@midscene/core 1.0.1-beta-20251120071640.0 → 1.0.1-beta-20251123160938.0

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.
Files changed (84) hide show
  1. package/dist/es/agent/agent.mjs +16 -8
  2. package/dist/es/agent/agent.mjs.map +1 -1
  3. package/dist/es/agent/index.mjs +2 -2
  4. package/dist/es/agent/task-builder.mjs +10 -13
  5. package/dist/es/agent/task-builder.mjs.map +1 -1
  6. package/dist/es/agent/tasks.mjs +14 -13
  7. package/dist/es/agent/tasks.mjs.map +1 -1
  8. package/dist/es/agent/ui-utils.mjs +47 -25
  9. package/dist/es/agent/ui-utils.mjs.map +1 -1
  10. package/dist/es/agent/utils.mjs +1 -1
  11. package/dist/es/agent/utils.mjs.map +1 -1
  12. package/dist/es/ai-model/index.mjs +1 -1
  13. package/dist/es/ai-model/inspect.mjs +2 -2
  14. package/dist/es/ai-model/inspect.mjs.map +1 -1
  15. package/dist/es/ai-model/llm-planning.mjs +4 -4
  16. package/dist/es/ai-model/llm-planning.mjs.map +1 -1
  17. package/dist/es/ai-model/prompt/llm-planning.mjs +14 -148
  18. package/dist/es/ai-model/prompt/llm-planning.mjs.map +1 -1
  19. package/dist/es/ai-model/service-caller/index.mjs +3 -3
  20. package/dist/es/ai-model/service-caller/index.mjs.map +1 -1
  21. package/dist/es/ai-model/ui-tars-planning.mjs +1 -1
  22. package/dist/es/ai-model/ui-tars-planning.mjs.map +1 -1
  23. package/dist/es/common.mjs.map +1 -0
  24. package/dist/es/device/index.mjs +21 -2
  25. package/dist/es/device/index.mjs.map +1 -1
  26. package/dist/es/service/index.mjs +1 -1
  27. package/dist/es/service/index.mjs.map +1 -1
  28. package/dist/es/types.mjs.map +1 -1
  29. package/dist/es/utils.mjs +2 -2
  30. package/dist/es/yaml/player.mjs.map +1 -1
  31. package/dist/es/yaml/utils.mjs.map +1 -1
  32. package/dist/lib/agent/agent.js +16 -8
  33. package/dist/lib/agent/agent.js.map +1 -1
  34. package/dist/lib/agent/index.js +4 -1
  35. package/dist/lib/agent/task-builder.js +10 -13
  36. package/dist/lib/agent/task-builder.js.map +1 -1
  37. package/dist/lib/agent/tasks.js +14 -13
  38. package/dist/lib/agent/tasks.js.map +1 -1
  39. package/dist/lib/agent/ui-utils.js +49 -24
  40. package/dist/lib/agent/ui-utils.js.map +1 -1
  41. package/dist/lib/agent/utils.js +1 -1
  42. package/dist/lib/agent/utils.js.map +1 -1
  43. package/dist/lib/ai-model/index.js +1 -1
  44. package/dist/lib/ai-model/inspect.js +2 -2
  45. package/dist/lib/ai-model/inspect.js.map +1 -1
  46. package/dist/lib/ai-model/llm-planning.js +4 -4
  47. package/dist/lib/ai-model/llm-planning.js.map +1 -1
  48. package/dist/lib/ai-model/prompt/llm-planning.js +14 -148
  49. package/dist/lib/ai-model/prompt/llm-planning.js.map +1 -1
  50. package/dist/lib/ai-model/service-caller/index.js +3 -3
  51. package/dist/lib/ai-model/service-caller/index.js.map +1 -1
  52. package/dist/lib/ai-model/ui-tars-planning.js +1 -1
  53. package/dist/lib/ai-model/ui-tars-planning.js.map +1 -1
  54. package/dist/lib/common.js.map +1 -0
  55. package/dist/lib/device/index.js +43 -15
  56. package/dist/lib/device/index.js.map +1 -1
  57. package/dist/lib/service/index.js +3 -3
  58. package/dist/lib/service/index.js.map +1 -1
  59. package/dist/lib/types.js.map +1 -1
  60. package/dist/lib/utils.js +2 -2
  61. package/dist/lib/yaml/player.js.map +1 -1
  62. package/dist/lib/yaml/utils.js.map +1 -1
  63. package/dist/types/agent/agent.d.ts +4 -2
  64. package/dist/types/agent/index.d.ts +1 -1
  65. package/dist/types/agent/task-builder.d.ts +1 -1
  66. package/dist/types/agent/tasks.d.ts +5 -5
  67. package/dist/types/agent/ui-utils.d.ts +7 -0
  68. package/dist/types/agent/utils.d.ts +1 -1
  69. package/dist/types/ai-model/index.d.ts +3 -3
  70. package/dist/types/ai-model/inspect.d.ts +1 -1
  71. package/dist/types/ai-model/llm-planning.d.ts +1 -0
  72. package/dist/types/ai-model/prompt/llm-planning.d.ts +2 -1
  73. package/dist/types/ai-model/service-caller/index.d.ts +1 -1
  74. package/dist/types/{ai-model/common.d.ts → common.d.ts} +2 -2
  75. package/dist/types/device/index.d.ts +28 -6
  76. package/dist/types/service/index.d.ts +1 -1
  77. package/dist/types/types.d.ts +4 -3
  78. package/dist/types/yaml/utils.d.ts +1 -1
  79. package/dist/types/yaml.d.ts +1 -1
  80. package/package.json +2 -2
  81. package/dist/es/ai-model/common.mjs.map +0 -1
  82. package/dist/lib/ai-model/common.js.map +0 -1
  83. /package/dist/es/{ai-model/common.mjs → common.mjs} +0 -0
  84. /package/dist/lib/{ai-model/common.js → common.js} +0 -0
@@ -34,8 +34,8 @@ class TaskExecutor {
34
34
  onTaskUpdate: null == (_this_hooks = this.hooks) ? void 0 : _this_hooks.onTaskUpdate
35
35
  });
36
36
  }
37
- async convertPlanToExecutable(plans, modelConfig, options) {
38
- return this.taskBuilder.build(plans, modelConfig, options);
37
+ async convertPlanToExecutable(plans, modelConfigForPlanning, modelConfigForDefaultIntent, options) {
38
+ return this.taskBuilder.build(plans, modelConfigForPlanning, modelConfigForDefaultIntent, options);
39
39
  }
40
40
  async loadYamlFlowAsPlanning(userInstruction, yamlString) {
41
41
  const session = this.createExecutionSession(taskTitleStr('Action', userInstruction));
@@ -73,7 +73,7 @@ class TaskExecutor {
73
73
  runner
74
74
  };
75
75
  }
76
- createPlanningTask(userInstruction, actionContext, modelConfig) {
76
+ createPlanningTask(userInstruction, actionContext, modelConfigForPlanning, modelConfigForDefaultIntent, includeBbox) {
77
77
  const task = {
78
78
  type: 'Planning',
79
79
  subType: 'Plan',
@@ -85,8 +85,8 @@ class TaskExecutor {
85
85
  const startTime = Date.now();
86
86
  const { uiContext } = executorContext;
87
87
  assert(uiContext, 'uiContext is required for Planning task');
88
- const { vlMode } = modelConfig;
89
- const uiTarsModelVersion = 'vlm-ui-tars' === vlMode ? modelConfig.uiTarsModelVersion : void 0;
88
+ const { vlMode } = modelConfigForPlanning;
89
+ const uiTarsModelVersion = 'vlm-ui-tars' === vlMode ? modelConfigForPlanning.uiTarsModelVersion : void 0;
90
90
  assert(this.interface.actionSpace, 'actionSpace for device is not implemented');
91
91
  const actionSpace = await this.interface.actionSpace();
92
92
  debug('actionSpace for this interface is:', actionSpace.map((action)=>action.name).join(', '));
@@ -97,8 +97,9 @@ class TaskExecutor {
97
97
  actionContext: param.aiActionContext,
98
98
  interfaceType: this.interface.interfaceType,
99
99
  actionSpace,
100
- modelConfig,
101
- conversationHistory: this.conversationHistory
100
+ modelConfig: modelConfigForPlanning,
101
+ conversationHistory: this.conversationHistory,
102
+ includeBbox
102
103
  });
103
104
  debug('planResult', JSON.stringify(planResult, null, 2));
104
105
  const { actions, log, more_actions_needed_by_instruction, error, usage, rawResponse, sleep } = planResult;
@@ -130,9 +131,9 @@ class TaskExecutor {
130
131
  };
131
132
  return task;
132
133
  }
133
- async runPlans(title, plans, modelConfig) {
134
+ async runPlans(title, plans, modelConfigForPlanning, modelConfigForDefaultIntent) {
134
135
  const session = this.createExecutionSession(title);
135
- const { tasks } = await this.convertPlanToExecutable(plans, modelConfig);
136
+ const { tasks } = await this.convertPlanToExecutable(plans, modelConfigForPlanning, modelConfigForDefaultIntent);
136
137
  const runner = session.getRunner();
137
138
  const result = await session.appendAndRun(tasks);
138
139
  const { output } = result ?? {};
@@ -144,26 +145,26 @@ class TaskExecutor {
144
145
  getReplanningCycleLimit(isVlmUiTars) {
145
146
  return this.replanningCycleLimit || globalConfigManager.getEnvConfigInNumber(MIDSCENE_REPLANNING_CYCLE_LIMIT) || (isVlmUiTars ? defaultVlmUiTarsReplanningCycleLimit : defaultReplanningCycleLimit);
146
147
  }
147
- async action(userPrompt, modelConfig, actionContext, cacheable) {
148
+ async action(userPrompt, modelConfigForPlanning, modelConfigForDefaultIntent, thinkingLevel, includeBboxInPlanning, actionContext, cacheable) {
148
149
  this.conversationHistory.reset();
149
150
  const session = this.createExecutionSession(taskTitleStr('Action', userPrompt));
150
151
  const runner = session.getRunner();
151
152
  let replanCount = 0;
152
153
  const yamlFlow = [];
153
- const replanningCycleLimit = this.getReplanningCycleLimit('vlm-ui-tars' === modelConfig.vlMode);
154
+ const replanningCycleLimit = this.getReplanningCycleLimit('vlm-ui-tars' === modelConfigForPlanning.vlMode);
154
155
  while(true){
155
156
  if (replanCount > replanningCycleLimit) {
156
157
  const errorMsg = `Replanning ${replanningCycleLimit} times, which is more than the limit, please split the task into multiple steps`;
157
158
  return session.appendErrorPlan(errorMsg);
158
159
  }
159
- const planningTask = this.createPlanningTask(userPrompt, actionContext, modelConfig);
160
+ const planningTask = this.createPlanningTask(userPrompt, actionContext, modelConfigForPlanning, modelConfigForDefaultIntent, includeBboxInPlanning);
160
161
  const result = await session.appendAndRun(planningTask);
161
162
  const planResult = null == result ? void 0 : result.output;
162
163
  const plans = (null == planResult ? void 0 : planResult.actions) || [];
163
164
  yamlFlow.push(...(null == planResult ? void 0 : planResult.yamlFlow) || []);
164
165
  let executables;
165
166
  try {
166
- executables = await this.convertPlanToExecutable(plans, modelConfig, {
167
+ executables = await this.convertPlanToExecutable(plans, modelConfigForPlanning, modelConfigForDefaultIntent, {
167
168
  cacheable,
168
169
  subTask: true
169
170
  });
@@ -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 Service from '@/service';\nimport type { TaskRunner } from '@/task-runner';\nimport { TaskExecutionError } from '@/task-runner';\nimport type {\n ExecutionTaskApply,\n ExecutionTaskInsightQueryApply,\n ExecutionTaskPlanningApply,\n ExecutionTaskProgressOptions,\n InterfaceType,\n MidsceneYamlFlowItem,\n PlanningAIResponse,\n PlanningAction,\n PlanningActionParamSleep,\n PlanningActionParamWaitFor,\n ServiceDump,\n ServiceExtractOption,\n ServiceExtractParam,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport {\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\ninterface TaskExecutorHooks {\n onTaskUpdate?: (\n runner: TaskRunner,\n error?: TaskExecutionError,\n ) => Promise<void> | void;\n}\n\nconst debug = getDebug('device-task-executor');\nconst defaultReplanningCycleLimit = 10;\nconst defaultVlmUiTarsReplanningCycleLimit = 40;\n\nexport { TaskExecutionError };\n\nexport class TaskExecutor {\n interface: AbstractInterface;\n\n service: Service;\n\n taskCache?: TaskCache;\n\n private readonly taskBuilder: TaskBuilder;\n\n private conversationHistory: ConversationHistory;\n\n onTaskStartCallback?: ExecutionTaskProgressOptions['onTaskStart'];\n\n private readonly hooks?: TaskExecutorHooks;\n\n replanningCycleLimit?: number;\n\n // @deprecated use .interface instead\n get page() {\n return this.interface;\n }\n\n constructor(\n interfaceInstance: AbstractInterface,\n service: Service,\n opts: {\n taskCache?: TaskCache;\n onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n replanningCycleLimit?: number;\n hooks?: TaskExecutorHooks;\n },\n ) {\n this.interface = interfaceInstance;\n this.service = service;\n this.taskCache = opts.taskCache;\n this.onTaskStartCallback = opts?.onTaskStart;\n this.replanningCycleLimit = opts.replanningCycleLimit;\n this.hooks = opts.hooks;\n this.conversationHistory = new ConversationHistory();\n this.taskBuilder = new TaskBuilder({\n interfaceInstance,\n service,\n taskCache: opts.taskCache,\n });\n }\n\n private createExecutionSession(\n title: string,\n options?: { tasks?: ExecutionTaskApply[] },\n ) {\n return new ExecutionSession(\n title,\n () => Promise.resolve(this.service.contextRetrieverFn()),\n {\n onTaskStart: this.onTaskStartCallback,\n tasks: options?.tasks,\n onTaskUpdate: this.hooks?.onTaskUpdate,\n },\n );\n }\n\n public async convertPlanToExecutable(\n plans: PlanningAction[],\n 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 param: {\n userInstruction,\n },\n executor: async (param, executorContext) => {\n const { uiContext } = executorContext;\n assert(uiContext, 'uiContext is required for Planning task');\n return {\n output: {\n actions: [],\n more_actions_needed_by_instruction: false,\n log: '',\n yamlString,\n },\n cache: {\n hit: true,\n },\n hitBy: {\n from: 'Cache',\n context: {\n yamlString,\n },\n },\n };\n },\n };\n const runner = session.getRunner();\n await session.appendAndRun(task);\n\n return {\n runner,\n };\n }\n\n 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 param: {\n userInstruction,\n aiActionContext: actionContext,\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: param.aiActionContext,\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 runner = session.getRunner();\n const result = await session.appendAndRun(tasks);\n const { output } = result ?? {};\n return {\n output,\n runner,\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 = result?.output as PlanningAIResponse | undefined;\n\n // Execute planned actions\n const plans = planResult?.actions || [];\n yamlFlow.push(...(planResult?.yamlFlow || []));\n\n let executables: Awaited<ReturnType<typeof this.convertPlanToExecutable>>;\n try {\n executables = await this.convertPlanToExecutable(plans, modelConfig, {\n cacheable,\n subTask: true,\n });\n } catch (error) {\n return session.appendErrorPlan(\n `Error converting plans to executable tasks: ${error}, plans: ${JSON.stringify(\n plans,\n )}`,\n );\n }\n await session.appendAndRun(executables.tasks);\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 const finalResult = {\n output: {\n yamlFlow,\n },\n runner,\n };\n return finalResult;\n }\n\n private createTypeQueryTask(\n type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert' | 'WaitFor',\n demand: ServiceExtractParam,\n modelConfig: IModelConfig,\n opt?: ServiceExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ) {\n const queryTask: ExecutionTaskInsightQueryApply = {\n type: 'Insight',\n subType: type,\n param: {\n dataDemand: multimodalPrompt\n ? ({\n demand,\n multimodalPrompt,\n } as never)\n : demand, // for user param presentation in report right sidebar\n },\n executor: async (param, taskContext) => {\n const { task } = taskContext;\n let queryDump: ServiceDump | undefined;\n const applyDump = (dump: ServiceDump) => {\n queryDump = dump;\n task.log = {\n dump,\n };\n };\n\n // Get context for query operations\n const uiContext = taskContext.uiContext;\n assert(uiContext, 'uiContext is required for Query task');\n\n const ifTypeRestricted = type !== 'Query';\n let demandInput = demand;\n let keyOfResult = 'result';\n if (ifTypeRestricted && (type === 'Assert' || type === 'WaitFor')) {\n keyOfResult = 'StatementIsTruthy';\n const booleanPrompt =\n type === 'Assert'\n ? `Boolean, whether the following statement is true: ${demand}`\n : `Boolean, the user wants to do some 'wait for' operation, please check whether the following statement is true: ${demand}`;\n demandInput = {\n [keyOfResult]: booleanPrompt,\n };\n } else if (ifTypeRestricted) {\n demandInput = {\n [keyOfResult]: `${type}, ${demand}`,\n };\n }\n\n let extractResult;\n\n let extraPageDescription = '';\n if (opt?.domIncluded && this.interface.getElementsNodeTree) {\n debug('appending tree info for page');\n const tree = await this.interface.getElementsNodeTree();\n extraPageDescription = await descriptionOfTree(\n tree,\n 200,\n false,\n opt?.domIncluded === 'visible-only',\n );\n }\n\n try {\n extractResult = await this.service.extract<any>(\n demandInput,\n modelConfig,\n opt,\n extraPageDescription,\n multimodalPrompt,\n );\n } catch (error) {\n if (error instanceof ServiceError) {\n applyDump(error.dump);\n }\n throw error;\n }\n\n const { data, usage, thought, dump } = extractResult;\n applyDump(dump);\n\n let outputResult = data;\n if (ifTypeRestricted) {\n // If AI returned a plain string instead of structured format, use it directly\n if (typeof data === 'string') {\n outputResult = data;\n } else if (type === 'WaitFor') {\n if (data === null || data === undefined) {\n outputResult = false;\n } else {\n outputResult = (data as any)[keyOfResult];\n }\n } else if (data === null || data === undefined) {\n outputResult = null;\n } else {\n assert(\n data?.[keyOfResult] !== undefined,\n 'No result in query data',\n );\n outputResult = (data as any)[keyOfResult];\n }\n }\n\n if (type === 'Assert' && !outputResult) {\n task.usage = usage;\n task.thought = thought;\n throw new Error(`Assertion failed: ${thought}`);\n }\n\n return {\n output: outputResult,\n log: queryDump,\n usage,\n thought,\n };\n },\n };\n\n return queryTask;\n }\n async createTypeQueryExecution<T>(\n type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert',\n demand: ServiceExtractParam,\n modelConfig: IModelConfig,\n opt?: ServiceExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ): Promise<ExecutionResult<T>> {\n const session = this.createExecutionSession(\n taskTitleStr(\n type,\n typeof demand === 'string' ? demand : JSON.stringify(demand),\n ),\n );\n\n const queryTask = await this.createTypeQueryTask(\n type,\n demand,\n modelConfig,\n opt,\n multimodalPrompt,\n );\n\n const runner = session.getRunner();\n const result = await session.appendAndRun(queryTask);\n\n if (!result) {\n throw new Error(\n 'result of taskExecutor.flush() is undefined in function createTypeQueryTask',\n );\n }\n\n const { output, thought } = result;\n\n return {\n output,\n thought,\n runner,\n };\n }\n\n private sleepPlan(timeMs: number): PlanningAction<PlanningActionParamSleep> {\n return {\n type: 'Sleep',\n param: {\n timeMs,\n },\n };\n }\n\n async taskForSleep(timeMs: number, _modelConfig: IModelConfig) {\n return this.taskBuilder.createSleepTask({\n timeMs,\n });\n }\n\n async waitFor(\n assertion: TUserPrompt,\n opt: PlanningActionParamWaitFor,\n modelConfig: IModelConfig,\n ): Promise<ExecutionResult<void>> {\n const { textPrompt, multimodalPrompt } = parsePrompt(assertion);\n\n const description = `waitFor: ${textPrompt}`;\n const session = this.createExecutionSession(\n taskTitleStr('WaitFor', description),\n );\n const runner = session.getRunner();\n const { timeoutMs, checkIntervalMs } = opt;\n\n assert(assertion, 'No assertion for waitFor');\n assert(timeoutMs, 'No timeoutMs for waitFor');\n assert(checkIntervalMs, 'No checkIntervalMs for waitFor');\n\n assert(\n checkIntervalMs <= timeoutMs,\n `wrong config for waitFor: checkIntervalMs must be less than timeoutMs, config: {checkIntervalMs: ${checkIntervalMs}, timeoutMs: ${timeoutMs}}`,\n );\n\n const overallStartTime = Date.now();\n let lastCheckStart = overallStartTime;\n let errorThought = '';\n // Continue checking as long as the previous iteration began within the timeout window.\n while (lastCheckStart - overallStartTime <= timeoutMs) {\n const currentCheckStart = Date.now();\n lastCheckStart = currentCheckStart;\n const queryTask = await this.createTypeQueryTask(\n 'WaitFor',\n textPrompt,\n modelConfig,\n undefined,\n multimodalPrompt,\n );\n\n const result = (await session.appendAndRun(queryTask)) as\n | {\n output: boolean;\n thought?: string;\n }\n | undefined;\n\n if (result?.output) {\n return {\n output: undefined,\n runner,\n };\n }\n\n errorThought =\n result?.thought ||\n (!result && `No result from assertion: ${textPrompt}`) ||\n `unknown error when waiting for assertion: ${textPrompt}`;\n const now = Date.now();\n if (now - currentCheckStart < checkIntervalMs) {\n const timeRemaining = checkIntervalMs - (now - currentCheckStart);\n const sleepTask = this.taskBuilder.createSleepTask({\n timeMs: timeRemaining,\n });\n await session.append(sleepTask);\n }\n }\n\n return session.appendErrorPlan(`waitFor timeout: ${errorThought}`);\n }\n}\n"],"names":["debug","getDebug","defaultReplanningCycleLimit","defaultVlmUiTarsReplanningCycleLimit","TaskExecutor","title","options","_this_hooks","ExecutionSession","Promise","plans","modelConfig","userInstruction","yamlString","session","taskTitleStr","task","param","executorContext","uiContext","assert","runner","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","replanCount","yamlFlow","replanningCycleLimit","errorMsg","planningTask","executables","finalResult","type","demand","opt","multimodalPrompt","queryTask","taskContext","queryDump","applyDump","dump","ifTypeRestricted","demandInput","keyOfResult","booleanPrompt","extractResult","extraPageDescription","tree","descriptionOfTree","ServiceError","data","thought","outputResult","Error","timeMs","_modelConfig","assertion","textPrompt","parsePrompt","description","timeoutMs","checkIntervalMs","overallStartTime","lastCheckStart","errorThought","currentCheckStart","now","sleepTask","interfaceInstance","service","opts","ConversationHistory","TaskBuilder"],"mappings":";;;;;;;;;;;;;;;;;;;;;AAkDA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,8BAA8B;AACpC,MAAMC,uCAAuC;AAItC,MAAMC;IAkBX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IA0BQ,uBACNC,KAAa,EACbC,OAA0C,EAC1C;YAOkBC;QANlB,OAAO,IAAIC,iBACTH,OACA,IAAMI,QAAQ,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,kBAAkB,KACrD;YACE,aAAa,IAAI,CAAC,mBAAmB;YACrC,OAAOH,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,KAAK;YACrB,cAAc,QAAAC,CAAAA,cAAAA,IAAI,CAAC,KAAK,AAAD,IAATA,KAAAA,IAAAA,YAAY,YAAY;QACxC;IAEJ;IAEA,MAAa,wBACXG,KAAuB,EACvBC,WAAyB,EACzBL,OAGC,EACD;QACA,OAAO,IAAI,CAAC,WAAW,CAAC,KAAK,CAACI,OAAOC,aAAaL;IACpD;IAEA,MAAM,uBAAuBM,eAAuB,EAAEC,UAAkB,EAAE;QACxE,MAAMC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUH;QAGzB,MAAMI,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,OAAO;gBACLJ;YACF;YACA,UAAU,OAAOK,OAAOC;gBACtB,MAAM,EAAEC,SAAS,EAAE,GAAGD;gBACtBE,OAAOD,WAAW;gBAClB,OAAO;oBACL,QAAQ;wBACN,SAAS,EAAE;wBACX,oCAAoC;wBACpC,KAAK;wBACLN;oBACF;oBACA,OAAO;wBACL,KAAK;oBACP;oBACA,OAAO;wBACL,MAAM;wBACN,SAAS;4BACPA;wBACF;oBACF;gBACF;YACF;QACF;QACA,MAAMQ,SAASP,QAAQ,SAAS;QAChC,MAAMA,QAAQ,YAAY,CAACE;QAE3B,OAAO;YACLK;QACF;IACF;IAEQ,mBACNT,eAAuB,EACvBU,aAAiC,EACjCX,WAAyB,EACG;QAC5B,MAAMK,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,OAAO;gBACLJ;gBACA,iBAAiBU;YACnB;YACA,UAAU,OAAOL,OAAOC;gBACtB,MAAMK,YAAYC,KAAK,GAAG;gBAC1B,MAAM,EAAEL,SAAS,EAAE,GAAGD;gBACtBE,OAAOD,WAAW;gBAClB,MAAM,EAAEM,MAAM,EAAE,GAAGd;gBACnB,MAAMe,qBACJD,AAAW,kBAAXA,SAA2Bd,YAAY,kBAAkB,GAAGgB;gBAE9DP,OACE,IAAI,CAAC,SAAS,CAAC,WAAW,EAC1B;gBAEF,MAAMQ,cAAc,MAAM,IAAI,CAAC,SAAS,CAAC,WAAW;gBACpD5B,MACE,sCACA4B,YAAY,GAAG,CAAC,CAACC,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;gBAEhDT,OAAOU,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,EACjEjB,MAAM,eAAe,EACrB;oBACE,SAASE;oBACT,eAAeF,MAAM,eAAe;oBACpC,eAAe,IAAI,CAAC,SAAS,CAAC,aAAa;oBAC3CW;oBACAjB;oBACA,qBAAqB,IAAI,CAAC,mBAAmB;gBAC/C;gBAEFX,MAAM,cAAcmC,KAAK,SAAS,CAACH,YAAY,MAAM;gBAErD,MAAM,EACJI,OAAO,EACPC,GAAG,EACHC,kCAAkC,EAClCC,KAAK,EACLC,KAAK,EACLC,WAAW,EACXC,KAAK,EACN,GAAGV;gBAEJd,gBAAgB,IAAI,CAAC,GAAG,GAAG;oBACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;oBAClCuB;gBACF;gBACAvB,gBAAgB,IAAI,CAAC,KAAK,GAAGsB;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,EACrBvB,OACE,CAACkB,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;oBACAb;gBACF;YACF;QACF;QAEA,OAAOH;IACT;IAEA,MAAM,SACJX,KAAa,EACbK,KAAuB,EACvBC,WAAyB,EACC;QAC1B,MAAMG,UAAU,IAAI,CAAC,sBAAsB,CAACT;QAC5C,MAAM,EAAEyC,KAAK,EAAE,GAAG,MAAM,IAAI,CAAC,uBAAuB,CAACpC,OAAOC;QAC5D,MAAMU,SAASP,QAAQ,SAAS;QAChC,MAAMiC,SAAS,MAAMjC,QAAQ,YAAY,CAACgC;QAC1C,MAAM,EAAEE,MAAM,EAAE,GAAGD,UAAU,CAAC;QAC9B,OAAO;YACLC;YACA3B;QACF;IACF;IAEQ,wBAAwB4B,WAAoB,EAAE;QACpD,OACE,IAAI,CAAC,oBAAoB,IACzBC,oBAAoB,oBAAoB,CACtCC,oCAEDF,CAAAA,cACG9C,uCACAD,2BAA0B;IAElC;IAEA,MAAM,OACJkD,UAAkB,EAClBzC,WAAyB,EACzBW,aAAsB,EACtB+B,SAAmB,EAQnB;QACA,IAAI,CAAC,mBAAmB,CAAC,KAAK;QAE9B,MAAMvC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUqC;QAEzB,MAAM/B,SAASP,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,CAC1CN,YACA9B,eACAX;YAGF,MAAMoC,SAAS,MAAMjC,QAAQ,YAAY,CAAC4C;YAC1C,MAAM1B,aAAae,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM;YAGjC,MAAMrC,QAAQsB,AAAAA,CAAAA,QAAAA,aAAAA,KAAAA,IAAAA,WAAY,OAAO,AAAD,KAAK,EAAE;YACvCuB,SAAS,IAAI,IAAKvB,AAAAA,CAAAA,QAAAA,aAAAA,KAAAA,IAAAA,WAAY,QAAQ,AAAD,KAAK,EAAE;YAE5C,IAAI2B;YACJ,IAAI;gBACFA,cAAc,MAAM,IAAI,CAAC,uBAAuB,CAACjD,OAAOC,aAAa;oBACnE0C;oBACA,SAAS;gBACX;YACF,EAAE,OAAOd,OAAO;gBACd,OAAOzB,QAAQ,eAAe,CAC5B,CAAC,4CAA4C,EAAEyB,MAAM,SAAS,EAAEJ,KAAK,SAAS,CAC5EzB,QACC;YAEP;YACA,MAAMI,QAAQ,YAAY,CAAC6C,YAAY,KAAK;YAG5C,IAAI,CAAC3B,CAAAA,QAAAA,aAAAA,KAAAA,IAAAA,WAAY,kCAAkC,AAAD,GAChD;YAIFsB;QACF;QAEA,MAAMM,cAAc;YAClB,QAAQ;gBACNL;YACF;YACAlC;QACF;QACA,OAAOuC;IACT;IAEQ,oBACNC,IAAsE,EACtEC,MAA2B,EAC3BnD,WAAyB,EACzBoD,GAA0B,EAC1BC,gBAAoC,EACpC;QACA,MAAMC,YAA4C;YAChD,MAAM;YACN,SAASJ;YACT,OAAO;gBACL,YAAYG,mBACP;oBACCF;oBACAE;gBACF,IACAF;YACN;YACA,UAAU,OAAO7C,OAAOiD;gBACtB,MAAM,EAAElD,IAAI,EAAE,GAAGkD;gBACjB,IAAIC;gBACJ,MAAMC,YAAY,CAACC;oBACjBF,YAAYE;oBACZrD,KAAK,GAAG,GAAG;wBACTqD;oBACF;gBACF;gBAGA,MAAMlD,YAAY+C,YAAY,SAAS;gBACvC9C,OAAOD,WAAW;gBAElB,MAAMmD,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;oBAC1D/D,MAAM;oBACN,MAAM4E,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,aACA5D,aACAoD,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;oBACL7D,OACE2D,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,IAAM,CAACP,YAAY,AAAD,MAAM7C,QACxB;oBAEFsD,eAAgBF,IAAY,CAACP,YAAY;gBAC3C;gBAGF,IAAIX,AAAS,aAATA,QAAqB,CAACoB,cAAc;oBACtCjE,KAAK,KAAK,GAAGwB;oBACbxB,KAAK,OAAO,GAAGgE;oBACf,MAAM,IAAIE,MAAM,CAAC,kBAAkB,EAAEF,SAAS;gBAChD;gBAEA,OAAO;oBACL,QAAQC;oBACR,KAAKd;oBACL3B;oBACAwC;gBACF;YACF;QACF;QAEA,OAAOf;IACT;IACA,MAAM,yBACJJ,IAA0D,EAC1DC,MAA2B,EAC3BnD,WAAyB,EACzBoD,GAA0B,EAC1BC,gBAAoC,EACP;QAC7B,MAAMlD,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aACE8C,MACA,AAAkB,YAAlB,OAAOC,SAAsBA,SAAS3B,KAAK,SAAS,CAAC2B;QAIzD,MAAMG,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9CJ,MACAC,QACAnD,aACAoD,KACAC;QAGF,MAAM3C,SAASP,QAAQ,SAAS;QAChC,MAAMiC,SAAS,MAAMjC,QAAQ,YAAY,CAACmD;QAE1C,IAAI,CAAClB,QACH,MAAM,IAAImC,MACR;QAIJ,MAAM,EAAElC,MAAM,EAAEgC,OAAO,EAAE,GAAGjC;QAE5B,OAAO;YACLC;YACAgC;YACA3D;QACF;IACF;IAEQ,UAAU8D,MAAc,EAA4C;QAC1E,OAAO;YACL,MAAM;YACN,OAAO;gBACLA;YACF;QACF;IACF;IAEA,MAAM,aAAaA,MAAc,EAAEC,YAA0B,EAAE;QAC7D,OAAO,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;YACtCD;QACF;IACF;IAEA,MAAM,QACJE,SAAsB,EACtBtB,GAA+B,EAC/BpD,WAAyB,EACO;QAChC,MAAM,EAAE2E,UAAU,EAAEtB,gBAAgB,EAAE,GAAGuB,YAAYF;QAErD,MAAMG,cAAc,CAAC,SAAS,EAAEF,YAAY;QAC5C,MAAMxE,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,WAAWyE;QAE1B,MAAMnE,SAASP,QAAQ,SAAS;QAChC,MAAM,EAAE2E,SAAS,EAAEC,eAAe,EAAE,GAAG3B;QAEvC3C,OAAOiE,WAAW;QAClBjE,OAAOqE,WAAW;QAClBrE,OAAOsE,iBAAiB;QAExBtE,OACEsE,mBAAmBD,WACnB,CAAC,iGAAiG,EAAEC,gBAAgB,aAAa,EAAED,UAAU,CAAC,CAAC;QAGjJ,MAAME,mBAAmBnE,KAAK,GAAG;QACjC,IAAIoE,iBAAiBD;QACrB,IAAIE,eAAe;QAEnB,MAAOD,iBAAiBD,oBAAoBF,UAAW;YACrD,MAAMK,oBAAoBtE,KAAK,GAAG;YAClCoE,iBAAiBE;YACjB,MAAM7B,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9C,WACAqB,YACA3E,aACAgB,QACAqC;YAGF,MAAMjB,SAAU,MAAMjC,QAAQ,YAAY,CAACmD;YAO3C,IAAIlB,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM,EAChB,OAAO;gBACL,QAAQpB;gBACRN;YACF;YAGFwE,eACE9C,AAAAA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,OAAO,AAAD,KACb,CAACA,UAAU,CAAC,0BAA0B,EAAEuC,YAAY,IACrD,CAAC,0CAA0C,EAAEA,YAAY;YAC3D,MAAMS,MAAMvE,KAAK,GAAG;YACpB,IAAIuE,MAAMD,oBAAoBJ,iBAAiB;gBAC7C,MAAM7C,gBAAgB6C,kBAAmBK,CAAAA,MAAMD,iBAAgB;gBAC/D,MAAME,YAAY,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;oBACjD,QAAQnD;gBACV;gBACA,MAAM/B,QAAQ,MAAM,CAACkF;YACvB;QACF;QAEA,OAAOlF,QAAQ,eAAe,CAAC,CAAC,iBAAiB,EAAE+E,cAAc;IACnE;IA/hBA,YACEI,iBAAoC,EACpCC,OAAgB,EAChBC,IAKC,CACD;QA9BF;QAEA;QAEA;QAEA,uBAAiB,eAAjB;QAEA,uBAAQ,uBAAR;QAEA;QAEA,uBAAiB,SAAjB;QAEA;QAiBE,IAAI,CAAC,SAAS,GAAGF;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC,KAAK,SAAS;QAC/B,IAAI,CAAC,mBAAmB,GAAGA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW;QAC5C,IAAI,CAAC,oBAAoB,GAAGA,KAAK,oBAAoB;QACrD,IAAI,CAAC,KAAK,GAAGA,KAAK,KAAK;QACvB,IAAI,CAAC,mBAAmB,GAAG,IAAIC;QAC/B,IAAI,CAAC,WAAW,GAAG,IAAIC,YAAY;YACjCJ;YACAC;YACA,WAAWC,KAAK,SAAS;QAC3B;IACF;AA0gBF"}
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 '@/common';\nimport type { AbstractInterface } from '@/device';\nimport type Service from '@/service';\nimport type { TaskRunner } from '@/task-runner';\nimport { TaskExecutionError } from '@/task-runner';\nimport type {\n ExecutionTaskApply,\n ExecutionTaskInsightQueryApply,\n ExecutionTaskPlanningApply,\n ExecutionTaskProgressOptions,\n InterfaceType,\n MidsceneYamlFlowItem,\n PlanningAIResponse,\n PlanningAction,\n PlanningActionParamSleep,\n PlanningActionParamWaitFor,\n ServiceDump,\n ServiceExtractOption,\n ServiceExtractParam,\n ThinkingLevel,\n} from '@/types';\nimport { ServiceError } 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\ninterface TaskExecutorHooks {\n onTaskUpdate?: (\n runner: TaskRunner,\n error?: TaskExecutionError,\n ) => Promise<void> | void;\n}\n\nconst debug = getDebug('device-task-executor');\nconst defaultReplanningCycleLimit = 10;\nconst defaultVlmUiTarsReplanningCycleLimit = 40;\n\nexport { TaskExecutionError };\n\nexport class TaskExecutor {\n interface: AbstractInterface;\n\n service: Service;\n\n taskCache?: TaskCache;\n\n private readonly taskBuilder: TaskBuilder;\n\n private conversationHistory: ConversationHistory;\n\n onTaskStartCallback?: ExecutionTaskProgressOptions['onTaskStart'];\n\n private readonly hooks?: TaskExecutorHooks;\n\n replanningCycleLimit?: number;\n\n // @deprecated use .interface instead\n get page() {\n return this.interface;\n }\n\n constructor(\n interfaceInstance: AbstractInterface,\n service: Service,\n opts: {\n taskCache?: TaskCache;\n onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n replanningCycleLimit?: number;\n hooks?: TaskExecutorHooks;\n },\n ) {\n this.interface = interfaceInstance;\n this.service = service;\n this.taskCache = opts.taskCache;\n this.onTaskStartCallback = opts?.onTaskStart;\n this.replanningCycleLimit = opts.replanningCycleLimit;\n this.hooks = opts.hooks;\n this.conversationHistory = new ConversationHistory();\n this.taskBuilder = new TaskBuilder({\n interfaceInstance,\n service,\n taskCache: opts.taskCache,\n });\n }\n\n private createExecutionSession(\n title: string,\n options?: { tasks?: ExecutionTaskApply[] },\n ) {\n return new ExecutionSession(\n title,\n () => Promise.resolve(this.service.contextRetrieverFn()),\n {\n onTaskStart: this.onTaskStartCallback,\n tasks: options?.tasks,\n onTaskUpdate: this.hooks?.onTaskUpdate,\n },\n );\n }\n\n public async convertPlanToExecutable(\n plans: PlanningAction[],\n modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n options?: {\n cacheable?: boolean;\n subTask?: boolean;\n },\n ) {\n return this.taskBuilder.build(\n plans,\n modelConfigForPlanning,\n modelConfigForDefaultIntent,\n options,\n );\n }\n\n async loadYamlFlowAsPlanning(userInstruction: string, yamlString: string) {\n const session = this.createExecutionSession(\n taskTitleStr('Action', userInstruction),\n );\n\n const task: ExecutionTaskPlanningApply = {\n type: 'Planning',\n subType: 'LoadYaml',\n param: {\n userInstruction,\n },\n executor: async (param, executorContext) => {\n const { uiContext } = executorContext;\n assert(uiContext, 'uiContext is required for Planning task');\n return {\n output: {\n actions: [],\n more_actions_needed_by_instruction: false,\n log: '',\n yamlString,\n },\n cache: {\n hit: true,\n },\n hitBy: {\n from: 'Cache',\n context: {\n yamlString,\n },\n },\n };\n },\n };\n const runner = session.getRunner();\n await session.appendAndRun(task);\n\n return {\n runner,\n };\n }\n\n private createPlanningTask(\n userInstruction: string,\n actionContext: string | undefined,\n modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n includeBbox: boolean,\n ): ExecutionTaskPlanningApply {\n const task: ExecutionTaskPlanningApply = {\n type: 'Planning',\n subType: 'Plan',\n param: {\n userInstruction,\n aiActionContext: actionContext,\n },\n executor: async (param, executorContext) => {\n const startTime = Date.now();\n const { uiContext } = executorContext;\n assert(uiContext, 'uiContext is required for Planning task');\n const { vlMode } = modelConfigForPlanning;\n const uiTarsModelVersion =\n vlMode === 'vlm-ui-tars'\n ? modelConfigForPlanning.uiTarsModelVersion\n : undefined;\n\n assert(\n this.interface.actionSpace,\n 'actionSpace for device is not implemented',\n );\n const actionSpace = await this.interface.actionSpace();\n debug(\n 'actionSpace for this interface is:',\n actionSpace.map((action) => action.name).join(', '),\n );\n assert(Array.isArray(actionSpace), 'actionSpace must be an array');\n if (actionSpace.length === 0) {\n console.warn(\n `ActionSpace for ${this.interface.interfaceType} is empty. This may lead to unexpected behavior.`,\n );\n }\n\n const planResult = await (uiTarsModelVersion ? uiTarsPlanning : plan)(\n param.userInstruction,\n {\n context: uiContext,\n actionContext: param.aiActionContext,\n interfaceType: this.interface.interfaceType as InterfaceType,\n actionSpace,\n modelConfig: modelConfigForPlanning,\n conversationHistory: this.conversationHistory,\n includeBbox,\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 modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n ): Promise<ExecutionResult> {\n const session = this.createExecutionSession(title);\n const { tasks } = await this.convertPlanToExecutable(\n plans,\n modelConfigForPlanning,\n modelConfigForDefaultIntent,\n );\n const runner = session.getRunner();\n const result = await session.appendAndRun(tasks);\n const { output } = result ?? {};\n return {\n output,\n runner,\n };\n }\n\n 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 modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n thinkingLevel: ThinkingLevel,\n includeBboxInPlanning: boolean,\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 modelConfigForPlanning.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 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 modelConfigForPlanning,\n modelConfigForDefaultIntent,\n includeBboxInPlanning,\n );\n\n const result = await session.appendAndRun(planningTask);\n const planResult = result?.output as PlanningAIResponse | undefined;\n\n // Execute planned actions\n const plans = planResult?.actions || [];\n yamlFlow.push(...(planResult?.yamlFlow || []));\n\n let executables: Awaited<ReturnType<typeof this.convertPlanToExecutable>>;\n try {\n executables = await this.convertPlanToExecutable(\n plans,\n modelConfigForPlanning,\n modelConfigForDefaultIntent,\n {\n cacheable,\n subTask: true,\n },\n );\n } catch (error) {\n return session.appendErrorPlan(\n `Error converting plans to executable tasks: ${error}, plans: ${JSON.stringify(\n plans,\n )}`,\n );\n }\n await session.appendAndRun(executables.tasks);\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 const finalResult = {\n output: {\n yamlFlow,\n },\n runner,\n };\n return finalResult;\n }\n\n private createTypeQueryTask(\n type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert' | 'WaitFor',\n demand: ServiceExtractParam,\n modelConfig: IModelConfig,\n opt?: ServiceExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ) {\n const queryTask: ExecutionTaskInsightQueryApply = {\n type: 'Insight',\n subType: type,\n param: {\n dataDemand: multimodalPrompt\n ? ({\n demand,\n multimodalPrompt,\n } as never)\n : demand, // for user param presentation in report right sidebar\n },\n executor: async (param, taskContext) => {\n const { task } = taskContext;\n let queryDump: ServiceDump | undefined;\n const applyDump = (dump: ServiceDump) => {\n queryDump = dump;\n task.log = {\n dump,\n };\n };\n\n // Get context for query operations\n const uiContext = taskContext.uiContext;\n assert(uiContext, 'uiContext is required for Query task');\n\n const ifTypeRestricted = type !== 'Query';\n let demandInput = demand;\n let keyOfResult = 'result';\n if (ifTypeRestricted && (type === 'Assert' || type === 'WaitFor')) {\n keyOfResult = 'StatementIsTruthy';\n const booleanPrompt =\n type === 'Assert'\n ? `Boolean, whether the following statement is true: ${demand}`\n : `Boolean, the user wants to do some 'wait for' operation, please check whether the following statement is true: ${demand}`;\n demandInput = {\n [keyOfResult]: booleanPrompt,\n };\n } else if (ifTypeRestricted) {\n demandInput = {\n [keyOfResult]: `${type}, ${demand}`,\n };\n }\n\n let extractResult;\n\n let extraPageDescription = '';\n if (opt?.domIncluded && this.interface.getElementsNodeTree) {\n debug('appending tree info for page');\n const tree = await this.interface.getElementsNodeTree();\n extraPageDescription = await descriptionOfTree(\n tree,\n 200,\n false,\n opt?.domIncluded === 'visible-only',\n );\n }\n\n try {\n extractResult = await this.service.extract<any>(\n demandInput,\n modelConfig,\n opt,\n extraPageDescription,\n multimodalPrompt,\n );\n } catch (error) {\n if (error instanceof ServiceError) {\n applyDump(error.dump);\n }\n throw error;\n }\n\n const { data, usage, thought, dump } = extractResult;\n applyDump(dump);\n\n let outputResult = data;\n if (ifTypeRestricted) {\n // If AI returned a plain string instead of structured format, use it directly\n if (typeof data === 'string') {\n outputResult = data;\n } else if (type === 'WaitFor') {\n if (data === null || data === undefined) {\n outputResult = false;\n } else {\n outputResult = (data as any)[keyOfResult];\n }\n } else if (data === null || data === undefined) {\n outputResult = null;\n } else {\n assert(\n data?.[keyOfResult] !== undefined,\n 'No result in query data',\n );\n outputResult = (data as any)[keyOfResult];\n }\n }\n\n if (type === 'Assert' && !outputResult) {\n task.usage = usage;\n task.thought = thought;\n throw new Error(`Assertion failed: ${thought}`);\n }\n\n return {\n output: outputResult,\n log: queryDump,\n usage,\n thought,\n };\n },\n };\n\n return queryTask;\n }\n async createTypeQueryExecution<T>(\n type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert',\n demand: ServiceExtractParam,\n modelConfig: IModelConfig,\n opt?: ServiceExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ): Promise<ExecutionResult<T>> {\n const session = this.createExecutionSession(\n taskTitleStr(\n type,\n typeof demand === 'string' ? demand : JSON.stringify(demand),\n ),\n );\n\n const queryTask = await this.createTypeQueryTask(\n type,\n demand,\n modelConfig,\n opt,\n multimodalPrompt,\n );\n\n const runner = session.getRunner();\n const result = await session.appendAndRun(queryTask);\n\n if (!result) {\n throw new Error(\n 'result of taskExecutor.flush() is undefined in function createTypeQueryTask',\n );\n }\n\n const { output, thought } = result;\n\n return {\n output,\n thought,\n runner,\n };\n }\n\n private sleepPlan(timeMs: number): PlanningAction<PlanningActionParamSleep> {\n return {\n type: 'Sleep',\n param: {\n timeMs,\n },\n };\n }\n\n async taskForSleep(timeMs: number, _modelConfig: IModelConfig) {\n return this.taskBuilder.createSleepTask({\n timeMs,\n });\n }\n\n async waitFor(\n assertion: TUserPrompt,\n opt: PlanningActionParamWaitFor,\n modelConfig: IModelConfig,\n ): Promise<ExecutionResult<void>> {\n const { textPrompt, multimodalPrompt } = parsePrompt(assertion);\n\n const description = `waitFor: ${textPrompt}`;\n const session = this.createExecutionSession(\n taskTitleStr('WaitFor', description),\n );\n const runner = session.getRunner();\n const { timeoutMs, checkIntervalMs } = opt;\n\n assert(assertion, 'No assertion for waitFor');\n assert(timeoutMs, 'No timeoutMs for waitFor');\n assert(checkIntervalMs, 'No checkIntervalMs for waitFor');\n\n assert(\n checkIntervalMs <= timeoutMs,\n `wrong config for waitFor: checkIntervalMs must be less than timeoutMs, config: {checkIntervalMs: ${checkIntervalMs}, timeoutMs: ${timeoutMs}}`,\n );\n\n const overallStartTime = Date.now();\n let lastCheckStart = overallStartTime;\n let errorThought = '';\n // Continue checking as long as the previous iteration began within the timeout window.\n while (lastCheckStart - overallStartTime <= timeoutMs) {\n const currentCheckStart = Date.now();\n lastCheckStart = currentCheckStart;\n const queryTask = await this.createTypeQueryTask(\n 'WaitFor',\n textPrompt,\n modelConfig,\n undefined,\n multimodalPrompt,\n );\n\n const result = (await session.appendAndRun(queryTask)) as\n | {\n output: boolean;\n thought?: string;\n }\n | undefined;\n\n if (result?.output) {\n return {\n output: undefined,\n runner,\n };\n }\n\n errorThought =\n result?.thought ||\n (!result && `No result from assertion: ${textPrompt}`) ||\n `unknown error when waiting for assertion: ${textPrompt}`;\n const now = Date.now();\n if (now - currentCheckStart < checkIntervalMs) {\n const timeRemaining = checkIntervalMs - (now - currentCheckStart);\n const sleepTask = this.taskBuilder.createSleepTask({\n timeMs: timeRemaining,\n });\n await session.append(sleepTask);\n }\n }\n\n return session.appendErrorPlan(`waitFor timeout: ${errorThought}`);\n }\n}\n"],"names":["debug","getDebug","defaultReplanningCycleLimit","defaultVlmUiTarsReplanningCycleLimit","TaskExecutor","title","options","_this_hooks","ExecutionSession","Promise","plans","modelConfigForPlanning","modelConfigForDefaultIntent","userInstruction","yamlString","session","taskTitleStr","task","param","executorContext","uiContext","assert","runner","actionContext","includeBbox","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","thinkingLevel","includeBboxInPlanning","cacheable","replanCount","yamlFlow","replanningCycleLimit","errorMsg","planningTask","executables","finalResult","type","demand","modelConfig","opt","multimodalPrompt","queryTask","taskContext","queryDump","applyDump","dump","ifTypeRestricted","demandInput","keyOfResult","booleanPrompt","extractResult","extraPageDescription","tree","descriptionOfTree","ServiceError","data","thought","outputResult","Error","timeMs","_modelConfig","assertion","textPrompt","parsePrompt","description","timeoutMs","checkIntervalMs","overallStartTime","lastCheckStart","errorThought","currentCheckStart","now","sleepTask","interfaceInstance","service","opts","ConversationHistory","TaskBuilder"],"mappings":";;;;;;;;;;;;;;;;;;;;;AAmDA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,8BAA8B;AACpC,MAAMC,uCAAuC;AAItC,MAAMC;IAkBX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IA0BQ,uBACNC,KAAa,EACbC,OAA0C,EAC1C;YAOkBC;QANlB,OAAO,IAAIC,iBACTH,OACA,IAAMI,QAAQ,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,kBAAkB,KACrD;YACE,aAAa,IAAI,CAAC,mBAAmB;YACrC,OAAOH,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,KAAK;YACrB,cAAc,QAAAC,CAAAA,cAAAA,IAAI,CAAC,KAAK,AAAD,IAATA,KAAAA,IAAAA,YAAY,YAAY;QACxC;IAEJ;IAEA,MAAa,wBACXG,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACzCN,OAGC,EACD;QACA,OAAO,IAAI,CAAC,WAAW,CAAC,KAAK,CAC3BI,OACAC,wBACAC,6BACAN;IAEJ;IAEA,MAAM,uBAAuBO,eAAuB,EAAEC,UAAkB,EAAE;QACxE,MAAMC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUH;QAGzB,MAAMI,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,OAAO;gBACLJ;YACF;YACA,UAAU,OAAOK,OAAOC;gBACtB,MAAM,EAAEC,SAAS,EAAE,GAAGD;gBACtBE,OAAOD,WAAW;gBAClB,OAAO;oBACL,QAAQ;wBACN,SAAS,EAAE;wBACX,oCAAoC;wBACpC,KAAK;wBACLN;oBACF;oBACA,OAAO;wBACL,KAAK;oBACP;oBACA,OAAO;wBACL,MAAM;wBACN,SAAS;4BACPA;wBACF;oBACF;gBACF;YACF;QACF;QACA,MAAMQ,SAASP,QAAQ,SAAS;QAChC,MAAMA,QAAQ,YAAY,CAACE;QAE3B,OAAO;YACLK;QACF;IACF;IAEQ,mBACNT,eAAuB,EACvBU,aAAiC,EACjCZ,sBAAoC,EACpCC,2BAAyC,EACzCY,WAAoB,EACQ;QAC5B,MAAMP,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,OAAO;gBACLJ;gBACA,iBAAiBU;YACnB;YACA,UAAU,OAAOL,OAAOC;gBACtB,MAAMM,YAAYC,KAAK,GAAG;gBAC1B,MAAM,EAAEN,SAAS,EAAE,GAAGD;gBACtBE,OAAOD,WAAW;gBAClB,MAAM,EAAEO,MAAM,EAAE,GAAGhB;gBACnB,MAAMiB,qBACJD,AAAW,kBAAXA,SACIhB,uBAAuB,kBAAkB,GACzCkB;gBAENR,OACE,IAAI,CAAC,SAAS,CAAC,WAAW,EAC1B;gBAEF,MAAMS,cAAc,MAAM,IAAI,CAAC,SAAS,CAAC,WAAW;gBACpD9B,MACE,sCACA8B,YAAY,GAAG,CAAC,CAACC,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;gBAEhDV,OAAOW,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,EACjElB,MAAM,eAAe,EACrB;oBACE,SAASE;oBACT,eAAeF,MAAM,eAAe;oBACpC,eAAe,IAAI,CAAC,SAAS,CAAC,aAAa;oBAC3CY;oBACA,aAAanB;oBACb,qBAAqB,IAAI,CAAC,mBAAmB;oBAC7Ca;gBACF;gBAEFxB,MAAM,cAAcqC,KAAK,SAAS,CAACH,YAAY,MAAM;gBAErD,MAAM,EACJI,OAAO,EACPC,GAAG,EACHC,kCAAkC,EAClCC,KAAK,EACLC,KAAK,EACLC,WAAW,EACXC,KAAK,EACN,GAAGV;gBAEJf,gBAAgB,IAAI,CAAC,GAAG,GAAG;oBACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;oBAClCwB;gBACF;gBACAxB,gBAAgB,IAAI,CAAC,KAAK,GAAGuB;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,EACrBxB,OACE,CAACmB,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;oBACAd;gBACF;YACF;QACF;QAEA,OAAOH;IACT;IAEA,MAAM,SACJZ,KAAa,EACbK,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACf;QAC1B,MAAMG,UAAU,IAAI,CAAC,sBAAsB,CAACV;QAC5C,MAAM,EAAE2C,KAAK,EAAE,GAAG,MAAM,IAAI,CAAC,uBAAuB,CAClDtC,OACAC,wBACAC;QAEF,MAAMU,SAASP,QAAQ,SAAS;QAChC,MAAMkC,SAAS,MAAMlC,QAAQ,YAAY,CAACiC;QAC1C,MAAM,EAAEE,MAAM,EAAE,GAAGD,UAAU,CAAC;QAC9B,OAAO;YACLC;YACA5B;QACF;IACF;IAEQ,wBAAwB6B,WAAoB,EAAE;QACpD,OACE,IAAI,CAAC,oBAAoB,IACzBC,oBAAoB,oBAAoB,CACtCC,oCAEDF,CAAAA,cACGhD,uCACAD,2BAA0B;IAElC;IAEA,MAAM,OACJoD,UAAkB,EAClB3C,sBAAoC,EACpCC,2BAAyC,EACzC2C,aAA4B,EAC5BC,qBAA8B,EAC9BjC,aAAsB,EACtBkC,SAAmB,EAQnB;QACA,IAAI,CAAC,mBAAmB,CAAC,KAAK;QAE9B,MAAM1C,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUsC;QAEzB,MAAMhC,SAASP,QAAQ,SAAS;QAEhC,IAAI2C,cAAc;QAClB,MAAMC,WAAmC,EAAE;QAC3C,MAAMC,uBAAuB,IAAI,CAAC,uBAAuB,CACvDjD,AAAkC,kBAAlCA,uBAAuB,MAAM;QAI/B,MAAO,KAAM;YACX,IAAI+C,cAAcE,sBAAsB;gBACtC,MAAMC,WAAW,CAAC,WAAW,EAAED,qBAAqB,+EAA+E,CAAC;gBACpI,OAAO7C,QAAQ,eAAe,CAAC8C;YACjC;YAGA,MAAMC,eAAe,IAAI,CAAC,kBAAkB,CAC1CR,YACA/B,eACAZ,wBACAC,6BACA4C;YAGF,MAAMP,SAAS,MAAMlC,QAAQ,YAAY,CAAC+C;YAC1C,MAAM5B,aAAae,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM;YAGjC,MAAMvC,QAAQwB,AAAAA,CAAAA,QAAAA,aAAAA,KAAAA,IAAAA,WAAY,OAAO,AAAD,KAAK,EAAE;YACvCyB,SAAS,IAAI,IAAKzB,AAAAA,CAAAA,QAAAA,aAAAA,KAAAA,IAAAA,WAAY,QAAQ,AAAD,KAAK,EAAE;YAE5C,IAAI6B;YACJ,IAAI;gBACFA,cAAc,MAAM,IAAI,CAAC,uBAAuB,CAC9CrD,OACAC,wBACAC,6BACA;oBACE6C;oBACA,SAAS;gBACX;YAEJ,EAAE,OAAOhB,OAAO;gBACd,OAAO1B,QAAQ,eAAe,CAC5B,CAAC,4CAA4C,EAAE0B,MAAM,SAAS,EAAEJ,KAAK,SAAS,CAC5E3B,QACC;YAEP;YACA,MAAMK,QAAQ,YAAY,CAACgD,YAAY,KAAK;YAG5C,IAAI,CAAC7B,CAAAA,QAAAA,aAAAA,KAAAA,IAAAA,WAAY,kCAAkC,AAAD,GAChD;YAIFwB;QACF;QAEA,MAAMM,cAAc;YAClB,QAAQ;gBACNL;YACF;YACArC;QACF;QACA,OAAO0C;IACT;IAEQ,oBACNC,IAAsE,EACtEC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACpC;QACA,MAAMC,YAA4C;YAChD,MAAM;YACN,SAASL;YACT,OAAO;gBACL,YAAYI,mBACP;oBACCH;oBACAG;gBACF,IACAH;YACN;YACA,UAAU,OAAOhD,OAAOqD;gBACtB,MAAM,EAAEtD,IAAI,EAAE,GAAGsD;gBACjB,IAAIC;gBACJ,MAAMC,YAAY,CAACC;oBACjBF,YAAYE;oBACZzD,KAAK,GAAG,GAAG;wBACTyD;oBACF;gBACF;gBAGA,MAAMtD,YAAYmD,YAAY,SAAS;gBACvClD,OAAOD,WAAW;gBAElB,MAAMuD,mBAAmBV,AAAS,YAATA;gBACzB,IAAIW,cAAcV;gBAClB,IAAIW,cAAc;gBAClB,IAAIF,oBAAqBV,CAAAA,AAAS,aAATA,QAAqBA,AAAS,cAATA,IAAiB,GAAI;oBACjEY,cAAc;oBACd,MAAMC,gBACJb,AAAS,aAATA,OACI,CAAC,kDAAkD,EAAEC,QAAQ,GAC7D,CAAC,+GAA+G,EAAEA,QAAQ;oBAChIU,cAAc;wBACZ,CAACC,YAAY,EAAEC;oBACjB;gBACF,OAAO,IAAIH,kBACTC,cAAc;oBACZ,CAACC,YAAY,EAAE,GAAGZ,KAAK,EAAE,EAAEC,QAAQ;gBACrC;gBAGF,IAAIa;gBAEJ,IAAIC,uBAAuB;gBAC3B,IAAIZ,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,KAAK,IAAI,CAAC,SAAS,CAAC,mBAAmB,EAAE;oBAC1DpE,MAAM;oBACN,MAAMiF,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,aACAT,aACAC,KACAY,sBACAX;gBAEJ,EAAE,OAAO5B,OAAO;oBACd,IAAIA,iBAAiB0C,cACnBV,UAAUhC,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAEA,MAAM,EAAE2C,IAAI,EAAE1C,KAAK,EAAE2C,OAAO,EAAEX,IAAI,EAAE,GAAGK;gBACvCN,UAAUC;gBAEV,IAAIY,eAAeF;gBACnB,IAAIT,kBAEF,IAAI,AAAgB,YAAhB,OAAOS,MACTE,eAAeF;qBACV,IAAInB,AAAS,cAATA,MAEPqB,eADEF,QAAAA,OACa,QAECA,IAAY,CAACP,YAAY;qBAEtC,IAAIO,QAAAA,MACTE,eAAe;qBACV;oBACLjE,OACE+D,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,IAAM,CAACP,YAAY,AAAD,MAAMhD,QACxB;oBAEFyD,eAAgBF,IAAY,CAACP,YAAY;gBAC3C;gBAGF,IAAIZ,AAAS,aAATA,QAAqB,CAACqB,cAAc;oBACtCrE,KAAK,KAAK,GAAGyB;oBACbzB,KAAK,OAAO,GAAGoE;oBACf,MAAM,IAAIE,MAAM,CAAC,kBAAkB,EAAEF,SAAS;gBAChD;gBAEA,OAAO;oBACL,QAAQC;oBACR,KAAKd;oBACL9B;oBACA2C;gBACF;YACF;QACF;QAEA,OAAOf;IACT;IACA,MAAM,yBACJL,IAA0D,EAC1DC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACP;QAC7B,MAAMtD,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aACEiD,MACA,AAAkB,YAAlB,OAAOC,SAAsBA,SAAS7B,KAAK,SAAS,CAAC6B;QAIzD,MAAMI,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9CL,MACAC,QACAC,aACAC,KACAC;QAGF,MAAM/C,SAASP,QAAQ,SAAS;QAChC,MAAMkC,SAAS,MAAMlC,QAAQ,YAAY,CAACuD;QAE1C,IAAI,CAACrB,QACH,MAAM,IAAIsC,MACR;QAIJ,MAAM,EAAErC,MAAM,EAAEmC,OAAO,EAAE,GAAGpC;QAE5B,OAAO;YACLC;YACAmC;YACA/D;QACF;IACF;IAEQ,UAAUkE,MAAc,EAA4C;QAC1E,OAAO;YACL,MAAM;YACN,OAAO;gBACLA;YACF;QACF;IACF;IAEA,MAAM,aAAaA,MAAc,EAAEC,YAA0B,EAAE;QAC7D,OAAO,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;YACtCD;QACF;IACF;IAEA,MAAM,QACJE,SAAsB,EACtBtB,GAA+B,EAC/BD,WAAyB,EACO;QAChC,MAAM,EAAEwB,UAAU,EAAEtB,gBAAgB,EAAE,GAAGuB,YAAYF;QAErD,MAAMG,cAAc,CAAC,SAAS,EAAEF,YAAY;QAC5C,MAAM5E,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,WAAW6E;QAE1B,MAAMvE,SAASP,QAAQ,SAAS;QAChC,MAAM,EAAE+E,SAAS,EAAEC,eAAe,EAAE,GAAG3B;QAEvC/C,OAAOqE,WAAW;QAClBrE,OAAOyE,WAAW;QAClBzE,OAAO0E,iBAAiB;QAExB1E,OACE0E,mBAAmBD,WACnB,CAAC,iGAAiG,EAAEC,gBAAgB,aAAa,EAAED,UAAU,CAAC,CAAC;QAGjJ,MAAME,mBAAmBtE,KAAK,GAAG;QACjC,IAAIuE,iBAAiBD;QACrB,IAAIE,eAAe;QAEnB,MAAOD,iBAAiBD,oBAAoBF,UAAW;YACrD,MAAMK,oBAAoBzE,KAAK,GAAG;YAClCuE,iBAAiBE;YACjB,MAAM7B,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9C,WACAqB,YACAxB,aACAtC,QACAwC;YAGF,MAAMpB,SAAU,MAAMlC,QAAQ,YAAY,CAACuD;YAO3C,IAAIrB,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM,EAChB,OAAO;gBACL,QAAQpB;gBACRP;YACF;YAGF4E,eACEjD,AAAAA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,OAAO,AAAD,KACb,CAACA,UAAU,CAAC,0BAA0B,EAAE0C,YAAY,IACrD,CAAC,0CAA0C,EAAEA,YAAY;YAC3D,MAAMS,MAAM1E,KAAK,GAAG;YACpB,IAAI0E,MAAMD,oBAAoBJ,iBAAiB;gBAC7C,MAAMhD,gBAAgBgD,kBAAmBK,CAAAA,MAAMD,iBAAgB;gBAC/D,MAAME,YAAY,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;oBACjD,QAAQtD;gBACV;gBACA,MAAMhC,QAAQ,MAAM,CAACsF;YACvB;QACF;QAEA,OAAOtF,QAAQ,eAAe,CAAC,CAAC,iBAAiB,EAAEmF,cAAc;IACnE;IAxjBA,YACEI,iBAAoC,EACpCC,OAAgB,EAChBC,IAKC,CACD;QA9BF;QAEA;QAEA;QAEA,uBAAiB,eAAjB;QAEA,uBAAQ,uBAAR;QAEA;QAEA,uBAAiB,SAAjB;QAEA;QAiBE,IAAI,CAAC,SAAS,GAAGF;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC,KAAK,SAAS;QAC/B,IAAI,CAAC,mBAAmB,GAAGA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW;QAC5C,IAAI,CAAC,oBAAoB,GAAGA,KAAK,oBAAoB;QACrD,IAAI,CAAC,KAAK,GAAGA,KAAK,KAAK;QACvB,IAAI,CAAC,mBAAmB,GAAG,IAAIC;QAC/B,IAAI,CAAC,WAAW,GAAG,IAAIC,YAAY;YACjCJ;YACAC;YACA,WAAWC,KAAK,SAAS;QAC3B;IACF;AAmiBF"}
@@ -5,18 +5,12 @@ function locateParamStr(locate) {
5
5
  if (!locate) return '';
6
6
  if ('string' == typeof locate) return locate;
7
7
  if ('object' == typeof locate) {
8
- if ('string' == typeof locate.prompt) return locate.prompt;
9
- if ('object' == typeof locate.prompt && locate.prompt.prompt) {
8
+ if ('object' == typeof locate.prompt && null !== locate.prompt && locate.prompt.prompt) {
10
9
  const prompt = locate.prompt.prompt;
11
- const images = locate.prompt.images || [];
12
- if (0 === images.length) return prompt;
13
- const imagesStr = images.map((image)=>{
14
- let url = image.url;
15
- if (url.startsWith('data:image/') || url.startsWith('data:') && url.includes('base64')) url = `${url.substring(0, 15)}...`;
16
- return `[${image.name}](${url})`;
17
- }).join(', ');
18
- return `${prompt}, ${imagesStr}`;
10
+ return prompt;
19
11
  }
12
+ if ('string' == typeof locate.prompt) return locate.prompt;
13
+ if ('string' == typeof locate.description) return locate.description;
20
14
  }
21
15
  return '';
22
16
  }
@@ -32,6 +26,37 @@ function pullParamStr(pullParam) {
32
26
  if (pullParam.duration) parts.push(`duration: ${pullParam.duration}ms`);
33
27
  return parts.join(', ');
34
28
  }
29
+ function extractInsightParam(taskParam) {
30
+ if (!taskParam) return {
31
+ content: ''
32
+ };
33
+ const extractImages = (source)=>{
34
+ var _source_multimodalPrompt;
35
+ return (null == source ? void 0 : null == (_source_multimodalPrompt = source.multimodalPrompt) ? void 0 : _source_multimodalPrompt.images) && Array.isArray(source.multimodalPrompt.images) ? source.multimodalPrompt.images : void 0;
36
+ };
37
+ const toContent = (value)=>'string' == typeof value ? value : JSON.stringify(value);
38
+ if (taskParam.demand) return {
39
+ content: toContent(taskParam.demand),
40
+ images: extractImages(taskParam)
41
+ };
42
+ if (taskParam.assertion) return {
43
+ content: toContent(taskParam.assertion),
44
+ images: extractImages(taskParam)
45
+ };
46
+ if (taskParam.dataDemand) {
47
+ const { dataDemand } = taskParam;
48
+ if ('string' == typeof dataDemand) return {
49
+ content: dataDemand
50
+ };
51
+ if ('object' == typeof dataDemand) return {
52
+ content: toContent(dataDemand.demand || dataDemand),
53
+ images: extractImages(dataDemand)
54
+ };
55
+ }
56
+ return {
57
+ content: ''
58
+ };
59
+ }
35
60
  function taskTitleStr(type, prompt) {
36
61
  if (prompt) return `${type} - ${prompt}`;
37
62
  return type;
@@ -43,23 +68,20 @@ function paramStr(task) {
43
68
  var _task_param;
44
69
  value = null == task ? void 0 : null == (_task_param = task.param) ? void 0 : _task_param.userInstruction;
45
70
  }
46
- if ('Insight' === task.type) {
47
- var _task_param1, _task_param2;
48
- value = (null == task ? void 0 : null == (_task_param1 = task.param) ? void 0 : _task_param1.dataDemand) || (null == task ? void 0 : null == (_task_param2 = task.param) ? void 0 : _task_param2.assertion);
49
- }
71
+ if ('Insight' === task.type) value = extractInsightParam(null == task ? void 0 : task.param).content;
50
72
  if ('Action Space' === task.type) {
51
- var _task_param3, _task_param4, _task_param5, _task_param6, _task_param7;
52
- const locate = null == task ? void 0 : null == (_task_param3 = task.param) ? void 0 : _task_param3.locate;
73
+ var _task_param1, _task_param2, _task_param3, _task_param4, _task_param5;
74
+ const locate = null == task ? void 0 : null == (_task_param1 = task.param) ? void 0 : _task_param1.locate;
53
75
  const locateStr = locate ? locateParamStr(locate) : '';
54
76
  value = task.thought || '';
55
- if ('number' == typeof (null == task ? void 0 : null == (_task_param4 = task.param) ? void 0 : _task_param4.timeMs)) {
56
- var _task_param8;
57
- value = `${null == task ? void 0 : null == (_task_param8 = task.param) ? void 0 : _task_param8.timeMs}ms`;
58
- } 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);
59
- 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);
60
- else if (void 0 !== (null == task ? void 0 : null == (_task_param7 = task.param) ? void 0 : _task_param7.value)) {
61
- var _task_param9;
62
- value = null == task ? void 0 : null == (_task_param9 = task.param) ? void 0 : _task_param9.value;
77
+ if ('number' == typeof (null == task ? void 0 : null == (_task_param2 = task.param) ? void 0 : _task_param2.timeMs)) {
78
+ var _task_param6;
79
+ value = `${null == task ? void 0 : null == (_task_param6 = task.param) ? void 0 : _task_param6.timeMs}ms`;
80
+ } else if ('string' == typeof (null == task ? void 0 : null == (_task_param3 = task.param) ? void 0 : _task_param3.scrollType)) value = scrollParamStr(null == task ? void 0 : task.param);
81
+ else if ('string' == typeof (null == task ? void 0 : null == (_task_param4 = task.param) ? void 0 : _task_param4.direction) && (null == task ? void 0 : task.subType) === 'AndroidPull') value = pullParamStr(null == task ? void 0 : task.param);
82
+ else if (void 0 !== (null == task ? void 0 : null == (_task_param5 = task.param) ? void 0 : _task_param5.value)) {
83
+ var _task_param7;
84
+ value = null == task ? void 0 : null == (_task_param7 = task.param) ? void 0 : _task_param7.value;
63
85
  } else if ((null == task ? void 0 : task.param) && 'object' == typeof (null == task ? void 0 : task.param) && Object.keys((null == task ? void 0 : task.param) || {}).length > 0) value = null == task ? void 0 : task.param;
64
86
  if (locateStr) value = value ? `${locateStr} - ${value}` : locateStr;
65
87
  }
@@ -68,6 +90,6 @@ function paramStr(task) {
68
90
  if ('object' == typeof value && locateParamStr(value)) return locateParamStr(value);
69
91
  return JSON.stringify(value, void 0, 2);
70
92
  }
71
- export { locateParamStr, paramStr, pullParamStr, scrollParamStr, taskTitleStr, typeStr };
93
+ export { extractInsightParam, locateParamStr, paramStr, pullParamStr, scrollParamStr, taskTitleStr, typeStr };
72
94
 
73
95
  //# sourceMappingURL=ui-utils.mjs.map
@@ -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 ExecutionTaskInsightQuery,\n ExecutionTaskPlanning,\n ExecutionTaskPlanningLocate,\n PullParam,\n ScrollParam,\n} from '@/types';\n\nexport function typeStr(task: ExecutionTask) {\n // // For Action tasks with subType, show \"Action Space / subType\"\n // if (task.type === 'Action' && task.subType) {\n // return `Action Space / ${task.subType}`;\n // }\n\n // // For all other cases with subType, show \"type / subType\"\n // if (task.subType) {\n // return `${task.type} / ${task.subType}`;\n // }\n\n // No subType, just show type\n return task.subType || task.type;\n}\n\nexport function locateParamStr(locate?: DetailedLocateParam | string): string {\n if (!locate) {\n return '';\n }\n\n if (typeof locate === 'string') {\n return locate;\n }\n\n if (typeof locate === 'object') {\n if (typeof locate.prompt === 'string') {\n return locate.prompt;\n }\n\n if (typeof locate.prompt === 'object' && locate.prompt.prompt) {\n const prompt = locate.prompt.prompt;\n const images = locate.prompt.images || [];\n\n if (images.length === 0) return prompt;\n\n const imagesStr = images\n .map((image) => {\n let url = image.url;\n if (\n url.startsWith('data:image/') ||\n (url.startsWith('data:') && url.includes('base64'))\n ) {\n url = `${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 || 'singleAction'}, ${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} - ${prompt}`;\n }\n return type;\n}\n\nexport function paramStr(task: ExecutionTask) {\n let value: string | undefined | object;\n if (task.type === 'Planning') {\n if (task.subType === 'Locate') {\n value = locateParamStr((task as ExecutionTaskPlanningLocate)?.param);\n } else {\n value = (task as ExecutionTaskPlanning)?.param?.userInstruction;\n }\n }\n\n if (task.type === 'Insight') {\n value =\n (task as ExecutionTaskInsightQuery)?.param?.dataDemand ||\n (task as ExecutionTaskInsightAssertion)?.param?.assertion;\n }\n\n if (task.type === 'Action Space') {\n const locate = (task as ExecutionTaskAction)?.param?.locate;\n const locateStr = locate ? locateParamStr(locate) : '';\n\n value = task.thought || '';\n if (typeof (task as ExecutionTaskAction)?.param?.timeMs === 'number') {\n value = `${(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 } else if (\n (task as ExecutionTaskAction)?.param &&\n typeof (task as ExecutionTaskAction)?.param === 'object' &&\n Object.keys((task as ExecutionTaskAction)?.param || {}).length > 0\n ) {\n // General parameter handling for actions with custom parameters\n // (e.g., runWdaRequest, runAdbShell)\n value = (task as ExecutionTaskAction)?.param;\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","_task_param7","locateStr","_task_param8","_task_param9","Object","JSON","undefined"],"mappings":"AAYO,SAASA,QAAQC,IAAmB;IAYzC,OAAOA,KAAK,OAAO,IAAIA,KAAK,IAAI;AAClC;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,eAAe,EAAE,EAAEA,YAAY,QAAQ,IAAI,oBAAoB;AACzI;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,EACX,IAAIA,AAAiB,aAAjBA,KAAK,OAAO,EACdgB,QAAQf,eAAgBD,QAAAA,OAAAA,KAAAA,IAAAA,KAAsC,KAAK;SAC9D;YACGiB;QAARD,QAAShB,QAAAA,OAAAA,KAAAA,IAAAA,QAADiB,CAAAA,cAACjB,KAAgC,KAAK,AAAD,IAArCiB,KAAAA,IAAAA,YAAwC,eAAe;IACjE;IAGF,IAAIjB,AAAc,cAAdA,KAAK,IAAI,EAAgB;YAEzBkB,cACAC;QAFFH,QACEE,AAAClB,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,mBAAdA,KAAK,IAAI,EAAqB;YACjBoB,cAIJC,cAGFC,cAIAC,cAKAC;QAhBT,MAAMtB,SAAUF,QAAAA,OAAAA,KAAAA,IAAAA,QAADoB,CAAAA,eAACpB,KAA8B,KAAK,AAAD,IAAnCoB,KAAAA,IAAAA,aAAsC,MAAM;QAC3D,MAAMK,YAAYvB,SAASD,eAAeC,UAAU;QAEpDc,QAAQhB,KAAK,OAAO,IAAI;QACxB,IAAI,AAAwD,YAAxD,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADqB,CAAAA,eAACrB,KAA8B,KAAK,AAAD,IAAnCqB,KAAAA,IAAAA,aAAsC,MAAM,AAAD,GAAgB;gBACzDK;YAAXV,QAAQ,GAAIhB,QAAAA,OAAAA,KAAAA,IAAAA,QAAD0B,CAAAA,eAAC1B,KAA8B,KAAK,AAAD,IAAnC0B,KAAAA,IAAAA,aAAsC,MAAM,CAAC,EAAE,CAAC;QAC7D,OAAO,IACL,AAA4D,YAA5D,OAAQ1B,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADsB,CAAAA,eAACtB,KAA8B,KAAK,AAAD,IAAnCsB,KAAAA,IAAAA,aAAsC,UAAU,AAAD,GAEtDN,QAAQR,eAAgBR,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;aACtD,IACL,AAA2D,YAA3D,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADuB,CAAAA,eAACvB,KAA8B,KAAK,AAAD,IAAnCuB,KAAAA,IAAAA,aAAsC,SAAS,AAAD,KACrD,AAACvB,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,QAADwB,CAAAA,eAACxB,KAA8B,KAAK,AAAD,IAAnCwB,KAAAA,IAAAA,aAAsC,KAAK,AAAD,GACjD;gBACQG;YAARX,QAAShB,QAAAA,OAAAA,KAAAA,IAAAA,QAAD2B,CAAAA,eAAC3B,KAA8B,KAAK,AAAD,IAAnC2B,KAAAA,IAAAA,aAAsC,KAAK;QACrD,OAAO,IACL,AAAC3B,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK,AAAD,KACnC,AAAgD,YAAhD,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK,AAAD,KAC1C4B,OAAO,IAAI,CAAC,AAAC5B,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK,AAAD,KAAK,CAAC,GAAG,MAAM,GAAG,GAIjEgB,QAAShB,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;QAG9C,IAAIyB,WAEAT,QADEA,QACM,GAAGS,UAAU,GAAG,EAAET,OAAO,GAEzBS;IAGd;IAEA,IAAI,AAAiB,WAAVT,OAAuB,OAAO;IAEzC,IAAI,AAAiB,YAAjB,OAAOA,OAAoB,OAAOA;IAEtC,IAAI,AAAiB,YAAjB,OAAOA,SAAsBf,eAAee,QAC9C,OAAOf,eAAee;IAGxB,OAAOa,KAAK,SAAS,CAACb,OAAOc,QAAW;AAC1C"}
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 ExecutionTaskInsightQuery,\n ExecutionTaskPlanning,\n ExecutionTaskPlanningLocate,\n PullParam,\n ScrollParam,\n} from '@/types';\n\nexport function typeStr(task: ExecutionTask) {\n // // For Action tasks with subType, show \"Action Space / subType\"\n // if (task.type === 'Action' && task.subType) {\n // return `Action Space / ${task.subType}`;\n // }\n\n // // For all other cases with subType, show \"type / subType\"\n // if (task.subType) {\n // return `${task.type} / ${task.subType}`;\n // }\n\n // No subType, just show type\n return task.subType || task.type;\n}\n\nexport function locateParamStr(locate?: DetailedLocateParam | string): string {\n if (!locate) {\n return '';\n }\n\n if (typeof locate === 'string') {\n return locate;\n }\n\n if (typeof locate === 'object') {\n // Check for nested prompt.prompt (Planning Locate tasks)\n if (\n typeof locate.prompt === 'object' &&\n locate.prompt !== null &&\n locate.prompt.prompt\n ) {\n const prompt = locate.prompt.prompt;\n return prompt;\n }\n\n // Check for direct prompt string\n if (typeof locate.prompt === 'string') {\n return locate.prompt;\n }\n\n // Check for description field (Action Space tasks like Tap, Hover)\n if (typeof (locate as any).description === 'string') {\n return (locate as any).description;\n }\n }\n\n return '';\n}\n\nexport function scrollParamStr(scrollParam?: ScrollParam) {\n if (!scrollParam) {\n return '';\n }\n return `${scrollParam.direction || 'down'}, ${scrollParam.scrollType || 'singleAction'}, ${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: ${pullParam.duration}ms`);\n }\n return parts.join(', ');\n}\n\nexport function extractInsightParam(taskParam: any): {\n content: string;\n images?: Array<{ name: string; url: string }>;\n} {\n if (!taskParam) {\n return { content: '' };\n }\n\n // Helper to extract images from multimodalPrompt\n const extractImages = (source: any) => {\n return source?.multimodalPrompt?.images &&\n Array.isArray(source.multimodalPrompt.images)\n ? source.multimodalPrompt.images\n : undefined;\n };\n\n // Helper to stringify if needed\n const toContent = (value: any) =>\n typeof value === 'string' ? value : JSON.stringify(value);\n\n // Extract from demand\n if (taskParam.demand) {\n return {\n content: toContent(taskParam.demand),\n images: extractImages(taskParam),\n };\n }\n\n // Extract from assertion\n if (taskParam.assertion) {\n return {\n content: toContent(taskParam.assertion),\n images: extractImages(taskParam),\n };\n }\n\n // Extract from dataDemand\n if (taskParam.dataDemand) {\n const { dataDemand } = taskParam;\n\n if (typeof dataDemand === 'string') {\n return { content: dataDemand };\n }\n\n if (typeof dataDemand === 'object') {\n return {\n content: toContent(dataDemand.demand || dataDemand),\n images: extractImages(dataDemand),\n };\n }\n }\n\n return { content: '' };\n}\n\nexport function taskTitleStr(\n type:\n | 'Tap'\n | 'Hover'\n | 'Input'\n | 'RightClick'\n | 'KeyboardPress'\n | 'Scroll'\n | 'Action'\n | 'Query'\n | 'Assert'\n | 'WaitFor'\n | 'Locate'\n | 'Boolean'\n | 'Number'\n | 'String',\n prompt: string,\n) {\n if (prompt) {\n return `${type} - ${prompt}`;\n }\n return type;\n}\n\nexport function paramStr(task: ExecutionTask) {\n let value: string | undefined | object;\n if (task.type === 'Planning') {\n if (task.subType === 'Locate') {\n value = locateParamStr((task as ExecutionTaskPlanningLocate)?.param);\n } else {\n value = (task as ExecutionTaskPlanning)?.param?.userInstruction;\n }\n }\n\n if (task.type === 'Insight') {\n value = extractInsightParam((task as any)?.param).content;\n }\n\n if (task.type === 'Action Space') {\n const locate = (task as ExecutionTaskAction)?.param?.locate;\n const locateStr = locate ? locateParamStr(locate) : '';\n\n value = task.thought || '';\n if (typeof (task as ExecutionTaskAction)?.param?.timeMs === 'number') {\n value = `${(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 } else if (\n (task as ExecutionTaskAction)?.param &&\n typeof (task as ExecutionTaskAction)?.param === 'object' &&\n Object.keys((task as ExecutionTaskAction)?.param || {}).length > 0\n ) {\n // General parameter handling for actions with custom parameters\n // (e.g., runWdaRequest, runAdbShell)\n value = (task as ExecutionTaskAction)?.param;\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","scrollParamStr","scrollParam","pullParamStr","pullParam","parts","extractInsightParam","taskParam","extractImages","source","_source_multimodalPrompt","Array","undefined","toContent","value","JSON","dataDemand","taskTitleStr","type","paramStr","_task_param","_task_param1","_task_param2","_task_param3","_task_param4","_task_param5","locateStr","_task_param6","_task_param7","Object"],"mappings":"AAYO,SAASA,QAAQC,IAAmB;IAYzC,OAAOA,KAAK,OAAO,IAAIA,KAAK,IAAI;AAClC;AAEO,SAASC,eAAeC,MAAqC;IAClE,IAAI,CAACA,QACH,OAAO;IAGT,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAGT,IAAI,AAAkB,YAAlB,OAAOA,QAAqB;QAE9B,IACE,AAAyB,YAAzB,OAAOA,OAAO,MAAM,IACpBA,AAAkB,SAAlBA,OAAO,MAAM,IACbA,OAAO,MAAM,CAAC,MAAM,EACpB;YACA,MAAMC,SAASD,OAAO,MAAM,CAAC,MAAM;YACnC,OAAOC;QACT;QAGA,IAAI,AAAyB,YAAzB,OAAOD,OAAO,MAAM,EACtB,OAAOA,OAAO,MAAM;QAItB,IAAI,AAAuC,YAAvC,OAAQA,OAAe,WAAW,EACpC,OAAQA,OAAe,WAAW;IAEtC;IAEA,OAAO;AACT;AAEO,SAASE,eAAeC,WAAyB;IACtD,IAAI,CAACA,aACH,OAAO;IAET,OAAO,GAAGA,YAAY,SAAS,IAAI,OAAO,EAAE,EAAEA,YAAY,UAAU,IAAI,eAAe,EAAE,EAAEA,YAAY,QAAQ,IAAI,oBAAoB;AACzI;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,oBAAoBC,SAAc;IAIhD,IAAI,CAACA,WACH,OAAO;QAAE,SAAS;IAAG;IAIvB,MAAMC,gBAAgB,CAACC;YACdC;QAAP,OAAOA,AAAAA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,QAAAA,CAAAA,2BAAAA,OAAQ,gBAAgB,AAAD,IAAvBA,KAAAA,IAAAA,yBAA0B,MAAM,AAAD,KACpCC,MAAM,OAAO,CAACF,OAAO,gBAAgB,CAAC,MAAM,IAC1CA,OAAO,gBAAgB,CAAC,MAAM,GAC9BG;IACN;IAGA,MAAMC,YAAY,CAACC,QACjB,AAAiB,YAAjB,OAAOA,QAAqBA,QAAQC,KAAK,SAAS,CAACD;IAGrD,IAAIP,UAAU,MAAM,EAClB,OAAO;QACL,SAASM,UAAUN,UAAU,MAAM;QACnC,QAAQC,cAAcD;IACxB;IAIF,IAAIA,UAAU,SAAS,EACrB,OAAO;QACL,SAASM,UAAUN,UAAU,SAAS;QACtC,QAAQC,cAAcD;IACxB;IAIF,IAAIA,UAAU,UAAU,EAAE;QACxB,MAAM,EAAES,UAAU,EAAE,GAAGT;QAEvB,IAAI,AAAsB,YAAtB,OAAOS,YACT,OAAO;YAAE,SAASA;QAAW;QAG/B,IAAI,AAAsB,YAAtB,OAAOA,YACT,OAAO;YACL,SAASH,UAAUG,WAAW,MAAM,IAAIA;YACxC,QAAQR,cAAcQ;QACxB;IAEJ;IAEA,OAAO;QAAE,SAAS;IAAG;AACvB;AAEO,SAASC,aACdC,IAcY,EACZlB,MAAc;IAEd,IAAIA,QACF,OAAO,GAAGkB,KAAK,GAAG,EAAElB,QAAQ;IAE9B,OAAOkB;AACT;AAEO,SAASC,SAAStB,IAAmB;IAC1C,IAAIiB;IACJ,IAAIjB,AAAc,eAAdA,KAAK,IAAI,EACX,IAAIA,AAAiB,aAAjBA,KAAK,OAAO,EACdiB,QAAQhB,eAAgBD,QAAAA,OAAAA,KAAAA,IAAAA,KAAsC,KAAK;SAC9D;YACGuB;QAARN,QAASjB,QAAAA,OAAAA,KAAAA,IAAAA,QAADuB,CAAAA,cAACvB,KAAgC,KAAK,AAAD,IAArCuB,KAAAA,IAAAA,YAAwC,eAAe;IACjE;IAGF,IAAIvB,AAAc,cAAdA,KAAK,IAAI,EACXiB,QAAQR,oBAAqBT,QAAAA,OAAAA,KAAAA,IAAAA,KAAc,KAAK,EAAE,OAAO;IAG3D,IAAIA,AAAc,mBAAdA,KAAK,IAAI,EAAqB;YACjBwB,cAIJC,cAGFC,cAIAC,cAKAC;QAhBT,MAAM1B,SAAUF,QAAAA,OAAAA,KAAAA,IAAAA,QAADwB,CAAAA,eAACxB,KAA8B,KAAK,AAAD,IAAnCwB,KAAAA,IAAAA,aAAsC,MAAM;QAC3D,MAAMK,YAAY3B,SAASD,eAAeC,UAAU;QAEpDe,QAAQjB,KAAK,OAAO,IAAI;QACxB,IAAI,AAAwD,YAAxD,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADyB,CAAAA,eAACzB,KAA8B,KAAK,AAAD,IAAnCyB,KAAAA,IAAAA,aAAsC,MAAM,AAAD,GAAgB;gBACzDK;YAAXb,QAAQ,GAAIjB,QAAAA,OAAAA,KAAAA,IAAAA,QAAD8B,CAAAA,eAAC9B,KAA8B,KAAK,AAAD,IAAnC8B,KAAAA,IAAAA,aAAsC,MAAM,CAAC,EAAE,CAAC;QAC7D,OAAO,IACL,AAA4D,YAA5D,OAAQ9B,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAAD0B,CAAAA,eAAC1B,KAA8B,KAAK,AAAD,IAAnC0B,KAAAA,IAAAA,aAAsC,UAAU,AAAD,GAEtDT,QAAQb,eAAgBJ,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;aACtD,IACL,AAA2D,YAA3D,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAAD2B,CAAAA,eAAC3B,KAA8B,KAAK,AAAD,IAAnC2B,KAAAA,IAAAA,aAAsC,SAAS,AAAD,KACrD,AAAC3B,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,OAAO,AAAD,MAAM,eAE3CiB,QAAQX,aAAcN,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;aACpD,IACL,AAAuD,WAA/CA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAAD4B,CAAAA,eAAC5B,KAA8B,KAAK,AAAD,IAAnC4B,KAAAA,IAAAA,aAAsC,KAAK,AAAD,GACjD;gBACQG;YAARd,QAASjB,QAAAA,OAAAA,KAAAA,IAAAA,QAAD+B,CAAAA,eAAC/B,KAA8B,KAAK,AAAD,IAAnC+B,KAAAA,IAAAA,aAAsC,KAAK;QACrD,OAAO,IACL,AAAC/B,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK,AAAD,KACnC,AAAgD,YAAhD,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK,AAAD,KAC1CgC,OAAO,IAAI,CAAC,AAAChC,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK,AAAD,KAAK,CAAC,GAAG,MAAM,GAAG,GAIjEiB,QAASjB,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;QAG9C,IAAI6B,WAEAZ,QADEA,QACM,GAAGY,UAAU,GAAG,EAAEZ,OAAO,GAEzBY;IAGd;IAEA,IAAI,AAAiB,WAAVZ,OAAuB,OAAO;IAEzC,IAAI,AAAiB,YAAjB,OAAOA,OAAoB,OAAOA;IAEtC,IAAI,AAAiB,YAAjB,OAAOA,SAAsBhB,eAAegB,QAC9C,OAAOhB,eAAegB;IAGxB,OAAOC,KAAK,SAAS,CAACD,OAAOF,QAAW;AAC1C"}
@@ -103,7 +103,7 @@ async function matchElementFromCache(context, cacheEntry, cachePrompt, cacheable
103
103
  return;
104
104
  }
105
105
  }
106
- const getMidsceneVersion = ()=>"1.0.1-beta-20251120071640.0";
106
+ const getMidsceneVersion = ()=>"1.0.1-beta-20251123160938.0";
107
107
  const parsePrompt = (prompt)=>{
108
108
  if ('string' == typeof prompt) return {
109
109
  textPrompt: prompt,
@@ -1 +1 @@
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 ifPlanLocateParamIsBbox(\n planLocateParam: PlanningLocateParam,\n): boolean {\n return !!(\n planLocateParam.bbox &&\n Array.isArray(planLocateParam.bbox) &&\n planLocateParam.bbox.length === 4\n );\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(\n centerPosition,\n typeof planLocateParam.prompt === 'string'\n ? planLocateParam.prompt\n : planLocateParam.prompt?.prompt || '',\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 center: [\n Math.round(rect.left + rect.width / 2),\n Math.round(rect.top + rect.height / 2),\n ],\n rect,\n description:\n typeof cachePrompt === 'string'\n ? cachePrompt\n : cachePrompt.prompt || '',\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","ifPlanLocateParamIsBbox","planLocateParam","Array","matchElementFromPlan","_planLocateParam_prompt","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,wBACdC,eAAoC;IAEpC,OAAO,CAAC,CACNA,CAAAA,gBAAgB,IAAI,IACpBC,MAAM,OAAO,CAACD,gBAAgB,IAAI,KAClCA,AAAgC,MAAhCA,gBAAgB,IAAI,CAAC,MAAM,AAAK;AAEpC;AAEO,SAASE,qBACdF,eAAoC;IAEpC,IAAI,CAACA,iBACH;IAGF,IAAIA,gBAAgB,IAAI,EAAE;YAUlBG;QATN,MAAMC,iBAAiB;YACrB,GAAGC,KAAK,KAAK,CAAEL,AAAAA,CAAAA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAC,IAAK;YACpE,GAAGK,KAAK,KAAK,CAAEL,AAAAA,CAAAA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAC,IAAK;QACtE;QAEA,MAAMM,UAAUC,0BACdH,gBACA,AAAkC,YAAlC,OAAOJ,gBAAgB,MAAM,GACzBA,gBAAgB,MAAM,GACtBG,AAAAA,SAAAA,CAAAA,0BAAAA,gBAAgB,MAAM,AAAD,IAArBA,KAAAA,IAAAA,wBAAwB,MAAM,AAAD,KAAK;QAExC,OAAOG;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,QAAQ;gBACND,KAAK,KAAK,CAACU,KAAK,IAAI,GAAGA,KAAK,KAAK,GAAG;gBACpCV,KAAK,KAAK,CAACU,KAAK,GAAG,GAAGA,KAAK,MAAM,GAAG;aACrC;YACDA;YACA,aACE,AAAuB,YAAvB,OAAOJ,cACHA,cACAA,YAAY,MAAM,IAAI;QAC9B;QAEAG,MAAW,yBAAyBH;QACpC,OAAOL;IACT,EAAE,OAAOxB,OAAO;QACdgC,MAAW,qCAAqChC;QAChD;IACF;AACF;AAIO,MAAMkC,qBAAqB,IAEvBC;AAUJ,MAAMC,cAAc,CACzBC;IAKA,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAO;QACL,YAAYA;QACZ,kBAAkBrB;IACpB;IAEF,OAAO;QACL,YAAYqB,OAAO,MAAM;QACzB,kBAAkBA,OAAO,MAAM,GAC3B;YACE,QAAQA,OAAO,MAAM;YACrB,yBAAyB,CAAC,CAACA,OAAO,uBAAuB;QAC3D,IACArB;IACN;AACF"}
1
+ {"version":3,"file":"agent/utils.mjs","sources":["webpack://@midscene/core/./src/agent/utils.ts"],"sourcesContent":["import type { TMultimodalPrompt, TUserPrompt } from '@/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 ifPlanLocateParamIsBbox(\n planLocateParam: PlanningLocateParam,\n): boolean {\n return !!(\n planLocateParam.bbox &&\n Array.isArray(planLocateParam.bbox) &&\n planLocateParam.bbox.length === 4\n );\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(\n centerPosition,\n typeof planLocateParam.prompt === 'string'\n ? planLocateParam.prompt\n : planLocateParam.prompt?.prompt || '',\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 center: [\n Math.round(rect.left + rect.width / 2),\n Math.round(rect.top + rect.height / 2),\n ],\n rect,\n description:\n typeof cachePrompt === 'string'\n ? cachePrompt\n : cachePrompt.prompt || '',\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","ifPlanLocateParamIsBbox","planLocateParam","Array","matchElementFromPlan","_planLocateParam_prompt","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,wBACdC,eAAoC;IAEpC,OAAO,CAAC,CACNA,CAAAA,gBAAgB,IAAI,IACpBC,MAAM,OAAO,CAACD,gBAAgB,IAAI,KAClCA,AAAgC,MAAhCA,gBAAgB,IAAI,CAAC,MAAM,AAAK;AAEpC;AAEO,SAASE,qBACdF,eAAoC;IAEpC,IAAI,CAACA,iBACH;IAGF,IAAIA,gBAAgB,IAAI,EAAE;YAUlBG;QATN,MAAMC,iBAAiB;YACrB,GAAGC,KAAK,KAAK,CAAEL,AAAAA,CAAAA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAC,IAAK;YACpE,GAAGK,KAAK,KAAK,CAAEL,AAAAA,CAAAA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAC,IAAK;QACtE;QAEA,MAAMM,UAAUC,0BACdH,gBACA,AAAkC,YAAlC,OAAOJ,gBAAgB,MAAM,GACzBA,gBAAgB,MAAM,GACtBG,AAAAA,SAAAA,CAAAA,0BAAAA,gBAAgB,MAAM,AAAD,IAArBA,KAAAA,IAAAA,wBAAwB,MAAM,AAAD,KAAK;QAExC,OAAOG;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,QAAQ;gBACND,KAAK,KAAK,CAACU,KAAK,IAAI,GAAGA,KAAK,KAAK,GAAG;gBACpCV,KAAK,KAAK,CAACU,KAAK,GAAG,GAAGA,KAAK,MAAM,GAAG;aACrC;YACDA;YACA,aACE,AAAuB,YAAvB,OAAOJ,cACHA,cACAA,YAAY,MAAM,IAAI;QAC9B;QAEAG,MAAW,yBAAyBH;QACpC,OAAOL;IACT,EAAE,OAAOxB,OAAO;QACdgC,MAAW,qCAAqChC;QAChD;IACF;AACF;AAIO,MAAMkC,qBAAqB,IAEvBC;AAUJ,MAAMC,cAAc,CACzBC;IAKA,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAO;QACL,YAAYA;QACZ,kBAAkBrB;IACpB;IAEF,OAAO;QACL,YAAYqB,OAAO,MAAM;QACzB,kBAAkBA,OAAO,MAAM,GAC3B;YACE,QAAQA,OAAO,MAAM;YACrB,yBAAyB,CAAC,CAACA,OAAO,uBAAuB;QAC3D,IACArB;IACN;AACF"}
@@ -5,7 +5,7 @@ import { generatePlaywrightTest, generatePlaywrightTestStream } from "./prompt/p
5
5
  import { generateYamlTest, generateYamlTestStream } from "./prompt/yaml-generator.mjs";
6
6
  import { AiExtractElementInfo, AiJudgeOrderSensitive, AiLocateElement, AiLocateSection } from "./inspect.mjs";
7
7
  import { plan } from "./llm-planning.mjs";
8
- import { AIActionType, PointSchema, RectSchema, SizeSchema, TMultimodalPromptSchema, TUserPromptSchema, adaptBboxToRect, dumpActionParam, findAllMidsceneLocatorField, getMidsceneLocationSchema, loadActionParam, parseActionParam } from "./common.mjs";
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
11
  export { AIActionType, AiExtractElementInfo, AiJudgeOrderSensitive, 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 };
@@ -2,7 +2,7 @@ import { generateElementByPosition } from "@midscene/shared/extractor/dom-util";
2
2
  import { cropByRect, paddingToMatchBlockByBase64, preProcessImageUrl } from "@midscene/shared/img";
3
3
  import { getDebug } from "@midscene/shared/logger";
4
4
  import { assert } from "@midscene/shared/utils";
5
- import { AIActionType, adaptBboxToRect, expandSearchArea, mergeRects } from "./common.mjs";
5
+ import { AIActionType, adaptBboxToRect, expandSearchArea, mergeRects } from "../common.mjs";
6
6
  import { extractDataQueryPrompt, systemPromptToExtract } from "./prompt/extraction.mjs";
7
7
  import { findElementPrompt, systemPromptToLocateElement } from "./prompt/llm-locator.mjs";
8
8
  import { sectionLocatorInstruction, systemPromptToLocateSection } from "./prompt/llm-section-locator.mjs";
@@ -34,7 +34,7 @@ const promptsToChatParam = async (multimodalPrompt)=>{
34
34
  content: [
35
35
  {
36
36
  type: 'text',
37
- text: `reference image ${item.name}:`
37
+ text: `this is the reference image named '${item.name}':`
38
38
  }
39
39
  ]
40
40
  });
@@ -1 +1 @@
1
- {"version":3,"file":"ai-model/inspect.mjs","sources":["webpack://@midscene/core/./src/ai-model/inspect.ts"],"sourcesContent":["import type {\n AIDataExtractionResponse,\n AIElementResponse,\n AISectionLocatorResponse,\n AIUsageInfo,\n Rect,\n ReferenceImage,\n ServiceExtractOption,\n UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { generateElementByPosition } from '@midscene/shared/extractor/dom-util';\nimport {\n cropByRect,\n paddingToMatchBlockByBase64,\n preProcessImageUrl,\n} from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport type { LocateResultElement } from '@midscene/shared/types';\nimport { assert } from '@midscene/shared/utils';\nimport type {\n ChatCompletionSystemMessageParam,\n ChatCompletionUserMessageParam,\n} from 'openai/resources/index';\nimport type { TMultimodalPrompt, TUserPrompt } from './common';\nimport {\n AIActionType,\n adaptBboxToRect,\n expandSearchArea,\n mergeRects,\n} from './common';\nimport {\n extractDataQueryPrompt,\n systemPromptToExtract,\n} from './prompt/extraction';\nimport {\n findElementPrompt,\n systemPromptToLocateElement,\n} from './prompt/llm-locator';\nimport {\n sectionLocatorInstruction,\n systemPromptToLocateSection,\n} from './prompt/llm-section-locator';\nimport {\n orderSensitiveJudgePrompt,\n systemPromptToJudgeOrderSensitive,\n} from './prompt/order-sensitive-judge';\nimport { callAIWithObjectResponse } from './service-caller/index';\n\nexport type AIArgs = [\n ChatCompletionSystemMessageParam,\n ...ChatCompletionUserMessageParam[],\n];\n\nconst debugInspect = getDebug('ai:inspect');\nconst debugSection = getDebug('ai:section');\n\nconst extraTextFromUserPrompt = (prompt: TUserPrompt): string => {\n if (typeof prompt === 'string') {\n return prompt;\n } else {\n return prompt.prompt;\n }\n};\n\nconst promptsToChatParam = async (\n multimodalPrompt: TMultimodalPrompt,\n): Promise<ChatCompletionUserMessageParam[]> => {\n const msgs: ChatCompletionUserMessageParam[] = [];\n if (multimodalPrompt?.images?.length) {\n msgs.push({\n role: 'user',\n content: [\n {\n type: 'text',\n text: 'Next, I will provide all the reference images.',\n },\n ],\n });\n\n for (const item of multimodalPrompt.images) {\n const base64 = await preProcessImageUrl(\n item.url,\n !!multimodalPrompt.convertHttpImage2Base64,\n );\n\n msgs.push({\n role: 'user',\n content: [\n {\n type: 'text',\n text: `reference image ${item.name}:`,\n },\n ],\n });\n\n msgs.push({\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: base64,\n detail: 'high',\n },\n },\n ],\n });\n }\n }\n return msgs;\n};\n\nexport async function AiLocateElement(options: {\n context: UIContext;\n targetElementDescription: TUserPrompt;\n referenceImage?: ReferenceImage;\n callAIFn: typeof callAIWithObjectResponse<\n AIElementResponse | [number, number]\n >;\n searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;\n modelConfig: IModelConfig;\n}): Promise<{\n parseResult: {\n elements: LocateResultElement[];\n errors?: string[];\n };\n rect?: Rect;\n rawResponse: string;\n usage?: AIUsageInfo;\n}> {\n const { context, targetElementDescription, callAIFn, modelConfig } = options;\n const { vlMode } = modelConfig;\n const { screenshotBase64 } = context;\n\n assert(\n targetElementDescription,\n 'cannot find the target element description',\n );\n const targetElementDescriptionText = extraTextFromUserPrompt(\n targetElementDescription,\n );\n const userInstructionPrompt = await findElementPrompt.format({\n targetElementDescription: targetElementDescriptionText,\n });\n const systemPrompt = systemPromptToLocateElement(vlMode);\n\n let imagePayload = screenshotBase64;\n let imageWidth = context.size.width;\n let imageHeight = context.size.height;\n let originalImageWidth = imageWidth;\n let originalImageHeight = imageHeight;\n\n if (options.searchConfig) {\n assert(\n options.searchConfig.rect,\n 'searchArea is provided but its rect cannot be found. Failed to locate element',\n );\n assert(\n options.searchConfig.imageBase64,\n 'searchArea is provided but its imageBase64 cannot be found. Failed to locate element',\n );\n\n imagePayload = options.searchConfig.imageBase64;\n imageWidth = options.searchConfig.rect?.width;\n imageHeight = options.searchConfig.rect?.height;\n originalImageWidth = imageWidth;\n originalImageHeight = imageHeight;\n } else if (vlMode === 'qwen2.5-vl') {\n const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n imageWidth = paddedResult.width;\n imageHeight = paddedResult.height;\n imagePayload = paddedResult.imageBase64;\n }\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n detail: 'high',\n },\n },\n {\n type: 'text',\n text: userInstructionPrompt,\n },\n ],\n },\n ];\n\n if (typeof targetElementDescription !== 'string') {\n const addOns = await promptsToChatParam({\n images: targetElementDescription.images,\n convertHttpImage2Base64: targetElementDescription.convertHttpImage2Base64,\n });\n msgs.push(...addOns);\n }\n\n const res = await callAIFn(msgs, AIActionType.INSPECT_ELEMENT, modelConfig);\n\n const rawResponse = JSON.stringify(res.content);\n\n let resRect: Rect | undefined;\n let matchedElements = 'elements' in res.content ? res.content.elements : [];\n let errors: string[] | undefined =\n 'errors' in res.content ? res.content.errors : [];\n try {\n if (\n 'bbox' in res.content &&\n Array.isArray(res.content.bbox) &&\n res.content.bbox.length >= 1\n ) {\n resRect = adaptBboxToRect(\n res.content.bbox,\n imageWidth,\n imageHeight,\n options.searchConfig?.rect?.left,\n options.searchConfig?.rect?.top,\n originalImageWidth,\n originalImageHeight,\n vlMode,\n );\n\n debugInspect('resRect', resRect);\n\n const rectCenter = {\n x: resRect.left + resRect.width / 2,\n y: resRect.top + resRect.height / 2,\n };\n\n const element: LocateResultElement = generateElementByPosition(\n rectCenter,\n targetElementDescriptionText as string,\n );\n errors = [];\n\n if (element) {\n matchedElements = [element];\n }\n }\n } catch (e) {\n const msg =\n e instanceof Error\n ? `Failed to parse bbox: ${e.message}`\n : 'unknown error in locate';\n if (!errors || errors?.length === 0) {\n errors = [msg];\n } else {\n errors.push(`(${msg})`);\n }\n }\n\n return {\n rect: resRect,\n parseResult: {\n elements: matchedElements as LocateResultElement[],\n errors: errors as string[],\n },\n rawResponse,\n usage: res.usage,\n };\n}\n\nexport async function AiLocateSection(options: {\n context: UIContext;\n sectionDescription: TUserPrompt;\n modelConfig: IModelConfig;\n}): Promise<{\n rect?: Rect;\n imageBase64?: string;\n error?: string;\n rawResponse: string;\n usage?: AIUsageInfo;\n}> {\n const { context, sectionDescription, modelConfig } = options;\n const { vlMode } = modelConfig;\n const { screenshotBase64 } = context;\n\n const systemPrompt = systemPromptToLocateSection(vlMode);\n const sectionLocatorInstructionText = await sectionLocatorInstruction.format({\n sectionDescription: extraTextFromUserPrompt(sectionDescription),\n });\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: screenshotBase64,\n detail: 'high',\n },\n },\n {\n type: 'text',\n text: sectionLocatorInstructionText,\n },\n ],\n },\n ];\n\n if (typeof sectionDescription !== 'string') {\n const addOns = await promptsToChatParam({\n images: sectionDescription.images,\n convertHttpImage2Base64: sectionDescription.convertHttpImage2Base64,\n });\n msgs.push(...addOns);\n }\n\n const result = await callAIWithObjectResponse<AISectionLocatorResponse>(\n msgs,\n AIActionType.EXTRACT_DATA,\n modelConfig,\n );\n\n let sectionRect: Rect | undefined;\n const sectionBbox = result.content.bbox;\n if (sectionBbox) {\n const targetRect = adaptBboxToRect(\n sectionBbox,\n context.size.width,\n context.size.height,\n 0,\n 0,\n context.size.width,\n context.size.height,\n vlMode,\n );\n debugSection('original targetRect %j', targetRect);\n\n const referenceBboxList = result.content.references_bbox || [];\n debugSection('referenceBboxList %j', referenceBboxList);\n\n const referenceRects = referenceBboxList\n .filter((bbox) => Array.isArray(bbox))\n .map((bbox) => {\n return adaptBboxToRect(\n bbox,\n context.size.width,\n context.size.height,\n 0,\n 0,\n context.size.width,\n context.size.height,\n vlMode,\n );\n });\n debugSection('referenceRects %j', referenceRects);\n\n // merge the sectionRect and referenceRects\n const mergedRect = mergeRects([targetRect, ...referenceRects]);\n debugSection('mergedRect %j', mergedRect);\n\n // expand search area to at least 200 x 200\n sectionRect = expandSearchArea(mergedRect, context.size, vlMode);\n debugSection('expanded sectionRect %j', sectionRect);\n }\n\n let imageBase64 = screenshotBase64;\n if (sectionRect) {\n const croppedResult = await cropByRect(\n screenshotBase64,\n sectionRect,\n vlMode === 'qwen2.5-vl',\n );\n imageBase64 = croppedResult.imageBase64;\n sectionRect.width = croppedResult.width;\n sectionRect.height = croppedResult.height;\n }\n\n return {\n rect: sectionRect,\n imageBase64,\n error: result.content.error,\n rawResponse: JSON.stringify(result.content),\n usage: result.usage,\n };\n}\n\nexport async function AiExtractElementInfo<T>(options: {\n dataQuery: string | Record<string, string>;\n multimodalPrompt?: TMultimodalPrompt;\n context: UIContext;\n pageDescription?: string;\n extractOption?: ServiceExtractOption;\n modelConfig: IModelConfig;\n}) {\n const { dataQuery, context, extractOption, multimodalPrompt, modelConfig } =\n options;\n const systemPrompt = systemPromptToExtract();\n const { screenshotBase64 } = context;\n\n const extractDataPromptText = await extractDataQueryPrompt(\n options.pageDescription || '',\n dataQuery,\n );\n\n const userContent: ChatCompletionUserMessageParam['content'] = [];\n\n if (extractOption?.screenshotIncluded !== false) {\n userContent.push({\n type: 'image_url',\n image_url: {\n url: screenshotBase64,\n detail: 'high',\n },\n });\n }\n\n userContent.push({\n type: 'text',\n text: extractDataPromptText,\n });\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: userContent,\n },\n ];\n\n if (multimodalPrompt) {\n const addOns = await promptsToChatParam({\n images: multimodalPrompt.images,\n convertHttpImage2Base64: multimodalPrompt.convertHttpImage2Base64,\n });\n msgs.push(...addOns);\n }\n\n const result = await callAIWithObjectResponse<AIDataExtractionResponse<T>>(\n msgs,\n AIActionType.EXTRACT_DATA,\n modelConfig,\n );\n return {\n parseResult: result.content,\n usage: result.usage,\n };\n}\n\nexport async function AiJudgeOrderSensitive(\n description: string,\n callAIFn: typeof callAIWithObjectResponse<{ isOrderSensitive: boolean }>,\n modelConfig: IModelConfig,\n): Promise<{\n isOrderSensitive: boolean;\n usage?: AIUsageInfo;\n}> {\n const systemPrompt = systemPromptToJudgeOrderSensitive();\n const userPrompt = orderSensitiveJudgePrompt(description);\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: userPrompt,\n },\n ];\n\n const result = await callAIFn(\n msgs,\n AIActionType.INSPECT_ELEMENT, // Reuse existing action type for now\n modelConfig,\n );\n\n return {\n isOrderSensitive: result.content.isOrderSensitive ?? false,\n usage: result.usage,\n };\n}\n"],"names":["debugInspect","getDebug","debugSection","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","_multimodalPrompt_images","msgs","item","base64","preProcessImageUrl","AiLocateElement","options","context","targetElementDescription","callAIFn","modelConfig","vlMode","screenshotBase64","assert","targetElementDescriptionText","userInstructionPrompt","findElementPrompt","systemPrompt","systemPromptToLocateElement","imagePayload","imageWidth","imageHeight","originalImageWidth","originalImageHeight","_options_searchConfig_rect","_options_searchConfig_rect1","paddedResult","paddingToMatchBlockByBase64","addOns","res","AIActionType","rawResponse","JSON","resRect","matchedElements","errors","Array","_options_searchConfig_rect2","_options_searchConfig_rect3","adaptBboxToRect","rectCenter","element","generateElementByPosition","e","msg","Error","AiLocateSection","sectionDescription","systemPromptToLocateSection","sectionLocatorInstructionText","sectionLocatorInstruction","result","callAIWithObjectResponse","sectionRect","sectionBbox","targetRect","referenceBboxList","referenceRects","bbox","mergedRect","mergeRects","expandSearchArea","imageBase64","croppedResult","cropByRect","AiExtractElementInfo","dataQuery","extractOption","systemPromptToExtract","extractDataPromptText","extractDataQueryPrompt","userContent","AiJudgeOrderSensitive","description","systemPromptToJudgeOrderSensitive","userPrompt","orderSensitiveJudgePrompt"],"mappings":";;;;;;;;;;AAsDA,MAAMA,eAAeC,SAAS;AAC9B,MAAMC,eAAeD,SAAS;AAE9B,MAAME,0BAA0B,CAACC;IAC/B,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAEP,OAAOA,OAAO,MAAM;AAExB;AAEA,MAAMC,qBAAqB,OACzBC;QAGIC;IADJ,MAAMC,OAAyC,EAAE;IACjD,IAAID,QAAAA,mBAAAA,KAAAA,IAAAA,QAAAA,CAAAA,2BAAAA,iBAAkB,MAAM,AAAD,IAAvBA,KAAAA,IAAAA,yBAA0B,MAAM,EAAE;QACpCC,KAAK,IAAI,CAAC;YACR,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM;gBACR;aACD;QACH;QAEA,KAAK,MAAMC,QAAQH,iBAAiB,MAAM,CAAE;YAC1C,MAAMI,SAAS,MAAMC,mBACnBF,KAAK,GAAG,EACR,CAAC,CAACH,iBAAiB,uBAAuB;YAG5CE,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,MAAM,CAAC,gBAAgB,EAAEC,KAAK,IAAI,CAAC,CAAC,CAAC;oBACvC;iBACD;YACH;YAEAD,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKE;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;QACF;IACF;IACA,OAAOF;AACT;AAEO,eAAeI,gBAAgBC,OASrC;IASC,MAAM,EAAEC,OAAO,EAAEC,wBAAwB,EAAEC,QAAQ,EAAEC,WAAW,EAAE,GAAGJ;IACrE,MAAM,EAAEK,MAAM,EAAE,GAAGD;IACnB,MAAM,EAAEE,gBAAgB,EAAE,GAAGL;IAE7BM,OACEL,0BACA;IAEF,MAAMM,+BAA+BlB,wBACnCY;IAEF,MAAMO,wBAAwB,MAAMC,kBAAkB,MAAM,CAAC;QAC3D,0BAA0BF;IAC5B;IACA,MAAMG,eAAeC,4BAA4BP;IAEjD,IAAIQ,eAAeP;IACnB,IAAIQ,aAAab,QAAQ,IAAI,CAAC,KAAK;IACnC,IAAIc,cAAcd,QAAQ,IAAI,CAAC,MAAM;IACrC,IAAIe,qBAAqBF;IACzB,IAAIG,sBAAsBF;IAE1B,IAAIf,QAAQ,YAAY,EAAE;YAWXkB,4BACCC;QAXdZ,OACEP,QAAQ,YAAY,CAAC,IAAI,EACzB;QAEFO,OACEP,QAAQ,YAAY,CAAC,WAAW,EAChC;QAGFa,eAAeb,QAAQ,YAAY,CAAC,WAAW;QAC/Cc,aAAa,QAAAI,CAAAA,6BAAAA,QAAQ,YAAY,CAAC,IAAI,AAAD,IAAxBA,KAAAA,IAAAA,2BAA2B,KAAK;QAC7CH,cAAc,QAAAI,CAAAA,8BAAAA,QAAQ,YAAY,CAAC,IAAI,AAAD,IAAxBA,KAAAA,IAAAA,4BAA2B,MAAM;QAC/CH,qBAAqBF;QACrBG,sBAAsBF;IACxB,OAAO,IAAIV,AAAW,iBAAXA,QAAyB;QAClC,MAAMe,eAAe,MAAMC,4BAA4BR;QACvDC,aAAaM,aAAa,KAAK;QAC/BL,cAAcK,aAAa,MAAM;QACjCP,eAAeO,aAAa,WAAW;IACzC;IAEA,MAAMzB,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKE;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMJ;gBACR;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAOP,0BAAuC;QAChD,MAAMoB,SAAS,MAAM9B,mBAAmB;YACtC,QAAQU,yBAAyB,MAAM;YACvC,yBAAyBA,yBAAyB,uBAAuB;QAC3E;QACAP,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAMC,MAAM,MAAMpB,SAASR,MAAM6B,aAAa,eAAe,EAAEpB;IAE/D,MAAMqB,cAAcC,KAAK,SAAS,CAACH,IAAI,OAAO;IAE9C,IAAII;IACJ,IAAIC,kBAAkB,cAAcL,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,QAAQ,GAAG,EAAE;IAC3E,IAAIM,SACF,YAAYN,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,MAAM,GAAG,EAAE;IACnD,IAAI;QACF,IACE,UAAUA,IAAI,OAAO,IACrBO,MAAM,OAAO,CAACP,IAAI,OAAO,CAAC,IAAI,KAC9BA,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,IAAI,GAC3B;gBAKEQ,6BAAAA,uBACAC,6BAAAA;YALFL,UAAUM,gBACRV,IAAI,OAAO,CAAC,IAAI,EAChBT,YACAC,aAAAA,QACAgB,CAAAA,wBAAAA,QAAQ,YAAY,AAAD,IAAnBA,KAAAA,IAAAA,QAAAA,CAAAA,8BAAAA,sBAAsB,IAAI,AAAD,IAAzBA,KAAAA,IAAAA,4BAA4B,IAAI,UAChCC,CAAAA,yBAAAA,QAAQ,YAAY,AAAD,IAAnBA,KAAAA,IAAAA,QAAAA,CAAAA,8BAAAA,uBAAsB,IAAI,AAAD,IAAzBA,KAAAA,IAAAA,4BAA4B,GAAG,EAC/BhB,oBACAC,qBACAZ;YAGFlB,aAAa,WAAWwC;YAExB,MAAMO,aAAa;gBACjB,GAAGP,QAAQ,IAAI,GAAGA,QAAQ,KAAK,GAAG;gBAClC,GAAGA,QAAQ,GAAG,GAAGA,QAAQ,MAAM,GAAG;YACpC;YAEA,MAAMQ,UAA+BC,0BACnCF,YACA1B;YAEFqB,SAAS,EAAE;YAEX,IAAIM,SACFP,kBAAkB;gBAACO;aAAQ;QAE/B;IACF,EAAE,OAAOE,GAAG;QACV,MAAMC,MACJD,aAAaE,QACT,CAAC,sBAAsB,EAAEF,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACR,UAAUA,AAAAA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM,AAAD,MAAM,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAES,IAAI,CAAC,CAAC;aAFtBT,SAAS;YAACS;SAAI;IAIlB;IAEA,OAAO;QACL,MAAMX;QACN,aAAa;YACX,UAAUC;YACV,QAAQC;QACV;QACAJ;QACA,OAAOF,IAAI,KAAK;IAClB;AACF;AAEO,eAAeiB,gBAAgBxC,OAIrC;IAOC,MAAM,EAAEC,OAAO,EAAEwC,kBAAkB,EAAErC,WAAW,EAAE,GAAGJ;IACrD,MAAM,EAAEK,MAAM,EAAE,GAAGD;IACnB,MAAM,EAAEE,gBAAgB,EAAE,GAAGL;IAE7B,MAAMU,eAAe+B,4BAA4BrC;IACjD,MAAMsC,gCAAgC,MAAMC,0BAA0B,MAAM,CAAC;QAC3E,oBAAoBtD,wBAAwBmD;IAC9C;IACA,MAAM9C,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKL;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMqC;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAMnB,SAAS,MAAM9B,mBAAmB;YACtC,QAAQiD,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACA9C,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAMuB,SAAS,MAAMC,yBACnBnD,MACA6B,aAAa,YAAY,EACzBpB;IAGF,IAAI2C;IACJ,MAAMC,cAAcH,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIG,aAAa;QACf,MAAMC,aAAahB,gBACjBe,aACA/C,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAEFhB,aAAa,0BAA0B4D;QAEvC,MAAMC,oBAAoBL,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9DxD,aAAa,wBAAwB6D;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAAStB,MAAM,OAAO,CAACsB,OAC/B,GAAG,CAAC,CAACA,OACGnB,gBACLmB,MACAnD,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAGNhB,aAAa,qBAAqB8D;QAGlC,MAAME,aAAaC,WAAW;YAACL;eAAeE;SAAe;QAC7D9D,aAAa,iBAAiBgE;QAG9BN,cAAcQ,iBAAiBF,YAAYpD,QAAQ,IAAI,EAAEI;QACzDhB,aAAa,2BAA2B0D;IAC1C;IAEA,IAAIS,cAAclD;IAClB,IAAIyC,aAAa;QACf,MAAMU,gBAAgB,MAAMC,WAC1BpD,kBACAyC,aACA1C,AAAW,iBAAXA;QAEFmD,cAAcC,cAAc,WAAW;QACvCV,YAAY,KAAK,GAAGU,cAAc,KAAK;QACvCV,YAAY,MAAM,GAAGU,cAAc,MAAM;IAC3C;IAEA,OAAO;QACL,MAAMV;QACNS;QACA,OAAOX,OAAO,OAAO,CAAC,KAAK;QAC3B,aAAanB,KAAK,SAAS,CAACmB,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAec,qBAAwB3D,OAO7C;IACC,MAAM,EAAE4D,SAAS,EAAE3D,OAAO,EAAE4D,aAAa,EAAEpE,gBAAgB,EAAEW,WAAW,EAAE,GACxEJ;IACF,MAAMW,eAAemD;IACrB,MAAM,EAAExD,gBAAgB,EAAE,GAAGL;IAE7B,MAAM8D,wBAAwB,MAAMC,uBAClChE,QAAQ,eAAe,IAAI,IAC3B4D;IAGF,MAAMK,cAAyD,EAAE;IAEjE,IAAIJ,AAAAA,CAAAA,QAAAA,gBAAAA,KAAAA,IAAAA,cAAe,kBAAkB,AAAD,MAAM,OACxCI,YAAY,IAAI,CAAC;QACf,MAAM;QACN,WAAW;YACT,KAAK3D;YACL,QAAQ;QACV;IACF;IAGF2D,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAMpE,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAASsD;QACX;KACD;IAED,IAAIxE,kBAAkB;QACpB,MAAM6B,SAAS,MAAM9B,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAE,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAMuB,SAAS,MAAMC,yBACnBnD,MACA6B,aAAa,YAAY,EACzBpB;IAEF,OAAO;QACL,aAAayC,OAAO,OAAO;QAC3B,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAeqB,sBACpBC,WAAmB,EACnBhE,QAAwE,EACxEC,WAAyB;IAKzB,MAAMO,eAAeyD;IACrB,MAAMC,aAAaC,0BAA0BH;IAE7C,MAAMxE,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS0D;QACX;KACD;IAED,MAAMxB,SAAS,MAAM1C,SACnBR,MACA6B,aAAa,eAAe,EAC5BpB;IAGF,OAAO;QACL,kBAAkByC,OAAO,OAAO,CAAC,gBAAgB,IAAI;QACrD,OAAOA,OAAO,KAAK;IACrB;AACF"}
1
+ {"version":3,"file":"ai-model/inspect.mjs","sources":["webpack://@midscene/core/./src/ai-model/inspect.ts"],"sourcesContent":["import type {\n AIDataExtractionResponse,\n AIElementResponse,\n AISectionLocatorResponse,\n AIUsageInfo,\n Rect,\n ReferenceImage,\n ServiceExtractOption,\n UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { generateElementByPosition } from '@midscene/shared/extractor/dom-util';\nimport {\n cropByRect,\n paddingToMatchBlockByBase64,\n preProcessImageUrl,\n} from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport type { LocateResultElement } from '@midscene/shared/types';\nimport { assert } from '@midscene/shared/utils';\nimport type {\n ChatCompletionSystemMessageParam,\n ChatCompletionUserMessageParam,\n} from 'openai/resources/index';\nimport type { TMultimodalPrompt, TUserPrompt } from '../common';\nimport {\n AIActionType,\n adaptBboxToRect,\n expandSearchArea,\n mergeRects,\n} from '../common';\nimport {\n extractDataQueryPrompt,\n systemPromptToExtract,\n} from './prompt/extraction';\nimport {\n findElementPrompt,\n systemPromptToLocateElement,\n} from './prompt/llm-locator';\nimport {\n sectionLocatorInstruction,\n systemPromptToLocateSection,\n} from './prompt/llm-section-locator';\nimport {\n orderSensitiveJudgePrompt,\n systemPromptToJudgeOrderSensitive,\n} from './prompt/order-sensitive-judge';\nimport { callAIWithObjectResponse } from './service-caller/index';\n\nexport type AIArgs = [\n ChatCompletionSystemMessageParam,\n ...ChatCompletionUserMessageParam[],\n];\n\nconst debugInspect = getDebug('ai:inspect');\nconst debugSection = getDebug('ai:section');\n\nconst extraTextFromUserPrompt = (prompt: TUserPrompt): string => {\n if (typeof prompt === 'string') {\n return prompt;\n } else {\n return prompt.prompt;\n }\n};\n\nconst promptsToChatParam = async (\n multimodalPrompt: TMultimodalPrompt,\n): Promise<ChatCompletionUserMessageParam[]> => {\n const msgs: ChatCompletionUserMessageParam[] = [];\n if (multimodalPrompt?.images?.length) {\n msgs.push({\n role: 'user',\n content: [\n {\n type: 'text',\n text: 'Next, I will provide all the reference images.',\n },\n ],\n });\n\n for (const item of multimodalPrompt.images) {\n const base64 = await preProcessImageUrl(\n item.url,\n !!multimodalPrompt.convertHttpImage2Base64,\n );\n\n msgs.push({\n role: 'user',\n content: [\n {\n type: 'text',\n text: `this is the reference image named '${item.name}':`,\n },\n ],\n });\n\n msgs.push({\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: base64,\n detail: 'high',\n },\n },\n ],\n });\n }\n }\n return msgs;\n};\n\nexport async function AiLocateElement(options: {\n context: UIContext;\n targetElementDescription: TUserPrompt;\n referenceImage?: ReferenceImage;\n callAIFn: typeof callAIWithObjectResponse<\n AIElementResponse | [number, number]\n >;\n searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;\n modelConfig: IModelConfig;\n}): Promise<{\n parseResult: {\n elements: LocateResultElement[];\n errors?: string[];\n };\n rect?: Rect;\n rawResponse: string;\n usage?: AIUsageInfo;\n}> {\n const { context, targetElementDescription, callAIFn, modelConfig } = options;\n const { vlMode } = modelConfig;\n const { screenshotBase64 } = context;\n\n assert(\n targetElementDescription,\n 'cannot find the target element description',\n );\n const targetElementDescriptionText = extraTextFromUserPrompt(\n targetElementDescription,\n );\n const userInstructionPrompt = await findElementPrompt.format({\n targetElementDescription: targetElementDescriptionText,\n });\n const systemPrompt = systemPromptToLocateElement(vlMode);\n\n let imagePayload = screenshotBase64;\n let imageWidth = context.size.width;\n let imageHeight = context.size.height;\n let originalImageWidth = imageWidth;\n let originalImageHeight = imageHeight;\n\n if (options.searchConfig) {\n assert(\n options.searchConfig.rect,\n 'searchArea is provided but its rect cannot be found. Failed to locate element',\n );\n assert(\n options.searchConfig.imageBase64,\n 'searchArea is provided but its imageBase64 cannot be found. Failed to locate element',\n );\n\n imagePayload = options.searchConfig.imageBase64;\n imageWidth = options.searchConfig.rect?.width;\n imageHeight = options.searchConfig.rect?.height;\n originalImageWidth = imageWidth;\n originalImageHeight = imageHeight;\n } else if (vlMode === 'qwen2.5-vl') {\n const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n imageWidth = paddedResult.width;\n imageHeight = paddedResult.height;\n imagePayload = paddedResult.imageBase64;\n }\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n detail: 'high',\n },\n },\n {\n type: 'text',\n text: userInstructionPrompt,\n },\n ],\n },\n ];\n\n if (typeof targetElementDescription !== 'string') {\n const addOns = await promptsToChatParam({\n images: targetElementDescription.images,\n convertHttpImage2Base64: targetElementDescription.convertHttpImage2Base64,\n });\n msgs.push(...addOns);\n }\n\n const res = await callAIFn(msgs, AIActionType.INSPECT_ELEMENT, modelConfig);\n\n const rawResponse = JSON.stringify(res.content);\n\n let resRect: Rect | undefined;\n let matchedElements = 'elements' in res.content ? res.content.elements : [];\n let errors: string[] | undefined =\n 'errors' in res.content ? res.content.errors : [];\n try {\n if (\n 'bbox' in res.content &&\n Array.isArray(res.content.bbox) &&\n res.content.bbox.length >= 1\n ) {\n resRect = adaptBboxToRect(\n res.content.bbox,\n imageWidth,\n imageHeight,\n options.searchConfig?.rect?.left,\n options.searchConfig?.rect?.top,\n originalImageWidth,\n originalImageHeight,\n vlMode,\n );\n\n debugInspect('resRect', resRect);\n\n const rectCenter = {\n x: resRect.left + resRect.width / 2,\n y: resRect.top + resRect.height / 2,\n };\n\n const element: LocateResultElement = generateElementByPosition(\n rectCenter,\n targetElementDescriptionText as string,\n );\n errors = [];\n\n if (element) {\n matchedElements = [element];\n }\n }\n } catch (e) {\n const msg =\n e instanceof Error\n ? `Failed to parse bbox: ${e.message}`\n : 'unknown error in locate';\n if (!errors || errors?.length === 0) {\n errors = [msg];\n } else {\n errors.push(`(${msg})`);\n }\n }\n\n return {\n rect: resRect,\n parseResult: {\n elements: matchedElements as LocateResultElement[],\n errors: errors as string[],\n },\n rawResponse,\n usage: res.usage,\n };\n}\n\nexport async function AiLocateSection(options: {\n context: UIContext;\n sectionDescription: TUserPrompt;\n modelConfig: IModelConfig;\n}): Promise<{\n rect?: Rect;\n imageBase64?: string;\n error?: string;\n rawResponse: string;\n usage?: AIUsageInfo;\n}> {\n const { context, sectionDescription, modelConfig } = options;\n const { vlMode } = modelConfig;\n const { screenshotBase64 } = context;\n\n const systemPrompt = systemPromptToLocateSection(vlMode);\n const sectionLocatorInstructionText = await sectionLocatorInstruction.format({\n sectionDescription: extraTextFromUserPrompt(sectionDescription),\n });\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: screenshotBase64,\n detail: 'high',\n },\n },\n {\n type: 'text',\n text: sectionLocatorInstructionText,\n },\n ],\n },\n ];\n\n if (typeof sectionDescription !== 'string') {\n const addOns = await promptsToChatParam({\n images: sectionDescription.images,\n convertHttpImage2Base64: sectionDescription.convertHttpImage2Base64,\n });\n msgs.push(...addOns);\n }\n\n const result = await callAIWithObjectResponse<AISectionLocatorResponse>(\n msgs,\n AIActionType.EXTRACT_DATA,\n modelConfig,\n );\n\n let sectionRect: Rect | undefined;\n const sectionBbox = result.content.bbox;\n if (sectionBbox) {\n const targetRect = adaptBboxToRect(\n sectionBbox,\n context.size.width,\n context.size.height,\n 0,\n 0,\n context.size.width,\n context.size.height,\n vlMode,\n );\n debugSection('original targetRect %j', targetRect);\n\n const referenceBboxList = result.content.references_bbox || [];\n debugSection('referenceBboxList %j', referenceBboxList);\n\n const referenceRects = referenceBboxList\n .filter((bbox) => Array.isArray(bbox))\n .map((bbox) => {\n return adaptBboxToRect(\n bbox,\n context.size.width,\n context.size.height,\n 0,\n 0,\n context.size.width,\n context.size.height,\n vlMode,\n );\n });\n debugSection('referenceRects %j', referenceRects);\n\n // merge the sectionRect and referenceRects\n const mergedRect = mergeRects([targetRect, ...referenceRects]);\n debugSection('mergedRect %j', mergedRect);\n\n // expand search area to at least 200 x 200\n sectionRect = expandSearchArea(mergedRect, context.size, vlMode);\n debugSection('expanded sectionRect %j', sectionRect);\n }\n\n let imageBase64 = screenshotBase64;\n if (sectionRect) {\n const croppedResult = await cropByRect(\n screenshotBase64,\n sectionRect,\n vlMode === 'qwen2.5-vl',\n );\n imageBase64 = croppedResult.imageBase64;\n sectionRect.width = croppedResult.width;\n sectionRect.height = croppedResult.height;\n }\n\n return {\n rect: sectionRect,\n imageBase64,\n error: result.content.error,\n rawResponse: JSON.stringify(result.content),\n usage: result.usage,\n };\n}\n\nexport async function AiExtractElementInfo<T>(options: {\n dataQuery: string | Record<string, string>;\n multimodalPrompt?: TMultimodalPrompt;\n context: UIContext;\n pageDescription?: string;\n extractOption?: ServiceExtractOption;\n modelConfig: IModelConfig;\n}) {\n const { dataQuery, context, extractOption, multimodalPrompt, modelConfig } =\n options;\n const systemPrompt = systemPromptToExtract();\n const { screenshotBase64 } = context;\n\n const extractDataPromptText = await extractDataQueryPrompt(\n options.pageDescription || '',\n dataQuery,\n );\n\n const userContent: ChatCompletionUserMessageParam['content'] = [];\n\n if (extractOption?.screenshotIncluded !== false) {\n userContent.push({\n type: 'image_url',\n image_url: {\n url: screenshotBase64,\n detail: 'high',\n },\n });\n }\n\n userContent.push({\n type: 'text',\n text: extractDataPromptText,\n });\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: userContent,\n },\n ];\n\n if (multimodalPrompt) {\n const addOns = await promptsToChatParam({\n images: multimodalPrompt.images,\n convertHttpImage2Base64: multimodalPrompt.convertHttpImage2Base64,\n });\n msgs.push(...addOns);\n }\n\n const result = await callAIWithObjectResponse<AIDataExtractionResponse<T>>(\n msgs,\n AIActionType.EXTRACT_DATA,\n modelConfig,\n );\n return {\n parseResult: result.content,\n usage: result.usage,\n };\n}\n\nexport async function AiJudgeOrderSensitive(\n description: string,\n callAIFn: typeof callAIWithObjectResponse<{ isOrderSensitive: boolean }>,\n modelConfig: IModelConfig,\n): Promise<{\n isOrderSensitive: boolean;\n usage?: AIUsageInfo;\n}> {\n const systemPrompt = systemPromptToJudgeOrderSensitive();\n const userPrompt = orderSensitiveJudgePrompt(description);\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: userPrompt,\n },\n ];\n\n const result = await callAIFn(\n msgs,\n AIActionType.INSPECT_ELEMENT, // Reuse existing action type for now\n modelConfig,\n );\n\n return {\n isOrderSensitive: result.content.isOrderSensitive ?? false,\n usage: result.usage,\n };\n}\n"],"names":["debugInspect","getDebug","debugSection","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","_multimodalPrompt_images","msgs","item","base64","preProcessImageUrl","AiLocateElement","options","context","targetElementDescription","callAIFn","modelConfig","vlMode","screenshotBase64","assert","targetElementDescriptionText","userInstructionPrompt","findElementPrompt","systemPrompt","systemPromptToLocateElement","imagePayload","imageWidth","imageHeight","originalImageWidth","originalImageHeight","_options_searchConfig_rect","_options_searchConfig_rect1","paddedResult","paddingToMatchBlockByBase64","addOns","res","AIActionType","rawResponse","JSON","resRect","matchedElements","errors","Array","_options_searchConfig_rect2","_options_searchConfig_rect3","adaptBboxToRect","rectCenter","element","generateElementByPosition","e","msg","Error","AiLocateSection","sectionDescription","systemPromptToLocateSection","sectionLocatorInstructionText","sectionLocatorInstruction","result","callAIWithObjectResponse","sectionRect","sectionBbox","targetRect","referenceBboxList","referenceRects","bbox","mergedRect","mergeRects","expandSearchArea","imageBase64","croppedResult","cropByRect","AiExtractElementInfo","dataQuery","extractOption","systemPromptToExtract","extractDataPromptText","extractDataQueryPrompt","userContent","AiJudgeOrderSensitive","description","systemPromptToJudgeOrderSensitive","userPrompt","orderSensitiveJudgePrompt"],"mappings":";;;;;;;;;;AAsDA,MAAMA,eAAeC,SAAS;AAC9B,MAAMC,eAAeD,SAAS;AAE9B,MAAME,0BAA0B,CAACC;IAC/B,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAEP,OAAOA,OAAO,MAAM;AAExB;AAEA,MAAMC,qBAAqB,OACzBC;QAGIC;IADJ,MAAMC,OAAyC,EAAE;IACjD,IAAID,QAAAA,mBAAAA,KAAAA,IAAAA,QAAAA,CAAAA,2BAAAA,iBAAkB,MAAM,AAAD,IAAvBA,KAAAA,IAAAA,yBAA0B,MAAM,EAAE;QACpCC,KAAK,IAAI,CAAC;YACR,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM;gBACR;aACD;QACH;QAEA,KAAK,MAAMC,QAAQH,iBAAiB,MAAM,CAAE;YAC1C,MAAMI,SAAS,MAAMC,mBACnBF,KAAK,GAAG,EACR,CAAC,CAACH,iBAAiB,uBAAuB;YAG5CE,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,MAAM,CAAC,mCAAmC,EAAEC,KAAK,IAAI,CAAC,EAAE,CAAC;oBAC3D;iBACD;YACH;YAEAD,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKE;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;QACF;IACF;IACA,OAAOF;AACT;AAEO,eAAeI,gBAAgBC,OASrC;IASC,MAAM,EAAEC,OAAO,EAAEC,wBAAwB,EAAEC,QAAQ,EAAEC,WAAW,EAAE,GAAGJ;IACrE,MAAM,EAAEK,MAAM,EAAE,GAAGD;IACnB,MAAM,EAAEE,gBAAgB,EAAE,GAAGL;IAE7BM,OACEL,0BACA;IAEF,MAAMM,+BAA+BlB,wBACnCY;IAEF,MAAMO,wBAAwB,MAAMC,kBAAkB,MAAM,CAAC;QAC3D,0BAA0BF;IAC5B;IACA,MAAMG,eAAeC,4BAA4BP;IAEjD,IAAIQ,eAAeP;IACnB,IAAIQ,aAAab,QAAQ,IAAI,CAAC,KAAK;IACnC,IAAIc,cAAcd,QAAQ,IAAI,CAAC,MAAM;IACrC,IAAIe,qBAAqBF;IACzB,IAAIG,sBAAsBF;IAE1B,IAAIf,QAAQ,YAAY,EAAE;YAWXkB,4BACCC;QAXdZ,OACEP,QAAQ,YAAY,CAAC,IAAI,EACzB;QAEFO,OACEP,QAAQ,YAAY,CAAC,WAAW,EAChC;QAGFa,eAAeb,QAAQ,YAAY,CAAC,WAAW;QAC/Cc,aAAa,QAAAI,CAAAA,6BAAAA,QAAQ,YAAY,CAAC,IAAI,AAAD,IAAxBA,KAAAA,IAAAA,2BAA2B,KAAK;QAC7CH,cAAc,QAAAI,CAAAA,8BAAAA,QAAQ,YAAY,CAAC,IAAI,AAAD,IAAxBA,KAAAA,IAAAA,4BAA2B,MAAM;QAC/CH,qBAAqBF;QACrBG,sBAAsBF;IACxB,OAAO,IAAIV,AAAW,iBAAXA,QAAyB;QAClC,MAAMe,eAAe,MAAMC,4BAA4BR;QACvDC,aAAaM,aAAa,KAAK;QAC/BL,cAAcK,aAAa,MAAM;QACjCP,eAAeO,aAAa,WAAW;IACzC;IAEA,MAAMzB,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKE;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMJ;gBACR;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAOP,0BAAuC;QAChD,MAAMoB,SAAS,MAAM9B,mBAAmB;YACtC,QAAQU,yBAAyB,MAAM;YACvC,yBAAyBA,yBAAyB,uBAAuB;QAC3E;QACAP,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAMC,MAAM,MAAMpB,SAASR,MAAM6B,aAAa,eAAe,EAAEpB;IAE/D,MAAMqB,cAAcC,KAAK,SAAS,CAACH,IAAI,OAAO;IAE9C,IAAII;IACJ,IAAIC,kBAAkB,cAAcL,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,QAAQ,GAAG,EAAE;IAC3E,IAAIM,SACF,YAAYN,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,MAAM,GAAG,EAAE;IACnD,IAAI;QACF,IACE,UAAUA,IAAI,OAAO,IACrBO,MAAM,OAAO,CAACP,IAAI,OAAO,CAAC,IAAI,KAC9BA,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,IAAI,GAC3B;gBAKEQ,6BAAAA,uBACAC,6BAAAA;YALFL,UAAUM,gBACRV,IAAI,OAAO,CAAC,IAAI,EAChBT,YACAC,aAAAA,QACAgB,CAAAA,wBAAAA,QAAQ,YAAY,AAAD,IAAnBA,KAAAA,IAAAA,QAAAA,CAAAA,8BAAAA,sBAAsB,IAAI,AAAD,IAAzBA,KAAAA,IAAAA,4BAA4B,IAAI,UAChCC,CAAAA,yBAAAA,QAAQ,YAAY,AAAD,IAAnBA,KAAAA,IAAAA,QAAAA,CAAAA,8BAAAA,uBAAsB,IAAI,AAAD,IAAzBA,KAAAA,IAAAA,4BAA4B,GAAG,EAC/BhB,oBACAC,qBACAZ;YAGFlB,aAAa,WAAWwC;YAExB,MAAMO,aAAa;gBACjB,GAAGP,QAAQ,IAAI,GAAGA,QAAQ,KAAK,GAAG;gBAClC,GAAGA,QAAQ,GAAG,GAAGA,QAAQ,MAAM,GAAG;YACpC;YAEA,MAAMQ,UAA+BC,0BACnCF,YACA1B;YAEFqB,SAAS,EAAE;YAEX,IAAIM,SACFP,kBAAkB;gBAACO;aAAQ;QAE/B;IACF,EAAE,OAAOE,GAAG;QACV,MAAMC,MACJD,aAAaE,QACT,CAAC,sBAAsB,EAAEF,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACR,UAAUA,AAAAA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM,AAAD,MAAM,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAES,IAAI,CAAC,CAAC;aAFtBT,SAAS;YAACS;SAAI;IAIlB;IAEA,OAAO;QACL,MAAMX;QACN,aAAa;YACX,UAAUC;YACV,QAAQC;QACV;QACAJ;QACA,OAAOF,IAAI,KAAK;IAClB;AACF;AAEO,eAAeiB,gBAAgBxC,OAIrC;IAOC,MAAM,EAAEC,OAAO,EAAEwC,kBAAkB,EAAErC,WAAW,EAAE,GAAGJ;IACrD,MAAM,EAAEK,MAAM,EAAE,GAAGD;IACnB,MAAM,EAAEE,gBAAgB,EAAE,GAAGL;IAE7B,MAAMU,eAAe+B,4BAA4BrC;IACjD,MAAMsC,gCAAgC,MAAMC,0BAA0B,MAAM,CAAC;QAC3E,oBAAoBtD,wBAAwBmD;IAC9C;IACA,MAAM9C,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKL;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMqC;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAMnB,SAAS,MAAM9B,mBAAmB;YACtC,QAAQiD,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACA9C,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAMuB,SAAS,MAAMC,yBACnBnD,MACA6B,aAAa,YAAY,EACzBpB;IAGF,IAAI2C;IACJ,MAAMC,cAAcH,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIG,aAAa;QACf,MAAMC,aAAahB,gBACjBe,aACA/C,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAEFhB,aAAa,0BAA0B4D;QAEvC,MAAMC,oBAAoBL,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9DxD,aAAa,wBAAwB6D;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAAStB,MAAM,OAAO,CAACsB,OAC/B,GAAG,CAAC,CAACA,OACGnB,gBACLmB,MACAnD,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAGNhB,aAAa,qBAAqB8D;QAGlC,MAAME,aAAaC,WAAW;YAACL;eAAeE;SAAe;QAC7D9D,aAAa,iBAAiBgE;QAG9BN,cAAcQ,iBAAiBF,YAAYpD,QAAQ,IAAI,EAAEI;QACzDhB,aAAa,2BAA2B0D;IAC1C;IAEA,IAAIS,cAAclD;IAClB,IAAIyC,aAAa;QACf,MAAMU,gBAAgB,MAAMC,WAC1BpD,kBACAyC,aACA1C,AAAW,iBAAXA;QAEFmD,cAAcC,cAAc,WAAW;QACvCV,YAAY,KAAK,GAAGU,cAAc,KAAK;QACvCV,YAAY,MAAM,GAAGU,cAAc,MAAM;IAC3C;IAEA,OAAO;QACL,MAAMV;QACNS;QACA,OAAOX,OAAO,OAAO,CAAC,KAAK;QAC3B,aAAanB,KAAK,SAAS,CAACmB,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAec,qBAAwB3D,OAO7C;IACC,MAAM,EAAE4D,SAAS,EAAE3D,OAAO,EAAE4D,aAAa,EAAEpE,gBAAgB,EAAEW,WAAW,EAAE,GACxEJ;IACF,MAAMW,eAAemD;IACrB,MAAM,EAAExD,gBAAgB,EAAE,GAAGL;IAE7B,MAAM8D,wBAAwB,MAAMC,uBAClChE,QAAQ,eAAe,IAAI,IAC3B4D;IAGF,MAAMK,cAAyD,EAAE;IAEjE,IAAIJ,AAAAA,CAAAA,QAAAA,gBAAAA,KAAAA,IAAAA,cAAe,kBAAkB,AAAD,MAAM,OACxCI,YAAY,IAAI,CAAC;QACf,MAAM;QACN,WAAW;YACT,KAAK3D;YACL,QAAQ;QACV;IACF;IAGF2D,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAMpE,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAASsD;QACX;KACD;IAED,IAAIxE,kBAAkB;QACpB,MAAM6B,SAAS,MAAM9B,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAE,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAMuB,SAAS,MAAMC,yBACnBnD,MACA6B,aAAa,YAAY,EACzBpB;IAEF,OAAO;QACL,aAAayC,OAAO,OAAO;QAC3B,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAeqB,sBACpBC,WAAmB,EACnBhE,QAAwE,EACxEC,WAAyB;IAKzB,MAAMO,eAAeyD;IACrB,MAAMC,aAAaC,0BAA0BH;IAE7C,MAAMxE,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS0D;QACX;KACD;IAED,MAAMxB,SAAS,MAAM1C,SACnBR,MACA6B,aAAa,eAAe,EAC5BpB;IAGF,OAAO;QACL,kBAAkByC,OAAO,OAAO,CAAC,gBAAgB,IAAI;QACrD,OAAOA,OAAO,KAAK;IACrB;AACF"}