model/prompt/util.mjs +3 -88
- package/dist/es/ai-model/prompt/util.mjs.map +1 -1
- package/dist/es/device/index.mjs +2 -2
- package/dist/es/device/index.mjs.map +1 -1
- package/dist/es/insight/index.mjs +6 -15
- package/dist/es/insight/index.mjs.map +1 -1
- package/dist/es/tree.mjs +13 -2
- package/dist/es/tree.mjs.map +1 -0
- package/dist/es/types.mjs.map +1 -1
- package/dist/es/utils.mjs +2 -2
- package/dist/es/yaml/utils.mjs +10 -4
- package/dist/es/yaml/utils.mjs.map +1 -1
- package/dist/lib/agent/agent.js +1 -2
- package/dist/lib/agent/agent.js.map +1 -1
- package/dist/lib/agent/index.js +0 -3
- package/dist/lib/agent/task-builder.js +11 -7
- package/dist/lib/agent/task-builder.js.map +1 -1
- package/dist/lib/agent/tasks.js +11 -2
- package/dist/lib/agent/tasks.js.map +1 -1
- package/dist/lib/agent/ui-utils.js +10 -10
- package/dist/lib/agent/ui-utils.js.map +1 -1
- package/dist/lib/agent/utils.js +5 -52
- package/dist/lib/agent/utils.js.map +1 -1
- package/dist/lib/ai-model/common.js.map +1 -1
- package/dist/lib/ai-model/index.js +11 -14
- package/dist/lib/ai-model/inspect.js +11 -30
- package/dist/lib/ai-model/inspect.js.map +1 -1
- package/dist/lib/ai-model/prompt/util.js +5 -93
- package/dist/lib/ai-model/prompt/util.js.map +1 -1
- package/dist/lib/device/index.js +2 -2
- package/dist/lib/device/index.js.map +1 -1
- package/dist/lib/insight/index.js +6 -15
- package/dist/lib/insight/index.js.map +1 -1
- package/dist/lib/tree.js +10 -1
- package/dist/lib/tree.js.map +1 -1
- package/dist/lib/types.js.map +1 -1
- package/dist/lib/utils.js +2 -2
- package/dist/lib/yaml/utils.js +10 -4
- package/dist/lib/yaml/utils.js.map +1 -1
- package/dist/types/agent/index.d.ts +1 -1
- package/dist/types/agent/utils.d.ts +2 -33
- package/dist/types/ai-model/index.d.ts +1 -1
- package/dist/types/ai-model/inspect.d.ts +12 -10
- package/dist/types/ai-model/prompt/util.d.ts +2 -34
- package/dist/types/insight/index.d.ts +6 -6
- package/dist/types/tree.d.ts +4 -1
- package/dist/types/types.d.ts +6 -30
- package/dist/types/yaml.d.ts +2 -2
- package/package.json +3 -3
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"agent/tasks.mjs","sources":["webpack://@midscene/core/./src/agent/tasks.ts"],"sourcesContent":["import { ConversationHistory, plan, uiTarsPlanning } from '@/ai-model';\nimport type { TMultimodalPrompt, TUserPrompt } from '@/ai-model/common';\nimport type { AbstractInterface } from '@/device';\nimport type Insight from '@/insight';\nimport type { TaskRunner } from '@/task-runner';\nimport type {\n ExecutionTaskApply,\n ExecutionTaskInsightQueryApply,\n ExecutionTaskPlanningApply,\n ExecutionTaskProgressOptions,\n InsightDump,\n InsightExtractOption,\n InsightExtractParam,\n InterfaceType,\n MidsceneYamlFlowItem,\n PlanningAIResponse,\n PlanningAction,\n PlanningActionParamSleep,\n PlanningActionParamWaitFor,\n} from '@/types';\nimport { InsightError } from '@/types';\nimport {\n type IModelConfig,\n MIDSCENE_REPLANNING_CYCLE_LIMIT,\n globalConfigManager,\n} 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 { taskTitleStr } from './ui-utils';\nimport { parsePrompt } from './utils';\n\ninterface ExecutionResult<OutputType = any> {\n output: OutputType;\n thought?: string;\n runner: TaskRunner;\n}\n\nconst debug = getDebug('device-task-executor');\nconst defaultReplanningCycleLimit = 10;\nconst defaultVlmUiTarsReplanningCycleLimit = 40;\n\nexport class TaskExecutor {\n interface: AbstractInterface;\n\n insight: Insight;\n\n taskCache?: TaskCache;\n\n private readonly taskBuilder: TaskBuilder;\n\n private conversationHistory: ConversationHistory;\n\n onTaskStartCallback?: ExecutionTaskProgressOptions['onTaskStart'];\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 insight: Insight,\n opts: {\n taskCache?: TaskCache;\n onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n replanningCycleLimit?: number;\n },\n ) {\n this.interface = interfaceInstance;\n this.insight = insight;\n this.taskCache = opts.taskCache;\n this.onTaskStartCallback = opts?.onTaskStart;\n this.replanningCycleLimit = opts.replanningCycleLimit;\n this.conversationHistory = new ConversationHistory();\n this.taskBuilder = new TaskBuilder({\n interfaceInstance,\n insight,\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.insight.contextRetrieverFn()),\n {\n onTaskStart: this.onTaskStartCallback,\n tasks: options?.tasks,\n },\n );\n }\n\n public async convertPlanToExecutable(\n plans: PlanningAction[],\n modelConfig: IModelConfig,\n options?: {\n cacheable?: boolean;\n subTask?: boolean;\n },\n ) {\n return this.taskBuilder.build(plans, modelConfig, options);\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 locate: null,\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 await session.appendAndRun(task);\n\n return {\n runner: session.getRunner(),\n };\n }\n\n private createPlanningTask(\n userInstruction: string,\n actionContext: string | undefined,\n modelConfig: IModelConfig,\n ): ExecutionTaskPlanningApply {\n const task: ExecutionTaskPlanningApply = {\n type: 'Planning',\n subType: 'Plan',\n locate: null,\n param: {\n userInstruction,\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 } = modelConfig;\n const uiTarsModelVersion =\n vlMode === 'vlm-ui-tars' ? modelConfig.uiTarsModelVersion : 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 ? uiTarsPlanning : plan)(\n param.userInstruction,\n {\n context: uiContext,\n actionContext,\n interfaceType: this.interface.interfaceType as InterfaceType,\n actionSpace,\n modelConfig,\n conversationHistory: this.conversationHistory,\n },\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\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 if (finalActions.length === 0) {\n assert(\n !more_actions_needed_by_instruction || sleep,\n error ? `Failed to plan: ${error}` : 'No plan found',\n );\n }\n\n return {\n output: {\n actions: finalActions,\n more_actions_needed_by_instruction,\n log,\n yamlFlow: planResult.yamlFlow,\n },\n cache: {\n hit: false,\n },\n uiContext,\n };\n },\n };\n\n return task;\n }\n\n async runPlans(\n title: string,\n plans: PlanningAction[],\n modelConfig: IModelConfig,\n ): Promise<ExecutionResult> {\n const session = this.createExecutionSession(title);\n const { tasks } = await this.convertPlanToExecutable(plans, modelConfig);\n const result = await session.appendAndRun(tasks);\n const { output } = result!;\n return {\n output,\n runner: session.getRunner(),\n };\n }\n\n private getReplanningCycleLimit(isVlmUiTars: boolean) {\n return (\n this.replanningCycleLimit ||\n globalConfigManager.getEnvConfigInNumber(\n MIDSCENE_REPLANNING_CYCLE_LIMIT,\n ) ||\n (isVlmUiTars\n ? defaultVlmUiTarsReplanningCycleLimit\n : defaultReplanningCycleLimit)\n );\n }\n\n async action(\n userPrompt: string,\n modelConfig: IModelConfig,\n actionContext?: string,\n cacheable?: boolean,\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 = this.getReplanningCycleLimit(\n modelConfig.vlMode === 'vlm-ui-tars',\n );\n\n // Main planning loop - unified plan/replan logic\n while (true) {\n if (replanCount > replanningCycleLimit) {\n const errorMsg = `Replanning ${replanningCycleLimit} times, which is more than the limit, please split the task into multiple steps`;\n\n return session.appendErrorPlan(errorMsg);\n }\n\n // Create planning task (automatically includes execution history if available)\n const planningTask = this.createPlanningTask(\n userPrompt,\n actionContext,\n modelConfig,\n );\n\n const result = await session.appendAndRun(planningTask);\n const planResult: PlanningAIResponse = result?.output;\n if (session.isInErrorState()) {\n return {\n output: planResult,\n runner,\n };\n }\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(plans, modelConfig, {\n cacheable,\n subTask: true,\n });\n await session.appendAndRun(executables.tasks);\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 (session.isInErrorState()) {\n return {\n output: undefined,\n runner,\n };\n }\n\n // Check if task is complete\n if (!planResult.more_actions_needed_by_instruction) {\n break;\n }\n\n // Increment replan count for next iteration\n replanCount++;\n }\n\n return {\n output: {\n yamlFlow,\n },\n runner,\n };\n }\n\n private createTypeQueryTask(\n type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert' | 'WaitFor',\n demand: InsightExtractParam,\n modelConfig: IModelConfig,\n opt?: InsightExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ) {\n const queryTask: ExecutionTaskInsightQueryApply = {\n type: 'Insight',\n subType: type,\n locate: null,\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: InsightDump | undefined;\n const applyDump = (dump: InsightDump) => {\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 try {\n extractResult = await this.insight.extract<any>(\n demandInput,\n modelConfig,\n opt,\n multimodalPrompt,\n );\n } catch (error) {\n if (error instanceof InsightError) {\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 {\n assert(\n type !== 'WaitFor' ? data?.[keyOfResult] !== undefined : true,\n 'No result in query data',\n );\n outputResult = (data as any)[keyOfResult];\n }\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: InsightExtractParam,\n modelConfig: IModelConfig,\n opt?: InsightExtractOption,\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 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: session.getRunner(),\n };\n }\n\n private sleepPlan(timeMs: number): PlanningAction<PlanningActionParamSleep> {\n return {\n type: 'Sleep',\n param: {\n timeMs,\n },\n locate: null,\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 startTime = Date.now();\n let errorThought = '';\n while (Date.now() - overallStartTime < timeoutMs) {\n startTime = Date.now();\n const queryTask = await this.createTypeQueryTask(\n 'WaitFor',\n textPrompt,\n modelConfig,\n {\n doNotThrowError: true,\n },\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 - startTime < checkIntervalMs) {\n const timeRemaining = checkIntervalMs - (now - startTime);\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","defaultReplanningCycleLimit","defaultVlmUiTarsReplanningCycleLimit","TaskExecutor","title","options","ExecutionSession","Promise","plans","modelConfig","userInstruction","yamlString","session","taskTitleStr","task","param","executorContext","uiContext","assert","actionContext","startTime","Date","vlMode","uiTarsModelVersion","undefined","actionSpace","action","Array","console","planResult","uiTarsPlanning","plan","JSON","actions","log","more_actions_needed_by_instruction","error","usage","rawResponse","sleep","finalActions","timeNow","timeRemaining","tasks","result","output","isVlmUiTars","globalConfigManager","MIDSCENE_REPLANNING_CYCLE_LIMIT","userPrompt","cacheable","runner","replanCount","yamlFlow","replanningCycleLimit","errorMsg","planningTask","executables","type","demand","opt","multimodalPrompt","queryTask","taskContext","queryDump","applyDump","dump","ifTypeRestricted","demandInput","keyOfResult","booleanPrompt","extractResult","InsightError","data","thought","outputResult","Error","timeMs","_modelConfig","assertion","textPrompt","parsePrompt","description","timeoutMs","checkIntervalMs","overallStartTime","errorThought","now","sleepTask","interfaceInstance","insight","opts","ConversationHistory","TaskBuilder"],"mappings":";;;;;;;;;;;;;;;;;;;AAyCA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,8BAA8B;AACpC,MAAMC,uCAAuC;AAEtC,MAAMC;IAgBX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IAwBQ,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,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,KAAK;QACvB;IAEJ;IAEA,MAAa,wBACXG,KAAuB,EACvBC,WAAyB,EACzBJ,OAGC,EACD;QACA,OAAO,IAAI,CAAC,WAAW,CAAC,KAAK,CAACG,OAAOC,aAAaJ;IACpD;IAEA,MAAM,uBAAuBK,eAAuB,EAAEC,UAAkB,EAAE;QACxE,MAAMC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUH;QAGzB,MAAMI,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,QAAQ;YACR,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,MAAMC,QAAQ,YAAY,CAACE;QAE3B,OAAO;YACL,QAAQF,QAAQ,SAAS;QAC3B;IACF;IAEQ,mBACNF,eAAuB,EACvBS,aAAiC,EACjCV,WAAyB,EACG;QAC5B,MAAMK,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,QAAQ;YACR,OAAO;gBACLJ;YACF;YACA,UAAU,OAAOK,OAAOC;gBACtB,MAAMI,YAAYC,KAAK,GAAG;gBAC1B,MAAM,EAAEJ,SAAS,EAAE,GAAGD;gBACtBE,OAAOD,WAAW;gBAClB,MAAM,EAAEK,MAAM,EAAE,GAAGb;gBACnB,MAAMc,qBACJD,AAAW,kBAAXA,SAA2Bb,YAAY,kBAAkB,GAAGe;gBAE9DN,OACE,IAAI,CAAC,SAAS,CAAC,WAAW,EAC1B;gBAEF,MAAMO,cAAc,MAAM,IAAI,CAAC,SAAS,CAAC,WAAW;gBACpD1B,MACE,sCACA0B,YAAY,GAAG,CAAC,CAACC,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;gBAEhDR,OAAOS,MAAM,OAAO,CAACF,cAAc;gBACnC,IAAIA,AAAuB,MAAvBA,YAAY,MAAM,EACpBG,QAAQ,IAAI,CACV,CAAC,gBAAgB,EAAE,IAAI,CAAC,SAAS,CAAC,aAAa,CAAC,gDAAgD,CAAC;gBAIrG,MAAMC,aAAa,MAAON,AAAAA,CAAAA,qBAAqBO,iBAAiBC,IAAG,EACjEhB,MAAM,eAAe,EACrB;oBACE,SAASE;oBACTE;oBACA,eAAe,IAAI,CAAC,SAAS,CAAC,aAAa;oBAC3CM;oBACAhB;oBACA,qBAAqB,IAAI,CAAC,mBAAmB;gBAC/C;gBAEFV,MAAM,cAAciC,KAAK,SAAS,CAACH,YAAY,MAAM;gBAErD,MAAM,EACJI,OAAO,EACPC,GAAG,EACHC,kCAAkC,EAClCC,KAAK,EACLC,KAAK,EACLC,WAAW,EACXC,KAAK,EACN,GAAGV;gBAEJb,gBAAgB,IAAI,CAAC,GAAG,GAAG;oBACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;oBAClCsB;gBACF;gBACAtB,gBAAgB,IAAI,CAAC,KAAK,GAAGqB;gBAE7B,MAAMG,eAAeP,WAAW,EAAE;gBAElC,IAAIM,OAAO;oBACT,MAAME,UAAUpB,KAAK,GAAG;oBACxB,MAAMqB,gBAAgBH,QAASE,CAAAA,UAAUrB,SAAQ;oBACjD,IAAIsB,gBAAgB,GAClBF,aAAa,IAAI,CAAC,IAAI,CAAC,SAAS,CAACE;gBAErC;gBAEA,IAAIF,AAAwB,MAAxBA,aAAa,MAAM,EACrBtB,OACE,CAACiB,sCAAsCI,OACvCH,QAAQ,CAAC,gBAAgB,EAAEA,OAAO,GAAG;gBAIzC,OAAO;oBACL,QAAQ;wBACN,SAASI;wBACTL;wBACAD;wBACA,UAAUL,WAAW,QAAQ;oBAC/B;oBACA,OAAO;wBACL,KAAK;oBACP;oBACAZ;gBACF;YACF;QACF;QAEA,OAAOH;IACT;IAEA,MAAM,SACJV,KAAa,EACbI,KAAuB,EACvBC,WAAyB,EACC;QAC1B,MAAMG,UAAU,IAAI,CAAC,sBAAsB,CAACR;QAC5C,MAAM,EAAEuC,KAAK,EAAE,GAAG,MAAM,IAAI,CAAC,uBAAuB,CAACnC,OAAOC;QAC5D,MAAMmC,SAAS,MAAMhC,QAAQ,YAAY,CAAC+B;QAC1C,MAAM,EAAEE,MAAM,EAAE,GAAGD;QACnB,OAAO;YACLC;YACA,QAAQjC,QAAQ,SAAS;QAC3B;IACF;IAEQ,wBAAwBkC,WAAoB,EAAE;QACpD,OACE,IAAI,CAAC,oBAAoB,IACzBC,oBAAoB,oBAAoB,CACtCC,oCAEDF,CAAAA,cACG5C,uCACAD,2BAA0B;IAElC;IAEA,MAAM,OACJgD,UAAkB,EAClBxC,WAAyB,EACzBU,aAAsB,EACtB+B,SAAmB,EAQnB;QACA,IAAI,CAAC,mBAAmB,CAAC,KAAK;QAE9B,MAAMtC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUoC;QAEzB,MAAME,SAASvC,QAAQ,SAAS;QAEhC,IAAIwC,cAAc;QAClB,MAAMC,WAAmC,EAAE;QAC3C,MAAMC,uBAAuB,IAAI,CAAC,uBAAuB,CACvD7C,AAAuB,kBAAvBA,YAAY,MAAM;QAIpB,MAAO,KAAM;YACX,IAAI2C,cAAcE,sBAAsB;gBACtC,MAAMC,WAAW,CAAC,WAAW,EAAED,qBAAqB,+EAA+E,CAAC;gBAEpI,OAAO1C,QAAQ,eAAe,CAAC2C;YACjC;YAGA,MAAMC,eAAe,IAAI,CAAC,kBAAkB,CAC1CP,YACA9B,eACAV;YAGF,MAAMmC,SAAS,MAAMhC,QAAQ,YAAY,CAAC4C;YAC1C,MAAM3B,aAAiCe,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM;YACrD,IAAIhC,QAAQ,cAAc,IACxB,OAAO;gBACL,QAAQiB;gBACRsB;YACF;YAIF,MAAM3C,QAAQqB,WAAW,OAAO,IAAI,EAAE;YACtCwB,SAAS,IAAI,IAAKxB,WAAW,QAAQ,IAAI,EAAE;YAE3C,IAAI4B;YACJ,IAAI;gBACFA,cAAc,MAAM,IAAI,CAAC,uBAAuB,CAACjD,OAAOC,aAAa;oBACnEyC;oBACA,SAAS;gBACX;gBACA,MAAMtC,QAAQ,YAAY,CAAC6C,YAAY,KAAK;YAC9C,EAAE,OAAOrB,OAAO;gBACd,OAAOxB,QAAQ,eAAe,CAC5B,CAAC,4CAA4C,EAAEwB,MAAM,SAAS,EAAEJ,KAAK,SAAS,CAC5ExB,QACC;YAEP;YACA,IAAII,QAAQ,cAAc,IACxB,OAAO;gBACL,QAAQY;gBACR2B;YACF;YAIF,IAAI,CAACtB,WAAW,kCAAkC,EAChD;YAIFuB;QACF;QAEA,OAAO;YACL,QAAQ;gBACNC;YACF;YACAF;QACF;IACF;IAEQ,oBACNO,IAAsE,EACtEC,MAA2B,EAC3BlD,WAAyB,EACzBmD,GAA0B,EAC1BC,gBAAoC,EACpC;QACA,MAAMC,YAA4C;YAChD,MAAM;YACN,SAASJ;YACT,QAAQ;YACR,OAAO;gBACL,YAAYG,mBACP;oBACCF;oBACAE;gBACF,IACAF;YACN;YACA,UAAU,OAAO5C,OAAOgD;gBACtB,MAAM,EAAEjD,IAAI,EAAE,GAAGiD;gBACjB,IAAIC;gBACJ,MAAMC,YAAY,CAACC;oBACjBF,YAAYE;oBACZpD,KAAK,GAAG,GAAG;wBACToD;oBACF;gBACF;gBAGA,MAAMjD,YAAY8C,YAAY,SAAS;gBACvC7C,OAAOD,WAAW;gBAElB,MAAMkD,mBAAmBT,AAAS,YAATA;gBACzB,IAAIU,cAAcT;gBAClB,IAAIU,cAAc;gBAClB,IAAIF,oBAAqBT,CAAAA,AAAS,aAATA,QAAqBA,AAAS,cAATA,IAAiB,GAAI;oBACjEW,cAAc;oBACd,MAAMC,gBACJZ,AAAS,aAATA,OACI,CAAC,kDAAkD,EAAEC,QAAQ,GAC7D,CAAC,+GAA+G,EAAEA,QAAQ;oBAChIS,cAAc;wBACZ,CAACC,YAAY,EAAEC;oBACjB;gBACF,OAAO,IAAIH,kBACTC,cAAc;oBACZ,CAACC,YAAY,EAAE,GAAGX,KAAK,EAAE,EAAEC,QAAQ;gBACrC;gBAGF,IAAIY;gBACJ,IAAI;oBACFA,gBAAgB,MAAM,IAAI,CAAC,OAAO,CAAC,OAAO,CACxCH,aACA3D,aACAmD,KACAC;gBAEJ,EAAE,OAAOzB,OAAO;oBACd,IAAIA,iBAAiBoC,cACnBP,UAAU7B,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAEA,MAAM,EAAEqC,IAAI,EAAEpC,KAAK,EAAEqC,OAAO,EAAER,IAAI,EAAE,GAAGK;gBACvCN,UAAUC;gBAEV,IAAIS,eAAeF;gBACnB,IAAIN,kBAEF,IAAI,AAAgB,YAAhB,OAAOM,MACTE,eAAeF;qBACV;oBACLvD,OACEwC,AAAS,cAATA,OAAqBe,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,IAAM,CAACJ,YAAY,AAAD,MAAM7C,SAAY,MACzD;oBAEFmD,eAAgBF,IAAY,CAACJ,YAAY;gBAC3C;gBAGF,OAAO;oBACL,QAAQM;oBACR,KAAKX;oBACL3B;oBACAqC;gBACF;YACF;QACF;QAEA,OAAOZ;IACT;IACA,MAAM,yBACJJ,IAA0D,EAC1DC,MAA2B,EAC3BlD,WAAyB,EACzBmD,GAA0B,EAC1BC,gBAAoC,EACP;QAC7B,MAAMjD,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aACE6C,MACA,AAAkB,YAAlB,OAAOC,SAAsBA,SAAS3B,KAAK,SAAS,CAAC2B;QAIzD,MAAMG,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9CJ,MACAC,QACAlD,aACAmD,KACAC;QAGF,MAAMjB,SAAS,MAAMhC,QAAQ,YAAY,CAACkD;QAE1C,IAAI,CAAClB,QACH,MAAM,IAAIgC,MACR;QAIJ,MAAM,EAAE/B,MAAM,EAAE6B,OAAO,EAAE,GAAG9B;QAE5B,OAAO;YACLC;YACA6B;YACA,QAAQ9D,QAAQ,SAAS;QAC3B;IACF;IAEQ,UAAUiE,MAAc,EAA4C;QAC1E,OAAO;YACL,MAAM;YACN,OAAO;gBACLA;YACF;YACA,QAAQ;QACV;IACF;IAEA,MAAM,aAAaA,MAAc,EAAEC,YAA0B,EAAE;QAC7D,OAAO,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;YACtCD;QACF;IACF;IAEA,MAAM,QACJE,SAAsB,EACtBnB,GAA+B,EAC/BnD,WAAyB,EACO;QAChC,MAAM,EAAEuE,UAAU,EAAEnB,gBAAgB,EAAE,GAAGoB,YAAYF;QAErD,MAAMG,cAAc,CAAC,SAAS,EAAEF,YAAY;QAC5C,MAAMpE,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,WAAWqE;QAE1B,MAAM/B,SAASvC,QAAQ,SAAS;QAChC,MAAM,EAAEuE,SAAS,EAAEC,eAAe,EAAE,GAAGxB;QAEvC1C,OAAO6D,WAAW;QAClB7D,OAAOiE,WAAW;QAClBjE,OAAOkE,iBAAiB;QAExBlE,OACEkE,mBAAmBD,WACnB,CAAC,iGAAiG,EAAEC,gBAAgB,aAAa,EAAED,UAAU,CAAC,CAAC;QAGjJ,MAAME,mBAAmBhE,KAAK,GAAG;QACjC,IAAID,YAAYC,KAAK,GAAG;QACxB,IAAIiE,eAAe;QACnB,MAAOjE,KAAK,GAAG,KAAKgE,mBAAmBF,UAAW;YAChD/D,YAAYC,KAAK,GAAG;YACpB,MAAMyC,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9C,WACAkB,YACAvE,aACA;gBACE,iBAAiB;YACnB,GACAoD;YAGF,MAAMjB,SAAU,MAAMhC,QAAQ,YAAY,CAACkD;YAO3C,IAAIlB,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM,EAChB,OAAO;gBACL,QAAQpB;gBACR2B;YACF;YAGFmC,eACE1C,AAAAA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,OAAO,AAAD,KACb,CAACA,UAAU,CAAC,0BAA0B,EAAEoC,YAAY,IACrD,CAAC,0CAA0C,EAAEA,YAAY;YAC3D,MAAMO,MAAMlE,KAAK,GAAG;YACpB,IAAIkE,MAAMnE,YAAYgE,iBAAiB;gBACrC,MAAM1C,gBAAgB0C,kBAAmBG,CAAAA,MAAMnE,SAAQ;gBACvD,MAAMoE,YAAY,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;oBACjD,QAAQ9C;gBACV;gBACA,MAAM9B,QAAQ,MAAM,CAAC4E;YACvB;QACF;QAEA,OAAO5E,QAAQ,eAAe,CAAC,CAAC,iBAAiB,EAAE0E,cAAc;IACnE;IA3gBA,YACEG,iBAAoC,EACpCC,OAAgB,EAChBC,IAIC,CACD;QA3BF;QAEA;QAEA;QAEA,uBAAiB,eAAjB;QAEA,uBAAQ,uBAAR;QAEA;QAEA;QAgBE,IAAI,CAAC,SAAS,GAAGF;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC,KAAK,SAAS;QAC/B,IAAI,CAAC,mBAAmB,GAAGA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW;QAC5C,IAAI,CAAC,oBAAoB,GAAGA,KAAK,oBAAoB;QACrD,IAAI,CAAC,mBAAmB,GAAG,IAAIC;QAC/B,IAAI,CAAC,WAAW,GAAG,IAAIC,YAAY;YACjCJ;YACAC;YACA,WAAWC,KAAK,SAAS;QAC3B;IACF;AAwfF"}
|
|
1
|
+
{"version":3,"file":"agent/tasks.mjs","sources":["webpack://@midscene/core/./src/agent/tasks.ts"],"sourcesContent":["import { ConversationHistory, plan, uiTarsPlanning } from '@/ai-model';\nimport type { TMultimodalPrompt, TUserPrompt } from '@/ai-model/common';\nimport type { AbstractInterface } from '@/device';\nimport type Insight from '@/insight';\nimport type { TaskRunner } from '@/task-runner';\nimport type {\n ExecutionTaskApply,\n ExecutionTaskInsightQueryApply,\n ExecutionTaskPlanningApply,\n ExecutionTaskProgressOptions,\n InsightDump,\n InsightExtractOption,\n InsightExtractParam,\n InterfaceType,\n MidsceneYamlFlowItem,\n PlanningAIResponse,\n PlanningAction,\n PlanningActionParamSleep,\n PlanningActionParamWaitFor,\n} from '@/types';\nimport { InsightError } from '@/types';\nimport {\n type IModelConfig,\n MIDSCENE_REPLANNING_CYCLE_LIMIT,\n globalConfigManager,\n} 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\nconst debug = getDebug('device-task-executor');\nconst defaultReplanningCycleLimit = 10;\nconst defaultVlmUiTarsReplanningCycleLimit = 40;\n\nexport class TaskExecutor {\n interface: AbstractInterface;\n\n insight: Insight;\n\n taskCache?: TaskCache;\n\n private readonly taskBuilder: TaskBuilder;\n\n private conversationHistory: ConversationHistory;\n\n onTaskStartCallback?: ExecutionTaskProgressOptions['onTaskStart'];\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 insight: Insight,\n opts: {\n taskCache?: TaskCache;\n onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n replanningCycleLimit?: number;\n },\n ) {\n this.interface = interfaceInstance;\n this.insight = insight;\n this.taskCache = opts.taskCache;\n this.onTaskStartCallback = opts?.onTaskStart;\n this.replanningCycleLimit = opts.replanningCycleLimit;\n this.conversationHistory = new ConversationHistory();\n this.taskBuilder = new TaskBuilder({\n interfaceInstance,\n insight,\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.insight.contextRetrieverFn()),\n {\n onTaskStart: this.onTaskStartCallback,\n tasks: options?.tasks,\n },\n );\n }\n\n public async convertPlanToExecutable(\n plans: PlanningAction[],\n modelConfig: IModelConfig,\n options?: {\n cacheable?: boolean;\n subTask?: boolean;\n },\n ) {\n return this.taskBuilder.build(plans, modelConfig, options);\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 locate: null,\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 await session.appendAndRun(task);\n\n return {\n runner: session.getRunner(),\n };\n }\n\n private createPlanningTask(\n userInstruction: string,\n actionContext: string | undefined,\n modelConfig: IModelConfig,\n ): ExecutionTaskPlanningApply {\n const task: ExecutionTaskPlanningApply = {\n type: 'Planning',\n subType: 'Plan',\n locate: null,\n param: {\n userInstruction,\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 } = modelConfig;\n const uiTarsModelVersion =\n vlMode === 'vlm-ui-tars' ? modelConfig.uiTarsModelVersion : 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 ? uiTarsPlanning : plan)(\n param.userInstruction,\n {\n context: uiContext,\n actionContext,\n interfaceType: this.interface.interfaceType as InterfaceType,\n actionSpace,\n modelConfig,\n conversationHistory: this.conversationHistory,\n },\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\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 if (finalActions.length === 0) {\n assert(\n !more_actions_needed_by_instruction || sleep,\n error ? `Failed to plan: ${error}` : 'No plan found',\n );\n }\n\n return {\n output: {\n actions: finalActions,\n more_actions_needed_by_instruction,\n log,\n yamlFlow: planResult.yamlFlow,\n },\n cache: {\n hit: false,\n },\n uiContext,\n };\n },\n };\n\n return task;\n }\n\n async runPlans(\n title: string,\n plans: PlanningAction[],\n modelConfig: IModelConfig,\n ): Promise<ExecutionResult> {\n const session = this.createExecutionSession(title);\n const { tasks } = await this.convertPlanToExecutable(plans, modelConfig);\n const result = await session.appendAndRun(tasks);\n const { output } = result!;\n return {\n output,\n runner: session.getRunner(),\n };\n }\n\n private getReplanningCycleLimit(isVlmUiTars: boolean) {\n return (\n this.replanningCycleLimit ||\n globalConfigManager.getEnvConfigInNumber(\n MIDSCENE_REPLANNING_CYCLE_LIMIT,\n ) ||\n (isVlmUiTars\n ? defaultVlmUiTarsReplanningCycleLimit\n : defaultReplanningCycleLimit)\n );\n }\n\n async action(\n userPrompt: string,\n modelConfig: IModelConfig,\n actionContext?: string,\n cacheable?: boolean,\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 = this.getReplanningCycleLimit(\n modelConfig.vlMode === 'vlm-ui-tars',\n );\n\n // Main planning loop - unified plan/replan logic\n while (true) {\n if (replanCount > replanningCycleLimit) {\n const errorMsg = `Replanning ${replanningCycleLimit} times, which is more than the limit, please split the task into multiple steps`;\n\n return session.appendErrorPlan(errorMsg);\n }\n\n // Create planning task (automatically includes execution history if available)\n const planningTask = this.createPlanningTask(\n userPrompt,\n actionContext,\n modelConfig,\n );\n\n const result = await session.appendAndRun(planningTask);\n const planResult: PlanningAIResponse = result?.output;\n if (session.isInErrorState()) {\n return {\n output: planResult,\n runner,\n };\n }\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(plans, modelConfig, {\n cacheable,\n subTask: true,\n });\n await session.appendAndRun(executables.tasks);\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 (session.isInErrorState()) {\n return {\n output: undefined,\n runner,\n };\n }\n\n // Check if task is complete\n if (!planResult.more_actions_needed_by_instruction) {\n break;\n }\n\n // Increment replan count for next iteration\n replanCount++;\n }\n\n return {\n output: {\n yamlFlow,\n },\n runner,\n };\n }\n\n private createTypeQueryTask(\n type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert' | 'WaitFor',\n demand: InsightExtractParam,\n modelConfig: IModelConfig,\n opt?: InsightExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ) {\n const queryTask: ExecutionTaskInsightQueryApply = {\n type: 'Insight',\n subType: type,\n locate: null,\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: InsightDump | undefined;\n const applyDump = (dump: InsightDump) => {\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.insight.extract<any>(\n demandInput,\n modelConfig,\n opt,\n extraPageDescription,\n multimodalPrompt,\n );\n } catch (error) {\n if (error instanceof InsightError) {\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 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: InsightExtractParam,\n modelConfig: IModelConfig,\n opt?: InsightExtractOption,\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 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: session.getRunner(),\n };\n }\n\n private sleepPlan(timeMs: number): PlanningAction<PlanningActionParamSleep> {\n return {\n type: 'Sleep',\n param: {\n timeMs,\n },\n locate: null,\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 startTime = Date.now();\n let errorThought = '';\n while (Date.now() - overallStartTime < timeoutMs) {\n startTime = Date.now();\n const queryTask = await this.createTypeQueryTask(\n 'WaitFor',\n textPrompt,\n modelConfig,\n {\n doNotThrowError: true,\n },\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 - startTime < checkIntervalMs) {\n const timeRemaining = checkIntervalMs - (now - startTime);\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","defaultReplanningCycleLimit","defaultVlmUiTarsReplanningCycleLimit","TaskExecutor","title","options","ExecutionSession","Promise","plans","modelConfig","userInstruction","yamlString","session","taskTitleStr","task","param","executorContext","uiContext","assert","actionContext","startTime","Date","vlMode","uiTarsModelVersion","undefined","actionSpace","action","Array","console","planResult","uiTarsPlanning","plan","JSON","actions","log","more_actions_needed_by_instruction","error","usage","rawResponse","sleep","finalActions","timeNow","timeRemaining","tasks","result","output","isVlmUiTars","globalConfigManager","MIDSCENE_REPLANNING_CYCLE_LIMIT","userPrompt","cacheable","runner","replanCount","yamlFlow","replanningCycleLimit","errorMsg","planningTask","executables","type","demand","opt","multimodalPrompt","queryTask","taskContext","queryDump","applyDump","dump","ifTypeRestricted","demandInput","keyOfResult","booleanPrompt","extractResult","extraPageDescription","tree","descriptionOfTree","InsightError","data","thought","outputResult","Error","timeMs","_modelConfig","assertion","textPrompt","parsePrompt","description","timeoutMs","checkIntervalMs","overallStartTime","errorThought","now","sleepTask","interfaceInstance","insight","opts","ConversationHistory","TaskBuilder"],"mappings":";;;;;;;;;;;;;;;;;;;;AA0CA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,8BAA8B;AACpC,MAAMC,uCAAuC;AAEtC,MAAMC;IAgBX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IAwBQ,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,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,KAAK;QACvB;IAEJ;IAEA,MAAa,wBACXG,KAAuB,EACvBC,WAAyB,EACzBJ,OAGC,EACD;QACA,OAAO,IAAI,CAAC,WAAW,CAAC,KAAK,CAACG,OAAOC,aAAaJ;IACpD;IAEA,MAAM,uBAAuBK,eAAuB,EAAEC,UAAkB,EAAE;QACxE,MAAMC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUH;QAGzB,MAAMI,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,QAAQ;YACR,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,MAAMC,QAAQ,YAAY,CAACE;QAE3B,OAAO;YACL,QAAQF,QAAQ,SAAS;QAC3B;IACF;IAEQ,mBACNF,eAAuB,EACvBS,aAAiC,EACjCV,WAAyB,EACG;QAC5B,MAAMK,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,QAAQ;YACR,OAAO;gBACLJ;YACF;YACA,UAAU,OAAOK,OAAOC;gBACtB,MAAMI,YAAYC,KAAK,GAAG;gBAC1B,MAAM,EAAEJ,SAAS,EAAE,GAAGD;gBACtBE,OAAOD,WAAW;gBAClB,MAAM,EAAEK,MAAM,EAAE,GAAGb;gBACnB,MAAMc,qBACJD,AAAW,kBAAXA,SAA2Bb,YAAY,kBAAkB,GAAGe;gBAE9DN,OACE,IAAI,CAAC,SAAS,CAAC,WAAW,EAC1B;gBAEF,MAAMO,cAAc,MAAM,IAAI,CAAC,SAAS,CAAC,WAAW;gBACpD1B,MACE,sCACA0B,YAAY,GAAG,CAAC,CAACC,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;gBAEhDR,OAAOS,MAAM,OAAO,CAACF,cAAc;gBACnC,IAAIA,AAAuB,MAAvBA,YAAY,MAAM,EACpBG,QAAQ,IAAI,CACV,CAAC,gBAAgB,EAAE,IAAI,CAAC,SAAS,CAAC,aAAa,CAAC,gDAAgD,CAAC;gBAIrG,MAAMC,aAAa,MAAON,AAAAA,CAAAA,qBAAqBO,iBAAiBC,IAAG,EACjEhB,MAAM,eAAe,EACrB;oBACE,SAASE;oBACTE;oBACA,eAAe,IAAI,CAAC,SAAS,CAAC,aAAa;oBAC3CM;oBACAhB;oBACA,qBAAqB,IAAI,CAAC,mBAAmB;gBAC/C;gBAEFV,MAAM,cAAciC,KAAK,SAAS,CAACH,YAAY,MAAM;gBAErD,MAAM,EACJI,OAAO,EACPC,GAAG,EACHC,kCAAkC,EAClCC,KAAK,EACLC,KAAK,EACLC,WAAW,EACXC,KAAK,EACN,GAAGV;gBAEJb,gBAAgB,IAAI,CAAC,GAAG,GAAG;oBACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;oBAClCsB;gBACF;gBACAtB,gBAAgB,IAAI,CAAC,KAAK,GAAGqB;gBAE7B,MAAMG,eAAeP,WAAW,EAAE;gBAElC,IAAIM,OAAO;oBACT,MAAME,UAAUpB,KAAK,GAAG;oBACxB,MAAMqB,gBAAgBH,QAASE,CAAAA,UAAUrB,SAAQ;oBACjD,IAAIsB,gBAAgB,GAClBF,aAAa,IAAI,CAAC,IAAI,CAAC,SAAS,CAACE;gBAErC;gBAEA,IAAIF,AAAwB,MAAxBA,aAAa,MAAM,EACrBtB,OACE,CAACiB,sCAAsCI,OACvCH,QAAQ,CAAC,gBAAgB,EAAEA,OAAO,GAAG;gBAIzC,OAAO;oBACL,QAAQ;wBACN,SAASI;wBACTL;wBACAD;wBACA,UAAUL,WAAW,QAAQ;oBAC/B;oBACA,OAAO;wBACL,KAAK;oBACP;oBACAZ;gBACF;YACF;QACF;QAEA,OAAOH;IACT;IAEA,MAAM,SACJV,KAAa,EACbI,KAAuB,EACvBC,WAAyB,EACC;QAC1B,MAAMG,UAAU,IAAI,CAAC,sBAAsB,CAACR;QAC5C,MAAM,EAAEuC,KAAK,EAAE,GAAG,MAAM,IAAI,CAAC,uBAAuB,CAACnC,OAAOC;QAC5D,MAAMmC,SAAS,MAAMhC,QAAQ,YAAY,CAAC+B;QAC1C,MAAM,EAAEE,MAAM,EAAE,GAAGD;QACnB,OAAO;YACLC;YACA,QAAQjC,QAAQ,SAAS;QAC3B;IACF;IAEQ,wBAAwBkC,WAAoB,EAAE;QACpD,OACE,IAAI,CAAC,oBAAoB,IACzBC,oBAAoB,oBAAoB,CACtCC,oCAEDF,CAAAA,cACG5C,uCACAD,2BAA0B;IAElC;IAEA,MAAM,OACJgD,UAAkB,EAClBxC,WAAyB,EACzBU,aAAsB,EACtB+B,SAAmB,EAQnB;QACA,IAAI,CAAC,mBAAmB,CAAC,KAAK;QAE9B,MAAMtC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUoC;QAEzB,MAAME,SAASvC,QAAQ,SAAS;QAEhC,IAAIwC,cAAc;QAClB,MAAMC,WAAmC,EAAE;QAC3C,MAAMC,uBAAuB,IAAI,CAAC,uBAAuB,CACvD7C,AAAuB,kBAAvBA,YAAY,MAAM;QAIpB,MAAO,KAAM;YACX,IAAI2C,cAAcE,sBAAsB;gBACtC,MAAMC,WAAW,CAAC,WAAW,EAAED,qBAAqB,+EAA+E,CAAC;gBAEpI,OAAO1C,QAAQ,eAAe,CAAC2C;YACjC;YAGA,MAAMC,eAAe,IAAI,CAAC,kBAAkB,CAC1CP,YACA9B,eACAV;YAGF,MAAMmC,SAAS,MAAMhC,QAAQ,YAAY,CAAC4C;YAC1C,MAAM3B,aAAiCe,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM;YACrD,IAAIhC,QAAQ,cAAc,IACxB,OAAO;gBACL,QAAQiB;gBACRsB;YACF;YAIF,MAAM3C,QAAQqB,WAAW,OAAO,IAAI,EAAE;YACtCwB,SAAS,IAAI,IAAKxB,WAAW,QAAQ,IAAI,EAAE;YAE3C,IAAI4B;YACJ,IAAI;gBACFA,cAAc,MAAM,IAAI,CAAC,uBAAuB,CAACjD,OAAOC,aAAa;oBACnEyC;oBACA,SAAS;gBACX;gBACA,MAAMtC,QAAQ,YAAY,CAAC6C,YAAY,KAAK;YAC9C,EAAE,OAAOrB,OAAO;gBACd,OAAOxB,QAAQ,eAAe,CAC5B,CAAC,4CAA4C,EAAEwB,MAAM,SAAS,EAAEJ,KAAK,SAAS,CAC5ExB,QACC;YAEP;YACA,IAAII,QAAQ,cAAc,IACxB,OAAO;gBACL,QAAQY;gBACR2B;YACF;YAIF,IAAI,CAACtB,WAAW,kCAAkC,EAChD;YAIFuB;QACF;QAEA,OAAO;YACL,QAAQ;gBACNC;YACF;YACAF;QACF;IACF;IAEQ,oBACNO,IAAsE,EACtEC,MAA2B,EAC3BlD,WAAyB,EACzBmD,GAA0B,EAC1BC,gBAAoC,EACpC;QACA,MAAMC,YAA4C;YAChD,MAAM;YACN,SAASJ;YACT,QAAQ;YACR,OAAO;gBACL,YAAYG,mBACP;oBACCF;oBACAE;gBACF,IACAF;YACN;YACA,UAAU,OAAO5C,OAAOgD;gBACtB,MAAM,EAAEjD,IAAI,EAAE,GAAGiD;gBACjB,IAAIC;gBACJ,MAAMC,YAAY,CAACC;oBACjBF,YAAYE;oBACZpD,KAAK,GAAG,GAAG;wBACToD;oBACF;gBACF;gBAGA,MAAMjD,YAAY8C,YAAY,SAAS;gBACvC7C,OAAOD,WAAW;gBAElB,MAAMkD,mBAAmBT,AAAS,YAATA;gBACzB,IAAIU,cAAcT;gBAClB,IAAIU,cAAc;gBAClB,IAAIF,oBAAqBT,CAAAA,AAAS,aAATA,QAAqBA,AAAS,cAATA,IAAiB,GAAI;oBACjEW,cAAc;oBACd,MAAMC,gBACJZ,AAAS,aAATA,OACI,CAAC,kDAAkD,EAAEC,QAAQ,GAC7D,CAAC,+GAA+G,EAAEA,QAAQ;oBAChIS,cAAc;wBACZ,CAACC,YAAY,EAAEC;oBACjB;gBACF,OAAO,IAAIH,kBACTC,cAAc;oBACZ,CAACC,YAAY,EAAE,GAAGX,KAAK,EAAE,EAAEC,QAAQ;gBACrC;gBAGF,IAAIY;gBAEJ,IAAIC,uBAAuB;gBAC3B,IAAIZ,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,KAAK,IAAI,CAAC,SAAS,CAAC,mBAAmB,EAAE;oBAC1D7D,MAAM;oBACN,MAAM0E,OAAO,MAAM,IAAI,CAAC,SAAS,CAAC,mBAAmB;oBACrDD,uBAAuB,MAAME,kBAC3BD,MACA,KACA,OACAb,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,MAAM;gBAEzB;gBAEA,IAAI;oBACFW,gBAAgB,MAAM,IAAI,CAAC,OAAO,CAAC,OAAO,CACxCH,aACA3D,aACAmD,KACAY,sBACAX;gBAEJ,EAAE,OAAOzB,OAAO;oBACd,IAAIA,iBAAiBuC,cACnBV,UAAU7B,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAEA,MAAM,EAAEwC,IAAI,EAAEvC,KAAK,EAAEwC,OAAO,EAAEX,IAAI,EAAE,GAAGK;gBACvCN,UAAUC;gBAEV,IAAIY,eAAeF;gBACnB,IAAIT,kBAEF,IAAI,AAAgB,YAAhB,OAAOS,MACTE,eAAeF;qBACV,IAAIlB,AAAS,cAATA,MAEPoB,eADEF,QAAAA,OACa,QAECA,IAAY,CAACP,YAAY;qBAEtC,IAAIO,QAAAA,MACTE,eAAe;qBACV;oBACL5D,OACE0D,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,IAAM,CAACP,YAAY,AAAD,MAAM7C,QACxB;oBAEFsD,eAAgBF,IAAY,CAACP,YAAY;gBAC3C;gBAGF,OAAO;oBACL,QAAQS;oBACR,KAAKd;oBACL3B;oBACAwC;gBACF;YACF;QACF;QAEA,OAAOf;IACT;IACA,MAAM,yBACJJ,IAA0D,EAC1DC,MAA2B,EAC3BlD,WAAyB,EACzBmD,GAA0B,EAC1BC,gBAAoC,EACP;QAC7B,MAAMjD,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aACE6C,MACA,AAAkB,YAAlB,OAAOC,SAAsBA,SAAS3B,KAAK,SAAS,CAAC2B;QAIzD,MAAMG,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9CJ,MACAC,QACAlD,aACAmD,KACAC;QAGF,MAAMjB,SAAS,MAAMhC,QAAQ,YAAY,CAACkD;QAE1C,IAAI,CAAClB,QACH,MAAM,IAAImC,MACR;QAIJ,MAAM,EAAElC,MAAM,EAAEgC,OAAO,EAAE,GAAGjC;QAE5B,OAAO;YACLC;YACAgC;YACA,QAAQjE,QAAQ,SAAS;QAC3B;IACF;IAEQ,UAAUoE,MAAc,EAA4C;QAC1E,OAAO;YACL,MAAM;YACN,OAAO;gBACLA;YACF;YACA,QAAQ;QACV;IACF;IAEA,MAAM,aAAaA,MAAc,EAAEC,YAA0B,EAAE;QAC7D,OAAO,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;YACtCD;QACF;IACF;IAEA,MAAM,QACJE,SAAsB,EACtBtB,GAA+B,EAC/BnD,WAAyB,EACO;QAChC,MAAM,EAAE0E,UAAU,EAAEtB,gBAAgB,EAAE,GAAGuB,YAAYF;QAErD,MAAMG,cAAc,CAAC,SAAS,EAAEF,YAAY;QAC5C,MAAMvE,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,WAAWwE;QAE1B,MAAMlC,SAASvC,QAAQ,SAAS;QAChC,MAAM,EAAE0E,SAAS,EAAEC,eAAe,EAAE,GAAG3B;QAEvC1C,OAAOgE,WAAW;QAClBhE,OAAOoE,WAAW;QAClBpE,OAAOqE,iBAAiB;QAExBrE,OACEqE,mBAAmBD,WACnB,CAAC,iGAAiG,EAAEC,gBAAgB,aAAa,EAAED,UAAU,CAAC,CAAC;QAGjJ,MAAME,mBAAmBnE,KAAK,GAAG;QACjC,IAAID,YAAYC,KAAK,GAAG;QACxB,IAAIoE,eAAe;QACnB,MAAOpE,KAAK,GAAG,KAAKmE,mBAAmBF,UAAW;YAChDlE,YAAYC,KAAK,GAAG;YACpB,MAAMyC,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9C,WACAqB,YACA1E,aACA;gBACE,iBAAiB;YACnB,GACAoD;YAGF,MAAMjB,SAAU,MAAMhC,QAAQ,YAAY,CAACkD;YAO3C,IAAIlB,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM,EAChB,OAAO;gBACL,QAAQpB;gBACR2B;YACF;YAGFsC,eACE7C,AAAAA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,OAAO,AAAD,KACb,CAACA,UAAU,CAAC,0BAA0B,EAAEuC,YAAY,IACrD,CAAC,0CAA0C,EAAEA,YAAY;YAC3D,MAAMO,MAAMrE,KAAK,GAAG;YACpB,IAAIqE,MAAMtE,YAAYmE,iBAAiB;gBACrC,MAAM7C,gBAAgB6C,kBAAmBG,CAAAA,MAAMtE,SAAQ;gBACvD,MAAMuE,YAAY,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;oBACjD,QAAQjD;gBACV;gBACA,MAAM9B,QAAQ,MAAM,CAAC+E;YACvB;QACF;QAEA,OAAO/E,QAAQ,eAAe,CAAC,CAAC,iBAAiB,EAAE6E,cAAc;IACnE;IAjiBA,YACEG,iBAAoC,EACpCC,OAAgB,EAChBC,IAIC,CACD;QA3BF;QAEA;QAEA;QAEA,uBAAiB,eAAjB;QAEA,uBAAQ,uBAAR;QAEA;QAEA;QAgBE,IAAI,CAAC,SAAS,GAAGF;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC,KAAK,SAAS;QAC/B,IAAI,CAAC,mBAAmB,GAAGA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW;QAC5C,IAAI,CAAC,oBAAoB,GAAGA,KAAK,oBAAoB;QACrD,IAAI,CAAC,mBAAmB,GAAG,IAAIC;QAC/B,IAAI,CAAC,WAAW,GAAG,IAAIC,YAAY;YACjCJ;YACAC;YACA,WAAWC,KAAK,SAAS;QAC3B;IACF;AA8gBF"}
|
|
@@ -43,22 +43,22 @@ function paramStr(task) {
|
|
|
43
43
|
value = null == task ? void 0 : null == (_task_param = task.param) ? void 0 : _task_param.userInstruction;
|
|
44
44
|
}
|
|
45
45
|
if ('Insight' === task.type) {
|
|
46
|
-
var _task_param1, _task_param2
|
|
47
|
-
value = locateParamStr(null == task ? void 0 : task.param) || (null == task ? void 0 : null == (_task_param1 = task.param) ? void 0 : _task_param1.
|
|
46
|
+
var _task_param1, _task_param2;
|
|
47
|
+
value = locateParamStr(null == task ? void 0 : task.param) || (null == task ? void 0 : null == (_task_param1 = task.param) ? void 0 : _task_param1.dataDemand) || (null == task ? void 0 : null == (_task_param2 = task.param) ? void 0 : _task_param2.assertion);
|
|
48
48
|
}
|
|
49
49
|
if ('Action' === task.type) {
|
|
50
|
-
var _task_param4, _task_param5, _task_param6
|
|
50
|
+
var _task_param3, _task_param4, _task_param5, _task_param6;
|
|
51
51
|
const locate = null == task ? void 0 : task.locate;
|
|
52
52
|
const locateStr = locate ? locateParamStr(locate) : '';
|
|
53
53
|
value = task.thought || '';
|
|
54
|
-
if ('number' == typeof (null == task ? void 0 : null == (
|
|
54
|
+
if ('number' == typeof (null == task ? void 0 : null == (_task_param3 = task.param) ? void 0 : _task_param3.timeMs)) {
|
|
55
|
+
var _task_param7;
|
|
56
|
+
value = `${null == task ? void 0 : null == (_task_param7 = task.param) ? void 0 : _task_param7.timeMs}ms`;
|
|
57
|
+
} else if ('string' == typeof (null == task ? void 0 : null == (_task_param4 = task.param) ? void 0 : _task_param4.scrollType)) value = scrollParamStr(null == task ? void 0 : task.param);
|
|
58
|
+
else if ('string' == typeof (null == task ? void 0 : null == (_task_param5 = task.param) ? void 0 : _task_param5.direction) && (null == task ? void 0 : task.subType) === 'AndroidPull') value = pullParamStr(null == task ? void 0 : task.param);
|
|
59
|
+
else if (void 0 !== (null == task ? void 0 : null == (_task_param6 = task.param) ? void 0 : _task_param6.value)) {
|
|
55
60
|
var _task_param8;
|
|
56
|
-
value =
|
|
57
|
-
} else if ('string' == typeof (null == task ? void 0 : null == (_task_param5 = task.param) ? void 0 : _task_param5.scrollType)) value = scrollParamStr(null == task ? void 0 : task.param);
|
|
58
|
-
else if ('string' == typeof (null == task ? void 0 : null == (_task_param6 = task.param) ? void 0 : _task_param6.direction) && (null == task ? void 0 : task.subType) === 'AndroidPull') value = pullParamStr(null == task ? void 0 : task.param);
|
|
59
|
-
else if (void 0 !== (null == task ? void 0 : null == (_task_param7 = task.param) ? void 0 : _task_param7.value)) {
|
|
60
|
-
var _task_param9;
|
|
61
|
-
value = null == task ? void 0 : null == (_task_param9 = task.param) ? void 0 : _task_param9.value;
|
|
61
|
+
value = null == task ? void 0 : null == (_task_param8 = task.param) ? void 0 : _task_param8.value;
|
|
62
62
|
}
|
|
63
63
|
if (locateStr) value = value ? `${locateStr} - ${value}` : locateStr;
|
|
64
64
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"agent/ui-utils.mjs","sources":["webpack://@midscene/core/./src/agent/ui-utils.ts"],"sourcesContent":["import type {\n DetailedLocateParam,\n ExecutionTask,\n ExecutionTaskAction,\n ExecutionTaskInsightAssertion,\n ExecutionTaskInsightLocate,\n ExecutionTaskInsightQuery,\n ExecutionTaskPlanning,\n PullParam,\n ScrollParam,\n} from '@/types';\n\nexport function typeStr(task: ExecutionTask) {\n return task.subType && task.subType !== 'Plan'\n ? `${task.type} / ${task.subType || ''}`\n : task.type;\n}\n\nexport function locateParamStr(locate?: DetailedLocateParam | string): string {\n if (!locate) {\n return '';\n }\n\n if (typeof locate === 'string') {\n return locate;\n }\n\n if (typeof locate === 'object') {\n if (typeof locate.prompt === 'string') {\n return locate.prompt;\n }\n\n if (typeof locate.prompt === 'object' && locate.prompt.prompt) {\n const prompt = locate.prompt.prompt;\n const images = locate.prompt.images || [];\n\n if (images.length === 0) return prompt;\n\n const imagesStr = images\n .map((image) => {\n let url = image.url;\n if (\n url.startsWith('data:image/') ||\n (url.startsWith('data:') && url.includes('base64'))\n ) {\n url = `${url.substring(0, 15)}...`;\n }\n return `[${image.name}](${url})`;\n })\n .join(', ');\n\n return `${prompt}, ${imagesStr}`;\n }\n }\n\n return '';\n}\n\nexport function scrollParamStr(scrollParam?: ScrollParam) {\n if (!scrollParam) {\n return '';\n }\n return `${scrollParam.direction || 'down'}, ${scrollParam.scrollType || 'once'}, ${scrollParam.distance || 'distance-not-set'}`;\n}\n\nexport function pullParamStr(pullParam?: PullParam) {\n if (!pullParam) {\n return '';\n }\n const parts: string[] = [];\n parts.push(`direction: ${pullParam.direction || 'down'}`);\n if (pullParam.distance) {\n parts.push(`distance: ${pullParam.distance}`);\n }\n if (pullParam.duration) {\n parts.push(`duration: ${pullParam.duration}ms`);\n }\n return parts.join(', ');\n}\n\nexport function taskTitleStr(\n type:\n | 'Tap'\n | 'Hover'\n | 'Input'\n | 'RightClick'\n | 'KeyboardPress'\n | 'Scroll'\n | 'Action'\n | 'Query'\n | 'Assert'\n | 'WaitFor'\n | 'Locate'\n | 'Boolean'\n | 'Number'\n | 'String',\n prompt: string,\n) {\n if (prompt) {\n return `${type} - ${prompt}`;\n }\n return type;\n}\n\nexport function paramStr(task: ExecutionTask) {\n let value: string | undefined | object;\n if (task.type === 'Planning') {\n value = (task as ExecutionTaskPlanning)?.param?.userInstruction;\n }\n\n if (task.type === 'Insight') {\n value =\n locateParamStr((task as ExecutionTaskInsightLocate)?.param) ||\n (task as
|
|
1
|
+
{"version":3,"file":"agent/ui-utils.mjs","sources":["webpack://@midscene/core/./src/agent/ui-utils.ts"],"sourcesContent":["import type {\n DetailedLocateParam,\n ExecutionTask,\n ExecutionTaskAction,\n ExecutionTaskInsightAssertion,\n ExecutionTaskInsightLocate,\n ExecutionTaskInsightQuery,\n ExecutionTaskPlanning,\n PullParam,\n ScrollParam,\n} from '@/types';\n\nexport function typeStr(task: ExecutionTask) {\n return task.subType && task.subType !== 'Plan'\n ? `${task.type} / ${task.subType || ''}`\n : task.type;\n}\n\nexport function locateParamStr(locate?: DetailedLocateParam | string): string {\n if (!locate) {\n return '';\n }\n\n if (typeof locate === 'string') {\n return locate;\n }\n\n if (typeof locate === 'object') {\n if (typeof locate.prompt === 'string') {\n return locate.prompt;\n }\n\n if (typeof locate.prompt === 'object' && locate.prompt.prompt) {\n const prompt = locate.prompt.prompt;\n const images = locate.prompt.images || [];\n\n if (images.length === 0) return prompt;\n\n const imagesStr = images\n .map((image) => {\n let url = image.url;\n if (\n url.startsWith('data:image/') ||\n (url.startsWith('data:') && url.includes('base64'))\n ) {\n url = `${url.substring(0, 15)}...`;\n }\n return `[${image.name}](${url})`;\n })\n .join(', ');\n\n return `${prompt}, ${imagesStr}`;\n }\n }\n\n return '';\n}\n\nexport function scrollParamStr(scrollParam?: ScrollParam) {\n if (!scrollParam) {\n return '';\n }\n return `${scrollParam.direction || 'down'}, ${scrollParam.scrollType || 'once'}, ${scrollParam.distance || 'distance-not-set'}`;\n}\n\nexport function pullParamStr(pullParam?: PullParam) {\n if (!pullParam) {\n return '';\n }\n const parts: string[] = [];\n parts.push(`direction: ${pullParam.direction || 'down'}`);\n if (pullParam.distance) {\n parts.push(`distance: ${pullParam.distance}`);\n }\n if (pullParam.duration) {\n parts.push(`duration: ${pullParam.duration}ms`);\n }\n return parts.join(', ');\n}\n\nexport function taskTitleStr(\n type:\n | 'Tap'\n | 'Hover'\n | 'Input'\n | 'RightClick'\n | 'KeyboardPress'\n | 'Scroll'\n | 'Action'\n | 'Query'\n | 'Assert'\n | 'WaitFor'\n | 'Locate'\n | 'Boolean'\n | 'Number'\n | 'String',\n prompt: string,\n) {\n if (prompt) {\n return `${type} - ${prompt}`;\n }\n return type;\n}\n\nexport function paramStr(task: ExecutionTask) {\n let value: string | undefined | object;\n if (task.type === 'Planning') {\n value = (task as ExecutionTaskPlanning)?.param?.userInstruction;\n }\n\n if (task.type === 'Insight') {\n value =\n locateParamStr((task as ExecutionTaskInsightLocate)?.param) ||\n (task as ExecutionTaskInsightQuery)?.param?.dataDemand ||\n (task as ExecutionTaskInsightAssertion)?.param?.assertion;\n }\n\n if (task.type === 'Action') {\n const locate = (task as ExecutionTaskAction)?.locate;\n const locateStr = locate ? locateParamStr(locate) : '';\n\n value = task.thought || '';\n if (typeof (task as ExecutionTaskAction)?.param?.timeMs === 'number') {\n value = `${(task as ExecutionTaskAction)?.param?.timeMs}ms`;\n } else if (\n typeof (task as ExecutionTaskAction)?.param?.scrollType === 'string'\n ) {\n value = scrollParamStr((task as ExecutionTaskAction)?.param);\n } else if (\n typeof (task as ExecutionTaskAction)?.param?.direction === 'string' &&\n (task as ExecutionTaskAction)?.subType === 'AndroidPull'\n ) {\n value = pullParamStr((task as ExecutionTaskAction)?.param);\n } else if (\n typeof (task as ExecutionTaskAction)?.param?.value !== 'undefined'\n ) {\n value = (task as ExecutionTaskAction)?.param?.value;\n }\n\n if (locateStr) {\n if (value) {\n value = `${locateStr} - ${value}`;\n } else {\n value = locateStr;\n }\n }\n }\n\n if (typeof value === 'undefined') return '';\n\n if (typeof value === 'string') return value;\n\n if (typeof value === 'object' && locateParamStr(value as any)) {\n return locateParamStr(value as any);\n }\n\n return JSON.stringify(value, undefined, 2);\n}\n"],"names":["typeStr","task","locateParamStr","locate","prompt","images","imagesStr","image","url","scrollParamStr","scrollParam","pullParamStr","pullParam","parts","taskTitleStr","type","paramStr","value","_task_param","_task_param1","_task_param2","_task_param3","_task_param4","_task_param5","_task_param6","locateStr","_task_param7","_task_param8","JSON","undefined"],"mappings":"AAYO,SAASA,QAAQC,IAAmB;IACzC,OAAOA,KAAK,OAAO,IAAIA,AAAiB,WAAjBA,KAAK,OAAO,GAC/B,GAAGA,KAAK,IAAI,CAAC,GAAG,EAAEA,KAAK,OAAO,IAAI,IAAI,GACtCA,KAAK,IAAI;AACf;AAEO,SAASC,eAAeC,MAAqC;IAClE,IAAI,CAACA,QACH,OAAO;IAGT,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAGT,IAAI,AAAkB,YAAlB,OAAOA,QAAqB;QAC9B,IAAI,AAAyB,YAAzB,OAAOA,OAAO,MAAM,EACtB,OAAOA,OAAO,MAAM;QAGtB,IAAI,AAAyB,YAAzB,OAAOA,OAAO,MAAM,IAAiBA,OAAO,MAAM,CAAC,MAAM,EAAE;YAC7D,MAAMC,SAASD,OAAO,MAAM,CAAC,MAAM;YACnC,MAAME,SAASF,OAAO,MAAM,CAAC,MAAM,IAAI,EAAE;YAEzC,IAAIE,AAAkB,MAAlBA,OAAO,MAAM,EAAQ,OAAOD;YAEhC,MAAME,YAAYD,OACf,GAAG,CAAC,CAACE;gBACJ,IAAIC,MAAMD,MAAM,GAAG;gBACnB,IACEC,IAAI,UAAU,CAAC,kBACdA,IAAI,UAAU,CAAC,YAAYA,IAAI,QAAQ,CAAC,WAEzCA,MAAM,GAAGA,IAAI,SAAS,CAAC,GAAG,IAAI,GAAG,CAAC;gBAEpC,OAAO,CAAC,CAAC,EAAED,MAAM,IAAI,CAAC,EAAE,EAAEC,IAAI,CAAC,CAAC;YAClC,GACC,IAAI,CAAC;YAER,OAAO,GAAGJ,OAAO,EAAE,EAAEE,WAAW;QAClC;IACF;IAEA,OAAO;AACT;AAEO,SAASG,eAAeC,WAAyB;IACtD,IAAI,CAACA,aACH,OAAO;IAET,OAAO,GAAGA,YAAY,SAAS,IAAI,OAAO,EAAE,EAAEA,YAAY,UAAU,IAAI,OAAO,EAAE,EAAEA,YAAY,QAAQ,IAAI,oBAAoB;AACjI;AAEO,SAASC,aAAaC,SAAqB;IAChD,IAAI,CAACA,WACH,OAAO;IAET,MAAMC,QAAkB,EAAE;IAC1BA,MAAM,IAAI,CAAC,CAAC,WAAW,EAAED,UAAU,SAAS,IAAI,QAAQ;IACxD,IAAIA,UAAU,QAAQ,EACpBC,MAAM,IAAI,CAAC,CAAC,UAAU,EAAED,UAAU,QAAQ,EAAE;IAE9C,IAAIA,UAAU,QAAQ,EACpBC,MAAM,IAAI,CAAC,CAAC,UAAU,EAAED,UAAU,QAAQ,CAAC,EAAE,CAAC;IAEhD,OAAOC,MAAM,IAAI,CAAC;AACpB;AAEO,SAASC,aACdC,IAcY,EACZX,MAAc;IAEd,IAAIA,QACF,OAAO,GAAGW,KAAK,GAAG,EAAEX,QAAQ;IAE9B,OAAOW;AACT;AAEO,SAASC,SAASf,IAAmB;IAC1C,IAAIgB;IACJ,IAAIhB,AAAc,eAAdA,KAAK,IAAI,EAAiB;YACpBiB;QAARD,QAAShB,QAAAA,OAAAA,KAAAA,IAAAA,QAADiB,CAAAA,cAACjB,KAAgC,KAAK,AAAD,IAArCiB,KAAAA,IAAAA,YAAwC,eAAe;IACjE;IAEA,IAAIjB,AAAc,cAAdA,KAAK,IAAI,EAAgB;YAGzBkB,cACAC;QAHFH,QACEf,eAAgBD,QAAAA,OAAAA,KAAAA,IAAAA,KAAqC,KAAK,KACzDA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADkB,CAAAA,eAAClB,KAAoC,KAAK,AAAD,IAAzCkB,KAAAA,IAAAA,aAA4C,UAAU,AAAD,KACpDlB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADmB,CAAAA,eAACnB,KAAwC,KAAK,AAAD,IAA7CmB,KAAAA,IAAAA,aAAgD,SAAS,AAAD;IAC5D;IAEA,IAAInB,AAAc,aAAdA,KAAK,IAAI,EAAe;YAKfoB,cAGFC,cAIAC,cAKAC;QAhBT,MAAMrB,SAAUF,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,MAAM;QACpD,MAAMwB,YAAYtB,SAASD,eAAeC,UAAU;QAEpDc,QAAQhB,KAAK,OAAO,IAAI;QACxB,IAAI,AAAwD,YAAxD,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADoB,CAAAA,eAACpB,KAA8B,KAAK,AAAD,IAAnCoB,KAAAA,IAAAA,aAAsC,MAAM,AAAD,GAAgB;gBACzDK;YAAXT,QAAQ,GAAIhB,QAAAA,OAAAA,KAAAA,IAAAA,QAADyB,CAAAA,eAACzB,KAA8B,KAAK,AAAD,IAAnCyB,KAAAA,IAAAA,aAAsC,MAAM,CAAC,EAAE,CAAC;QAC7D,OAAO,IACL,AAA4D,YAA5D,OAAQzB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADqB,CAAAA,eAACrB,KAA8B,KAAK,AAAD,IAAnCqB,KAAAA,IAAAA,aAAsC,UAAU,AAAD,GAEtDL,QAAQR,eAAgBR,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;aACtD,IACL,AAA2D,YAA3D,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADsB,CAAAA,eAACtB,KAA8B,KAAK,AAAD,IAAnCsB,KAAAA,IAAAA,aAAsC,SAAS,AAAD,KACrD,AAACtB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,OAAO,AAAD,MAAM,eAE3CgB,QAAQN,aAAcV,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;aACpD,IACL,AAAuD,WAA/CA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADuB,CAAAA,eAACvB,KAA8B,KAAK,AAAD,IAAnCuB,KAAAA,IAAAA,aAAsC,KAAK,AAAD,GACjD;gBACQG;YAARV,QAAShB,QAAAA,OAAAA,KAAAA,IAAAA,QAAD0B,CAAAA,eAAC1B,KAA8B,KAAK,AAAD,IAAnC0B,KAAAA,IAAAA,aAAsC,KAAK;QACrD;QAEA,IAAIF,WAEAR,QADEA,QACM,GAAGQ,UAAU,GAAG,EAAER,OAAO,GAEzBQ;IAGd;IAEA,IAAI,AAAiB,WAAVR,OAAuB,OAAO;IAEzC,IAAI,AAAiB,YAAjB,OAAOA,OAAoB,OAAOA;IAEtC,IAAI,AAAiB,YAAjB,OAAOA,SAAsBf,eAAee,QAC9C,OAAOf,eAAee;IAGxB,OAAOW,KAAK,SAAS,CAACX,OAAOY,QAAW;AAC1C"}
|
package/dist/es/agent/utils.mjs
CHANGED
|
@@ -1,8 +1,6 @@
|
|
|
1
|
-
import { elementByPositionWithElementInfo } from "../ai-model/index.mjs";
|
|
2
1
|
import { uploadTestInfoToServer } from "../utils.mjs";
|
|
3
|
-
import { NodeType } from "@midscene/shared/constants";
|
|
4
2
|
import { MIDSCENE_REPORT_TAG_NAME, globalConfigManager } from "@midscene/shared/env";
|
|
5
|
-
import { generateElementByPosition
|
|
3
|
+
import { generateElementByPosition } from "@midscene/shared/extractor";
|
|
6
4
|
import { getDebug } from "@midscene/shared/logger";
|
|
7
5
|
import { assert, logMsg, uuid } from "@midscene/shared/utils";
|
|
8
6
|
import dayjs from "dayjs";
|
|
@@ -25,10 +23,6 @@ async function commonContextParser(interfaceInstance, _opt) {
|
|
|
25
23
|
const size = await interfaceInstance.size();
|
|
26
24
|
debugProfile(`size: ${size.width}x${size.height} dpr: ${size.dpr}`);
|
|
27
25
|
return {
|
|
28
|
-
tree: {
|
|
29
|
-
node: null,
|
|
30
|
-
children: []
|
|
31
|
-
},
|
|
32
26
|
size,
|
|
33
27
|
screenshotBase64: screenshotBase64
|
|
34
28
|
};
|
|
@@ -71,16 +65,14 @@ function generateCacheId(fileName) {
|
|
|
71
65
|
} else testFileIndex.set(taskFile, 1);
|
|
72
66
|
return `${taskFile}-${testFileIndex.get(taskFile)}`;
|
|
73
67
|
}
|
|
74
|
-
function matchElementFromPlan(planLocateParam
|
|
68
|
+
function matchElementFromPlan(planLocateParam) {
|
|
75
69
|
if (!planLocateParam) return;
|
|
76
|
-
if (planLocateParam.id) return getNodeFromCacheList(planLocateParam.id);
|
|
77
70
|
if (planLocateParam.bbox) {
|
|
78
71
|
const centerPosition = {
|
|
79
72
|
x: Math.floor((planLocateParam.bbox[0] + planLocateParam.bbox[2]) / 2),
|
|
80
73
|
y: Math.floor((planLocateParam.bbox[1] + planLocateParam.bbox[3]) / 2)
|
|
81
74
|
};
|
|
82
|
-
|
|
83
|
-
if (!element) element = generateElementByPosition(centerPosition);
|
|
75
|
+
const element = generateElementByPosition(centerPosition);
|
|
84
76
|
return element;
|
|
85
77
|
}
|
|
86
78
|
}
|
|
@@ -98,11 +90,7 @@ async function matchElementFromCache(context, cacheEntry, cachePrompt, cacheable
|
|
|
98
90
|
Math.round(rect.left + rect.width / 2),
|
|
99
91
|
Math.round(rect.top + rect.height / 2)
|
|
100
92
|
],
|
|
101
|
-
rect
|
|
102
|
-
xpaths: [],
|
|
103
|
-
attributes: {
|
|
104
|
-
nodeType: NodeType.POSITION
|
|
105
|
-
}
|
|
93
|
+
rect
|
|
106
94
|
};
|
|
107
95
|
debug('cache hit, prompt: %s', cachePrompt);
|
|
108
96
|
return element;
|
|
@@ -111,39 +99,7 @@ async function matchElementFromCache(context, cacheEntry, cachePrompt, cacheable
|
|
|
111
99
|
return;
|
|
112
100
|
}
|
|
113
101
|
}
|
|
114
|
-
|
|
115
|
-
function filterVisibleTree(node) {
|
|
116
|
-
if (!node) return null;
|
|
117
|
-
const filteredChildren = Array.isArray(node.children) ? node.children.map(filterVisibleTree).filter((child)=>null !== child) : [];
|
|
118
|
-
if (node.node && true === node.node.isVisible) return {
|
|
119
|
-
...node,
|
|
120
|
-
children: filteredChildren
|
|
121
|
-
};
|
|
122
|
-
if (filteredChildren.length > 0) return {
|
|
123
|
-
node: null,
|
|
124
|
-
children: filteredChildren
|
|
125
|
-
};
|
|
126
|
-
return null;
|
|
127
|
-
}
|
|
128
|
-
return {
|
|
129
|
-
...execution,
|
|
130
|
-
tasks: Array.isArray(execution.tasks) ? execution.tasks.map((task)=>{
|
|
131
|
-
var _task_uiContext;
|
|
132
|
-
const newTask = {
|
|
133
|
-
...task
|
|
134
|
-
};
|
|
135
|
-
if (null == (_task_uiContext = task.uiContext) ? void 0 : _task_uiContext.tree) newTask.uiContext = {
|
|
136
|
-
...task.uiContext,
|
|
137
|
-
tree: filterVisibleTree(task.uiContext.tree) || {
|
|
138
|
-
node: null,
|
|
139
|
-
children: []
|
|
140
|
-
}
|
|
141
|
-
};
|
|
142
|
-
return newTask;
|
|
143
|
-
}) : execution.tasks
|
|
144
|
-
};
|
|
145
|
-
}
|
|
146
|
-
const getMidsceneVersion = ()=>"1.0.1-beta-20251022061922.0";
|
|
102
|
+
const getMidsceneVersion = ()=>"1.0.1-beta-20251024064637.0";
|
|
147
103
|
const parsePrompt = (prompt)=>{
|
|
148
104
|
if ('string' == typeof prompt) return {
|
|
149
105
|
textPrompt: prompt,
|
|
@@ -157,6 +113,6 @@ const parsePrompt = (prompt)=>{
|
|
|
157
113
|
} : void 0
|
|
158
114
|
};
|
|
159
115
|
};
|
|
160
|
-
export { commonContextParser, generateCacheId, getCurrentExecutionFile, getMidsceneVersion, getReportFileName, matchElementFromCache, matchElementFromPlan, parsePrompt, printReportMsg
|
|
116
|
+
export { commonContextParser, generateCacheId, getCurrentExecutionFile, getMidsceneVersion, getReportFileName, matchElementFromCache, matchElementFromPlan, parsePrompt, printReportMsg };
|
|
161
117
|
|
|
162
118
|
//# sourceMappingURL=utils.mjs.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"agent/utils.mjs","sources":["webpack://@midscene/core/./src/agent/utils.ts"],"sourcesContent":["import { elementByPositionWithElementInfo } from '@/ai-model';\nimport type { TMultimodalPrompt, TUserPrompt } from '@/ai-model/common';\nimport type { AbstractInterface } from '@/device';\nimport type {\n BaseElement,\n ElementCacheFeature,\n ElementTreeNode,\n ExecutionDump,\n ExecutionTask,\n ExecutorContext,\n LocateResultElement,\n PlanningLocateParam,\n UIContext,\n} from '@/types';\nimport { uploadTestInfoToServer } from '@/utils';\nimport { NodeType } from '@midscene/shared/constants';\nimport {\n MIDSCENE_REPORT_TAG_NAME,\n globalConfigManager,\n} from '@midscene/shared/env';\nimport {\n generateElementByPosition,\n getNodeFromCacheList,\n} from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { assert, logMsg, uuid } from '@midscene/shared/utils';\nimport dayjs from 'dayjs';\nimport type { TaskCache } from './task-cache';\nimport { debug as cacheDebug } from './task-cache';\n\nconst debugProfile = getDebug('web:tool:profile');\n\nexport async function commonContextParser(\n interfaceInstance: AbstractInterface,\n _opt: { uploadServerUrl?: string },\n): Promise<UIContext> {\n assert(interfaceInstance, 'interfaceInstance is required');\n\n debugProfile('Getting interface description');\n const description = interfaceInstance.describe?.() || '';\n debugProfile('Interface description end');\n\n debugProfile('Uploading test info to server');\n uploadTestInfoToServer({\n testUrl: description,\n serverUrl: _opt.uploadServerUrl,\n });\n debugProfile('UploadTestInfoToServer end');\n\n const screenshotBase64 = await interfaceInstance.screenshotBase64();\n assert(screenshotBase64!, 'screenshotBase64 is required');\n\n const size = await interfaceInstance.size();\n debugProfile(`size: ${size.width}x${size.height} dpr: ${size.dpr}`);\n\n return {\n tree: {\n node: null,\n children: [],\n },\n size,\n screenshotBase64: screenshotBase64!,\n };\n}\n\nexport function getReportFileName(tag = 'web') {\n const reportTagName = globalConfigManager.getEnvConfigValue(\n MIDSCENE_REPORT_TAG_NAME,\n );\n const dateTimeInFileName = dayjs().format('YYYY-MM-DD_HH-mm-ss');\n // ensure uniqueness at the same time\n const uniqueId = uuid().substring(0, 8);\n return `${reportTagName || tag}-${dateTimeInFileName}-${uniqueId}`;\n}\n\nexport function printReportMsg(filepath: string) {\n logMsg(`Midscene - report file updated: ${filepath}`);\n}\n\n/**\n * Get the current execution file name\n * @returns The name of the current execution file\n */\nexport function getCurrentExecutionFile(trace?: string): string | false {\n const error = new Error();\n const stackTrace = trace || error.stack;\n const pkgDir = process.cwd() || '';\n if (stackTrace) {\n const stackLines = stackTrace.split('\\n');\n for (const line of stackLines) {\n if (\n line.includes('.spec.') ||\n line.includes('.test.') ||\n line.includes('.ts') ||\n line.includes('.js')\n ) {\n const match = line.match(/(?:at\\s+)?(.*?\\.(?:spec|test)\\.[jt]s)/);\n if (match?.[1]) {\n const targetFileName = match[1]\n .replace(pkgDir, '')\n .trim()\n .replace('at ', '');\n return targetFileName;\n }\n }\n }\n }\n return false;\n}\n\nconst testFileIndex = new Map<string, number>();\n\nexport function generateCacheId(fileName?: string): string {\n let taskFile = fileName || getCurrentExecutionFile();\n if (!taskFile) {\n taskFile = uuid();\n console.warn(\n 'Midscene - using random UUID for cache id. Cache may be invalid.',\n );\n }\n\n if (testFileIndex.has(taskFile)) {\n const currentIndex = testFileIndex.get(taskFile);\n if (currentIndex !== undefined) {\n testFileIndex.set(taskFile, currentIndex + 1);\n }\n } else {\n testFileIndex.set(taskFile, 1);\n }\n return `${taskFile}-${testFileIndex.get(taskFile)}`;\n}\n\nexport function matchElementFromPlan(\n planLocateParam: PlanningLocateParam,\n tree: ElementTreeNode<BaseElement>,\n) {\n if (!planLocateParam) {\n return undefined;\n }\n if (planLocateParam.id) {\n return getNodeFromCacheList(planLocateParam.id);\n }\n\n if (planLocateParam.bbox) {\n const centerPosition = {\n x: Math.floor((planLocateParam.bbox[0] + planLocateParam.bbox[2]) / 2),\n y: Math.floor((planLocateParam.bbox[1] + planLocateParam.bbox[3]) / 2),\n };\n let element = elementByPositionWithElementInfo(tree, centerPosition);\n\n if (!element) {\n element = generateElementByPosition(centerPosition) as BaseElement;\n }\n\n return element;\n }\n\n return undefined;\n}\n\nexport async function matchElementFromCache(\n context: {\n taskCache?: TaskCache;\n interfaceInstance: AbstractInterface;\n },\n cacheEntry: ElementCacheFeature | undefined,\n cachePrompt: TUserPrompt,\n cacheable: boolean | undefined,\n): Promise<LocateResultElement | undefined> {\n if (!cacheEntry) {\n return undefined;\n }\n\n if (cacheable === false) {\n cacheDebug('cache disabled for prompt: %s', cachePrompt);\n return undefined;\n }\n\n if (!context.taskCache?.isCacheResultUsed) {\n return undefined;\n }\n\n if (!context.interfaceInstance.rectMatchesCacheFeature) {\n cacheDebug(\n 'interface does not implement rectMatchesCacheFeature, skip cache',\n );\n return undefined;\n }\n\n try {\n const rect =\n await context.interfaceInstance.rectMatchesCacheFeature(cacheEntry);\n const element: LocateResultElement = {\n id: uuid(),\n center: [\n Math.round(rect.left + rect.width / 2),\n Math.round(rect.top + rect.height / 2),\n ],\n rect,\n xpaths: [],\n attributes: {\n nodeType: NodeType.POSITION,\n },\n };\n\n cacheDebug('cache hit, prompt: %s', cachePrompt);\n return element;\n } catch (error) {\n cacheDebug('rectMatchesCacheFeature error: %s', error);\n return undefined;\n }\n}\n\nexport function trimContextByViewport(execution: ExecutionDump) {\n function filterVisibleTree(\n node: ElementTreeNode<BaseElement>,\n ): ElementTreeNode<BaseElement> | null {\n if (!node) return null;\n\n // recursively process all children\n const filteredChildren = Array.isArray(node.children)\n ? (node.children\n .map(filterVisibleTree)\n .filter((child) => child !== null) as ElementTreeNode<BaseElement>[])\n : [];\n\n // if the current node is visible, keep it and the filtered children\n if (node.node && node.node.isVisible === true) {\n return {\n ...node,\n children: filteredChildren,\n };\n }\n\n // if the current node is invisible, but has visible children, create an empty node to include these children\n if (filteredChildren.length > 0) {\n return {\n node: null,\n children: filteredChildren,\n };\n }\n\n // if the current node is invisible and has no visible children, return null\n return null;\n }\n\n return {\n ...execution,\n tasks: Array.isArray(execution.tasks)\n ? execution.tasks.map((task: ExecutionTask) => {\n const newTask = { ...task };\n if (task.uiContext?.tree) {\n newTask.uiContext = {\n ...task.uiContext,\n tree: filterVisibleTree(task.uiContext.tree) || {\n node: null,\n children: [],\n },\n };\n }\n return newTask;\n })\n : execution.tasks,\n };\n}\n\ndeclare const __VERSION__: string | undefined;\n\nexport const getMidsceneVersion = (): string => {\n if (typeof __VERSION__ !== 'undefined') {\n return __VERSION__;\n } else if (\n process.env.__VERSION__ &&\n process.env.__VERSION__ !== 'undefined'\n ) {\n return process.env.__VERSION__;\n }\n throw new Error('__VERSION__ inject failed during build');\n};\n\nexport const parsePrompt = (\n prompt: TUserPrompt,\n): {\n textPrompt: string;\n multimodalPrompt?: TMultimodalPrompt;\n} => {\n if (typeof prompt === 'string') {\n return {\n textPrompt: prompt,\n multimodalPrompt: undefined,\n };\n }\n return {\n textPrompt: prompt.prompt,\n multimodalPrompt: prompt.images\n ? {\n images: prompt.images,\n convertHttpImage2Base64: !!prompt.convertHttpImage2Base64,\n }\n : undefined,\n };\n};\n"],"names":["debugProfile","getDebug","commonContextParser","interfaceInstance","_opt","assert","description","uploadTestInfoToServer","screenshotBase64","size","getReportFileName","tag","reportTagName","globalConfigManager","MIDSCENE_REPORT_TAG_NAME","dateTimeInFileName","dayjs","uniqueId","uuid","printReportMsg","filepath","logMsg","getCurrentExecutionFile","trace","error","Error","stackTrace","pkgDir","process","stackLines","line","match","targetFileName","testFileIndex","Map","generateCacheId","fileName","taskFile","console","currentIndex","undefined","matchElementFromPlan","planLocateParam","tree","getNodeFromCacheList","centerPosition","Math","element","elementByPositionWithElementInfo","generateElementByPosition","matchElementFromCache","context","cacheEntry","cachePrompt","cacheable","_context_taskCache","cacheDebug","rect","NodeType","trimContextByViewport","execution","filterVisibleTree","node","filteredChildren","Array","child","task","_task_uiContext","newTask","getMidsceneVersion","__VERSION__","parsePrompt","prompt"],"mappings":";;;;;;;;;AA+BA,MAAMA,eAAeC,SAAS;AAEvB,eAAeC,oBACpBC,iBAAoC,EACpCC,IAAkC;QAKdD;IAHpBE,OAAOF,mBAAmB;IAE1BH,aAAa;IACb,MAAMM,cAAcH,AAAAA,SAAAA,CAAAA,8BAAAA,kBAAkB,QAAQ,AAAD,IAAzBA,KAAAA,IAAAA,4BAAAA,IAAAA,CAAAA,kBAAiB,KAAiB;IACtDH,aAAa;IAEbA,aAAa;IACbO,uBAAuB;QACrB,SAASD;QACT,WAAWF,KAAK,eAAe;IACjC;IACAJ,aAAa;IAEb,MAAMQ,mBAAmB,MAAML,kBAAkB,gBAAgB;IACjEE,OAAOG,kBAAmB;IAE1B,MAAMC,OAAO,MAAMN,kBAAkB,IAAI;IACzCH,aAAa,CAAC,MAAM,EAAES,KAAK,KAAK,CAAC,CAAC,EAAEA,KAAK,MAAM,CAAC,MAAM,EAAEA,KAAK,GAAG,EAAE;IAElE,OAAO;QACL,MAAM;YACJ,MAAM;YACN,UAAU,EAAE;QACd;QACAA;QACA,kBAAkBD;IACpB;AACF;AAEO,SAASE,kBAAkBC,MAAM,KAAK;IAC3C,MAAMC,gBAAgBC,oBAAoB,iBAAiB,CACzDC;IAEF,MAAMC,qBAAqBC,QAAQ,MAAM,CAAC;IAE1C,MAAMC,WAAWC,OAAO,SAAS,CAAC,GAAG;IACrC,OAAO,GAAGN,iBAAiBD,IAAI,CAAC,EAAEI,mBAAmB,CAAC,EAAEE,UAAU;AACpE;AAEO,SAASE,eAAeC,QAAgB;IAC7CC,OAAO,CAAC,gCAAgC,EAAED,UAAU;AACtD;AAMO,SAASE,wBAAwBC,KAAc;IACpD,MAAMC,QAAQ,IAAIC;IAClB,MAAMC,aAAaH,SAASC,MAAM,KAAK;IACvC,MAAMG,SAASC,QAAQ,GAAG,MAAM;IAChC,IAAIF,YAAY;QACd,MAAMG,aAAaH,WAAW,KAAK,CAAC;QACpC,KAAK,MAAMI,QAAQD,WACjB,IACEC,KAAK,QAAQ,CAAC,aACdA,KAAK,QAAQ,CAAC,aACdA,KAAK,QAAQ,CAAC,UACdA,KAAK,QAAQ,CAAC,QACd;YACA,MAAMC,QAAQD,KAAK,KAAK,CAAC;YACzB,IAAIC,QAAAA,QAAAA,KAAAA,IAAAA,KAAO,CAAC,EAAE,EAAE;gBACd,MAAMC,iBAAiBD,KAAK,CAAC,EAAE,CAC5B,OAAO,CAACJ,QAAQ,IAChB,IAAI,GACJ,OAAO,CAAC,OAAO;gBAClB,OAAOK;YACT;QACF;IAEJ;IACA,OAAO;AACT;AAEA,MAAMC,gBAAgB,IAAIC;AAEnB,SAASC,gBAAgBC,QAAiB;IAC/C,IAAIC,WAAWD,YAAYd;IAC3B,IAAI,CAACe,UAAU;QACbA,WAAWnB;QACXoB,QAAQ,IAAI,CACV;IAEJ;IAEA,IAAIL,cAAc,GAAG,CAACI,WAAW;QAC/B,MAAME,eAAeN,cAAc,GAAG,CAACI;QACvC,IAAIE,AAAiBC,WAAjBD,cACFN,cAAc,GAAG,CAACI,UAAUE,eAAe;IAE/C,OACEN,cAAc,GAAG,CAACI,UAAU;IAE9B,OAAO,GAAGA,SAAS,CAAC,EAAEJ,cAAc,GAAG,CAACI,WAAW;AACrD;AAEO,SAASI,qBACdC,eAAoC,EACpCC,IAAkC;IAElC,IAAI,CAACD,iBACH;IAEF,IAAIA,gBAAgB,EAAE,EACpB,OAAOE,qBAAqBF,gBAAgB,EAAE;IAGhD,IAAIA,gBAAgB,IAAI,EAAE;QACxB,MAAMG,iBAAiB;YACrB,GAAGC,KAAK,KAAK,CAAEJ,AAAAA,CAAAA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAC,IAAK;YACpE,GAAGI,KAAK,KAAK,CAAEJ,AAAAA,CAAAA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAC,IAAK;QACtE;QACA,IAAIK,UAAUC,iCAAiCL,MAAME;QAErD,IAAI,CAACE,SACHA,UAAUE,0BAA0BJ;QAGtC,OAAOE;IACT;AAGF;AAEO,eAAeG,sBACpBC,OAGC,EACDC,UAA2C,EAC3CC,WAAwB,EACxBC,SAA8B;QAWzBC;IATL,IAAI,CAACH,YACH;IAGF,IAAIE,AAAc,UAAdA,WAAqB,YACvBE,MAAW,iCAAiCH;IAI9C,IAAI,UAACE,CAAAA,qBAAAA,QAAQ,SAAS,AAAD,IAAhBA,KAAAA,IAAAA,mBAAmB,iBAAiB,AAAD,GACtC;IAGF,IAAI,CAACJ,QAAQ,iBAAiB,CAAC,uBAAuB,EAAE,YACtDK,MACE;IAKJ,IAAI;QACF,MAAMC,OACJ,MAAMN,QAAQ,iBAAiB,CAAC,uBAAuB,CAACC;QAC1D,MAAML,UAA+B;YACnC,IAAI7B;YACJ,QAAQ;gBACN4B,KAAK,KAAK,CAACW,KAAK,IAAI,GAAGA,KAAK,KAAK,GAAG;gBACpCX,KAAK,KAAK,CAACW,KAAK,GAAG,GAAGA,KAAK,MAAM,GAAG;aACrC;YACDA;YACA,QAAQ,EAAE;YACV,YAAY;gBACV,UAAUC,SAAS,QAAQ;YAC7B;QACF;QAEAF,MAAW,yBAAyBH;QACpC,OAAON;IACT,EAAE,OAAOvB,OAAO;QACdgC,MAAW,qCAAqChC;QAChD;IACF;AACF;AAEO,SAASmC,sBAAsBC,SAAwB;IAC5D,SAASC,kBACPC,IAAkC;QAElC,IAAI,CAACA,MAAM,OAAO;QAGlB,MAAMC,mBAAmBC,MAAM,OAAO,CAACF,KAAK,QAAQ,IAC/CA,KAAK,QAAQ,CACX,GAAG,CAACD,mBACJ,MAAM,CAAC,CAACI,QAAUA,AAAU,SAAVA,SACrB,EAAE;QAGN,IAAIH,KAAK,IAAI,IAAIA,AAAwB,SAAxBA,KAAK,IAAI,CAAC,SAAS,EAClC,OAAO;YACL,GAAGA,IAAI;YACP,UAAUC;QACZ;QAIF,IAAIA,iBAAiB,MAAM,GAAG,GAC5B,OAAO;YACL,MAAM;YACN,UAAUA;QACZ;QAIF,OAAO;IACT;IAEA,OAAO;QACL,GAAGH,SAAS;QACZ,OAAOI,MAAM,OAAO,CAACJ,UAAU,KAAK,IAChCA,UAAU,KAAK,CAAC,GAAG,CAAC,CAACM;gBAEfC;YADJ,MAAMC,UAAU;gBAAE,GAAGF,IAAI;YAAC;YAC1B,IAAI,QAAAC,CAAAA,kBAAAA,KAAK,SAAS,AAAD,IAAbA,KAAAA,IAAAA,gBAAgB,IAAI,EACtBC,QAAQ,SAAS,GAAG;gBAClB,GAAGF,KAAK,SAAS;gBACjB,MAAML,kBAAkBK,KAAK,SAAS,CAAC,IAAI,KAAK;oBAC9C,MAAM;oBACN,UAAU,EAAE;gBACd;YACF;YAEF,OAAOE;QACT,KACAR,UAAU,KAAK;IACrB;AACF;AAIO,MAAMS,qBAAqB,IAEvBC;AAUJ,MAAMC,cAAc,CACzBC;IAKA,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAO;QACL,YAAYA;QACZ,kBAAkBhC;IACpB;IAEF,OAAO;QACL,YAAYgC,OAAO,MAAM;QACzB,kBAAkBA,OAAO,MAAM,GAC3B;YACE,QAAQA,OAAO,MAAM;YACrB,yBAAyB,CAAC,CAACA,OAAO,uBAAuB;QAC3D,IACAhC;IACN;AACF"}
|
|
1
|
+
{"version":3,"file":"agent/utils.mjs","sources":["webpack://@midscene/core/./src/agent/utils.ts"],"sourcesContent":["import type { TMultimodalPrompt, TUserPrompt } from '@/ai-model/common';\nimport type { AbstractInterface } from '@/device';\nimport type {\n ElementCacheFeature,\n LocateResultElement,\n PlanningLocateParam,\n UIContext,\n} from '@/types';\nimport { uploadTestInfoToServer } from '@/utils';\nimport {\n MIDSCENE_REPORT_TAG_NAME,\n globalConfigManager,\n} from '@midscene/shared/env';\nimport { generateElementByPosition } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { assert, logMsg, uuid } from '@midscene/shared/utils';\nimport dayjs from 'dayjs';\nimport type { TaskCache } from './task-cache';\nimport { debug as cacheDebug } from './task-cache';\n\nconst debugProfile = getDebug('web:tool:profile');\n\nexport async function commonContextParser(\n interfaceInstance: AbstractInterface,\n _opt: { uploadServerUrl?: string },\n): Promise<UIContext> {\n assert(interfaceInstance, 'interfaceInstance is required');\n\n debugProfile('Getting interface description');\n const description = interfaceInstance.describe?.() || '';\n debugProfile('Interface description end');\n\n debugProfile('Uploading test info to server');\n uploadTestInfoToServer({\n testUrl: description,\n serverUrl: _opt.uploadServerUrl,\n });\n debugProfile('UploadTestInfoToServer end');\n\n const screenshotBase64 = await interfaceInstance.screenshotBase64();\n assert(screenshotBase64!, 'screenshotBase64 is required');\n\n const size = await interfaceInstance.size();\n debugProfile(`size: ${size.width}x${size.height} dpr: ${size.dpr}`);\n\n return {\n size,\n screenshotBase64: screenshotBase64!,\n };\n}\n\nexport function getReportFileName(tag = 'web') {\n const reportTagName = globalConfigManager.getEnvConfigValue(\n MIDSCENE_REPORT_TAG_NAME,\n );\n const dateTimeInFileName = dayjs().format('YYYY-MM-DD_HH-mm-ss');\n // ensure uniqueness at the same time\n const uniqueId = uuid().substring(0, 8);\n return `${reportTagName || tag}-${dateTimeInFileName}-${uniqueId}`;\n}\n\nexport function printReportMsg(filepath: string) {\n logMsg(`Midscene - report file updated: ${filepath}`);\n}\n\n/**\n * Get the current execution file name\n * @returns The name of the current execution file\n */\nexport function getCurrentExecutionFile(trace?: string): string | false {\n const error = new Error();\n const stackTrace = trace || error.stack;\n const pkgDir = process.cwd() || '';\n if (stackTrace) {\n const stackLines = stackTrace.split('\\n');\n for (const line of stackLines) {\n if (\n line.includes('.spec.') ||\n line.includes('.test.') ||\n line.includes('.ts') ||\n line.includes('.js')\n ) {\n const match = line.match(/(?:at\\s+)?(.*?\\.(?:spec|test)\\.[jt]s)/);\n if (match?.[1]) {\n const targetFileName = match[1]\n .replace(pkgDir, '')\n .trim()\n .replace('at ', '');\n return targetFileName;\n }\n }\n }\n }\n return false;\n}\n\nconst testFileIndex = new Map<string, number>();\n\nexport function generateCacheId(fileName?: string): string {\n let taskFile = fileName || getCurrentExecutionFile();\n if (!taskFile) {\n taskFile = uuid();\n console.warn(\n 'Midscene - using random UUID for cache id. Cache may be invalid.',\n );\n }\n\n if (testFileIndex.has(taskFile)) {\n const currentIndex = testFileIndex.get(taskFile);\n if (currentIndex !== undefined) {\n testFileIndex.set(taskFile, currentIndex + 1);\n }\n } else {\n testFileIndex.set(taskFile, 1);\n }\n return `${taskFile}-${testFileIndex.get(taskFile)}`;\n}\n\nexport function matchElementFromPlan(\n planLocateParam: PlanningLocateParam,\n): LocateResultElement | undefined {\n if (!planLocateParam) {\n return undefined;\n }\n\n if (planLocateParam.bbox) {\n const centerPosition = {\n x: Math.floor((planLocateParam.bbox[0] + planLocateParam.bbox[2]) / 2),\n y: Math.floor((planLocateParam.bbox[1] + planLocateParam.bbox[3]) / 2),\n };\n\n const element = generateElementByPosition(centerPosition);\n return element;\n }\n\n return undefined;\n}\n\nexport async function matchElementFromCache(\n context: {\n taskCache?: TaskCache;\n interfaceInstance: AbstractInterface;\n },\n cacheEntry: ElementCacheFeature | undefined,\n cachePrompt: TUserPrompt,\n cacheable: boolean | undefined,\n): Promise<LocateResultElement | undefined> {\n if (!cacheEntry) {\n return undefined;\n }\n\n if (cacheable === false) {\n cacheDebug('cache disabled for prompt: %s', cachePrompt);\n return undefined;\n }\n\n if (!context.taskCache?.isCacheResultUsed) {\n return undefined;\n }\n\n if (!context.interfaceInstance.rectMatchesCacheFeature) {\n cacheDebug(\n 'interface does not implement rectMatchesCacheFeature, skip cache',\n );\n return undefined;\n }\n\n try {\n const rect =\n await context.interfaceInstance.rectMatchesCacheFeature(cacheEntry);\n const element: LocateResultElement = {\n id: uuid(),\n center: [\n Math.round(rect.left + rect.width / 2),\n Math.round(rect.top + rect.height / 2),\n ],\n rect,\n };\n\n cacheDebug('cache hit, prompt: %s', cachePrompt);\n return element;\n } catch (error) {\n cacheDebug('rectMatchesCacheFeature error: %s', error);\n return undefined;\n }\n}\n\ndeclare const __VERSION__: string | undefined;\n\nexport const getMidsceneVersion = (): string => {\n if (typeof __VERSION__ !== 'undefined') {\n return __VERSION__;\n } else if (\n process.env.__VERSION__ &&\n process.env.__VERSION__ !== 'undefined'\n ) {\n return process.env.__VERSION__;\n }\n throw new Error('__VERSION__ inject failed during build');\n};\n\nexport const parsePrompt = (\n prompt: TUserPrompt,\n): {\n textPrompt: string;\n multimodalPrompt?: TMultimodalPrompt;\n} => {\n if (typeof prompt === 'string') {\n return {\n textPrompt: prompt,\n multimodalPrompt: undefined,\n };\n }\n return {\n textPrompt: prompt.prompt,\n multimodalPrompt: prompt.images\n ? {\n images: prompt.images,\n convertHttpImage2Base64: !!prompt.convertHttpImage2Base64,\n }\n : undefined,\n };\n};\n"],"names":["debugProfile","getDebug","commonContextParser","interfaceInstance","_opt","assert","description","uploadTestInfoToServer","screenshotBase64","size","getReportFileName","tag","reportTagName","globalConfigManager","MIDSCENE_REPORT_TAG_NAME","dateTimeInFileName","dayjs","uniqueId","uuid","printReportMsg","filepath","logMsg","getCurrentExecutionFile","trace","error","Error","stackTrace","pkgDir","process","stackLines","line","match","targetFileName","testFileIndex","Map","generateCacheId","fileName","taskFile","console","currentIndex","undefined","matchElementFromPlan","planLocateParam","centerPosition","Math","element","generateElementByPosition","matchElementFromCache","context","cacheEntry","cachePrompt","cacheable","_context_taskCache","cacheDebug","rect","getMidsceneVersion","__VERSION__","parsePrompt","prompt"],"mappings":";;;;;;;AAqBA,MAAMA,eAAeC,SAAS;AAEvB,eAAeC,oBACpBC,iBAAoC,EACpCC,IAAkC;QAKdD;IAHpBE,OAAOF,mBAAmB;IAE1BH,aAAa;IACb,MAAMM,cAAcH,AAAAA,SAAAA,CAAAA,8BAAAA,kBAAkB,QAAQ,AAAD,IAAzBA,KAAAA,IAAAA,4BAAAA,IAAAA,CAAAA,kBAAiB,KAAiB;IACtDH,aAAa;IAEbA,aAAa;IACbO,uBAAuB;QACrB,SAASD;QACT,WAAWF,KAAK,eAAe;IACjC;IACAJ,aAAa;IAEb,MAAMQ,mBAAmB,MAAML,kBAAkB,gBAAgB;IACjEE,OAAOG,kBAAmB;IAE1B,MAAMC,OAAO,MAAMN,kBAAkB,IAAI;IACzCH,aAAa,CAAC,MAAM,EAAES,KAAK,KAAK,CAAC,CAAC,EAAEA,KAAK,MAAM,CAAC,MAAM,EAAEA,KAAK,GAAG,EAAE;IAElE,OAAO;QACLA;QACA,kBAAkBD;IACpB;AACF;AAEO,SAASE,kBAAkBC,MAAM,KAAK;IAC3C,MAAMC,gBAAgBC,oBAAoB,iBAAiB,CACzDC;IAEF,MAAMC,qBAAqBC,QAAQ,MAAM,CAAC;IAE1C,MAAMC,WAAWC,OAAO,SAAS,CAAC,GAAG;IACrC,OAAO,GAAGN,iBAAiBD,IAAI,CAAC,EAAEI,mBAAmB,CAAC,EAAEE,UAAU;AACpE;AAEO,SAASE,eAAeC,QAAgB;IAC7CC,OAAO,CAAC,gCAAgC,EAAED,UAAU;AACtD;AAMO,SAASE,wBAAwBC,KAAc;IACpD,MAAMC,QAAQ,IAAIC;IAClB,MAAMC,aAAaH,SAASC,MAAM,KAAK;IACvC,MAAMG,SAASC,QAAQ,GAAG,MAAM;IAChC,IAAIF,YAAY;QACd,MAAMG,aAAaH,WAAW,KAAK,CAAC;QACpC,KAAK,MAAMI,QAAQD,WACjB,IACEC,KAAK,QAAQ,CAAC,aACdA,KAAK,QAAQ,CAAC,aACdA,KAAK,QAAQ,CAAC,UACdA,KAAK,QAAQ,CAAC,QACd;YACA,MAAMC,QAAQD,KAAK,KAAK,CAAC;YACzB,IAAIC,QAAAA,QAAAA,KAAAA,IAAAA,KAAO,CAAC,EAAE,EAAE;gBACd,MAAMC,iBAAiBD,KAAK,CAAC,EAAE,CAC5B,OAAO,CAACJ,QAAQ,IAChB,IAAI,GACJ,OAAO,CAAC,OAAO;gBAClB,OAAOK;YACT;QACF;IAEJ;IACA,OAAO;AACT;AAEA,MAAMC,gBAAgB,IAAIC;AAEnB,SAASC,gBAAgBC,QAAiB;IAC/C,IAAIC,WAAWD,YAAYd;IAC3B,IAAI,CAACe,UAAU;QACbA,WAAWnB;QACXoB,QAAQ,IAAI,CACV;IAEJ;IAEA,IAAIL,cAAc,GAAG,CAACI,WAAW;QAC/B,MAAME,eAAeN,cAAc,GAAG,CAACI;QACvC,IAAIE,AAAiBC,WAAjBD,cACFN,cAAc,GAAG,CAACI,UAAUE,eAAe;IAE/C,OACEN,cAAc,GAAG,CAACI,UAAU;IAE9B,OAAO,GAAGA,SAAS,CAAC,EAAEJ,cAAc,GAAG,CAACI,WAAW;AACrD;AAEO,SAASI,qBACdC,eAAoC;IAEpC,IAAI,CAACA,iBACH;IAGF,IAAIA,gBAAgB,IAAI,EAAE;QACxB,MAAMC,iBAAiB;YACrB,GAAGC,KAAK,KAAK,CAAEF,AAAAA,CAAAA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAC,IAAK;YACpE,GAAGE,KAAK,KAAK,CAAEF,AAAAA,CAAAA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAC,IAAK;QACtE;QAEA,MAAMG,UAAUC,0BAA0BH;QAC1C,OAAOE;IACT;AAGF;AAEO,eAAeE,sBACpBC,OAGC,EACDC,UAA2C,EAC3CC,WAAwB,EACxBC,SAA8B;QAWzBC;IATL,IAAI,CAACH,YACH;IAGF,IAAIE,AAAc,UAAdA,WAAqB,YACvBE,MAAW,iCAAiCH;IAI9C,IAAI,UAACE,CAAAA,qBAAAA,QAAQ,SAAS,AAAD,IAAhBA,KAAAA,IAAAA,mBAAmB,iBAAiB,AAAD,GACtC;IAGF,IAAI,CAACJ,QAAQ,iBAAiB,CAAC,uBAAuB,EAAE,YACtDK,MACE;IAKJ,IAAI;QACF,MAAMC,OACJ,MAAMN,QAAQ,iBAAiB,CAAC,uBAAuB,CAACC;QAC1D,MAAMJ,UAA+B;YACnC,IAAI3B;YACJ,QAAQ;gBACN0B,KAAK,KAAK,CAACU,KAAK,IAAI,GAAGA,KAAK,KAAK,GAAG;gBACpCV,KAAK,KAAK,CAACU,KAAK,GAAG,GAAGA,KAAK,MAAM,GAAG;aACrC;YACDA;QACF;QAEAD,MAAW,yBAAyBH;QACpC,OAAOL;IACT,EAAE,OAAOrB,OAAO;QACd6B,MAAW,qCAAqC7B;QAChD;IACF;AACF;AAIO,MAAM+B,qBAAqB,IAEvBC;AAUJ,MAAMC,cAAc,CACzBC;IAKA,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAO;QACL,YAAYA;QACZ,kBAAkBlB;IACpB;IAEF,OAAO;QACL,YAAYkB,OAAO,MAAM;QACzB,kBAAkBA,OAAO,MAAM,GAC3B;YACE,QAAQA,OAAO,MAAM;YACrB,yBAAyB,CAAC,CAACA,OAAO,uBAAuB;QAC3D,IACAlB;IACN;AACF"}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"ai-model/common.mjs","sources":["webpack://@midscene/core/./src/ai-model/common.ts"],"sourcesContent":["import type {\n AIUsageInfo,\n BaseElement,\n DeviceAction,\n ElementTreeNode,\n MidsceneYamlFlowItem,\n PlanningAction,\n Rect,\n Size,\n} from '@/types';\nimport { assert } from '@midscene/shared/utils';\n\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\n\nimport type { PlanningLocateParam } from '@/types';\nimport { NodeType } from '@midscene/shared/constants';\nimport type { TVlModeTypes } from '@midscene/shared/env';\nimport { treeToList } from '@midscene/shared/extractor';\nimport { compositeElementInfoImg } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { z } from 'zod';\n\nexport type AIArgs = ChatCompletionMessageParam[];\n\nexport enum AIActionType {\n ASSERT = 0,\n INSPECT_ELEMENT = 1,\n EXTRACT_DATA = 2,\n PLAN = 3,\n DESCRIBE_ELEMENT = 4,\n TEXT = 5,\n}\n\nconst defaultBboxSize = 20; // must be even number\nconst debugInspectUtils = getDebug('ai:common');\n\n// transform the param of locate from qwen mode\nexport function fillBboxParam(\n locate: PlanningLocateParam,\n width: number,\n height: number,\n rightLimit: number,\n bottomLimit: number,\n vlMode: TVlModeTypes | undefined,\n) {\n // The Qwen model might have hallucinations of naming bbox as bbox_2d.\n if ((locate as any).bbox_2d && !locate?.bbox) {\n locate.bbox = (locate as any).bbox_2d;\n // biome-ignore lint/performance/noDelete: <explanation>\n delete (locate as any).bbox_2d;\n }\n\n if (locate?.bbox) {\n locate.bbox = adaptBbox(\n locate.bbox,\n width,\n height,\n rightLimit,\n bottomLimit,\n vlMode,\n );\n }\n\n return locate;\n}\n\nexport function adaptQwenBbox(\n bbox: number[],\n): [number, number, number, number] {\n if (bbox.length < 2) {\n const msg = `invalid bbox data for qwen-vl mode: ${JSON.stringify(bbox)} `;\n throw new Error(msg);\n }\n\n const result: [number, number, number, number] = [\n Math.round(bbox[0]),\n Math.round(bbox[1]),\n typeof bbox[2] === 'number'\n ? Math.round(bbox[2])\n : Math.round(bbox[0] + defaultBboxSize),\n typeof bbox[3] === 'number'\n ? Math.round(bbox[3])\n : Math.round(bbox[1] + defaultBboxSize),\n ];\n return result;\n}\n\nexport function adaptDoubaoBbox(\n bbox: string[] | number[] | string,\n width: number,\n height: number,\n): [number, number, number, number] {\n assert(\n width > 0 && height > 0,\n 'width and height must be greater than 0 in doubao mode',\n );\n\n if (typeof bbox === 'string') {\n assert(\n /^(\\d+)\\s(\\d+)\\s(\\d+)\\s(\\d+)$/.test(bbox.trim()),\n `invalid bbox data string for doubao-vision mode: ${bbox}`,\n );\n const splitted = bbox.split(' ');\n if (splitted.length === 4) {\n return [\n Math.round((Number(splitted[0]) * width) / 1000),\n Math.round((Number(splitted[1]) * height) / 1000),\n Math.round((Number(splitted[2]) * width) / 1000),\n Math.round((Number(splitted[3]) * height) / 1000),\n ];\n }\n throw new Error(`invalid bbox data string for doubao-vision mode: ${bbox}`);\n }\n\n if (Array.isArray(bbox) && Array.isArray(bbox[0])) {\n bbox = bbox[0];\n }\n\n let bboxList: number[] = [];\n if (Array.isArray(bbox) && typeof bbox[0] === 'string') {\n bbox.forEach((item) => {\n if (typeof item === 'string' && item.includes(',')) {\n const [x, y] = item.split(',');\n bboxList.push(Number(x.trim()), Number(y.trim()));\n } else if (typeof item === 'string' && item.includes(' ')) {\n const [x, y] = item.split(' ');\n bboxList.push(Number(x.trim()), Number(y.trim()));\n } else {\n bboxList.push(Number(item));\n }\n });\n } else {\n bboxList = bbox as any;\n }\n\n if (bboxList.length === 4 || bboxList.length === 5) {\n return [\n Math.round((bboxList[0] * width) / 1000),\n Math.round((bboxList[1] * height) / 1000),\n Math.round((bboxList[2] * width) / 1000),\n Math.round((bboxList[3] * height) / 1000),\n ];\n }\n\n // treat the bbox as a center point\n if (\n bboxList.length === 6 ||\n bboxList.length === 2 ||\n bboxList.length === 3 ||\n bboxList.length === 7\n ) {\n return [\n Math.max(\n 0,\n Math.round((bboxList[0] * width) / 1000) - defaultBboxSize / 2,\n ),\n Math.max(\n 0,\n Math.round((bboxList[1] * height) / 1000) - defaultBboxSize / 2,\n ),\n Math.min(\n width,\n Math.round((bboxList[0] * width) / 1000) + defaultBboxSize / 2,\n ),\n Math.min(\n height,\n Math.round((bboxList[1] * height) / 1000) + defaultBboxSize / 2,\n ),\n ];\n }\n\n if (bbox.length === 8) {\n return [\n Math.round((bboxList[0] * width) / 1000),\n Math.round((bboxList[1] * height) / 1000),\n Math.round((bboxList[4] * width) / 1000),\n Math.round((bboxList[5] * height) / 1000),\n ];\n }\n\n const msg = `invalid bbox data for doubao-vision mode: ${JSON.stringify(bbox)} `;\n throw new Error(msg);\n}\n\nexport function adaptBbox(\n bbox: number[],\n width: number,\n height: number,\n rightLimit: number,\n bottomLimit: number,\n vlMode: TVlModeTypes | undefined,\n): [number, number, number, number] {\n let result: [number, number, number, number] = [0, 0, 0, 0];\n if (vlMode === 'doubao-vision' || vlMode === 'vlm-ui-tars') {\n result = adaptDoubaoBbox(bbox, width, height);\n } else if (vlMode === 'gemini') {\n result = adaptGeminiBbox(bbox, width, height);\n } else if (vlMode === 'qwen3-vl') {\n result = normalized01000(bbox, width, height);\n } else {\n result = adaptQwenBbox(bbox);\n }\n\n result[2] = Math.min(result[2], rightLimit);\n result[3] = Math.min(result[3], bottomLimit);\n\n return result;\n}\n\n// x1, y1, x2, y2 -> 0-1000\nexport function normalized01000(\n bbox: number[],\n width: number,\n height: number,\n): [number, number, number, number] {\n return [\n Math.round((bbox[0] * width) / 1000),\n Math.round((bbox[1] * height) / 1000),\n Math.round((bbox[2] * width) / 1000),\n Math.round((bbox[3] * height) / 1000),\n ];\n}\n\n// y1, x1, y2, x2 -> 0-1000\nexport function adaptGeminiBbox(\n bbox: number[],\n width: number,\n height: number,\n): [number, number, number, number] {\n const left = Math.round((bbox[1] * width) / 1000);\n const top = Math.round((bbox[0] * height) / 1000);\n const right = Math.round((bbox[3] * width) / 1000);\n const bottom = Math.round((bbox[2] * height) / 1000);\n return [left, top, right, bottom];\n}\n\nexport function adaptBboxToRect(\n bbox: number[],\n width: number,\n height: number,\n offsetX = 0,\n offsetY = 0,\n rightLimit = width,\n bottomLimit = height,\n vlMode?: TVlModeTypes | undefined,\n): Rect {\n debugInspectUtils(\n 'adaptBboxToRect',\n bbox,\n width,\n height,\n 'offset',\n offsetX,\n offsetY,\n 'limit',\n rightLimit,\n bottomLimit,\n 'vlMode',\n vlMode,\n );\n const [left, top, right, bottom] = adaptBbox(\n bbox,\n width,\n height,\n rightLimit,\n bottomLimit,\n vlMode,\n );\n\n // Calculate initial rect dimensions\n const rectLeft = left;\n const rectTop = top;\n let rectWidth = right - left;\n let rectHeight = bottom - top;\n\n // Ensure the rect doesn't exceed image boundaries\n // If right edge exceeds width, adjust the width\n if (rectLeft + rectWidth > width) {\n rectWidth = width - rectLeft;\n }\n\n // If bottom edge exceeds height, adjust the height\n if (rectTop + rectHeight > height) {\n rectHeight = height - rectTop;\n }\n\n // Ensure minimum dimensions (width and height should be at least 1)\n rectWidth = Math.max(1, rectWidth);\n rectHeight = Math.max(1, rectHeight);\n\n const rect = {\n left: rectLeft + offsetX,\n top: rectTop + offsetY,\n width: rectWidth,\n height: rectHeight,\n };\n debugInspectUtils('adaptBboxToRect, result=', rect);\n\n return rect;\n}\n\nexport function mergeRects(rects: Rect[]) {\n const minLeft = Math.min(...rects.map((r) => r.left));\n const minTop = Math.min(...rects.map((r) => r.top));\n const maxRight = Math.max(...rects.map((r) => r.left + r.width));\n const maxBottom = Math.max(...rects.map((r) => r.top + r.height));\n return {\n left: minLeft,\n top: minTop,\n width: maxRight - minLeft,\n height: maxBottom - minTop,\n };\n}\n\n// expand the search area to at least 300 x 300, or add a default padding\nexport function expandSearchArea(\n rect: Rect,\n screenSize: Size,\n vlMode: TVlModeTypes | undefined,\n) {\n const minEdgeSize = vlMode === 'doubao-vision' ? 500 : 300;\n const defaultPadding = 160;\n\n // Calculate padding needed to reach minimum edge size\n const paddingSizeHorizontal =\n rect.width < minEdgeSize\n ? Math.ceil((minEdgeSize - rect.width) / 2)\n : defaultPadding;\n const paddingSizeVertical =\n rect.height < minEdgeSize\n ? Math.ceil((minEdgeSize - rect.height) / 2)\n : defaultPadding;\n\n // Calculate new dimensions (ensure minimum edge size)\n let newWidth = Math.max(minEdgeSize, rect.width + paddingSizeHorizontal * 2);\n let newHeight = Math.max(minEdgeSize, rect.height + paddingSizeVertical * 2);\n\n // Calculate initial position with padding\n let newLeft = rect.left - paddingSizeHorizontal;\n let newTop = rect.top - paddingSizeVertical;\n\n // Ensure the rect doesn't exceed screen boundaries by adjusting position\n // If the rect goes beyond the right edge, shift it left\n if (newLeft + newWidth > screenSize.width) {\n newLeft = screenSize.width - newWidth;\n }\n\n // If the rect goes beyond the bottom edge, shift it up\n if (newTop + newHeight > screenSize.height) {\n newTop = screenSize.height - newHeight;\n }\n\n // Ensure the rect doesn't go beyond the left/top edges\n newLeft = Math.max(0, newLeft);\n newTop = Math.max(0, newTop);\n\n // If after position adjustment, the rect still exceeds screen boundaries,\n // clamp the dimensions to fit within screen\n if (newLeft + newWidth > screenSize.width) {\n newWidth = screenSize.width - newLeft;\n }\n if (newTop + newHeight > screenSize.height) {\n newHeight = screenSize.height - newTop;\n }\n\n rect.left = newLeft;\n rect.top = newTop;\n rect.width = newWidth;\n rect.height = newHeight;\n\n return rect;\n}\n\nexport async function markupImageForLLM(\n screenshotBase64: string,\n tree: ElementTreeNode<BaseElement>,\n size: Size,\n) {\n const elementsInfo = treeToList(tree);\n const elementsPositionInfoWithoutText = elementsInfo!.filter(\n (elementInfo) => {\n if (elementInfo.attributes.nodeType === NodeType.TEXT) {\n return false;\n }\n return true;\n },\n );\n\n const imagePayload = await compositeElementInfoImg({\n inputImgBase64: screenshotBase64,\n elementsPositionInfo: elementsPositionInfoWithoutText,\n size,\n });\n return imagePayload;\n}\n\nexport function buildYamlFlowFromPlans(\n plans: PlanningAction[],\n actionSpace: DeviceAction<any>[],\n sleep?: number,\n): MidsceneYamlFlowItem[] {\n const flow: MidsceneYamlFlowItem[] = [];\n\n for (const plan of plans) {\n const verb = plan.type;\n\n const action = actionSpace.find((action) => action.name === verb);\n if (!action) {\n console.warn(\n `Cannot convert action ${verb} to yaml flow. Will ignore it.`,\n );\n continue;\n }\n\n const flowKey = action.interfaceAlias || verb;\n const flowParam = action.paramSchema\n ? dumpActionParam(plan.param || {}, action.paramSchema)\n : {};\n\n const flowItem: MidsceneYamlFlowItem = {\n [flowKey]: '',\n ...flowParam,\n };\n\n flow.push(flowItem);\n }\n\n if (sleep) {\n flow.push({\n sleep,\n });\n }\n\n return flow;\n}\n\n// Zod schemas for shared types\nexport const PointSchema = z.object({\n left: z.number(),\n top: z.number(),\n});\n\nexport const SizeSchema = z.object({\n width: z.number(),\n height: z.number(),\n dpr: z.number().optional(),\n});\n\nexport const RectSchema = PointSchema.and(SizeSchema).and(\n z.object({\n zoom: z.number().optional(),\n }),\n);\n\n// Zod schema for TMultimodalPrompt\nexport const TMultimodalPromptSchema = z.object({\n images: z\n .array(\n z.object({\n name: z.string(),\n url: z.string(),\n }),\n )\n .optional(),\n convertHttpImage2Base64: z.boolean().optional(),\n});\n\n// Zod schema for TUserPrompt\nexport const TUserPromptSchema = z.union([\n z.string(),\n z\n .object({\n prompt: z.string(),\n })\n .and(TMultimodalPromptSchema.partial()),\n]);\n\n// Generate TypeScript types from Zod schemas\nexport type TMultimodalPrompt = z.infer<typeof TMultimodalPromptSchema>;\nexport type TUserPrompt = z.infer<typeof TUserPromptSchema>;\n\nconst locateFieldFlagName = 'midscene_location_field_flag';\n\n// Schema for locator field input (when users provide locate parameters)\nconst MidsceneLocationInput = z\n .object({\n prompt: TUserPromptSchema,\n deepThink: z.boolean().optional(),\n cacheable: z.boolean().optional(),\n xpath: z.union([z.string(), z.boolean()]).optional(),\n })\n .passthrough();\n\n// Schema for locator field result (when AI returns locate results)\nconst MidsceneLocationResult = z\n .object({\n [locateFieldFlagName]: z.literal(true),\n prompt: TUserPromptSchema,\n\n // optional fields\n deepThink: z.boolean().optional(), // only available in vl model\n cacheable: z.boolean().optional(),\n xpath: z.boolean().optional(), // preset result for xpath\n\n // these two fields will only appear in the result\n center: z.tuple([z.number(), z.number()]),\n rect: RectSchema,\n })\n .passthrough();\n\n// Export the result type - this is used for runtime results that include center and rect\nexport type MidsceneLocationResultType = z.infer<typeof MidsceneLocationResult>;\n\n// Export the input type - this is the inferred type from getMidsceneLocationSchema()\nexport type MidsceneLocationInputType = z.infer<typeof MidsceneLocationInput>;\n\n/**\n * Returns the schema for locator fields.\n * This now returns the input schema which is more permissive and suitable for validation.\n */\nexport const getMidsceneLocationSchema = () => {\n return MidsceneLocationInput;\n};\n\nexport const ifMidsceneLocatorField = (field: any): boolean => {\n // Handle optional fields by getting the inner type\n let actualField = field;\n if (actualField._def?.typeName === 'ZodOptional') {\n actualField = actualField._def.innerType;\n }\n\n // Check if this is a ZodObject\n if (actualField._def?.typeName === 'ZodObject') {\n const shape = actualField._def.shape();\n\n // Method 1: Check for the location field flag (for result schema)\n if (locateFieldFlagName in shape) {\n return true;\n }\n\n // Method 2: Check if it's the input schema by checking for 'prompt' field\n // Input schema has 'prompt' as a required field\n if ('prompt' in shape && shape.prompt) {\n return true;\n }\n }\n\n return false;\n};\n\nexport const dumpMidsceneLocatorField = (field: any): string => {\n assert(\n ifMidsceneLocatorField(field),\n 'field is not a midscene locator field',\n );\n\n // If field is a string, return it directly\n if (typeof field === 'string') {\n return field;\n }\n\n // If field is an object with prompt property\n if (field && typeof field === 'object' && field.prompt) {\n // If prompt is a string, return it directly\n if (typeof field.prompt === 'string') {\n return field.prompt;\n }\n // If prompt is a TUserPrompt object, extract the prompt string\n if (typeof field.prompt === 'object' && field.prompt.prompt) {\n return field.prompt.prompt; // TODO: dump images if necessary\n }\n }\n\n // Fallback: try to convert to string\n return String(field);\n};\n\nexport const findAllMidsceneLocatorField = (\n zodType?: z.ZodType<any>,\n requiredOnly?: boolean,\n): string[] => {\n if (!zodType) {\n return [];\n }\n\n // Check if this is a ZodObject by checking if it has a shape property\n const zodObject = zodType as any;\n if (zodObject._def?.typeName === 'ZodObject' && zodObject.shape) {\n const keys = Object.keys(zodObject.shape);\n return keys.filter((key) => {\n const field = zodObject.shape[key];\n if (!ifMidsceneLocatorField(field)) {\n return false;\n }\n\n // If requiredOnly is true, filter out optional fields\n if (requiredOnly) {\n return field._def?.typeName !== 'ZodOptional';\n }\n\n return true;\n });\n }\n\n // For other ZodType instances, we can't extract field names\n return [];\n};\n\nexport const dumpActionParam = (\n jsonObject: Record<string, any>,\n zodSchema: z.ZodType<any>,\n): Record<string, any> => {\n const locatorFields = findAllMidsceneLocatorField(zodSchema);\n const result = { ...jsonObject };\n\n for (const fieldName of locatorFields) {\n const fieldValue = result[fieldName];\n if (fieldValue) {\n // If it's already a string, keep it as is\n if (typeof fieldValue === 'string') {\n result[fieldName] = fieldValue;\n } else if (typeof fieldValue === 'object') {\n // Check if this field is actually a MidsceneLocationType object\n if (fieldValue.prompt) {\n // If prompt is a string, use it directly\n if (typeof fieldValue.prompt === 'string') {\n result[fieldName] = fieldValue.prompt;\n } else if (\n typeof fieldValue.prompt === 'object' &&\n fieldValue.prompt.prompt\n ) {\n // If prompt is a TUserPrompt object, extract the prompt string\n result[fieldName] = fieldValue.prompt.prompt;\n }\n }\n }\n }\n }\n\n return result;\n};\n\nexport const loadActionParam = (\n jsonObject: Record<string, any>,\n zodSchema: z.ZodType<any>,\n): Record<string, any> => {\n const locatorFields = findAllMidsceneLocatorField(zodSchema);\n const result = { ...jsonObject };\n\n for (const fieldName of locatorFields) {\n const fieldValue = result[fieldName];\n if (fieldValue && typeof fieldValue === 'string') {\n result[fieldName] = {\n [locateFieldFlagName]: true,\n prompt: fieldValue,\n };\n }\n }\n\n return result;\n};\n\n/**\n * Parse and validate action parameters using Zod schema.\n * All fields are validated through Zod, EXCEPT locator fields which are skipped.\n * Default values defined in the schema are automatically applied.\n *\n * Locator fields are special business logic fields with complex validation requirements,\n * so they are intentionally excluded from Zod parsing and use existing validation logic.\n */\nexport const parseActionParam = (\n rawParam: Record<string, any>,\n zodSchema: z.ZodType<any>,\n): Record<string, any> => {\n // Handle undefined or null rawParam by providing an empty object\n const param = rawParam ?? {};\n\n // Find all locate fields in the schema\n const locateFields = findAllMidsceneLocatorField(zodSchema);\n\n // If there are no locate fields, just do normal validation\n if (locateFields.length === 0) {\n return zodSchema.parse(param);\n }\n\n // Extract locate field values to restore later\n const locateFieldValues: Record<string, any> = {};\n for (const fieldName of locateFields) {\n if (fieldName in param) {\n locateFieldValues[fieldName] = param[fieldName];\n }\n }\n\n // Build params for validation - skip locate fields and use dummy values\n const paramsForValidation: Record<string, any> = {};\n for (const key in param) {\n if (locateFields.includes(key)) {\n // Use dummy value to satisfy schema validation\n paramsForValidation[key] = { prompt: '_dummy_' };\n } else {\n paramsForValidation[key] = param[key];\n }\n }\n\n // Validate with dummy locate values\n const validated = zodSchema.parse(paramsForValidation);\n\n // Restore the actual locate field values (unvalidated, as per business requirement)\n for (const fieldName in locateFieldValues) {\n validated[fieldName] = locateFieldValues[fieldName];\n }\n\n return validated;\n};\n"],"names":["AIActionType","defaultBboxSize","debugInspectUtils","getDebug","fillBboxParam","locate","width","height","rightLimit","bottomLimit","vlMode","adaptBbox","adaptQwenBbox","bbox","msg","JSON","Error","result","Math","adaptDoubaoBbox","assert","splitted","Number","Array","bboxList","item","x","y","adaptGeminiBbox","normalized01000","left","top","right","bottom","adaptBboxToRect","offsetX","offsetY","rectLeft","rectTop","rectWidth","rectHeight","rect","mergeRects","rects","minLeft","r","minTop","maxRight","maxBottom","expandSearchArea","screenSize","minEdgeSize","defaultPadding","paddingSizeHorizontal","paddingSizeVertical","newWidth","newHeight","newLeft","newTop","markupImageForLLM","screenshotBase64","tree","size","elementsInfo","treeToList","elementsPositionInfoWithoutText","elementInfo","NodeType","imagePayload","compositeElementInfoImg","buildYamlFlowFromPlans","plans","actionSpace","sleep","flow","plan","verb","action","console","flowKey","flowParam","dumpActionParam","flowItem","PointSchema","z","SizeSchema","RectSchema","TMultimodalPromptSchema","TUserPromptSchema","locateFieldFlagName","MidsceneLocationInput","getMidsceneLocationSchema","ifMidsceneLocatorField","field","_actualField__def","_actualField__def1","actualField","shape","dumpMidsceneLocatorField","String","findAllMidsceneLocatorField","zodType","requiredOnly","_zodObject__def","zodObject","keys","Object","key","_field__def","jsonObject","zodSchema","locatorFields","fieldName","fieldValue","loadActionParam","parseActionParam","rawParam","param","locateFields","locateFieldValues","paramsForValidation","validated"],"mappings":";;;;;;AAwBO,IAAKA,sBAAYA,WAAAA,GAAAA,SAAZA,YAAY;;;;;;;WAAZA;;AASZ,MAAMC,kBAAkB;AACxB,MAAMC,oBAAoBC,SAAS;AAG5B,SAASC,cACdC,MAA2B,EAC3BC,KAAa,EACbC,MAAc,EACdC,UAAkB,EAClBC,WAAmB,EACnBC,MAAgC;IAGhC,IAAKL,OAAe,OAAO,IAAI,CAACA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,IAAI,AAAD,GAAG;QAC5CA,OAAO,IAAI,GAAIA,OAAe,OAAO;QAErC,OAAQA,OAAe,OAAO;IAChC;IAEA,IAAIA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,IAAI,EACdA,OAAO,IAAI,GAAGM,UACZN,OAAO,IAAI,EACXC,OACAC,QACAC,YACAC,aACAC;IAIJ,OAAOL;AACT;AAEO,SAASO,cACdC,IAAc;IAEd,IAAIA,KAAK,MAAM,GAAG,GAAG;QACnB,MAAMC,MAAM,CAAC,oCAAoC,EAAEC,KAAK,SAAS,CAACF,MAAM,CAAC,CAAC;QAC1E,MAAM,IAAIG,MAAMF;IAClB;IAEA,MAAMG,SAA2C;QAC/CC,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE;QAClBK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE;QACC,YAAnB,OAAOA,IAAI,CAAC,EAAE,GACVK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,IAClBK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,GAAGZ;QACN,YAAnB,OAAOY,IAAI,CAAC,EAAE,GACVK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,IAClBK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,GAAGZ;KAC1B;IACD,OAAOgB;AACT;AAEO,SAASE,gBACdN,IAAkC,EAClCP,KAAa,EACbC,MAAc;IAEda,OACEd,QAAQ,KAAKC,SAAS,GACtB;IAGF,IAAI,AAAgB,YAAhB,OAAOM,MAAmB;QAC5BO,OACE,+BAA+B,IAAI,CAACP,KAAK,IAAI,KAC7C,CAAC,iDAAiD,EAAEA,MAAM;QAE5D,MAAMQ,WAAWR,KAAK,KAAK,CAAC;QAC5B,IAAIQ,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACLH,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAIf,QAAS;YAC3CY,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAId,SAAU;YAC5CW,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAIf,QAAS;YAC3CY,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAId,SAAU;SAC7C;QAEH,MAAM,IAAIS,MAAM,CAAC,iDAAiD,EAAEH,MAAM;IAC5E;IAEA,IAAIU,MAAM,OAAO,CAACV,SAASU,MAAM,OAAO,CAACV,IAAI,CAAC,EAAE,GAC9CA,OAAOA,IAAI,CAAC,EAAE;IAGhB,IAAIW,WAAqB,EAAE;IAC3B,IAAID,MAAM,OAAO,CAACV,SAAS,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EACvCA,KAAK,OAAO,CAAC,CAACY;QACZ,IAAI,AAAgB,YAAhB,OAAOA,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YAClD,MAAM,CAACC,GAAGC,EAAE,GAAGF,KAAK,KAAK,CAAC;YAC1BD,SAAS,IAAI,CAACF,OAAOI,EAAE,IAAI,KAAKJ,OAAOK,EAAE,IAAI;QAC/C,OAAO,IAAI,AAAgB,YAAhB,OAAOF,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YACzD,MAAM,CAACC,GAAGC,EAAE,GAAGF,KAAK,KAAK,CAAC;YAC1BD,SAAS,IAAI,CAACF,OAAOI,EAAE,IAAI,KAAKJ,OAAOK,EAAE,IAAI;QAC/C,OACEH,SAAS,IAAI,CAACF,OAAOG;IAEzB;SAEAD,WAAWX;IAGb,IAAIW,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAO;QACLN,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGlB,QAAS;QACnCY,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGjB,SAAU;QACpCW,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGlB,QAAS;QACnCY,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGjB,SAAU;KACrC;IAIH,IACEiB,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAO;QACLN,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGlB,QAAS,QAAQL,kBAAkB;QAE/DiB,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGjB,SAAU,QAAQN,kBAAkB;QAEhEiB,KAAK,GAAG,CACNZ,OACAY,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGlB,QAAS,QAAQL,kBAAkB;QAE/DiB,KAAK,GAAG,CACNX,QACAW,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGjB,SAAU,QAAQN,kBAAkB;KAEjE;IAGH,IAAIY,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAO;QACLK,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGlB,QAAS;QACnCY,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGjB,SAAU;QACpCW,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGlB,QAAS;QACnCY,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGjB,SAAU;KACrC;IAGH,MAAMO,MAAM,CAAC,0CAA0C,EAAEC,KAAK,SAAS,CAACF,MAAM,CAAC,CAAC;IAChF,MAAM,IAAIG,MAAMF;AAClB;AAEO,SAASH,UACdE,IAAc,EACdP,KAAa,EACbC,MAAc,EACdC,UAAkB,EAClBC,WAAmB,EACnBC,MAAgC;IAEhC,IAAIO,SAA2C;QAAC;QAAG;QAAG;QAAG;KAAE;IAEzDA,SADEP,AAAW,oBAAXA,UAA8BA,AAAW,kBAAXA,SACvBS,gBAAgBN,MAAMP,OAAOC,UAC7BG,AAAW,aAAXA,SACAkB,gBAAgBf,MAAMP,OAAOC,UAC7BG,AAAW,eAAXA,SACAmB,gBAAgBhB,MAAMP,OAAOC,UAE7BK,cAAcC;IAGzBI,MAAM,CAAC,EAAE,GAAGC,KAAK,GAAG,CAACD,MAAM,CAAC,EAAE,EAAET;IAChCS,MAAM,CAAC,EAAE,GAAGC,KAAK,GAAG,CAACD,MAAM,CAAC,EAAE,EAAER;IAEhC,OAAOQ;AACT;AAGO,SAASY,gBACdhB,IAAc,EACdP,KAAa,EACbC,MAAc;IAEd,OAAO;QACLW,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;QAC/BY,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;QAChCW,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;QAC/BY,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;KACjC;AACH;AAGO,SAASqB,gBACdf,IAAc,EACdP,KAAa,EACbC,MAAc;IAEd,MAAMuB,OAAOZ,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;IAC5C,MAAMyB,MAAMb,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;IAC5C,MAAMyB,QAAQd,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;IAC7C,MAAM2B,SAASf,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;IAC/C,OAAO;QAACuB;QAAMC;QAAKC;QAAOC;KAAO;AACnC;AAEO,SAASC,gBACdrB,IAAc,EACdP,KAAa,EACbC,MAAc,EACd4B,UAAU,CAAC,EACXC,UAAU,CAAC,EACX5B,aAAaF,KAAK,EAClBG,cAAcF,MAAM,EACpBG,MAAiC;IAEjCR,kBACE,mBACAW,MACAP,OACAC,QACA,UACA4B,SACAC,SACA,SACA5B,YACAC,aACA,UACAC;IAEF,MAAM,CAACoB,MAAMC,KAAKC,OAAOC,OAAO,GAAGtB,UACjCE,MACAP,OACAC,QACAC,YACAC,aACAC;IAIF,MAAM2B,WAAWP;IACjB,MAAMQ,UAAUP;IAChB,IAAIQ,YAAYP,QAAQF;IACxB,IAAIU,aAAaP,SAASF;IAI1B,IAAIM,WAAWE,YAAYjC,OACzBiC,YAAYjC,QAAQ+B;IAItB,IAAIC,UAAUE,aAAajC,QACzBiC,aAAajC,SAAS+B;IAIxBC,YAAYrB,KAAK,GAAG,CAAC,GAAGqB;IACxBC,aAAatB,KAAK,GAAG,CAAC,GAAGsB;IAEzB,MAAMC,OAAO;QACX,MAAMJ,WAAWF;QACjB,KAAKG,UAAUF;QACf,OAAOG;QACP,QAAQC;IACV;IACAtC,kBAAkB,4BAA4BuC;IAE9C,OAAOA;AACT;AAEO,SAASC,WAAWC,KAAa;IACtC,MAAMC,UAAU1B,KAAK,GAAG,IAAIyB,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,IAAI;IACnD,MAAMC,SAAS5B,KAAK,GAAG,IAAIyB,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,GAAG;IACjD,MAAME,WAAW7B,KAAK,GAAG,IAAIyB,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,IAAI,GAAGA,EAAE,KAAK;IAC9D,MAAMG,YAAY9B,KAAK,GAAG,IAAIyB,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,GAAG,GAAGA,EAAE,MAAM;IAC/D,OAAO;QACL,MAAMD;QACN,KAAKE;QACL,OAAOC,WAAWH;QAClB,QAAQI,YAAYF;IACtB;AACF;AAGO,SAASG,iBACdR,IAAU,EACVS,UAAgB,EAChBxC,MAAgC;IAEhC,MAAMyC,cAAczC,AAAW,oBAAXA,SAA6B,MAAM;IACvD,MAAM0C,iBAAiB;IAGvB,MAAMC,wBACJZ,KAAK,KAAK,GAAGU,cACTjC,KAAK,IAAI,CAAEiC,AAAAA,CAAAA,cAAcV,KAAK,KAAI,IAAK,KACvCW;IACN,MAAME,sBACJb,KAAK,MAAM,GAAGU,cACVjC,KAAK,IAAI,CAAEiC,AAAAA,CAAAA,cAAcV,KAAK,MAAK,IAAK,KACxCW;IAGN,IAAIG,WAAWrC,KAAK,GAAG,CAACiC,aAAaV,KAAK,KAAK,GAAGY,AAAwB,IAAxBA;IAClD,IAAIG,YAAYtC,KAAK,GAAG,CAACiC,aAAaV,KAAK,MAAM,GAAGa,AAAsB,IAAtBA;IAGpD,IAAIG,UAAUhB,KAAK,IAAI,GAAGY;IAC1B,IAAIK,SAASjB,KAAK,GAAG,GAAGa;IAIxB,IAAIG,UAAUF,WAAWL,WAAW,KAAK,EACvCO,UAAUP,WAAW,KAAK,GAAGK;IAI/B,IAAIG,SAASF,YAAYN,WAAW,MAAM,EACxCQ,SAASR,WAAW,MAAM,GAAGM;IAI/BC,UAAUvC,KAAK,GAAG,CAAC,GAAGuC;IACtBC,SAASxC,KAAK,GAAG,CAAC,GAAGwC;IAIrB,IAAID,UAAUF,WAAWL,WAAW,KAAK,EACvCK,WAAWL,WAAW,KAAK,GAAGO;IAEhC,IAAIC,SAASF,YAAYN,WAAW,MAAM,EACxCM,YAAYN,WAAW,MAAM,GAAGQ;IAGlCjB,KAAK,IAAI,GAAGgB;IACZhB,KAAK,GAAG,GAAGiB;IACXjB,KAAK,KAAK,GAAGc;IACbd,KAAK,MAAM,GAAGe;IAEd,OAAOf;AACT;AAEO,eAAekB,kBACpBC,gBAAwB,EACxBC,IAAkC,EAClCC,IAAU;IAEV,MAAMC,eAAeC,WAAWH;IAChC,MAAMI,kCAAkCF,aAAc,MAAM,CAC1D,CAACG;QACC,IAAIA,YAAY,UAAU,CAAC,QAAQ,KAAKC,SAAS,IAAI,EACnD,OAAO;QAET,OAAO;IACT;IAGF,MAAMC,eAAe,MAAMC,wBAAwB;QACjD,gBAAgBT;QAChB,sBAAsBK;QACtBH;IACF;IACA,OAAOM;AACT;AAEO,SAASE,uBACdC,KAAuB,EACvBC,WAAgC,EAChCC,KAAc;IAEd,MAAMC,OAA+B,EAAE;IAEvC,KAAK,MAAMC,QAAQJ,MAAO;QACxB,MAAMK,OAAOD,KAAK,IAAI;QAEtB,MAAME,SAASL,YAAY,IAAI,CAAC,CAACK,SAAWA,OAAO,IAAI,KAAKD;QAC5D,IAAI,CAACC,QAAQ;YACXC,QAAQ,IAAI,CACV,CAAC,sBAAsB,EAAEF,KAAK,8BAA8B,CAAC;YAE/D;QACF;QAEA,MAAMG,UAAUF,OAAO,cAAc,IAAID;QACzC,MAAMI,YAAYH,OAAO,WAAW,GAChCI,gBAAgBN,KAAK,KAAK,IAAI,CAAC,GAAGE,OAAO,WAAW,IACpD,CAAC;QAEL,MAAMK,WAAiC;YACrC,CAACH,QAAQ,EAAE;YACX,GAAGC,SAAS;QACd;QAEAN,KAAK,IAAI,CAACQ;IACZ;IAEA,IAAIT,OACFC,KAAK,IAAI,CAAC;QACRD;IACF;IAGF,OAAOC;AACT;AAGO,MAAMS,cAAcC,EAAE,MAAM,CAAC;IAClC,MAAMA,EAAE,MAAM;IACd,KAAKA,EAAE,MAAM;AACf;AAEO,MAAMC,aAAaD,EAAE,MAAM,CAAC;IACjC,OAAOA,EAAE,MAAM;IACf,QAAQA,EAAE,MAAM;IAChB,KAAKA,EAAE,MAAM,GAAG,QAAQ;AAC1B;AAEO,MAAME,aAAaH,YAAY,GAAG,CAACE,YAAY,GAAG,CACvDD,EAAE,MAAM,CAAC;IACP,MAAMA,EAAE,MAAM,GAAG,QAAQ;AAC3B;AAIK,MAAMG,0BAA0BH,EAAE,MAAM,CAAC;IAC9C,QAAQA,EAAAA,KACA,CACJA,EAAE,MAAM,CAAC;QACP,MAAMA,EAAE,MAAM;QACd,KAAKA,EAAE,MAAM;IACf,IAED,QAAQ;IACX,yBAAyBA,EAAE,OAAO,GAAG,QAAQ;AAC/C;AAGO,MAAMI,oBAAoBJ,EAAE,KAAK,CAAC;IACvCA,EAAE,MAAM;IACRA,EAAAA,MACS,CAAC;QACN,QAAQA,EAAE,MAAM;IAClB,GACC,GAAG,CAACG,wBAAwB,OAAO;CACvC;AAMD,MAAME,sBAAsB;AAG5B,MAAMC,wBAAwBN,EAAAA,MACrB,CAAC;IACN,QAAQI;IACR,WAAWJ,EAAE,OAAO,GAAG,QAAQ;IAC/B,WAAWA,EAAE,OAAO,GAAG,QAAQ;IAC/B,OAAOA,EAAE,KAAK,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,OAAO;KAAG,EAAE,QAAQ;AACpD,GACC,WAAW;AAGiBA,EAAAA,MACtB,CAAC;IACN,CAACK,oBAAoB,EAAEL,EAAE,OAAO,CAAC;IACjC,QAAQI;IAGR,WAAWJ,EAAE,OAAO,GAAG,QAAQ;IAC/B,WAAWA,EAAE,OAAO,GAAG,QAAQ;IAC/B,OAAOA,EAAE,OAAO,GAAG,QAAQ;IAG3B,QAAQA,EAAE,KAAK,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,MAAM;KAAG;IACxC,MAAME;AACR,GACC,WAAW;AAYP,MAAMK,4BAA4B,IAChCD;AAGF,MAAME,yBAAyB,CAACC;QAGjCC,mBAKAC;IANJ,IAAIC,cAAcH;IAClB,IAAIC,AAAAA,SAAAA,CAAAA,oBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,kBAAkB,QAAQ,AAAD,MAAM,eACjCE,cAAcA,YAAY,IAAI,CAAC,SAAS;IAI1C,IAAID,AAAAA,SAAAA,CAAAA,qBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,mBAAkB,QAAQ,AAAD,MAAM,aAAa;QAC9C,MAAME,QAAQD,YAAY,IAAI,CAAC,KAAK;QAGpC,IAAIP,uBAAuBQ,OACzB,OAAO;QAKT,IAAI,YAAYA,SAASA,MAAM,MAAM,EACnC,OAAO;IAEX;IAEA,OAAO;AACT;AAEO,MAAMC,2BAA2B,CAACL;IACvCzE,OACEwE,uBAAuBC,QACvB;IAIF,IAAI,AAAiB,YAAjB,OAAOA,OACT,OAAOA;IAIT,IAAIA,SAAS,AAAiB,YAAjB,OAAOA,SAAsBA,MAAM,MAAM,EAAE;QAEtD,IAAI,AAAwB,YAAxB,OAAOA,MAAM,MAAM,EACrB,OAAOA,MAAM,MAAM;QAGrB,IAAI,AAAwB,YAAxB,OAAOA,MAAM,MAAM,IAAiBA,MAAM,MAAM,CAAC,MAAM,EACzD,OAAOA,MAAM,MAAM,CAAC,MAAM;IAE9B;IAGA,OAAOM,OAAON;AAChB;AAEO,MAAMO,8BAA8B,CACzCC,SACAC;QAQIC;IANJ,IAAI,CAACF,SACH,OAAO,EAAE;IAIX,MAAMG,YAAYH;IAClB,IAAIE,AAAAA,SAAAA,CAAAA,kBAAAA,UAAU,IAAI,AAAD,IAAbA,KAAAA,IAAAA,gBAAgB,QAAQ,AAAD,MAAM,eAAeC,UAAU,KAAK,EAAE;QAC/D,MAAMC,OAAOC,OAAO,IAAI,CAACF,UAAU,KAAK;QACxC,OAAOC,KAAK,MAAM,CAAC,CAACE;YAClB,MAAMd,QAAQW,UAAU,KAAK,CAACG,IAAI;YAClC,IAAI,CAACf,uBAAuBC,QAC1B,OAAO;YAIT,IAAIS,cAAc;oBACTM;gBAAP,OAAOA,AAAAA,SAAAA,CAAAA,cAAAA,MAAM,IAAI,AAAD,IAATA,KAAAA,IAAAA,YAAY,QAAQ,AAAD,MAAM;YAClC;YAEA,OAAO;QACT;IACF;IAGA,OAAO,EAAE;AACX;AAEO,MAAM3B,kBAAkB,CAC7B4B,YACAC;IAEA,MAAMC,gBAAgBX,4BAA4BU;IAClD,MAAM7F,SAAS;QAAE,GAAG4F,UAAU;IAAC;IAE/B,KAAK,MAAMG,aAAaD,cAAe;QACrC,MAAME,aAAahG,MAAM,CAAC+F,UAAU;QACpC,IAAIC,YAEF;YAAA,IAAI,AAAsB,YAAtB,OAAOA,YACThG,MAAM,CAAC+F,UAAU,GAAGC;iBACf,IAAI,AAAsB,YAAtB,OAAOA,YAEhB;gBAAA,IAAIA,WAAW,MAAM,EAEnB;oBAAA,IAAI,AAA6B,YAA7B,OAAOA,WAAW,MAAM,EAC1BhG,MAAM,CAAC+F,UAAU,GAAGC,WAAW,MAAM;yBAChC,IACL,AAA6B,YAA7B,OAAOA,WAAW,MAAM,IACxBA,WAAW,MAAM,CAAC,MAAM,EAGxBhG,MAAM,CAAC+F,UAAU,GAAGC,WAAW,MAAM,CAAC,MAAM;gBAC9C;YACF;QACF;IAEJ;IAEA,OAAOhG;AACT;AAEO,MAAMiG,kBAAkB,CAC7BL,YACAC;IAEA,MAAMC,gBAAgBX,4BAA4BU;IAClD,MAAM7F,SAAS;QAAE,GAAG4F,UAAU;IAAC;IAE/B,KAAK,MAAMG,aAAaD,cAAe;QACrC,MAAME,aAAahG,MAAM,CAAC+F,UAAU;QACpC,IAAIC,cAAc,AAAsB,YAAtB,OAAOA,YACvBhG,MAAM,CAAC+F,UAAU,GAAG;YAClB,CAACvB,oBAAoB,EAAE;YACvB,QAAQwB;QACV;IAEJ;IAEA,OAAOhG;AACT;AAUO,MAAMkG,mBAAmB,CAC9BC,UACAN;IAGA,MAAMO,QAAQD,YAAY,CAAC;IAG3B,MAAME,eAAelB,4BAA4BU;IAGjD,IAAIQ,AAAwB,MAAxBA,aAAa,MAAM,EACrB,OAAOR,UAAU,KAAK,CAACO;IAIzB,MAAME,oBAAyC,CAAC;IAChD,KAAK,MAAMP,aAAaM,aACtB,IAAIN,aAAaK,OACfE,iBAAiB,CAACP,UAAU,GAAGK,KAAK,CAACL,UAAU;IAKnD,MAAMQ,sBAA2C,CAAC;IAClD,IAAK,MAAMb,OAAOU,MAChB,IAAIC,aAAa,QAAQ,CAACX,MAExBa,mBAAmB,CAACb,IAAI,GAAG;QAAE,QAAQ;IAAU;SAE/Ca,mBAAmB,CAACb,IAAI,GAAGU,KAAK,CAACV,IAAI;IAKzC,MAAMc,YAAYX,UAAU,KAAK,CAACU;IAGlC,IAAK,MAAMR,aAAaO,kBACtBE,SAAS,CAACT,UAAU,GAAGO,iBAAiB,CAACP,UAAU;IAGrD,OAAOS;AACT"}
|
|
1
|
+
{"version":3,"file":"ai-model/common.mjs","sources":["webpack://@midscene/core/./src/ai-model/common.ts"],"sourcesContent":["import type {\n BaseElement,\n DeviceAction,\n ElementTreeNode,\n MidsceneYamlFlowItem,\n PlanningAction,\n Rect,\n Size,\n} from '@/types';\nimport { assert } from '@midscene/shared/utils';\n\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\n\nimport type { PlanningLocateParam } from '@/types';\nimport { NodeType } from '@midscene/shared/constants';\nimport type { TVlModeTypes } from '@midscene/shared/env';\nimport { treeToList } from '@midscene/shared/extractor';\nimport { compositeElementInfoImg } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { z } from 'zod';\n\nexport type AIArgs = ChatCompletionMessageParam[];\n\nexport enum AIActionType {\n ASSERT = 0,\n INSPECT_ELEMENT = 1,\n EXTRACT_DATA = 2,\n PLAN = 3,\n DESCRIBE_ELEMENT = 4,\n TEXT = 5,\n}\n\nconst defaultBboxSize = 20; // must be even number\nconst debugInspectUtils = getDebug('ai:common');\n\n// transform the param of locate from qwen mode\nexport function fillBboxParam(\n locate: PlanningLocateParam,\n width: number,\n height: number,\n rightLimit: number,\n bottomLimit: number,\n vlMode: TVlModeTypes | undefined,\n) {\n // The Qwen model might have hallucinations of naming bbox as bbox_2d.\n if ((locate as any).bbox_2d && !locate?.bbox) {\n locate.bbox = (locate as any).bbox_2d;\n // biome-ignore lint/performance/noDelete: <explanation>\n delete (locate as any).bbox_2d;\n }\n\n if (locate?.bbox) {\n locate.bbox = adaptBbox(\n locate.bbox,\n width,\n height,\n rightLimit,\n bottomLimit,\n vlMode,\n );\n }\n\n return locate;\n}\n\nexport function adaptQwenBbox(\n bbox: number[],\n): [number, number, number, number] {\n if (bbox.length < 2) {\n const msg = `invalid bbox data for qwen-vl mode: ${JSON.stringify(bbox)} `;\n throw new Error(msg);\n }\n\n const result: [number, number, number, number] = [\n Math.round(bbox[0]),\n Math.round(bbox[1]),\n typeof bbox[2] === 'number'\n ? Math.round(bbox[2])\n : Math.round(bbox[0] + defaultBboxSize),\n typeof bbox[3] === 'number'\n ? Math.round(bbox[3])\n : Math.round(bbox[1] + defaultBboxSize),\n ];\n return result;\n}\n\nexport function adaptDoubaoBbox(\n bbox: string[] | number[] | string,\n width: number,\n height: number,\n): [number, number, number, number] {\n assert(\n width > 0 && height > 0,\n 'width and height must be greater than 0 in doubao mode',\n );\n\n if (typeof bbox === 'string') {\n assert(\n /^(\\d+)\\s(\\d+)\\s(\\d+)\\s(\\d+)$/.test(bbox.trim()),\n `invalid bbox data string for doubao-vision mode: ${bbox}`,\n );\n const splitted = bbox.split(' ');\n if (splitted.length === 4) {\n return [\n Math.round((Number(splitted[0]) * width) / 1000),\n Math.round((Number(splitted[1]) * height) / 1000),\n Math.round((Number(splitted[2]) * width) / 1000),\n Math.round((Number(splitted[3]) * height) / 1000),\n ];\n }\n throw new Error(`invalid bbox data string for doubao-vision mode: ${bbox}`);\n }\n\n if (Array.isArray(bbox) && Array.isArray(bbox[0])) {\n bbox = bbox[0];\n }\n\n let bboxList: number[] = [];\n if (Array.isArray(bbox) && typeof bbox[0] === 'string') {\n bbox.forEach((item) => {\n if (typeof item === 'string' && item.includes(',')) {\n const [x, y] = item.split(',');\n bboxList.push(Number(x.trim()), Number(y.trim()));\n } else if (typeof item === 'string' && item.includes(' ')) {\n const [x, y] = item.split(' ');\n bboxList.push(Number(x.trim()), Number(y.trim()));\n } else {\n bboxList.push(Number(item));\n }\n });\n } else {\n bboxList = bbox as any;\n }\n\n if (bboxList.length === 4 || bboxList.length === 5) {\n return [\n Math.round((bboxList[0] * width) / 1000),\n Math.round((bboxList[1] * height) / 1000),\n Math.round((bboxList[2] * width) / 1000),\n Math.round((bboxList[3] * height) / 1000),\n ];\n }\n\n // treat the bbox as a center point\n if (\n bboxList.length === 6 ||\n bboxList.length === 2 ||\n bboxList.length === 3 ||\n bboxList.length === 7\n ) {\n return [\n Math.max(\n 0,\n Math.round((bboxList[0] * width) / 1000) - defaultBboxSize / 2,\n ),\n Math.max(\n 0,\n Math.round((bboxList[1] * height) / 1000) - defaultBboxSize / 2,\n ),\n Math.min(\n width,\n Math.round((bboxList[0] * width) / 1000) + defaultBboxSize / 2,\n ),\n Math.min(\n height,\n Math.round((bboxList[1] * height) / 1000) + defaultBboxSize / 2,\n ),\n ];\n }\n\n if (bbox.length === 8) {\n return [\n Math.round((bboxList[0] * width) / 1000),\n Math.round((bboxList[1] * height) / 1000),\n Math.round((bboxList[4] * width) / 1000),\n Math.round((bboxList[5] * height) / 1000),\n ];\n }\n\n const msg = `invalid bbox data for doubao-vision mode: ${JSON.stringify(bbox)} `;\n throw new Error(msg);\n}\n\nexport function adaptBbox(\n bbox: number[],\n width: number,\n height: number,\n rightLimit: number,\n bottomLimit: number,\n vlMode: TVlModeTypes | undefined,\n): [number, number, number, number] {\n let result: [number, number, number, number] = [0, 0, 0, 0];\n if (vlMode === 'doubao-vision' || vlMode === 'vlm-ui-tars') {\n result = adaptDoubaoBbox(bbox, width, height);\n } else if (vlMode === 'gemini') {\n result = adaptGeminiBbox(bbox, width, height);\n } else if (vlMode === 'qwen3-vl') {\n result = normalized01000(bbox, width, height);\n } else {\n result = adaptQwenBbox(bbox);\n }\n\n result[2] = Math.min(result[2], rightLimit);\n result[3] = Math.min(result[3], bottomLimit);\n\n return result;\n}\n\n// x1, y1, x2, y2 -> 0-1000\nexport function normalized01000(\n bbox: number[],\n width: number,\n height: number,\n): [number, number, number, number] {\n return [\n Math.round((bbox[0] * width) / 1000),\n Math.round((bbox[1] * height) / 1000),\n Math.round((bbox[2] * width) / 1000),\n Math.round((bbox[3] * height) / 1000),\n ];\n}\n\n// y1, x1, y2, x2 -> 0-1000\nexport function adaptGeminiBbox(\n bbox: number[],\n width: number,\n height: number,\n): [number, number, number, number] {\n const left = Math.round((bbox[1] * width) / 1000);\n const top = Math.round((bbox[0] * height) / 1000);\n const right = Math.round((bbox[3] * width) / 1000);\n const bottom = Math.round((bbox[2] * height) / 1000);\n return [left, top, right, bottom];\n}\n\nexport function adaptBboxToRect(\n bbox: number[],\n width: number,\n height: number,\n offsetX = 0,\n offsetY = 0,\n rightLimit = width,\n bottomLimit = height,\n vlMode?: TVlModeTypes | undefined,\n): Rect {\n debugInspectUtils(\n 'adaptBboxToRect',\n bbox,\n width,\n height,\n 'offset',\n offsetX,\n offsetY,\n 'limit',\n rightLimit,\n bottomLimit,\n 'vlMode',\n vlMode,\n );\n const [left, top, right, bottom] = adaptBbox(\n bbox,\n width,\n height,\n rightLimit,\n bottomLimit,\n vlMode,\n );\n\n // Calculate initial rect dimensions\n const rectLeft = left;\n const rectTop = top;\n let rectWidth = right - left;\n let rectHeight = bottom - top;\n\n // Ensure the rect doesn't exceed image boundaries\n // If right edge exceeds width, adjust the width\n if (rectLeft + rectWidth > width) {\n rectWidth = width - rectLeft;\n }\n\n // If bottom edge exceeds height, adjust the height\n if (rectTop + rectHeight > height) {\n rectHeight = height - rectTop;\n }\n\n // Ensure minimum dimensions (width and height should be at least 1)\n rectWidth = Math.max(1, rectWidth);\n rectHeight = Math.max(1, rectHeight);\n\n const rect = {\n left: rectLeft + offsetX,\n top: rectTop + offsetY,\n width: rectWidth,\n height: rectHeight,\n };\n debugInspectUtils('adaptBboxToRect, result=', rect);\n\n return rect;\n}\n\nexport function mergeRects(rects: Rect[]) {\n const minLeft = Math.min(...rects.map((r) => r.left));\n const minTop = Math.min(...rects.map((r) => r.top));\n const maxRight = Math.max(...rects.map((r) => r.left + r.width));\n const maxBottom = Math.max(...rects.map((r) => r.top + r.height));\n return {\n left: minLeft,\n top: minTop,\n width: maxRight - minLeft,\n height: maxBottom - minTop,\n };\n}\n\n// expand the search area to at least 300 x 300, or add a default padding\nexport function expandSearchArea(\n rect: Rect,\n screenSize: Size,\n vlMode: TVlModeTypes | undefined,\n) {\n const minEdgeSize = vlMode === 'doubao-vision' ? 500 : 300;\n const defaultPadding = 160;\n\n // Calculate padding needed to reach minimum edge size\n const paddingSizeHorizontal =\n rect.width < minEdgeSize\n ? Math.ceil((minEdgeSize - rect.width) / 2)\n : defaultPadding;\n const paddingSizeVertical =\n rect.height < minEdgeSize\n ? Math.ceil((minEdgeSize - rect.height) / 2)\n : defaultPadding;\n\n // Calculate new dimensions (ensure minimum edge size)\n let newWidth = Math.max(minEdgeSize, rect.width + paddingSizeHorizontal * 2);\n let newHeight = Math.max(minEdgeSize, rect.height + paddingSizeVertical * 2);\n\n // Calculate initial position with padding\n let newLeft = rect.left - paddingSizeHorizontal;\n let newTop = rect.top - paddingSizeVertical;\n\n // Ensure the rect doesn't exceed screen boundaries by adjusting position\n // If the rect goes beyond the right edge, shift it left\n if (newLeft + newWidth > screenSize.width) {\n newLeft = screenSize.width - newWidth;\n }\n\n // If the rect goes beyond the bottom edge, shift it up\n if (newTop + newHeight > screenSize.height) {\n newTop = screenSize.height - newHeight;\n }\n\n // Ensure the rect doesn't go beyond the left/top edges\n newLeft = Math.max(0, newLeft);\n newTop = Math.max(0, newTop);\n\n // If after position adjustment, the rect still exceeds screen boundaries,\n // clamp the dimensions to fit within screen\n if (newLeft + newWidth > screenSize.width) {\n newWidth = screenSize.width - newLeft;\n }\n if (newTop + newHeight > screenSize.height) {\n newHeight = screenSize.height - newTop;\n }\n\n rect.left = newLeft;\n rect.top = newTop;\n rect.width = newWidth;\n rect.height = newHeight;\n\n return rect;\n}\n\nexport async function markupImageForLLM(\n screenshotBase64: string,\n tree: ElementTreeNode<BaseElement>,\n size: Size,\n) {\n const elementsInfo = treeToList(tree);\n const elementsPositionInfoWithoutText = elementsInfo!.filter(\n (elementInfo) => {\n if (elementInfo.attributes.nodeType === NodeType.TEXT) {\n return false;\n }\n return true;\n },\n );\n\n const imagePayload = await compositeElementInfoImg({\n inputImgBase64: screenshotBase64,\n elementsPositionInfo: elementsPositionInfoWithoutText,\n size,\n });\n return imagePayload;\n}\n\nexport function buildYamlFlowFromPlans(\n plans: PlanningAction[],\n actionSpace: DeviceAction<any>[],\n sleep?: number,\n): MidsceneYamlFlowItem[] {\n const flow: MidsceneYamlFlowItem[] = [];\n\n for (const plan of plans) {\n const verb = plan.type;\n\n const action = actionSpace.find((action) => action.name === verb);\n if (!action) {\n console.warn(\n `Cannot convert action ${verb} to yaml flow. Will ignore it.`,\n );\n continue;\n }\n\n const flowKey = action.interfaceAlias || verb;\n const flowParam = action.paramSchema\n ? dumpActionParam(plan.param || {}, action.paramSchema)\n : {};\n\n const flowItem: MidsceneYamlFlowItem = {\n [flowKey]: '',\n ...flowParam,\n };\n\n flow.push(flowItem);\n }\n\n if (sleep) {\n flow.push({\n sleep,\n });\n }\n\n return flow;\n}\n\n// Zod schemas for shared types\nexport const PointSchema = z.object({\n left: z.number(),\n top: z.number(),\n});\n\nexport const SizeSchema = z.object({\n width: z.number(),\n height: z.number(),\n dpr: z.number().optional(),\n});\n\nexport const RectSchema = PointSchema.and(SizeSchema).and(\n z.object({\n zoom: z.number().optional(),\n }),\n);\n\n// Zod schema for TMultimodalPrompt\nexport const TMultimodalPromptSchema = z.object({\n images: z\n .array(\n z.object({\n name: z.string(),\n url: z.string(),\n }),\n )\n .optional(),\n convertHttpImage2Base64: z.boolean().optional(),\n});\n\n// Zod schema for TUserPrompt\nexport const TUserPromptSchema = z.union([\n z.string(),\n z\n .object({\n prompt: z.string(),\n })\n .and(TMultimodalPromptSchema.partial()),\n]);\n\n// Generate TypeScript types from Zod schemas\nexport type TMultimodalPrompt = z.infer<typeof TMultimodalPromptSchema>;\nexport type TUserPrompt = z.infer<typeof TUserPromptSchema>;\n\nconst locateFieldFlagName = 'midscene_location_field_flag';\n\n// Schema for locator field input (when users provide locate parameters)\nconst MidsceneLocationInput = z\n .object({\n prompt: TUserPromptSchema,\n deepThink: z.boolean().optional(),\n cacheable: z.boolean().optional(),\n xpath: z.union([z.string(), z.boolean()]).optional(),\n })\n .passthrough();\n\n// Schema for locator field result (when AI returns locate results)\nconst MidsceneLocationResult = z\n .object({\n [locateFieldFlagName]: z.literal(true),\n prompt: TUserPromptSchema,\n\n // optional fields\n deepThink: z.boolean().optional(), // only available in vl model\n cacheable: z.boolean().optional(),\n xpath: z.boolean().optional(), // preset result for xpath\n\n // these two fields will only appear in the result\n center: z.tuple([z.number(), z.number()]),\n rect: RectSchema,\n })\n .passthrough();\n\n// Export the result type - this is used for runtime results that include center and rect\nexport type MidsceneLocationResultType = z.infer<typeof MidsceneLocationResult>;\n\n// Export the input type - this is the inferred type from getMidsceneLocationSchema()\nexport type MidsceneLocationInputType = z.infer<typeof MidsceneLocationInput>;\n\n/**\n * Returns the schema for locator fields.\n * This now returns the input schema which is more permissive and suitable for validation.\n */\nexport const getMidsceneLocationSchema = () => {\n return MidsceneLocationInput;\n};\n\nexport const ifMidsceneLocatorField = (field: any): boolean => {\n // Handle optional fields by getting the inner type\n let actualField = field;\n if (actualField._def?.typeName === 'ZodOptional') {\n actualField = actualField._def.innerType;\n }\n\n // Check if this is a ZodObject\n if (actualField._def?.typeName === 'ZodObject') {\n const shape = actualField._def.shape();\n\n // Method 1: Check for the location field flag (for result schema)\n if (locateFieldFlagName in shape) {\n return true;\n }\n\n // Method 2: Check if it's the input schema by checking for 'prompt' field\n // Input schema has 'prompt' as a required field\n if ('prompt' in shape && shape.prompt) {\n return true;\n }\n }\n\n return false;\n};\n\nexport const dumpMidsceneLocatorField = (field: any): string => {\n assert(\n ifMidsceneLocatorField(field),\n 'field is not a midscene locator field',\n );\n\n // If field is a string, return it directly\n if (typeof field === 'string') {\n return field;\n }\n\n // If field is an object with prompt property\n if (field && typeof field === 'object' && field.prompt) {\n // If prompt is a string, return it directly\n if (typeof field.prompt === 'string') {\n return field.prompt;\n }\n // If prompt is a TUserPrompt object, extract the prompt string\n if (typeof field.prompt === 'object' && field.prompt.prompt) {\n return field.prompt.prompt; // TODO: dump images if necessary\n }\n }\n\n // Fallback: try to convert to string\n return String(field);\n};\n\nexport const findAllMidsceneLocatorField = (\n zodType?: z.ZodType<any>,\n requiredOnly?: boolean,\n): string[] => {\n if (!zodType) {\n return [];\n }\n\n // Check if this is a ZodObject by checking if it has a shape property\n const zodObject = zodType as any;\n if (zodObject._def?.typeName === 'ZodObject' && zodObject.shape) {\n const keys = Object.keys(zodObject.shape);\n return keys.filter((key) => {\n const field = zodObject.shape[key];\n if (!ifMidsceneLocatorField(field)) {\n return false;\n }\n\n // If requiredOnly is true, filter out optional fields\n if (requiredOnly) {\n return field._def?.typeName !== 'ZodOptional';\n }\n\n return true;\n });\n }\n\n // For other ZodType instances, we can't extract field names\n return [];\n};\n\nexport const dumpActionParam = (\n jsonObject: Record<string, any>,\n zodSchema: z.ZodType<any>,\n): Record<string, any> => {\n const locatorFields = findAllMidsceneLocatorField(zodSchema);\n const result = { ...jsonObject };\n\n for (const fieldName of locatorFields) {\n const fieldValue = result[fieldName];\n if (fieldValue) {\n // If it's already a string, keep it as is\n if (typeof fieldValue === 'string') {\n result[fieldName] = fieldValue;\n } else if (typeof fieldValue === 'object') {\n // Check if this field is actually a MidsceneLocationType object\n if (fieldValue.prompt) {\n // If prompt is a string, use it directly\n if (typeof fieldValue.prompt === 'string') {\n result[fieldName] = fieldValue.prompt;\n } else if (\n typeof fieldValue.prompt === 'object' &&\n fieldValue.prompt.prompt\n ) {\n // If prompt is a TUserPrompt object, extract the prompt string\n result[fieldName] = fieldValue.prompt.prompt;\n }\n }\n }\n }\n }\n\n return result;\n};\n\nexport const loadActionParam = (\n jsonObject: Record<string, any>,\n zodSchema: z.ZodType<any>,\n): Record<string, any> => {\n const locatorFields = findAllMidsceneLocatorField(zodSchema);\n const result = { ...jsonObject };\n\n for (const fieldName of locatorFields) {\n const fieldValue = result[fieldName];\n if (fieldValue && typeof fieldValue === 'string') {\n result[fieldName] = {\n [locateFieldFlagName]: true,\n prompt: fieldValue,\n };\n }\n }\n\n return result;\n};\n\n/**\n * Parse and validate action parameters using Zod schema.\n * All fields are validated through Zod, EXCEPT locator fields which are skipped.\n * Default values defined in the schema are automatically applied.\n *\n * Locator fields are special business logic fields with complex validation requirements,\n * so they are intentionally excluded from Zod parsing and use existing validation logic.\n */\nexport const parseActionParam = (\n rawParam: Record<string, any>,\n zodSchema: z.ZodType<any>,\n): Record<string, any> => {\n // Handle undefined or null rawParam by providing an empty object\n const param = rawParam ?? {};\n\n // Find all locate fields in the schema\n const locateFields = findAllMidsceneLocatorField(zodSchema);\n\n // If there are no locate fields, just do normal validation\n if (locateFields.length === 0) {\n return zodSchema.parse(param);\n }\n\n // Extract locate field values to restore later\n const locateFieldValues: Record<string, any> = {};\n for (const fieldName of locateFields) {\n if (fieldName in param) {\n locateFieldValues[fieldName] = param[fieldName];\n }\n }\n\n // Build params for validation - skip locate fields and use dummy values\n const paramsForValidation: Record<string, any> = {};\n for (const key in param) {\n if (locateFields.includes(key)) {\n // Use dummy value to satisfy schema validation\n paramsForValidation[key] = { prompt: '_dummy_' };\n } else {\n paramsForValidation[key] = param[key];\n }\n }\n\n // Validate with dummy locate values\n const validated = zodSchema.parse(paramsForValidation);\n\n // Restore the actual locate field values (unvalidated, as per business requirement)\n for (const fieldName in locateFieldValues) {\n validated[fieldName] = locateFieldValues[fieldName];\n }\n\n return validated;\n};\n"],"names":["AIActionType","defaultBboxSize","debugInspectUtils","getDebug","fillBboxParam","locate","width","height","rightLimit","bottomLimit","vlMode","adaptBbox","adaptQwenBbox","bbox","msg","JSON","Error","result","Math","adaptDoubaoBbox","assert","splitted","Number","Array","bboxList","item","x","y","adaptGeminiBbox","normalized01000","left","top","right","bottom","adaptBboxToRect","offsetX","offsetY","rectLeft","rectTop","rectWidth","rectHeight","rect","mergeRects","rects","minLeft","r","minTop","maxRight","maxBottom","expandSearchArea","screenSize","minEdgeSize","defaultPadding","paddingSizeHorizontal","paddingSizeVertical","newWidth","newHeight","newLeft","newTop","markupImageForLLM","screenshotBase64","tree","size","elementsInfo","treeToList","elementsPositionInfoWithoutText","elementInfo","NodeType","imagePayload","compositeElementInfoImg","buildYamlFlowFromPlans","plans","actionSpace","sleep","flow","plan","verb","action","console","flowKey","flowParam","dumpActionParam","flowItem","PointSchema","z","SizeSchema","RectSchema","TMultimodalPromptSchema","TUserPromptSchema","locateFieldFlagName","MidsceneLocationInput","getMidsceneLocationSchema","ifMidsceneLocatorField","field","_actualField__def","_actualField__def1","actualField","shape","dumpMidsceneLocatorField","String","findAllMidsceneLocatorField","zodType","requiredOnly","_zodObject__def","zodObject","keys","Object","key","_field__def","jsonObject","zodSchema","locatorFields","fieldName","fieldValue","loadActionParam","parseActionParam","rawParam","param","locateFields","locateFieldValues","paramsForValidation","validated"],"mappings":";;;;;;AAuBO,IAAKA,sBAAYA,WAAAA,GAAAA,SAAZA,YAAY;;;;;;;WAAZA;;AASZ,MAAMC,kBAAkB;AACxB,MAAMC,oBAAoBC,SAAS;AAG5B,SAASC,cACdC,MAA2B,EAC3BC,KAAa,EACbC,MAAc,EACdC,UAAkB,EAClBC,WAAmB,EACnBC,MAAgC;IAGhC,IAAKL,OAAe,OAAO,IAAI,CAACA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,IAAI,AAAD,GAAG;QAC5CA,OAAO,IAAI,GAAIA,OAAe,OAAO;QAErC,OAAQA,OAAe,OAAO;IAChC;IAEA,IAAIA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,IAAI,EACdA,OAAO,IAAI,GAAGM,UACZN,OAAO,IAAI,EACXC,OACAC,QACAC,YACAC,aACAC;IAIJ,OAAOL;AACT;AAEO,SAASO,cACdC,IAAc;IAEd,IAAIA,KAAK,MAAM,GAAG,GAAG;QACnB,MAAMC,MAAM,CAAC,oCAAoC,EAAEC,KAAK,SAAS,CAACF,MAAM,CAAC,CAAC;QAC1E,MAAM,IAAIG,MAAMF;IAClB;IAEA,MAAMG,SAA2C;QAC/CC,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE;QAClBK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE;QACC,YAAnB,OAAOA,IAAI,CAAC,EAAE,GACVK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,IAClBK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,GAAGZ;QACN,YAAnB,OAAOY,IAAI,CAAC,EAAE,GACVK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,IAClBK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,GAAGZ;KAC1B;IACD,OAAOgB;AACT;AAEO,SAASE,gBACdN,IAAkC,EAClCP,KAAa,EACbC,MAAc;IAEda,OACEd,QAAQ,KAAKC,SAAS,GACtB;IAGF,IAAI,AAAgB,YAAhB,OAAOM,MAAmB;QAC5BO,OACE,+BAA+B,IAAI,CAACP,KAAK,IAAI,KAC7C,CAAC,iDAAiD,EAAEA,MAAM;QAE5D,MAAMQ,WAAWR,KAAK,KAAK,CAAC;QAC5B,IAAIQ,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACLH,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAIf,QAAS;YAC3CY,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAId,SAAU;YAC5CW,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAIf,QAAS;YAC3CY,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAId,SAAU;SAC7C;QAEH,MAAM,IAAIS,MAAM,CAAC,iDAAiD,EAAEH,MAAM;IAC5E;IAEA,IAAIU,MAAM,OAAO,CAACV,SAASU,MAAM,OAAO,CAACV,IAAI,CAAC,EAAE,GAC9CA,OAAOA,IAAI,CAAC,EAAE;IAGhB,IAAIW,WAAqB,EAAE;IAC3B,IAAID,MAAM,OAAO,CAACV,SAAS,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EACvCA,KAAK,OAAO,CAAC,CAACY;QACZ,IAAI,AAAgB,YAAhB,OAAOA,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YAClD,MAAM,CAACC,GAAGC,EAAE,GAAGF,KAAK,KAAK,CAAC;YAC1BD,SAAS,IAAI,CAACF,OAAOI,EAAE,IAAI,KAAKJ,OAAOK,EAAE,IAAI;QAC/C,OAAO,IAAI,AAAgB,YAAhB,OAAOF,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YACzD,MAAM,CAACC,GAAGC,EAAE,GAAGF,KAAK,KAAK,CAAC;YAC1BD,SAAS,IAAI,CAACF,OAAOI,EAAE,IAAI,KAAKJ,OAAOK,EAAE,IAAI;QAC/C,OACEH,SAAS,IAAI,CAACF,OAAOG;IAEzB;SAEAD,WAAWX;IAGb,IAAIW,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAO;QACLN,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGlB,QAAS;QACnCY,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGjB,SAAU;QACpCW,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGlB,QAAS;QACnCY,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGjB,SAAU;KACrC;IAIH,IACEiB,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAO;QACLN,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGlB,QAAS,QAAQL,kBAAkB;QAE/DiB,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGjB,SAAU,QAAQN,kBAAkB;QAEhEiB,KAAK,GAAG,CACNZ,OACAY,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGlB,QAAS,QAAQL,kBAAkB;QAE/DiB,KAAK,GAAG,CACNX,QACAW,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGjB,SAAU,QAAQN,kBAAkB;KAEjE;IAGH,IAAIY,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAO;QACLK,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGlB,QAAS;QACnCY,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGjB,SAAU;QACpCW,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGlB,QAAS;QACnCY,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGjB,SAAU;KACrC;IAGH,MAAMO,MAAM,CAAC,0CAA0C,EAAEC,KAAK,SAAS,CAACF,MAAM,CAAC,CAAC;IAChF,MAAM,IAAIG,MAAMF;AAClB;AAEO,SAASH,UACdE,IAAc,EACdP,KAAa,EACbC,MAAc,EACdC,UAAkB,EAClBC,WAAmB,EACnBC,MAAgC;IAEhC,IAAIO,SAA2C;QAAC;QAAG;QAAG;QAAG;KAAE;IAEzDA,SADEP,AAAW,oBAAXA,UAA8BA,AAAW,kBAAXA,SACvBS,gBAAgBN,MAAMP,OAAOC,UAC7BG,AAAW,aAAXA,SACAkB,gBAAgBf,MAAMP,OAAOC,UAC7BG,AAAW,eAAXA,SACAmB,gBAAgBhB,MAAMP,OAAOC,UAE7BK,cAAcC;IAGzBI,MAAM,CAAC,EAAE,GAAGC,KAAK,GAAG,CAACD,MAAM,CAAC,EAAE,EAAET;IAChCS,MAAM,CAAC,EAAE,GAAGC,KAAK,GAAG,CAACD,MAAM,CAAC,EAAE,EAAER;IAEhC,OAAOQ;AACT;AAGO,SAASY,gBACdhB,IAAc,EACdP,KAAa,EACbC,MAAc;IAEd,OAAO;QACLW,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;QAC/BY,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;QAChCW,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;QAC/BY,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;KACjC;AACH;AAGO,SAASqB,gBACdf,IAAc,EACdP,KAAa,EACbC,MAAc;IAEd,MAAMuB,OAAOZ,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;IAC5C,MAAMyB,MAAMb,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;IAC5C,MAAMyB,QAAQd,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;IAC7C,MAAM2B,SAASf,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;IAC/C,OAAO;QAACuB;QAAMC;QAAKC;QAAOC;KAAO;AACnC;AAEO,SAASC,gBACdrB,IAAc,EACdP,KAAa,EACbC,MAAc,EACd4B,UAAU,CAAC,EACXC,UAAU,CAAC,EACX5B,aAAaF,KAAK,EAClBG,cAAcF,MAAM,EACpBG,MAAiC;IAEjCR,kBACE,mBACAW,MACAP,OACAC,QACA,UACA4B,SACAC,SACA,SACA5B,YACAC,aACA,UACAC;IAEF,MAAM,CAACoB,MAAMC,KAAKC,OAAOC,OAAO,GAAGtB,UACjCE,MACAP,OACAC,QACAC,YACAC,aACAC;IAIF,MAAM2B,WAAWP;IACjB,MAAMQ,UAAUP;IAChB,IAAIQ,YAAYP,QAAQF;IACxB,IAAIU,aAAaP,SAASF;IAI1B,IAAIM,WAAWE,YAAYjC,OACzBiC,YAAYjC,QAAQ+B;IAItB,IAAIC,UAAUE,aAAajC,QACzBiC,aAAajC,SAAS+B;IAIxBC,YAAYrB,KAAK,GAAG,CAAC,GAAGqB;IACxBC,aAAatB,KAAK,GAAG,CAAC,GAAGsB;IAEzB,MAAMC,OAAO;QACX,MAAMJ,WAAWF;QACjB,KAAKG,UAAUF;QACf,OAAOG;QACP,QAAQC;IACV;IACAtC,kBAAkB,4BAA4BuC;IAE9C,OAAOA;AACT;AAEO,SAASC,WAAWC,KAAa;IACtC,MAAMC,UAAU1B,KAAK,GAAG,IAAIyB,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,IAAI;IACnD,MAAMC,SAAS5B,KAAK,GAAG,IAAIyB,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,GAAG;IACjD,MAAME,WAAW7B,KAAK,GAAG,IAAIyB,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,IAAI,GAAGA,EAAE,KAAK;IAC9D,MAAMG,YAAY9B,KAAK,GAAG,IAAIyB,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,GAAG,GAAGA,EAAE,MAAM;IAC/D,OAAO;QACL,MAAMD;QACN,KAAKE;QACL,OAAOC,WAAWH;QAClB,QAAQI,YAAYF;IACtB;AACF;AAGO,SAASG,iBACdR,IAAU,EACVS,UAAgB,EAChBxC,MAAgC;IAEhC,MAAMyC,cAAczC,AAAW,oBAAXA,SAA6B,MAAM;IACvD,MAAM0C,iBAAiB;IAGvB,MAAMC,wBACJZ,KAAK,KAAK,GAAGU,cACTjC,KAAK,IAAI,CAAEiC,AAAAA,CAAAA,cAAcV,KAAK,KAAI,IAAK,KACvCW;IACN,MAAME,sBACJb,KAAK,MAAM,GAAGU,cACVjC,KAAK,IAAI,CAAEiC,AAAAA,CAAAA,cAAcV,KAAK,MAAK,IAAK,KACxCW;IAGN,IAAIG,WAAWrC,KAAK,GAAG,CAACiC,aAAaV,KAAK,KAAK,GAAGY,AAAwB,IAAxBA;IAClD,IAAIG,YAAYtC,KAAK,GAAG,CAACiC,aAAaV,KAAK,MAAM,GAAGa,AAAsB,IAAtBA;IAGpD,IAAIG,UAAUhB,KAAK,IAAI,GAAGY;IAC1B,IAAIK,SAASjB,KAAK,GAAG,GAAGa;IAIxB,IAAIG,UAAUF,WAAWL,WAAW,KAAK,EACvCO,UAAUP,WAAW,KAAK,GAAGK;IAI/B,IAAIG,SAASF,YAAYN,WAAW,MAAM,EACxCQ,SAASR,WAAW,MAAM,GAAGM;IAI/BC,UAAUvC,KAAK,GAAG,CAAC,GAAGuC;IACtBC,SAASxC,KAAK,GAAG,CAAC,GAAGwC;IAIrB,IAAID,UAAUF,WAAWL,WAAW,KAAK,EACvCK,WAAWL,WAAW,KAAK,GAAGO;IAEhC,IAAIC,SAASF,YAAYN,WAAW,MAAM,EACxCM,YAAYN,WAAW,MAAM,GAAGQ;IAGlCjB,KAAK,IAAI,GAAGgB;IACZhB,KAAK,GAAG,GAAGiB;IACXjB,KAAK,KAAK,GAAGc;IACbd,KAAK,MAAM,GAAGe;IAEd,OAAOf;AACT;AAEO,eAAekB,kBACpBC,gBAAwB,EACxBC,IAAkC,EAClCC,IAAU;IAEV,MAAMC,eAAeC,WAAWH;IAChC,MAAMI,kCAAkCF,aAAc,MAAM,CAC1D,CAACG;QACC,IAAIA,YAAY,UAAU,CAAC,QAAQ,KAAKC,SAAS,IAAI,EACnD,OAAO;QAET,OAAO;IACT;IAGF,MAAMC,eAAe,MAAMC,wBAAwB;QACjD,gBAAgBT;QAChB,sBAAsBK;QACtBH;IACF;IACA,OAAOM;AACT;AAEO,SAASE,uBACdC,KAAuB,EACvBC,WAAgC,EAChCC,KAAc;IAEd,MAAMC,OAA+B,EAAE;IAEvC,KAAK,MAAMC,QAAQJ,MAAO;QACxB,MAAMK,OAAOD,KAAK,IAAI;QAEtB,MAAME,SAASL,YAAY,IAAI,CAAC,CAACK,SAAWA,OAAO,IAAI,KAAKD;QAC5D,IAAI,CAACC,QAAQ;YACXC,QAAQ,IAAI,CACV,CAAC,sBAAsB,EAAEF,KAAK,8BAA8B,CAAC;YAE/D;QACF;QAEA,MAAMG,UAAUF,OAAO,cAAc,IAAID;QACzC,MAAMI,YAAYH,OAAO,WAAW,GAChCI,gBAAgBN,KAAK,KAAK,IAAI,CAAC,GAAGE,OAAO,WAAW,IACpD,CAAC;QAEL,MAAMK,WAAiC;YACrC,CAACH,QAAQ,EAAE;YACX,GAAGC,SAAS;QACd;QAEAN,KAAK,IAAI,CAACQ;IACZ;IAEA,IAAIT,OACFC,KAAK,IAAI,CAAC;QACRD;IACF;IAGF,OAAOC;AACT;AAGO,MAAMS,cAAcC,EAAE,MAAM,CAAC;IAClC,MAAMA,EAAE,MAAM;IACd,KAAKA,EAAE,MAAM;AACf;AAEO,MAAMC,aAAaD,EAAE,MAAM,CAAC;IACjC,OAAOA,EAAE,MAAM;IACf,QAAQA,EAAE,MAAM;IAChB,KAAKA,EAAE,MAAM,GAAG,QAAQ;AAC1B;AAEO,MAAME,aAAaH,YAAY,GAAG,CAACE,YAAY,GAAG,CACvDD,EAAE,MAAM,CAAC;IACP,MAAMA,EAAE,MAAM,GAAG,QAAQ;AAC3B;AAIK,MAAMG,0BAA0BH,EAAE,MAAM,CAAC;IAC9C,QAAQA,EAAAA,KACA,CACJA,EAAE,MAAM,CAAC;QACP,MAAMA,EAAE,MAAM;QACd,KAAKA,EAAE,MAAM;IACf,IAED,QAAQ;IACX,yBAAyBA,EAAE,OAAO,GAAG,QAAQ;AAC/C;AAGO,MAAMI,oBAAoBJ,EAAE,KAAK,CAAC;IACvCA,EAAE,MAAM;IACRA,EAAAA,MACS,CAAC;QACN,QAAQA,EAAE,MAAM;IAClB,GACC,GAAG,CAACG,wBAAwB,OAAO;CACvC;AAMD,MAAME,sBAAsB;AAG5B,MAAMC,wBAAwBN,EAAAA,MACrB,CAAC;IACN,QAAQI;IACR,WAAWJ,EAAE,OAAO,GAAG,QAAQ;IAC/B,WAAWA,EAAE,OAAO,GAAG,QAAQ;IAC/B,OAAOA,EAAE,KAAK,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,OAAO;KAAG,EAAE,QAAQ;AACpD,GACC,WAAW;AAGiBA,EAAAA,MACtB,CAAC;IACN,CAACK,oBAAoB,EAAEL,EAAE,OAAO,CAAC;IACjC,QAAQI;IAGR,WAAWJ,EAAE,OAAO,GAAG,QAAQ;IAC/B,WAAWA,EAAE,OAAO,GAAG,QAAQ;IAC/B,OAAOA,EAAE,OAAO,GAAG,QAAQ;IAG3B,QAAQA,EAAE,KAAK,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,MAAM;KAAG;IACxC,MAAME;AACR,GACC,WAAW;AAYP,MAAMK,4BAA4B,IAChCD;AAGF,MAAME,yBAAyB,CAACC;QAGjCC,mBAKAC;IANJ,IAAIC,cAAcH;IAClB,IAAIC,AAAAA,SAAAA,CAAAA,oBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,kBAAkB,QAAQ,AAAD,MAAM,eACjCE,cAAcA,YAAY,IAAI,CAAC,SAAS;IAI1C,IAAID,AAAAA,SAAAA,CAAAA,qBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,mBAAkB,QAAQ,AAAD,MAAM,aAAa;QAC9C,MAAME,QAAQD,YAAY,IAAI,CAAC,KAAK;QAGpC,IAAIP,uBAAuBQ,OACzB,OAAO;QAKT,IAAI,YAAYA,SAASA,MAAM,MAAM,EACnC,OAAO;IAEX;IAEA,OAAO;AACT;AAEO,MAAMC,2BAA2B,CAACL;IACvCzE,OACEwE,uBAAuBC,QACvB;IAIF,IAAI,AAAiB,YAAjB,OAAOA,OACT,OAAOA;IAIT,IAAIA,SAAS,AAAiB,YAAjB,OAAOA,SAAsBA,MAAM,MAAM,EAAE;QAEtD,IAAI,AAAwB,YAAxB,OAAOA,MAAM,MAAM,EACrB,OAAOA,MAAM,MAAM;QAGrB,IAAI,AAAwB,YAAxB,OAAOA,MAAM,MAAM,IAAiBA,MAAM,MAAM,CAAC,MAAM,EACzD,OAAOA,MAAM,MAAM,CAAC,MAAM;IAE9B;IAGA,OAAOM,OAAON;AAChB;AAEO,MAAMO,8BAA8B,CACzCC,SACAC;QAQIC;IANJ,IAAI,CAACF,SACH,OAAO,EAAE;IAIX,MAAMG,YAAYH;IAClB,IAAIE,AAAAA,SAAAA,CAAAA,kBAAAA,UAAU,IAAI,AAAD,IAAbA,KAAAA,IAAAA,gBAAgB,QAAQ,AAAD,MAAM,eAAeC,UAAU,KAAK,EAAE;QAC/D,MAAMC,OAAOC,OAAO,IAAI,CAACF,UAAU,KAAK;QACxC,OAAOC,KAAK,MAAM,CAAC,CAACE;YAClB,MAAMd,QAAQW,UAAU,KAAK,CAACG,IAAI;YAClC,IAAI,CAACf,uBAAuBC,QAC1B,OAAO;YAIT,IAAIS,cAAc;oBACTM;gBAAP,OAAOA,AAAAA,SAAAA,CAAAA,cAAAA,MAAM,IAAI,AAAD,IAATA,KAAAA,IAAAA,YAAY,QAAQ,AAAD,MAAM;YAClC;YAEA,OAAO;QACT;IACF;IAGA,OAAO,EAAE;AACX;AAEO,MAAM3B,kBAAkB,CAC7B4B,YACAC;IAEA,MAAMC,gBAAgBX,4BAA4BU;IAClD,MAAM7F,SAAS;QAAE,GAAG4F,UAAU;IAAC;IAE/B,KAAK,MAAMG,aAAaD,cAAe;QACrC,MAAME,aAAahG,MAAM,CAAC+F,UAAU;QACpC,IAAIC,YAEF;YAAA,IAAI,AAAsB,YAAtB,OAAOA,YACThG,MAAM,CAAC+F,UAAU,GAAGC;iBACf,IAAI,AAAsB,YAAtB,OAAOA,YAEhB;gBAAA,IAAIA,WAAW,MAAM,EAEnB;oBAAA,IAAI,AAA6B,YAA7B,OAAOA,WAAW,MAAM,EAC1BhG,MAAM,CAAC+F,UAAU,GAAGC,WAAW,MAAM;yBAChC,IACL,AAA6B,YAA7B,OAAOA,WAAW,MAAM,IACxBA,WAAW,MAAM,CAAC,MAAM,EAGxBhG,MAAM,CAAC+F,UAAU,GAAGC,WAAW,MAAM,CAAC,MAAM;gBAC9C;YACF;QACF;IAEJ;IAEA,OAAOhG;AACT;AAEO,MAAMiG,kBAAkB,CAC7BL,YACAC;IAEA,MAAMC,gBAAgBX,4BAA4BU;IAClD,MAAM7F,SAAS;QAAE,GAAG4F,UAAU;IAAC;IAE/B,KAAK,MAAMG,aAAaD,cAAe;QACrC,MAAME,aAAahG,MAAM,CAAC+F,UAAU;QACpC,IAAIC,cAAc,AAAsB,YAAtB,OAAOA,YACvBhG,MAAM,CAAC+F,UAAU,GAAG;YAClB,CAACvB,oBAAoB,EAAE;YACvB,QAAQwB;QACV;IAEJ;IAEA,OAAOhG;AACT;AAUO,MAAMkG,mBAAmB,CAC9BC,UACAN;IAGA,MAAMO,QAAQD,YAAY,CAAC;IAG3B,MAAME,eAAelB,4BAA4BU;IAGjD,IAAIQ,AAAwB,MAAxBA,aAAa,MAAM,EACrB,OAAOR,UAAU,KAAK,CAACO;IAIzB,MAAME,oBAAyC,CAAC;IAChD,KAAK,MAAMP,aAAaM,aACtB,IAAIN,aAAaK,OACfE,iBAAiB,CAACP,UAAU,GAAGK,KAAK,CAACL,UAAU;IAKnD,MAAMQ,sBAA2C,CAAC;IAClD,IAAK,MAAMb,OAAOU,MAChB,IAAIC,aAAa,QAAQ,CAACX,MAExBa,mBAAmB,CAACb,IAAI,GAAG;QAAE,QAAQ;IAAU;SAE/Ca,mBAAmB,CAACb,IAAI,GAAGU,KAAK,CAACV,IAAI;IAKzC,MAAMc,YAAYX,UAAU,KAAK,CAACU;IAGlC,IAAK,MAAMR,aAAaO,kBACtBE,SAAS,CAACT,UAAU,GAAGO,iBAAiB,CAACP,UAAU;IAGrD,OAAOS;AACT"}
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
import { callAI, callAIWithObjectResponse, callAIWithStringResponse } from "./service-caller/index.mjs";
|
|
2
2
|
import { systemPromptToLocateElement } from "./prompt/llm-locator.mjs";
|
|
3
|
-
import { describeUserPage
|
|
3
|
+
import { describeUserPage } from "./prompt/util.mjs";
|
|
4
4
|
import { generatePlaywrightTest, generatePlaywrightTestStream } from "./prompt/playwright-generator.mjs";
|
|
5
5
|
import { generateYamlTest, generateYamlTestStream } from "./prompt/yaml-generator.mjs";
|
|
6
6
|
import { AiExtractElementInfo, AiLocateElement, AiLocateSection } from "./inspect.mjs";
|
|
@@ -8,4 +8,4 @@ import { plan } from "./llm-planning.mjs";
|
|
|
8
8
|
import { AIActionType, PointSchema, RectSchema, SizeSchema, TMultimodalPromptSchema, TUserPromptSchema, adaptBboxToRect, dumpActionParam, findAllMidsceneLocatorField, getMidsceneLocationSchema, loadActionParam, parseActionParam } from "./common.mjs";
|
|
9
9
|
import { resizeImageForUiTars, uiTarsPlanning } from "./ui-tars-planning.mjs";
|
|
10
10
|
import { ConversationHistory } from "./conversation-history.mjs";
|
|
11
|
-
export { AIActionType, AiExtractElementInfo, AiLocateElement, AiLocateSection, ConversationHistory, PointSchema, RectSchema, SizeSchema, TMultimodalPromptSchema, TUserPromptSchema, adaptBboxToRect, callAI, callAIWithObjectResponse, callAIWithStringResponse, describeUserPage, dumpActionParam,
|
|
11
|
+
export { AIActionType, AiExtractElementInfo, AiLocateElement, AiLocateSection, ConversationHistory, PointSchema, RectSchema, SizeSchema, TMultimodalPromptSchema, TUserPromptSchema, adaptBboxToRect, callAI, callAIWithObjectResponse, callAIWithStringResponse, describeUserPage, dumpActionParam, findAllMidsceneLocatorField, generatePlaywrightTest, generatePlaywrightTestStream, generateYamlTest, generateYamlTestStream, getMidsceneLocationSchema, loadActionParam, parseActionParam, plan, resizeImageForUiTars, systemPromptToLocateElement, uiTarsPlanning };
|