@midscene/core 1.0.1-beta-20251028021317.0 → 1.0.1-beta-20251028121806.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/es/agent/agent.mjs +15 -15
- package/dist/es/agent/agent.mjs.map +1 -1
- package/dist/es/agent/task-builder.mjs +9 -9
- package/dist/es/agent/task-builder.mjs.map +1 -1
- package/dist/es/agent/tasks.mjs +8 -8
- package/dist/es/agent/tasks.mjs.map +1 -1
- package/dist/es/agent/ui-utils.mjs +0 -1
- package/dist/es/agent/ui-utils.mjs.map +1 -1
- package/dist/es/agent/utils.mjs +1 -1
- package/dist/es/ai-model/inspect.mjs.map +1 -1
- package/dist/es/index.mjs +4 -4
- package/dist/es/index.mjs.map +1 -1
- package/dist/es/{insight → service}/index.mjs +13 -13
- package/dist/es/{insight → service}/index.mjs.map +1 -1
- package/dist/es/{insight → service}/utils.mjs +2 -2
- package/dist/es/service/utils.mjs.map +1 -0
- package/dist/es/task-runner.mjs +1 -1
- package/dist/es/task-runner.mjs.map +1 -1
- package/dist/es/types.mjs +3 -3
- package/dist/es/types.mjs.map +1 -1
- package/dist/es/utils.mjs +2 -2
- package/dist/lib/agent/agent.js +14 -14
- package/dist/lib/agent/agent.js.map +1 -1
- package/dist/lib/agent/task-builder.js +8 -8
- package/dist/lib/agent/task-builder.js.map +1 -1
- package/dist/lib/agent/tasks.js +7 -7
- package/dist/lib/agent/tasks.js.map +1 -1
- package/dist/lib/agent/ui-utils.js +0 -1
- package/dist/lib/agent/ui-utils.js.map +1 -1
- package/dist/lib/agent/utils.js +1 -1
- package/dist/lib/ai-model/inspect.js.map +1 -1
- package/dist/lib/index.js +10 -10
- package/dist/lib/index.js.map +1 -1
- package/dist/lib/{insight → service}/index.js +11 -11
- package/dist/lib/{insight → service}/index.js.map +1 -1
- package/dist/lib/{insight → service}/utils.js +4 -4
- package/dist/lib/{insight → service}/utils.js.map +1 -1
- package/dist/lib/task-runner.js +1 -1
- package/dist/lib/task-runner.js.map +1 -1
- package/dist/lib/types.js +7 -7
- package/dist/lib/types.js.map +1 -1
- package/dist/lib/utils.js +2 -2
- package/dist/types/agent/agent.d.ts +9 -9
- package/dist/types/agent/task-builder.d.ts +4 -4
- package/dist/types/agent/tasks.d.ts +5 -5
- package/dist/types/ai-model/inspect.d.ts +2 -2
- package/dist/types/index.d.ts +4 -4
- package/dist/types/{insight → service}/index.d.ts +8 -8
- package/dist/types/service/utils.d.ts +2 -0
- package/dist/types/types.d.ts +18 -18
- package/dist/types/yaml.d.ts +6 -6
- package/package.json +3 -3
- package/dist/es/insight/utils.mjs.map +0 -1
- package/dist/types/insight/utils.d.ts +0 -2
package/dist/es/agent/tasks.mjs
CHANGED
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
import { ConversationHistory, plan, uiTarsPlanning } from "../ai-model/index.mjs";
|
|
2
|
-
import {
|
|
2
|
+
import { ServiceError } from "../types.mjs";
|
|
3
3
|
import { MIDSCENE_REPLANNING_CYCLE_LIMIT, globalConfigManager } from "@midscene/shared/env";
|
|
4
4
|
import { getDebug } from "@midscene/shared/logger";
|
|
5
5
|
import { assert } from "@midscene/shared/utils";
|
|
@@ -26,7 +26,7 @@ class TaskExecutor {
|
|
|
26
26
|
return this.interface;
|
|
27
27
|
}
|
|
28
28
|
createExecutionSession(title, options) {
|
|
29
|
-
return new ExecutionSession(title, ()=>Promise.resolve(this.
|
|
29
|
+
return new ExecutionSession(title, ()=>Promise.resolve(this.service.contextRetrieverFn()), {
|
|
30
30
|
onTaskStart: this.onTaskStartCallback,
|
|
31
31
|
tasks: null == options ? void 0 : options.tasks
|
|
32
32
|
});
|
|
@@ -227,9 +227,9 @@ class TaskExecutor {
|
|
|
227
227
|
extraPageDescription = await descriptionOfTree(tree, 200, false, (null == opt ? void 0 : opt.domIncluded) === 'visible-only');
|
|
228
228
|
}
|
|
229
229
|
try {
|
|
230
|
-
extractResult = await this.
|
|
230
|
+
extractResult = await this.service.extract(demandInput, modelConfig, opt, extraPageDescription, multimodalPrompt);
|
|
231
231
|
} catch (error) {
|
|
232
|
-
if (error instanceof
|
|
232
|
+
if (error instanceof ServiceError) applyDump(error.dump);
|
|
233
233
|
throw error;
|
|
234
234
|
}
|
|
235
235
|
const { data, usage, thought, dump } = extractResult;
|
|
@@ -313,23 +313,23 @@ class TaskExecutor {
|
|
|
313
313
|
}
|
|
314
314
|
return session.appendErrorPlan(`waitFor timeout: ${errorThought}`);
|
|
315
315
|
}
|
|
316
|
-
constructor(interfaceInstance,
|
|
316
|
+
constructor(interfaceInstance, service, opts){
|
|
317
317
|
_define_property(this, "interface", void 0);
|
|
318
|
-
_define_property(this, "
|
|
318
|
+
_define_property(this, "service", void 0);
|
|
319
319
|
_define_property(this, "taskCache", void 0);
|
|
320
320
|
_define_property(this, "taskBuilder", void 0);
|
|
321
321
|
_define_property(this, "conversationHistory", void 0);
|
|
322
322
|
_define_property(this, "onTaskStartCallback", void 0);
|
|
323
323
|
_define_property(this, "replanningCycleLimit", void 0);
|
|
324
324
|
this.interface = interfaceInstance;
|
|
325
|
-
this.
|
|
325
|
+
this.service = service;
|
|
326
326
|
this.taskCache = opts.taskCache;
|
|
327
327
|
this.onTaskStartCallback = null == opts ? void 0 : opts.onTaskStart;
|
|
328
328
|
this.replanningCycleLimit = opts.replanningCycleLimit;
|
|
329
329
|
this.conversationHistory = new ConversationHistory();
|
|
330
330
|
this.taskBuilder = new TaskBuilder({
|
|
331
331
|
interfaceInstance,
|
|
332
|
-
|
|
332
|
+
service,
|
|
333
333
|
taskCache: opts.taskCache
|
|
334
334
|
});
|
|
335
335
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"agent/tasks.mjs","sources":["webpack://@midscene/core/./src/agent/tasks.ts"],"sourcesContent":["import { ConversationHistory, plan, uiTarsPlanning } from '@/ai-model';\nimport type { TMultimodalPrompt, TUserPrompt } from '@/ai-model/common';\nimport type { AbstractInterface } from '@/device';\nimport type Insight from '@/insight';\nimport type { TaskRunner } from '@/task-runner';\nimport type {\n ExecutionTaskApply,\n ExecutionTaskInsightQueryApply,\n ExecutionTaskPlanningApply,\n ExecutionTaskProgressOptions,\n InsightDump,\n InsightExtractOption,\n InsightExtractParam,\n InterfaceType,\n MidsceneYamlFlowItem,\n PlanningAIResponse,\n PlanningAction,\n PlanningActionParamSleep,\n PlanningActionParamWaitFor,\n} from '@/types';\nimport { InsightError } from '@/types';\nimport {\n type IModelConfig,\n MIDSCENE_REPLANNING_CYCLE_LIMIT,\n globalConfigManager,\n} from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { ExecutionSession } from './execution-session';\nimport { TaskBuilder } from './task-builder';\nimport type { TaskCache } from './task-cache';\nexport { locatePlanForLocate } from './task-builder';\nimport { descriptionOfTree } from '@midscene/shared/extractor';\nimport { taskTitleStr } from './ui-utils';\nimport { parsePrompt } from './utils';\n\ninterface ExecutionResult<OutputType = any> {\n output: OutputType;\n thought?: string;\n runner: TaskRunner;\n}\n\nconst debug = getDebug('device-task-executor');\nconst defaultReplanningCycleLimit = 10;\nconst defaultVlmUiTarsReplanningCycleLimit = 40;\n\nexport class TaskExecutor {\n interface: AbstractInterface;\n\n insight: Insight;\n\n taskCache?: TaskCache;\n\n private readonly taskBuilder: TaskBuilder;\n\n private conversationHistory: ConversationHistory;\n\n onTaskStartCallback?: ExecutionTaskProgressOptions['onTaskStart'];\n\n replanningCycleLimit?: number;\n\n // @deprecated use .interface instead\n get page() {\n return this.interface;\n }\n\n constructor(\n interfaceInstance: AbstractInterface,\n insight: Insight,\n opts: {\n taskCache?: TaskCache;\n onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n replanningCycleLimit?: number;\n },\n ) {\n this.interface = interfaceInstance;\n this.insight = insight;\n this.taskCache = opts.taskCache;\n this.onTaskStartCallback = opts?.onTaskStart;\n this.replanningCycleLimit = opts.replanningCycleLimit;\n this.conversationHistory = new ConversationHistory();\n this.taskBuilder = new TaskBuilder({\n interfaceInstance,\n insight,\n taskCache: opts.taskCache,\n });\n }\n\n private createExecutionSession(\n title: string,\n options?: { tasks?: ExecutionTaskApply[] },\n ) {\n return new ExecutionSession(\n title,\n () => Promise.resolve(this.insight.contextRetrieverFn()),\n {\n onTaskStart: this.onTaskStartCallback,\n tasks: options?.tasks,\n },\n );\n }\n\n public async convertPlanToExecutable(\n plans: PlanningAction[],\n modelConfig: IModelConfig,\n options?: {\n cacheable?: boolean;\n subTask?: boolean;\n },\n ) {\n return this.taskBuilder.build(plans, modelConfig, options);\n }\n\n async loadYamlFlowAsPlanning(userInstruction: string, yamlString: string) {\n const session = this.createExecutionSession(\n taskTitleStr('Action', userInstruction),\n );\n\n const task: ExecutionTaskPlanningApply = {\n type: 'Planning',\n subType: 'LoadYaml',\n locate: null,\n param: {\n userInstruction,\n },\n executor: async (param, executorContext) => {\n const { uiContext } = executorContext;\n assert(uiContext, 'uiContext is required for Planning task');\n return {\n output: {\n actions: [],\n more_actions_needed_by_instruction: false,\n log: '',\n yamlString,\n },\n cache: {\n hit: true,\n },\n hitBy: {\n from: 'Cache',\n context: {\n yamlString,\n },\n },\n };\n },\n };\n await session.appendAndRun(task);\n\n return {\n runner: session.getRunner(),\n };\n }\n\n private createPlanningTask(\n userInstruction: string,\n actionContext: string | undefined,\n modelConfig: IModelConfig,\n ): ExecutionTaskPlanningApply {\n const task: ExecutionTaskPlanningApply = {\n type: 'Planning',\n subType: 'Plan',\n locate: null,\n param: {\n userInstruction,\n },\n executor: async (param, executorContext) => {\n const startTime = Date.now();\n const { uiContext } = executorContext;\n assert(uiContext, 'uiContext is required for Planning task');\n const { vlMode } = modelConfig;\n const uiTarsModelVersion =\n vlMode === 'vlm-ui-tars' ? modelConfig.uiTarsModelVersion : undefined;\n\n assert(\n this.interface.actionSpace,\n 'actionSpace for device is not implemented',\n );\n const actionSpace = await this.interface.actionSpace();\n debug(\n 'actionSpace for this interface is:',\n actionSpace.map((action) => action.name).join(', '),\n );\n assert(Array.isArray(actionSpace), 'actionSpace must be an array');\n if (actionSpace.length === 0) {\n console.warn(\n `ActionSpace for ${this.interface.interfaceType} is empty. This may lead to unexpected behavior.`,\n );\n }\n\n const planResult = await (uiTarsModelVersion ? uiTarsPlanning : plan)(\n param.userInstruction,\n {\n context: uiContext,\n actionContext,\n interfaceType: this.interface.interfaceType as InterfaceType,\n actionSpace,\n modelConfig,\n conversationHistory: this.conversationHistory,\n },\n );\n debug('planResult', JSON.stringify(planResult, null, 2));\n\n const {\n actions,\n log,\n more_actions_needed_by_instruction,\n error,\n usage,\n rawResponse,\n sleep,\n } = planResult;\n\n executorContext.task.log = {\n ...(executorContext.task.log || {}),\n rawResponse,\n };\n executorContext.task.usage = usage;\n\n const finalActions = actions || [];\n\n if (sleep) {\n const timeNow = Date.now();\n const timeRemaining = sleep - (timeNow - startTime);\n if (timeRemaining > 0) {\n finalActions.push(this.sleepPlan(timeRemaining));\n }\n }\n\n if (finalActions.length === 0) {\n assert(\n !more_actions_needed_by_instruction || sleep,\n error ? `Failed to plan: ${error}` : 'No plan found',\n );\n }\n\n return {\n output: {\n actions: finalActions,\n more_actions_needed_by_instruction,\n log,\n yamlFlow: planResult.yamlFlow,\n },\n cache: {\n hit: false,\n },\n uiContext,\n };\n },\n };\n\n return task;\n }\n\n async runPlans(\n title: string,\n plans: PlanningAction[],\n modelConfig: IModelConfig,\n ): Promise<ExecutionResult> {\n const session = this.createExecutionSession(title);\n const { tasks } = await this.convertPlanToExecutable(plans, modelConfig);\n const result = await session.appendAndRun(tasks);\n const { output } = result!;\n return {\n output,\n runner: session.getRunner(),\n };\n }\n\n private getReplanningCycleLimit(isVlmUiTars: boolean) {\n return (\n this.replanningCycleLimit ||\n globalConfigManager.getEnvConfigInNumber(\n MIDSCENE_REPLANNING_CYCLE_LIMIT,\n ) ||\n (isVlmUiTars\n ? defaultVlmUiTarsReplanningCycleLimit\n : defaultReplanningCycleLimit)\n );\n }\n\n async action(\n userPrompt: string,\n modelConfig: IModelConfig,\n actionContext?: string,\n cacheable?: boolean,\n ): Promise<\n ExecutionResult<\n | {\n yamlFlow?: MidsceneYamlFlowItem[]; // for cache use\n }\n | undefined\n >\n > {\n this.conversationHistory.reset();\n\n const session = this.createExecutionSession(\n taskTitleStr('Action', userPrompt),\n );\n const runner = session.getRunner();\n\n let replanCount = 0;\n const yamlFlow: MidsceneYamlFlowItem[] = [];\n const replanningCycleLimit = this.getReplanningCycleLimit(\n modelConfig.vlMode === 'vlm-ui-tars',\n );\n\n // Main planning loop - unified plan/replan logic\n while (true) {\n if (replanCount > replanningCycleLimit) {\n const errorMsg = `Replanning ${replanningCycleLimit} times, which is more than the limit, please split the task into multiple steps`;\n\n return session.appendErrorPlan(errorMsg);\n }\n\n // Create planning task (automatically includes execution history if available)\n const planningTask = this.createPlanningTask(\n userPrompt,\n actionContext,\n modelConfig,\n );\n\n const result = await session.appendAndRun(planningTask);\n const planResult: PlanningAIResponse = result?.output;\n if (session.isInErrorState()) {\n return {\n output: planResult,\n runner,\n };\n }\n\n // Execute planned actions\n const plans = planResult.actions || [];\n yamlFlow.push(...(planResult.yamlFlow || []));\n\n let executables: Awaited<ReturnType<typeof this.convertPlanToExecutable>>;\n try {\n executables = await this.convertPlanToExecutable(plans, modelConfig, {\n cacheable,\n subTask: true,\n });\n await session.appendAndRun(executables.tasks);\n } catch (error) {\n return session.appendErrorPlan(\n `Error converting plans to executable tasks: ${error}, plans: ${JSON.stringify(\n plans,\n )}`,\n );\n }\n if (session.isInErrorState()) {\n return {\n output: undefined,\n runner,\n };\n }\n\n // Check if task is complete\n if (!planResult.more_actions_needed_by_instruction) {\n break;\n }\n\n // Increment replan count for next iteration\n replanCount++;\n }\n\n return {\n output: {\n yamlFlow,\n },\n runner,\n };\n }\n\n private createTypeQueryTask(\n type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert' | 'WaitFor',\n demand: InsightExtractParam,\n modelConfig: IModelConfig,\n opt?: InsightExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ) {\n const queryTask: ExecutionTaskInsightQueryApply = {\n type: 'Insight',\n subType: type,\n locate: null,\n param: {\n dataDemand: multimodalPrompt\n ? ({\n demand,\n multimodalPrompt,\n } as never)\n : demand, // for user param presentation in report right sidebar\n },\n executor: async (param, taskContext) => {\n const { task } = taskContext;\n let queryDump: InsightDump | undefined;\n const applyDump = (dump: InsightDump) => {\n queryDump = dump;\n task.log = {\n dump,\n };\n };\n\n // Get context for query operations\n const uiContext = taskContext.uiContext;\n assert(uiContext, 'uiContext is required for Query task');\n\n const ifTypeRestricted = type !== 'Query';\n let demandInput = demand;\n let keyOfResult = 'result';\n if (ifTypeRestricted && (type === 'Assert' || type === 'WaitFor')) {\n keyOfResult = 'StatementIsTruthy';\n const booleanPrompt =\n type === 'Assert'\n ? `Boolean, whether the following statement is true: ${demand}`\n : `Boolean, the user wants to do some 'wait for' operation, please check whether the following statement is true: ${demand}`;\n demandInput = {\n [keyOfResult]: booleanPrompt,\n };\n } else if (ifTypeRestricted) {\n demandInput = {\n [keyOfResult]: `${type}, ${demand}`,\n };\n }\n\n let extractResult;\n\n let extraPageDescription = '';\n if (opt?.domIncluded && this.interface.getElementsNodeTree) {\n debug('appending tree info for page');\n const tree = await this.interface.getElementsNodeTree();\n extraPageDescription = await descriptionOfTree(\n tree,\n 200,\n false,\n opt?.domIncluded === 'visible-only',\n );\n }\n\n try {\n extractResult = await this.insight.extract<any>(\n demandInput,\n modelConfig,\n opt,\n extraPageDescription,\n multimodalPrompt,\n );\n } catch (error) {\n if (error instanceof InsightError) {\n applyDump(error.dump);\n }\n throw error;\n }\n\n const { data, usage, thought, dump } = extractResult;\n applyDump(dump);\n\n let outputResult = data;\n if (ifTypeRestricted) {\n // If AI returned a plain string instead of structured format, use it directly\n if (typeof data === 'string') {\n outputResult = data;\n } else if (type === 'WaitFor') {\n if (data === null || data === undefined) {\n outputResult = false;\n } else {\n outputResult = (data as any)[keyOfResult];\n }\n } else if (data === null || data === undefined) {\n outputResult = null;\n } else {\n assert(\n data?.[keyOfResult] !== undefined,\n 'No result in query data',\n );\n outputResult = (data as any)[keyOfResult];\n }\n }\n\n return {\n output: outputResult,\n log: queryDump,\n usage,\n thought,\n };\n },\n };\n\n return queryTask;\n }\n async createTypeQueryExecution<T>(\n type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert',\n demand: InsightExtractParam,\n modelConfig: IModelConfig,\n opt?: InsightExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ): Promise<ExecutionResult<T>> {\n const session = this.createExecutionSession(\n taskTitleStr(\n type,\n typeof demand === 'string' ? demand : JSON.stringify(demand),\n ),\n );\n\n const queryTask = await this.createTypeQueryTask(\n type,\n demand,\n modelConfig,\n opt,\n multimodalPrompt,\n );\n\n const result = await session.appendAndRun(queryTask);\n\n if (!result) {\n throw new Error(\n 'result of taskExecutor.flush() is undefined in function createTypeQueryTask',\n );\n }\n\n const { output, thought } = result;\n\n return {\n output,\n thought,\n runner: session.getRunner(),\n };\n }\n\n private sleepPlan(timeMs: number): PlanningAction<PlanningActionParamSleep> {\n return {\n type: 'Sleep',\n param: {\n timeMs,\n },\n locate: null,\n };\n }\n\n async taskForSleep(timeMs: number, _modelConfig: IModelConfig) {\n return this.taskBuilder.createSleepTask({\n timeMs,\n });\n }\n\n async waitFor(\n assertion: TUserPrompt,\n opt: PlanningActionParamWaitFor,\n modelConfig: IModelConfig,\n ): Promise<ExecutionResult<void>> {\n const { textPrompt, multimodalPrompt } = parsePrompt(assertion);\n\n const description = `waitFor: ${textPrompt}`;\n const session = this.createExecutionSession(\n taskTitleStr('WaitFor', description),\n );\n const runner = session.getRunner();\n const { timeoutMs, checkIntervalMs } = opt;\n\n assert(assertion, 'No assertion for waitFor');\n assert(timeoutMs, 'No timeoutMs for waitFor');\n assert(checkIntervalMs, 'No checkIntervalMs for waitFor');\n\n assert(\n checkIntervalMs <= timeoutMs,\n `wrong config for waitFor: checkIntervalMs must be less than timeoutMs, config: {checkIntervalMs: ${checkIntervalMs}, timeoutMs: ${timeoutMs}}`,\n );\n\n const overallStartTime = Date.now();\n let startTime = Date.now();\n let errorThought = '';\n while (Date.now() - overallStartTime < timeoutMs) {\n startTime = Date.now();\n const queryTask = await this.createTypeQueryTask(\n 'WaitFor',\n textPrompt,\n modelConfig,\n {\n doNotThrowError: true,\n },\n multimodalPrompt,\n );\n\n const result = (await session.appendAndRun(queryTask)) as\n | {\n output: boolean;\n thought?: string;\n }\n | undefined;\n\n if (result?.output) {\n return {\n output: undefined,\n runner,\n };\n }\n\n errorThought =\n result?.thought ||\n (!result && `No result from assertion: ${textPrompt}`) ||\n `unknown error when waiting for assertion: ${textPrompt}`;\n const now = Date.now();\n if (now - startTime < checkIntervalMs) {\n const timeRemaining = checkIntervalMs - (now - startTime);\n const sleepTask = this.taskBuilder.createSleepTask({\n timeMs: timeRemaining,\n });\n await session.append(sleepTask);\n }\n }\n\n return session.appendErrorPlan(`waitFor timeout: ${errorThought}`);\n }\n}\n"],"names":["debug","getDebug","defaultReplanningCycleLimit","defaultVlmUiTarsReplanningCycleLimit","TaskExecutor","title","options","ExecutionSession","Promise","plans","modelConfig","userInstruction","yamlString","session","taskTitleStr","task","param","executorContext","uiContext","assert","actionContext","startTime","Date","vlMode","uiTarsModelVersion","undefined","actionSpace","action","Array","console","planResult","uiTarsPlanning","plan","JSON","actions","log","more_actions_needed_by_instruction","error","usage","rawResponse","sleep","finalActions","timeNow","timeRemaining","tasks","result","output","isVlmUiTars","globalConfigManager","MIDSCENE_REPLANNING_CYCLE_LIMIT","userPrompt","cacheable","runner","replanCount","yamlFlow","replanningCycleLimit","errorMsg","planningTask","executables","type","demand","opt","multimodalPrompt","queryTask","taskContext","queryDump","applyDump","dump","ifTypeRestricted","demandInput","keyOfResult","booleanPrompt","extractResult","extraPageDescription","tree","descriptionOfTree","InsightError","data","thought","outputResult","Error","timeMs","_modelConfig","assertion","textPrompt","parsePrompt","description","timeoutMs","checkIntervalMs","overallStartTime","errorThought","now","sleepTask","interfaceInstance","insight","opts","ConversationHistory","TaskBuilder"],"mappings":";;;;;;;;;;;;;;;;;;;;AA0CA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,8BAA8B;AACpC,MAAMC,uCAAuC;AAEtC,MAAMC;IAgBX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IAwBQ,uBACNC,KAAa,EACbC,OAA0C,EAC1C;QACA,OAAO,IAAIC,iBACTF,OACA,IAAMG,QAAQ,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,kBAAkB,KACrD;YACE,aAAa,IAAI,CAAC,mBAAmB;YACrC,OAAOF,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,KAAK;QACvB;IAEJ;IAEA,MAAa,wBACXG,KAAuB,EACvBC,WAAyB,EACzBJ,OAGC,EACD;QACA,OAAO,IAAI,CAAC,WAAW,CAAC,KAAK,CAACG,OAAOC,aAAaJ;IACpD;IAEA,MAAM,uBAAuBK,eAAuB,EAAEC,UAAkB,EAAE;QACxE,MAAMC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUH;QAGzB,MAAMI,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,QAAQ;YACR,OAAO;gBACLJ;YACF;YACA,UAAU,OAAOK,OAAOC;gBACtB,MAAM,EAAEC,SAAS,EAAE,GAAGD;gBACtBE,OAAOD,WAAW;gBAClB,OAAO;oBACL,QAAQ;wBACN,SAAS,EAAE;wBACX,oCAAoC;wBACpC,KAAK;wBACLN;oBACF;oBACA,OAAO;wBACL,KAAK;oBACP;oBACA,OAAO;wBACL,MAAM;wBACN,SAAS;4BACPA;wBACF;oBACF;gBACF;YACF;QACF;QACA,MAAMC,QAAQ,YAAY,CAACE;QAE3B,OAAO;YACL,QAAQF,QAAQ,SAAS;QAC3B;IACF;IAEQ,mBACNF,eAAuB,EACvBS,aAAiC,EACjCV,WAAyB,EACG;QAC5B,MAAMK,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,QAAQ;YACR,OAAO;gBACLJ;YACF;YACA,UAAU,OAAOK,OAAOC;gBACtB,MAAMI,YAAYC,KAAK,GAAG;gBAC1B,MAAM,EAAEJ,SAAS,EAAE,GAAGD;gBACtBE,OAAOD,WAAW;gBAClB,MAAM,EAAEK,MAAM,EAAE,GAAGb;gBACnB,MAAMc,qBACJD,AAAW,kBAAXA,SAA2Bb,YAAY,kBAAkB,GAAGe;gBAE9DN,OACE,IAAI,CAAC,SAAS,CAAC,WAAW,EAC1B;gBAEF,MAAMO,cAAc,MAAM,IAAI,CAAC,SAAS,CAAC,WAAW;gBACpD1B,MACE,sCACA0B,YAAY,GAAG,CAAC,CAACC,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;gBAEhDR,OAAOS,MAAM,OAAO,CAACF,cAAc;gBACnC,IAAIA,AAAuB,MAAvBA,YAAY,MAAM,EACpBG,QAAQ,IAAI,CACV,CAAC,gBAAgB,EAAE,IAAI,CAAC,SAAS,CAAC,aAAa,CAAC,gDAAgD,CAAC;gBAIrG,MAAMC,aAAa,MAAON,AAAAA,CAAAA,qBAAqBO,iBAAiBC,IAAG,EACjEhB,MAAM,eAAe,EACrB;oBACE,SAASE;oBACTE;oBACA,eAAe,IAAI,CAAC,SAAS,CAAC,aAAa;oBAC3CM;oBACAhB;oBACA,qBAAqB,IAAI,CAAC,mBAAmB;gBAC/C;gBAEFV,MAAM,cAAciC,KAAK,SAAS,CAACH,YAAY,MAAM;gBAErD,MAAM,EACJI,OAAO,EACPC,GAAG,EACHC,kCAAkC,EAClCC,KAAK,EACLC,KAAK,EACLC,WAAW,EACXC,KAAK,EACN,GAAGV;gBAEJb,gBAAgB,IAAI,CAAC,GAAG,GAAG;oBACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;oBAClCsB;gBACF;gBACAtB,gBAAgB,IAAI,CAAC,KAAK,GAAGqB;gBAE7B,MAAMG,eAAeP,WAAW,EAAE;gBAElC,IAAIM,OAAO;oBACT,MAAME,UAAUpB,KAAK,GAAG;oBACxB,MAAMqB,gBAAgBH,QAASE,CAAAA,UAAUrB,SAAQ;oBACjD,IAAIsB,gBAAgB,GAClBF,aAAa,IAAI,CAAC,IAAI,CAAC,SAAS,CAACE;gBAErC;gBAEA,IAAIF,AAAwB,MAAxBA,aAAa,MAAM,EACrBtB,OACE,CAACiB,sCAAsCI,OACvCH,QAAQ,CAAC,gBAAgB,EAAEA,OAAO,GAAG;gBAIzC,OAAO;oBACL,QAAQ;wBACN,SAASI;wBACTL;wBACAD;wBACA,UAAUL,WAAW,QAAQ;oBAC/B;oBACA,OAAO;wBACL,KAAK;oBACP;oBACAZ;gBACF;YACF;QACF;QAEA,OAAOH;IACT;IAEA,MAAM,SACJV,KAAa,EACbI,KAAuB,EACvBC,WAAyB,EACC;QAC1B,MAAMG,UAAU,IAAI,CAAC,sBAAsB,CAACR;QAC5C,MAAM,EAAEuC,KAAK,EAAE,GAAG,MAAM,IAAI,CAAC,uBAAuB,CAACnC,OAAOC;QAC5D,MAAMmC,SAAS,MAAMhC,QAAQ,YAAY,CAAC+B;QAC1C,MAAM,EAAEE,MAAM,EAAE,GAAGD;QACnB,OAAO;YACLC;YACA,QAAQjC,QAAQ,SAAS;QAC3B;IACF;IAEQ,wBAAwBkC,WAAoB,EAAE;QACpD,OACE,IAAI,CAAC,oBAAoB,IACzBC,oBAAoB,oBAAoB,CACtCC,oCAEDF,CAAAA,cACG5C,uCACAD,2BAA0B;IAElC;IAEA,MAAM,OACJgD,UAAkB,EAClBxC,WAAyB,EACzBU,aAAsB,EACtB+B,SAAmB,EAQnB;QACA,IAAI,CAAC,mBAAmB,CAAC,KAAK;QAE9B,MAAMtC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUoC;QAEzB,MAAME,SAASvC,QAAQ,SAAS;QAEhC,IAAIwC,cAAc;QAClB,MAAMC,WAAmC,EAAE;QAC3C,MAAMC,uBAAuB,IAAI,CAAC,uBAAuB,CACvD7C,AAAuB,kBAAvBA,YAAY,MAAM;QAIpB,MAAO,KAAM;YACX,IAAI2C,cAAcE,sBAAsB;gBACtC,MAAMC,WAAW,CAAC,WAAW,EAAED,qBAAqB,+EAA+E,CAAC;gBAEpI,OAAO1C,QAAQ,eAAe,CAAC2C;YACjC;YAGA,MAAMC,eAAe,IAAI,CAAC,kBAAkB,CAC1CP,YACA9B,eACAV;YAGF,MAAMmC,SAAS,MAAMhC,QAAQ,YAAY,CAAC4C;YAC1C,MAAM3B,aAAiCe,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM;YACrD,IAAIhC,QAAQ,cAAc,IACxB,OAAO;gBACL,QAAQiB;gBACRsB;YACF;YAIF,MAAM3C,QAAQqB,WAAW,OAAO,IAAI,EAAE;YACtCwB,SAAS,IAAI,IAAKxB,WAAW,QAAQ,IAAI,EAAE;YAE3C,IAAI4B;YACJ,IAAI;gBACFA,cAAc,MAAM,IAAI,CAAC,uBAAuB,CAACjD,OAAOC,aAAa;oBACnEyC;oBACA,SAAS;gBACX;gBACA,MAAMtC,QAAQ,YAAY,CAAC6C,YAAY,KAAK;YAC9C,EAAE,OAAOrB,OAAO;gBACd,OAAOxB,QAAQ,eAAe,CAC5B,CAAC,4CAA4C,EAAEwB,MAAM,SAAS,EAAEJ,KAAK,SAAS,CAC5ExB,QACC;YAEP;YACA,IAAII,QAAQ,cAAc,IACxB,OAAO;gBACL,QAAQY;gBACR2B;YACF;YAIF,IAAI,CAACtB,WAAW,kCAAkC,EAChD;YAIFuB;QACF;QAEA,OAAO;YACL,QAAQ;gBACNC;YACF;YACAF;QACF;IACF;IAEQ,oBACNO,IAAsE,EACtEC,MAA2B,EAC3BlD,WAAyB,EACzBmD,GAA0B,EAC1BC,gBAAoC,EACpC;QACA,MAAMC,YAA4C;YAChD,MAAM;YACN,SAASJ;YACT,QAAQ;YACR,OAAO;gBACL,YAAYG,mBACP;oBACCF;oBACAE;gBACF,IACAF;YACN;YACA,UAAU,OAAO5C,OAAOgD;gBACtB,MAAM,EAAEjD,IAAI,EAAE,GAAGiD;gBACjB,IAAIC;gBACJ,MAAMC,YAAY,CAACC;oBACjBF,YAAYE;oBACZpD,KAAK,GAAG,GAAG;wBACToD;oBACF;gBACF;gBAGA,MAAMjD,YAAY8C,YAAY,SAAS;gBACvC7C,OAAOD,WAAW;gBAElB,MAAMkD,mBAAmBT,AAAS,YAATA;gBACzB,IAAIU,cAAcT;gBAClB,IAAIU,cAAc;gBAClB,IAAIF,oBAAqBT,CAAAA,AAAS,aAATA,QAAqBA,AAAS,cAATA,IAAiB,GAAI;oBACjEW,cAAc;oBACd,MAAMC,gBACJZ,AAAS,aAATA,OACI,CAAC,kDAAkD,EAAEC,QAAQ,GAC7D,CAAC,+GAA+G,EAAEA,QAAQ;oBAChIS,cAAc;wBACZ,CAACC,YAAY,EAAEC;oBACjB;gBACF,OAAO,IAAIH,kBACTC,cAAc;oBACZ,CAACC,YAAY,EAAE,GAAGX,KAAK,EAAE,EAAEC,QAAQ;gBACrC;gBAGF,IAAIY;gBAEJ,IAAIC,uBAAuB;gBAC3B,IAAIZ,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,KAAK,IAAI,CAAC,SAAS,CAAC,mBAAmB,EAAE;oBAC1D7D,MAAM;oBACN,MAAM0E,OAAO,MAAM,IAAI,CAAC,SAAS,CAAC,mBAAmB;oBACrDD,uBAAuB,MAAME,kBAC3BD,MACA,KACA,OACAb,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,MAAM;gBAEzB;gBAEA,IAAI;oBACFW,gBAAgB,MAAM,IAAI,CAAC,OAAO,CAAC,OAAO,CACxCH,aACA3D,aACAmD,KACAY,sBACAX;gBAEJ,EAAE,OAAOzB,OAAO;oBACd,IAAIA,iBAAiBuC,cACnBV,UAAU7B,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAEA,MAAM,EAAEwC,IAAI,EAAEvC,KAAK,EAAEwC,OAAO,EAAEX,IAAI,EAAE,GAAGK;gBACvCN,UAAUC;gBAEV,IAAIY,eAAeF;gBACnB,IAAIT,kBAEF,IAAI,AAAgB,YAAhB,OAAOS,MACTE,eAAeF;qBACV,IAAIlB,AAAS,cAATA,MAEPoB,eADEF,QAAAA,OACa,QAECA,IAAY,CAACP,YAAY;qBAEtC,IAAIO,QAAAA,MACTE,eAAe;qBACV;oBACL5D,OACE0D,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,IAAM,CAACP,YAAY,AAAD,MAAM7C,QACxB;oBAEFsD,eAAgBF,IAAY,CAACP,YAAY;gBAC3C;gBAGF,OAAO;oBACL,QAAQS;oBACR,KAAKd;oBACL3B;oBACAwC;gBACF;YACF;QACF;QAEA,OAAOf;IACT;IACA,MAAM,yBACJJ,IAA0D,EAC1DC,MAA2B,EAC3BlD,WAAyB,EACzBmD,GAA0B,EAC1BC,gBAAoC,EACP;QAC7B,MAAMjD,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aACE6C,MACA,AAAkB,YAAlB,OAAOC,SAAsBA,SAAS3B,KAAK,SAAS,CAAC2B;QAIzD,MAAMG,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9CJ,MACAC,QACAlD,aACAmD,KACAC;QAGF,MAAMjB,SAAS,MAAMhC,QAAQ,YAAY,CAACkD;QAE1C,IAAI,CAAClB,QACH,MAAM,IAAImC,MACR;QAIJ,MAAM,EAAElC,MAAM,EAAEgC,OAAO,EAAE,GAAGjC;QAE5B,OAAO;YACLC;YACAgC;YACA,QAAQjE,QAAQ,SAAS;QAC3B;IACF;IAEQ,UAAUoE,MAAc,EAA4C;QAC1E,OAAO;YACL,MAAM;YACN,OAAO;gBACLA;YACF;YACA,QAAQ;QACV;IACF;IAEA,MAAM,aAAaA,MAAc,EAAEC,YAA0B,EAAE;QAC7D,OAAO,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;YACtCD;QACF;IACF;IAEA,MAAM,QACJE,SAAsB,EACtBtB,GAA+B,EAC/BnD,WAAyB,EACO;QAChC,MAAM,EAAE0E,UAAU,EAAEtB,gBAAgB,EAAE,GAAGuB,YAAYF;QAErD,MAAMG,cAAc,CAAC,SAAS,EAAEF,YAAY;QAC5C,MAAMvE,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,WAAWwE;QAE1B,MAAMlC,SAASvC,QAAQ,SAAS;QAChC,MAAM,EAAE0E,SAAS,EAAEC,eAAe,EAAE,GAAG3B;QAEvC1C,OAAOgE,WAAW;QAClBhE,OAAOoE,WAAW;QAClBpE,OAAOqE,iBAAiB;QAExBrE,OACEqE,mBAAmBD,WACnB,CAAC,iGAAiG,EAAEC,gBAAgB,aAAa,EAAED,UAAU,CAAC,CAAC;QAGjJ,MAAME,mBAAmBnE,KAAK,GAAG;QACjC,IAAID,YAAYC,KAAK,GAAG;QACxB,IAAIoE,eAAe;QACnB,MAAOpE,KAAK,GAAG,KAAKmE,mBAAmBF,UAAW;YAChDlE,YAAYC,KAAK,GAAG;YACpB,MAAMyC,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9C,WACAqB,YACA1E,aACA;gBACE,iBAAiB;YACnB,GACAoD;YAGF,MAAMjB,SAAU,MAAMhC,QAAQ,YAAY,CAACkD;YAO3C,IAAIlB,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM,EAChB,OAAO;gBACL,QAAQpB;gBACR2B;YACF;YAGFsC,eACE7C,AAAAA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,OAAO,AAAD,KACb,CAACA,UAAU,CAAC,0BAA0B,EAAEuC,YAAY,IACrD,CAAC,0CAA0C,EAAEA,YAAY;YAC3D,MAAMO,MAAMrE,KAAK,GAAG;YACpB,IAAIqE,MAAMtE,YAAYmE,iBAAiB;gBACrC,MAAM7C,gBAAgB6C,kBAAmBG,CAAAA,MAAMtE,SAAQ;gBACvD,MAAMuE,YAAY,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;oBACjD,QAAQjD;gBACV;gBACA,MAAM9B,QAAQ,MAAM,CAAC+E;YACvB;QACF;QAEA,OAAO/E,QAAQ,eAAe,CAAC,CAAC,iBAAiB,EAAE6E,cAAc;IACnE;IAjiBA,YACEG,iBAAoC,EACpCC,OAAgB,EAChBC,IAIC,CACD;QA3BF;QAEA;QAEA;QAEA,uBAAiB,eAAjB;QAEA,uBAAQ,uBAAR;QAEA;QAEA;QAgBE,IAAI,CAAC,SAAS,GAAGF;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC,KAAK,SAAS;QAC/B,IAAI,CAAC,mBAAmB,GAAGA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW;QAC5C,IAAI,CAAC,oBAAoB,GAAGA,KAAK,oBAAoB;QACrD,IAAI,CAAC,mBAAmB,GAAG,IAAIC;QAC/B,IAAI,CAAC,WAAW,GAAG,IAAIC,YAAY;YACjCJ;YACAC;YACA,WAAWC,KAAK,SAAS;QAC3B;IACF;AA8gBF"}
|
|
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 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\nconst debug = getDebug('device-task-executor');\nconst defaultReplanningCycleLimit = 10;\nconst defaultVlmUiTarsReplanningCycleLimit = 40;\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 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 },\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.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 },\n );\n }\n\n public async convertPlanToExecutable(\n plans: PlanningAction[],\n modelConfig: IModelConfig,\n options?: {\n cacheable?: boolean;\n subTask?: boolean;\n },\n ) {\n return this.taskBuilder.build(plans, modelConfig, options);\n }\n\n async loadYamlFlowAsPlanning(userInstruction: string, yamlString: string) {\n const session = this.createExecutionSession(\n taskTitleStr('Action', userInstruction),\n );\n\n const task: ExecutionTaskPlanningApply = {\n type: 'Planning',\n subType: 'LoadYaml',\n locate: null,\n param: {\n userInstruction,\n },\n executor: async (param, executorContext) => {\n const { uiContext } = executorContext;\n assert(uiContext, 'uiContext is required for Planning task');\n return {\n output: {\n actions: [],\n more_actions_needed_by_instruction: false,\n log: '',\n yamlString,\n },\n cache: {\n hit: true,\n },\n hitBy: {\n from: 'Cache',\n context: {\n yamlString,\n },\n },\n };\n },\n };\n await session.appendAndRun(task);\n\n return {\n runner: session.getRunner(),\n };\n }\n\n private createPlanningTask(\n userInstruction: string,\n actionContext: string | undefined,\n modelConfig: IModelConfig,\n ): ExecutionTaskPlanningApply {\n const task: ExecutionTaskPlanningApply = {\n type: 'Planning',\n subType: 'Plan',\n locate: null,\n param: {\n userInstruction,\n },\n executor: async (param, executorContext) => {\n const startTime = Date.now();\n const { uiContext } = executorContext;\n assert(uiContext, 'uiContext is required for Planning task');\n const { vlMode } = modelConfig;\n const uiTarsModelVersion =\n vlMode === 'vlm-ui-tars' ? modelConfig.uiTarsModelVersion : undefined;\n\n assert(\n this.interface.actionSpace,\n 'actionSpace for device is not implemented',\n );\n const actionSpace = await this.interface.actionSpace();\n debug(\n 'actionSpace for this interface is:',\n actionSpace.map((action) => action.name).join(', '),\n );\n assert(Array.isArray(actionSpace), 'actionSpace must be an array');\n if (actionSpace.length === 0) {\n console.warn(\n `ActionSpace for ${this.interface.interfaceType} is empty. This may lead to unexpected behavior.`,\n );\n }\n\n const planResult = await (uiTarsModelVersion ? uiTarsPlanning : plan)(\n param.userInstruction,\n {\n context: uiContext,\n actionContext,\n interfaceType: this.interface.interfaceType as InterfaceType,\n actionSpace,\n modelConfig,\n conversationHistory: this.conversationHistory,\n },\n );\n debug('planResult', JSON.stringify(planResult, null, 2));\n\n const {\n actions,\n log,\n more_actions_needed_by_instruction,\n error,\n usage,\n rawResponse,\n sleep,\n } = planResult;\n\n executorContext.task.log = {\n ...(executorContext.task.log || {}),\n rawResponse,\n };\n executorContext.task.usage = usage;\n\n const finalActions = actions || [];\n\n if (sleep) {\n const timeNow = Date.now();\n const timeRemaining = sleep - (timeNow - startTime);\n if (timeRemaining > 0) {\n finalActions.push(this.sleepPlan(timeRemaining));\n }\n }\n\n if (finalActions.length === 0) {\n assert(\n !more_actions_needed_by_instruction || sleep,\n error ? `Failed to plan: ${error}` : 'No plan found',\n );\n }\n\n return {\n output: {\n actions: finalActions,\n more_actions_needed_by_instruction,\n log,\n yamlFlow: planResult.yamlFlow,\n },\n cache: {\n hit: false,\n },\n uiContext,\n };\n },\n };\n\n return task;\n }\n\n async runPlans(\n title: string,\n plans: PlanningAction[],\n modelConfig: IModelConfig,\n ): Promise<ExecutionResult> {\n const session = this.createExecutionSession(title);\n const { tasks } = await this.convertPlanToExecutable(plans, modelConfig);\n const result = await session.appendAndRun(tasks);\n const { output } = result!;\n return {\n output,\n runner: session.getRunner(),\n };\n }\n\n private getReplanningCycleLimit(isVlmUiTars: boolean) {\n return (\n this.replanningCycleLimit ||\n globalConfigManager.getEnvConfigInNumber(\n MIDSCENE_REPLANNING_CYCLE_LIMIT,\n ) ||\n (isVlmUiTars\n ? defaultVlmUiTarsReplanningCycleLimit\n : defaultReplanningCycleLimit)\n );\n }\n\n async action(\n userPrompt: string,\n modelConfig: IModelConfig,\n actionContext?: string,\n cacheable?: boolean,\n ): Promise<\n ExecutionResult<\n | {\n yamlFlow?: MidsceneYamlFlowItem[]; // for cache use\n }\n | undefined\n >\n > {\n this.conversationHistory.reset();\n\n const session = this.createExecutionSession(\n taskTitleStr('Action', userPrompt),\n );\n const runner = session.getRunner();\n\n let replanCount = 0;\n const yamlFlow: MidsceneYamlFlowItem[] = [];\n const replanningCycleLimit = this.getReplanningCycleLimit(\n modelConfig.vlMode === 'vlm-ui-tars',\n );\n\n // Main planning loop - unified plan/replan logic\n while (true) {\n if (replanCount > replanningCycleLimit) {\n const errorMsg = `Replanning ${replanningCycleLimit} times, which is more than the limit, please split the task into multiple steps`;\n\n return session.appendErrorPlan(errorMsg);\n }\n\n // Create planning task (automatically includes execution history if available)\n const planningTask = this.createPlanningTask(\n userPrompt,\n actionContext,\n modelConfig,\n );\n\n const result = await session.appendAndRun(planningTask);\n const planResult: PlanningAIResponse = result?.output;\n if (session.isInErrorState()) {\n return {\n output: planResult,\n runner,\n };\n }\n\n // Execute planned actions\n const plans = planResult.actions || [];\n yamlFlow.push(...(planResult.yamlFlow || []));\n\n let executables: Awaited<ReturnType<typeof this.convertPlanToExecutable>>;\n try {\n executables = await this.convertPlanToExecutable(plans, modelConfig, {\n cacheable,\n subTask: true,\n });\n await session.appendAndRun(executables.tasks);\n } catch (error) {\n return session.appendErrorPlan(\n `Error converting plans to executable tasks: ${error}, plans: ${JSON.stringify(\n plans,\n )}`,\n );\n }\n if (session.isInErrorState()) {\n return {\n output: undefined,\n runner,\n };\n }\n\n // Check if task is complete\n if (!planResult.more_actions_needed_by_instruction) {\n break;\n }\n\n // Increment replan count for next iteration\n replanCount++;\n }\n\n return {\n output: {\n yamlFlow,\n },\n runner,\n };\n }\n\n private createTypeQueryTask(\n type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert' | 'WaitFor',\n demand: ServiceExtractParam,\n modelConfig: IModelConfig,\n opt?: ServiceExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ) {\n const queryTask: ExecutionTaskInsightQueryApply = {\n type: 'Insight',\n subType: type,\n locate: null,\n param: {\n dataDemand: multimodalPrompt\n ? ({\n demand,\n multimodalPrompt,\n } as never)\n : demand, // for user param presentation in report right sidebar\n },\n executor: async (param, taskContext) => {\n const { task } = taskContext;\n let queryDump: 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 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 result = await session.appendAndRun(queryTask);\n\n if (!result) {\n throw new Error(\n 'result of taskExecutor.flush() is undefined in function createTypeQueryTask',\n );\n }\n\n const { output, thought } = result;\n\n return {\n output,\n thought,\n runner: session.getRunner(),\n };\n }\n\n private sleepPlan(timeMs: number): PlanningAction<PlanningActionParamSleep> {\n return {\n type: 'Sleep',\n param: {\n timeMs,\n },\n locate: null,\n };\n }\n\n async taskForSleep(timeMs: number, _modelConfig: IModelConfig) {\n return this.taskBuilder.createSleepTask({\n timeMs,\n });\n }\n\n async waitFor(\n assertion: TUserPrompt,\n opt: PlanningActionParamWaitFor,\n modelConfig: IModelConfig,\n ): Promise<ExecutionResult<void>> {\n const { textPrompt, multimodalPrompt } = parsePrompt(assertion);\n\n const description = `waitFor: ${textPrompt}`;\n const session = this.createExecutionSession(\n taskTitleStr('WaitFor', description),\n );\n const runner = session.getRunner();\n const { timeoutMs, checkIntervalMs } = opt;\n\n assert(assertion, 'No assertion for waitFor');\n assert(timeoutMs, 'No timeoutMs for waitFor');\n assert(checkIntervalMs, 'No checkIntervalMs for waitFor');\n\n assert(\n checkIntervalMs <= timeoutMs,\n `wrong config for waitFor: checkIntervalMs must be less than timeoutMs, config: {checkIntervalMs: ${checkIntervalMs}, timeoutMs: ${timeoutMs}}`,\n );\n\n const overallStartTime = Date.now();\n let startTime = Date.now();\n let errorThought = '';\n while (Date.now() - overallStartTime < timeoutMs) {\n startTime = Date.now();\n const queryTask = await this.createTypeQueryTask(\n 'WaitFor',\n textPrompt,\n modelConfig,\n {\n doNotThrowError: true,\n },\n multimodalPrompt,\n );\n\n const result = (await session.appendAndRun(queryTask)) as\n | {\n output: boolean;\n thought?: string;\n }\n | undefined;\n\n if (result?.output) {\n return {\n output: undefined,\n runner,\n };\n }\n\n errorThought =\n result?.thought ||\n (!result && `No result from assertion: ${textPrompt}`) ||\n `unknown error when waiting for assertion: ${textPrompt}`;\n const now = Date.now();\n if (now - startTime < checkIntervalMs) {\n const timeRemaining = checkIntervalMs - (now - startTime);\n const sleepTask = this.taskBuilder.createSleepTask({\n timeMs: timeRemaining,\n });\n await session.append(sleepTask);\n }\n }\n\n return session.appendErrorPlan(`waitFor timeout: ${errorThought}`);\n }\n}\n"],"names":["debug","getDebug","defaultReplanningCycleLimit","defaultVlmUiTarsReplanningCycleLimit","TaskExecutor","title","options","ExecutionSession","Promise","plans","modelConfig","userInstruction","yamlString","session","taskTitleStr","task","param","executorContext","uiContext","assert","actionContext","startTime","Date","vlMode","uiTarsModelVersion","undefined","actionSpace","action","Array","console","planResult","uiTarsPlanning","plan","JSON","actions","log","more_actions_needed_by_instruction","error","usage","rawResponse","sleep","finalActions","timeNow","timeRemaining","tasks","result","output","isVlmUiTars","globalConfigManager","MIDSCENE_REPLANNING_CYCLE_LIMIT","userPrompt","cacheable","runner","replanCount","yamlFlow","replanningCycleLimit","errorMsg","planningTask","executables","type","demand","opt","multimodalPrompt","queryTask","taskContext","queryDump","applyDump","dump","ifTypeRestricted","demandInput","keyOfResult","booleanPrompt","extractResult","extraPageDescription","tree","descriptionOfTree","ServiceError","data","thought","outputResult","Error","timeMs","_modelConfig","assertion","textPrompt","parsePrompt","description","timeoutMs","checkIntervalMs","overallStartTime","errorThought","now","sleepTask","interfaceInstance","service","opts","ConversationHistory","TaskBuilder"],"mappings":";;;;;;;;;;;;;;;;;;;;AA0CA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,8BAA8B;AACpC,MAAMC,uCAAuC;AAEtC,MAAMC;IAgBX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IAwBQ,uBACNC,KAAa,EACbC,OAA0C,EAC1C;QACA,OAAO,IAAIC,iBACTF,OACA,IAAMG,QAAQ,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,kBAAkB,KACrD;YACE,aAAa,IAAI,CAAC,mBAAmB;YACrC,OAAOF,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,KAAK;QACvB;IAEJ;IAEA,MAAa,wBACXG,KAAuB,EACvBC,WAAyB,EACzBJ,OAGC,EACD;QACA,OAAO,IAAI,CAAC,WAAW,CAAC,KAAK,CAACG,OAAOC,aAAaJ;IACpD;IAEA,MAAM,uBAAuBK,eAAuB,EAAEC,UAAkB,EAAE;QACxE,MAAMC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUH;QAGzB,MAAMI,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,QAAQ;YACR,OAAO;gBACLJ;YACF;YACA,UAAU,OAAOK,OAAOC;gBACtB,MAAM,EAAEC,SAAS,EAAE,GAAGD;gBACtBE,OAAOD,WAAW;gBAClB,OAAO;oBACL,QAAQ;wBACN,SAAS,EAAE;wBACX,oCAAoC;wBACpC,KAAK;wBACLN;oBACF;oBACA,OAAO;wBACL,KAAK;oBACP;oBACA,OAAO;wBACL,MAAM;wBACN,SAAS;4BACPA;wBACF;oBACF;gBACF;YACF;QACF;QACA,MAAMC,QAAQ,YAAY,CAACE;QAE3B,OAAO;YACL,QAAQF,QAAQ,SAAS;QAC3B;IACF;IAEQ,mBACNF,eAAuB,EACvBS,aAAiC,EACjCV,WAAyB,EACG;QAC5B,MAAMK,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,QAAQ;YACR,OAAO;gBACLJ;YACF;YACA,UAAU,OAAOK,OAAOC;gBACtB,MAAMI,YAAYC,KAAK,GAAG;gBAC1B,MAAM,EAAEJ,SAAS,EAAE,GAAGD;gBACtBE,OAAOD,WAAW;gBAClB,MAAM,EAAEK,MAAM,EAAE,GAAGb;gBACnB,MAAMc,qBACJD,AAAW,kBAAXA,SAA2Bb,YAAY,kBAAkB,GAAGe;gBAE9DN,OACE,IAAI,CAAC,SAAS,CAAC,WAAW,EAC1B;gBAEF,MAAMO,cAAc,MAAM,IAAI,CAAC,SAAS,CAAC,WAAW;gBACpD1B,MACE,sCACA0B,YAAY,GAAG,CAAC,CAACC,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;gBAEhDR,OAAOS,MAAM,OAAO,CAACF,cAAc;gBACnC,IAAIA,AAAuB,MAAvBA,YAAY,MAAM,EACpBG,QAAQ,IAAI,CACV,CAAC,gBAAgB,EAAE,IAAI,CAAC,SAAS,CAAC,aAAa,CAAC,gDAAgD,CAAC;gBAIrG,MAAMC,aAAa,MAAON,AAAAA,CAAAA,qBAAqBO,iBAAiBC,IAAG,EACjEhB,MAAM,eAAe,EACrB;oBACE,SAASE;oBACTE;oBACA,eAAe,IAAI,CAAC,SAAS,CAAC,aAAa;oBAC3CM;oBACAhB;oBACA,qBAAqB,IAAI,CAAC,mBAAmB;gBAC/C;gBAEFV,MAAM,cAAciC,KAAK,SAAS,CAACH,YAAY,MAAM;gBAErD,MAAM,EACJI,OAAO,EACPC,GAAG,EACHC,kCAAkC,EAClCC,KAAK,EACLC,KAAK,EACLC,WAAW,EACXC,KAAK,EACN,GAAGV;gBAEJb,gBAAgB,IAAI,CAAC,GAAG,GAAG;oBACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;oBAClCsB;gBACF;gBACAtB,gBAAgB,IAAI,CAAC,KAAK,GAAGqB;gBAE7B,MAAMG,eAAeP,WAAW,EAAE;gBAElC,IAAIM,OAAO;oBACT,MAAME,UAAUpB,KAAK,GAAG;oBACxB,MAAMqB,gBAAgBH,QAASE,CAAAA,UAAUrB,SAAQ;oBACjD,IAAIsB,gBAAgB,GAClBF,aAAa,IAAI,CAAC,IAAI,CAAC,SAAS,CAACE;gBAErC;gBAEA,IAAIF,AAAwB,MAAxBA,aAAa,MAAM,EACrBtB,OACE,CAACiB,sCAAsCI,OACvCH,QAAQ,CAAC,gBAAgB,EAAEA,OAAO,GAAG;gBAIzC,OAAO;oBACL,QAAQ;wBACN,SAASI;wBACTL;wBACAD;wBACA,UAAUL,WAAW,QAAQ;oBAC/B;oBACA,OAAO;wBACL,KAAK;oBACP;oBACAZ;gBACF;YACF;QACF;QAEA,OAAOH;IACT;IAEA,MAAM,SACJV,KAAa,EACbI,KAAuB,EACvBC,WAAyB,EACC;QAC1B,MAAMG,UAAU,IAAI,CAAC,sBAAsB,CAACR;QAC5C,MAAM,EAAEuC,KAAK,EAAE,GAAG,MAAM,IAAI,CAAC,uBAAuB,CAACnC,OAAOC;QAC5D,MAAMmC,SAAS,MAAMhC,QAAQ,YAAY,CAAC+B;QAC1C,MAAM,EAAEE,MAAM,EAAE,GAAGD;QACnB,OAAO;YACLC;YACA,QAAQjC,QAAQ,SAAS;QAC3B;IACF;IAEQ,wBAAwBkC,WAAoB,EAAE;QACpD,OACE,IAAI,CAAC,oBAAoB,IACzBC,oBAAoB,oBAAoB,CACtCC,oCAEDF,CAAAA,cACG5C,uCACAD,2BAA0B;IAElC;IAEA,MAAM,OACJgD,UAAkB,EAClBxC,WAAyB,EACzBU,aAAsB,EACtB+B,SAAmB,EAQnB;QACA,IAAI,CAAC,mBAAmB,CAAC,KAAK;QAE9B,MAAMtC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,UAAUoC;QAEzB,MAAME,SAASvC,QAAQ,SAAS;QAEhC,IAAIwC,cAAc;QAClB,MAAMC,WAAmC,EAAE;QAC3C,MAAMC,uBAAuB,IAAI,CAAC,uBAAuB,CACvD7C,AAAuB,kBAAvBA,YAAY,MAAM;QAIpB,MAAO,KAAM;YACX,IAAI2C,cAAcE,sBAAsB;gBACtC,MAAMC,WAAW,CAAC,WAAW,EAAED,qBAAqB,+EAA+E,CAAC;gBAEpI,OAAO1C,QAAQ,eAAe,CAAC2C;YACjC;YAGA,MAAMC,eAAe,IAAI,CAAC,kBAAkB,CAC1CP,YACA9B,eACAV;YAGF,MAAMmC,SAAS,MAAMhC,QAAQ,YAAY,CAAC4C;YAC1C,MAAM3B,aAAiCe,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM;YACrD,IAAIhC,QAAQ,cAAc,IACxB,OAAO;gBACL,QAAQiB;gBACRsB;YACF;YAIF,MAAM3C,QAAQqB,WAAW,OAAO,IAAI,EAAE;YACtCwB,SAAS,IAAI,IAAKxB,WAAW,QAAQ,IAAI,EAAE;YAE3C,IAAI4B;YACJ,IAAI;gBACFA,cAAc,MAAM,IAAI,CAAC,uBAAuB,CAACjD,OAAOC,aAAa;oBACnEyC;oBACA,SAAS;gBACX;gBACA,MAAMtC,QAAQ,YAAY,CAAC6C,YAAY,KAAK;YAC9C,EAAE,OAAOrB,OAAO;gBACd,OAAOxB,QAAQ,eAAe,CAC5B,CAAC,4CAA4C,EAAEwB,MAAM,SAAS,EAAEJ,KAAK,SAAS,CAC5ExB,QACC;YAEP;YACA,IAAII,QAAQ,cAAc,IACxB,OAAO;gBACL,QAAQY;gBACR2B;YACF;YAIF,IAAI,CAACtB,WAAW,kCAAkC,EAChD;YAIFuB;QACF;QAEA,OAAO;YACL,QAAQ;gBACNC;YACF;YACAF;QACF;IACF;IAEQ,oBACNO,IAAsE,EACtEC,MAA2B,EAC3BlD,WAAyB,EACzBmD,GAA0B,EAC1BC,gBAAoC,EACpC;QACA,MAAMC,YAA4C;YAChD,MAAM;YACN,SAASJ;YACT,QAAQ;YACR,OAAO;gBACL,YAAYG,mBACP;oBACCF;oBACAE;gBACF,IACAF;YACN;YACA,UAAU,OAAO5C,OAAOgD;gBACtB,MAAM,EAAEjD,IAAI,EAAE,GAAGiD;gBACjB,IAAIC;gBACJ,MAAMC,YAAY,CAACC;oBACjBF,YAAYE;oBACZpD,KAAK,GAAG,GAAG;wBACToD;oBACF;gBACF;gBAGA,MAAMjD,YAAY8C,YAAY,SAAS;gBACvC7C,OAAOD,WAAW;gBAElB,MAAMkD,mBAAmBT,AAAS,YAATA;gBACzB,IAAIU,cAAcT;gBAClB,IAAIU,cAAc;gBAClB,IAAIF,oBAAqBT,CAAAA,AAAS,aAATA,QAAqBA,AAAS,cAATA,IAAiB,GAAI;oBACjEW,cAAc;oBACd,MAAMC,gBACJZ,AAAS,aAATA,OACI,CAAC,kDAAkD,EAAEC,QAAQ,GAC7D,CAAC,+GAA+G,EAAEA,QAAQ;oBAChIS,cAAc;wBACZ,CAACC,YAAY,EAAEC;oBACjB;gBACF,OAAO,IAAIH,kBACTC,cAAc;oBACZ,CAACC,YAAY,EAAE,GAAGX,KAAK,EAAE,EAAEC,QAAQ;gBACrC;gBAGF,IAAIY;gBAEJ,IAAIC,uBAAuB;gBAC3B,IAAIZ,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,KAAK,IAAI,CAAC,SAAS,CAAC,mBAAmB,EAAE;oBAC1D7D,MAAM;oBACN,MAAM0E,OAAO,MAAM,IAAI,CAAC,SAAS,CAAC,mBAAmB;oBACrDD,uBAAuB,MAAME,kBAC3BD,MACA,KACA,OACAb,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,MAAM;gBAEzB;gBAEA,IAAI;oBACFW,gBAAgB,MAAM,IAAI,CAAC,OAAO,CAAC,OAAO,CACxCH,aACA3D,aACAmD,KACAY,sBACAX;gBAEJ,EAAE,OAAOzB,OAAO;oBACd,IAAIA,iBAAiBuC,cACnBV,UAAU7B,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAEA,MAAM,EAAEwC,IAAI,EAAEvC,KAAK,EAAEwC,OAAO,EAAEX,IAAI,EAAE,GAAGK;gBACvCN,UAAUC;gBAEV,IAAIY,eAAeF;gBACnB,IAAIT,kBAEF,IAAI,AAAgB,YAAhB,OAAOS,MACTE,eAAeF;qBACV,IAAIlB,AAAS,cAATA,MAEPoB,eADEF,QAAAA,OACa,QAECA,IAAY,CAACP,YAAY;qBAEtC,IAAIO,QAAAA,MACTE,eAAe;qBACV;oBACL5D,OACE0D,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,IAAM,CAACP,YAAY,AAAD,MAAM7C,QACxB;oBAEFsD,eAAgBF,IAAY,CAACP,YAAY;gBAC3C;gBAGF,OAAO;oBACL,QAAQS;oBACR,KAAKd;oBACL3B;oBACAwC;gBACF;YACF;QACF;QAEA,OAAOf;IACT;IACA,MAAM,yBACJJ,IAA0D,EAC1DC,MAA2B,EAC3BlD,WAAyB,EACzBmD,GAA0B,EAC1BC,gBAAoC,EACP;QAC7B,MAAMjD,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aACE6C,MACA,AAAkB,YAAlB,OAAOC,SAAsBA,SAAS3B,KAAK,SAAS,CAAC2B;QAIzD,MAAMG,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9CJ,MACAC,QACAlD,aACAmD,KACAC;QAGF,MAAMjB,SAAS,MAAMhC,QAAQ,YAAY,CAACkD;QAE1C,IAAI,CAAClB,QACH,MAAM,IAAImC,MACR;QAIJ,MAAM,EAAElC,MAAM,EAAEgC,OAAO,EAAE,GAAGjC;QAE5B,OAAO;YACLC;YACAgC;YACA,QAAQjE,QAAQ,SAAS;QAC3B;IACF;IAEQ,UAAUoE,MAAc,EAA4C;QAC1E,OAAO;YACL,MAAM;YACN,OAAO;gBACLA;YACF;YACA,QAAQ;QACV;IACF;IAEA,MAAM,aAAaA,MAAc,EAAEC,YAA0B,EAAE;QAC7D,OAAO,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;YACtCD;QACF;IACF;IAEA,MAAM,QACJE,SAAsB,EACtBtB,GAA+B,EAC/BnD,WAAyB,EACO;QAChC,MAAM,EAAE0E,UAAU,EAAEtB,gBAAgB,EAAE,GAAGuB,YAAYF;QAErD,MAAMG,cAAc,CAAC,SAAS,EAAEF,YAAY;QAC5C,MAAMvE,UAAU,IAAI,CAAC,sBAAsB,CACzCC,aAAa,WAAWwE;QAE1B,MAAMlC,SAASvC,QAAQ,SAAS;QAChC,MAAM,EAAE0E,SAAS,EAAEC,eAAe,EAAE,GAAG3B;QAEvC1C,OAAOgE,WAAW;QAClBhE,OAAOoE,WAAW;QAClBpE,OAAOqE,iBAAiB;QAExBrE,OACEqE,mBAAmBD,WACnB,CAAC,iGAAiG,EAAEC,gBAAgB,aAAa,EAAED,UAAU,CAAC,CAAC;QAGjJ,MAAME,mBAAmBnE,KAAK,GAAG;QACjC,IAAID,YAAYC,KAAK,GAAG;QACxB,IAAIoE,eAAe;QACnB,MAAOpE,KAAK,GAAG,KAAKmE,mBAAmBF,UAAW;YAChDlE,YAAYC,KAAK,GAAG;YACpB,MAAMyC,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9C,WACAqB,YACA1E,aACA;gBACE,iBAAiB;YACnB,GACAoD;YAGF,MAAMjB,SAAU,MAAMhC,QAAQ,YAAY,CAACkD;YAO3C,IAAIlB,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM,EAChB,OAAO;gBACL,QAAQpB;gBACR2B;YACF;YAGFsC,eACE7C,AAAAA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,OAAO,AAAD,KACb,CAACA,UAAU,CAAC,0BAA0B,EAAEuC,YAAY,IACrD,CAAC,0CAA0C,EAAEA,YAAY;YAC3D,MAAMO,MAAMrE,KAAK,GAAG;YACpB,IAAIqE,MAAMtE,YAAYmE,iBAAiB;gBACrC,MAAM7C,gBAAgB6C,kBAAmBG,CAAAA,MAAMtE,SAAQ;gBACvD,MAAMuE,YAAY,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;oBACjD,QAAQjD;gBACV;gBACA,MAAM9B,QAAQ,MAAM,CAAC+E;YACvB;QACF;QAEA,OAAO/E,QAAQ,eAAe,CAAC,CAAC,iBAAiB,EAAE6E,cAAc;IACnE;IAjiBA,YACEG,iBAAoC,EACpCC,OAAgB,EAChBC,IAIC,CACD;QA3BF;QAEA;QAEA;QAEA,uBAAiB,eAAjB;QAEA,uBAAQ,uBAAR;QAEA;QAEA;QAgBE,IAAI,CAAC,SAAS,GAAGF;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC,KAAK,SAAS;QAC/B,IAAI,CAAC,mBAAmB,GAAGA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW;QAC5C,IAAI,CAAC,oBAAoB,GAAGA,KAAK,oBAAoB;QACrD,IAAI,CAAC,mBAAmB,GAAG,IAAIC;QAC/B,IAAI,CAAC,WAAW,GAAG,IAAIC,YAAY;YACjCJ;YACAC;YACA,WAAWC,KAAK,SAAS;QAC3B;IACF;AA8gBF"}
|
|
@@ -1,5 +1,4 @@
|
|
|
1
1
|
function typeStr(task) {
|
|
2
|
-
if ('Insight' === task.type && ('Query' === task.subType || 'Assert' === task.subType)) return task.type;
|
|
3
2
|
if ('Action' === task.type && task.subType) return `Action Space / ${task.subType}`;
|
|
4
3
|
if (task.subType) return `${task.type} / ${task.subType}`;
|
|
5
4
|
return task.type;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"agent/ui-utils.mjs","sources":["webpack://@midscene/core/./src/agent/ui-utils.ts"],"sourcesContent":["import type {\n DetailedLocateParam,\n ExecutionTask,\n ExecutionTaskAction,\n ExecutionTaskInsightAssertion,\n ExecutionTaskInsightLocate,\n ExecutionTaskInsightQuery,\n ExecutionTaskPlanning,\n PullParam,\n ScrollParam,\n} from '@/types';\n\nexport function typeStr(task: ExecutionTask) {\n // For
|
|
1
|
+
{"version":3,"file":"agent/ui-utils.mjs","sources":["webpack://@midscene/core/./src/agent/ui-utils.ts"],"sourcesContent":["import type {\n DetailedLocateParam,\n ExecutionTask,\n ExecutionTaskAction,\n ExecutionTaskInsightAssertion,\n ExecutionTaskInsightLocate,\n ExecutionTaskInsightQuery,\n ExecutionTaskPlanning,\n PullParam,\n ScrollParam,\n} from '@/types';\n\nexport function typeStr(task: ExecutionTask) {\n // 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} / ${task.subType}`;\n }\n\n // No subType, just show type\n return task.type;\n}\n\nexport function locateParamStr(locate?: DetailedLocateParam | string): string {\n if (!locate) {\n return '';\n }\n\n if (typeof locate === 'string') {\n return locate;\n }\n\n if (typeof locate === 'object') {\n if (typeof locate.prompt === 'string') {\n return locate.prompt;\n }\n\n if (typeof locate.prompt === 'object' && locate.prompt.prompt) {\n const prompt = locate.prompt.prompt;\n const images = locate.prompt.images || [];\n\n if (images.length === 0) return prompt;\n\n const imagesStr = images\n .map((image) => {\n let url = image.url;\n if (\n url.startsWith('data:image/') ||\n (url.startsWith('data:') && url.includes('base64'))\n ) {\n url = `${url.substring(0, 15)}...`;\n }\n return `[${image.name}](${url})`;\n })\n .join(', ');\n\n return `${prompt}, ${imagesStr}`;\n }\n }\n\n return '';\n}\n\nexport function scrollParamStr(scrollParam?: ScrollParam) {\n if (!scrollParam) {\n return '';\n }\n return `${scrollParam.direction || 'down'}, ${scrollParam.scrollType || 'once'}, ${scrollParam.distance || 'distance-not-set'}`;\n}\n\nexport function pullParamStr(pullParam?: PullParam) {\n if (!pullParam) {\n return '';\n }\n const parts: string[] = [];\n parts.push(`direction: ${pullParam.direction || 'down'}`);\n if (pullParam.distance) {\n parts.push(`distance: ${pullParam.distance}`);\n }\n if (pullParam.duration) {\n parts.push(`duration: ${pullParam.duration}ms`);\n }\n return parts.join(', ');\n}\n\nexport function taskTitleStr(\n type:\n | 'Tap'\n | 'Hover'\n | 'Input'\n | 'RightClick'\n | 'KeyboardPress'\n | 'Scroll'\n | 'Action'\n | 'Query'\n | 'Assert'\n | 'WaitFor'\n | 'Locate'\n | 'Boolean'\n | 'Number'\n | 'String',\n prompt: string,\n) {\n if (prompt) {\n return `${type} - ${prompt}`;\n }\n return type;\n}\n\nexport function paramStr(task: ExecutionTask) {\n let value: string | undefined | object;\n if (task.type === 'Planning') {\n value = (task as ExecutionTaskPlanning)?.param?.userInstruction;\n }\n\n if (task.type === 'Insight') {\n value =\n locateParamStr((task as ExecutionTaskInsightLocate)?.param) ||\n (task as ExecutionTaskInsightQuery)?.param?.dataDemand ||\n (task as ExecutionTaskInsightAssertion)?.param?.assertion;\n }\n\n if (task.type === 'Action') {\n const locate = (task as ExecutionTaskAction)?.locate;\n const locateStr = locate ? locateParamStr(locate) : '';\n\n value = task.thought || '';\n if (typeof (task as ExecutionTaskAction)?.param?.timeMs === 'number') {\n value = `${(task as ExecutionTaskAction)?.param?.timeMs}ms`;\n } else if (\n typeof (task as ExecutionTaskAction)?.param?.scrollType === 'string'\n ) {\n value = scrollParamStr((task as ExecutionTaskAction)?.param);\n } else if (\n typeof (task as ExecutionTaskAction)?.param?.direction === 'string' &&\n (task as ExecutionTaskAction)?.subType === 'AndroidPull'\n ) {\n value = pullParamStr((task as ExecutionTaskAction)?.param);\n } else if (\n typeof (task as ExecutionTaskAction)?.param?.value !== 'undefined'\n ) {\n value = (task as ExecutionTaskAction)?.param?.value;\n }\n\n if (locateStr) {\n if (value) {\n value = `${locateStr} - ${value}`;\n } else {\n value = locateStr;\n }\n }\n }\n\n if (typeof value === 'undefined') return '';\n\n if (typeof value === 'string') return value;\n\n if (typeof value === 'object' && locateParamStr(value as any)) {\n return locateParamStr(value as any);\n }\n\n return JSON.stringify(value, undefined, 2);\n}\n"],"names":["typeStr","task","locateParamStr","locate","prompt","images","imagesStr","image","url","scrollParamStr","scrollParam","pullParamStr","pullParam","parts","taskTitleStr","type","paramStr","value","_task_param","_task_param1","_task_param2","_task_param3","_task_param4","_task_param5","_task_param6","locateStr","_task_param7","_task_param8","JSON","undefined"],"mappings":"AAYO,SAASA,QAAQC,IAAmB;IAEzC,IAAIA,AAAc,aAAdA,KAAK,IAAI,IAAiBA,KAAK,OAAO,EACxC,OAAO,CAAC,eAAe,EAAEA,KAAK,OAAO,EAAE;IAIzC,IAAIA,KAAK,OAAO,EACd,OAAO,GAAGA,KAAK,IAAI,CAAC,GAAG,EAAEA,KAAK,OAAO,EAAE;IAIzC,OAAOA,KAAK,IAAI;AAClB;AAEO,SAASC,eAAeC,MAAqC;IAClE,IAAI,CAACA,QACH,OAAO;IAGT,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAGT,IAAI,AAAkB,YAAlB,OAAOA,QAAqB;QAC9B,IAAI,AAAyB,YAAzB,OAAOA,OAAO,MAAM,EACtB,OAAOA,OAAO,MAAM;QAGtB,IAAI,AAAyB,YAAzB,OAAOA,OAAO,MAAM,IAAiBA,OAAO,MAAM,CAAC,MAAM,EAAE;YAC7D,MAAMC,SAASD,OAAO,MAAM,CAAC,MAAM;YACnC,MAAME,SAASF,OAAO,MAAM,CAAC,MAAM,IAAI,EAAE;YAEzC,IAAIE,AAAkB,MAAlBA,OAAO,MAAM,EAAQ,OAAOD;YAEhC,MAAME,YAAYD,OACf,GAAG,CAAC,CAACE;gBACJ,IAAIC,MAAMD,MAAM,GAAG;gBACnB,IACEC,IAAI,UAAU,CAAC,kBACdA,IAAI,UAAU,CAAC,YAAYA,IAAI,QAAQ,CAAC,WAEzCA,MAAM,GAAGA,IAAI,SAAS,CAAC,GAAG,IAAI,GAAG,CAAC;gBAEpC,OAAO,CAAC,CAAC,EAAED,MAAM,IAAI,CAAC,EAAE,EAAEC,IAAI,CAAC,CAAC;YAClC,GACC,IAAI,CAAC;YAER,OAAO,GAAGJ,OAAO,EAAE,EAAEE,WAAW;QAClC;IACF;IAEA,OAAO;AACT;AAEO,SAASG,eAAeC,WAAyB;IACtD,IAAI,CAACA,aACH,OAAO;IAET,OAAO,GAAGA,YAAY,SAAS,IAAI,OAAO,EAAE,EAAEA,YAAY,UAAU,IAAI,OAAO,EAAE,EAAEA,YAAY,QAAQ,IAAI,oBAAoB;AACjI;AAEO,SAASC,aAAaC,SAAqB;IAChD,IAAI,CAACA,WACH,OAAO;IAET,MAAMC,QAAkB,EAAE;IAC1BA,MAAM,IAAI,CAAC,CAAC,WAAW,EAAED,UAAU,SAAS,IAAI,QAAQ;IACxD,IAAIA,UAAU,QAAQ,EACpBC,MAAM,IAAI,CAAC,CAAC,UAAU,EAAED,UAAU,QAAQ,EAAE;IAE9C,IAAIA,UAAU,QAAQ,EACpBC,MAAM,IAAI,CAAC,CAAC,UAAU,EAAED,UAAU,QAAQ,CAAC,EAAE,CAAC;IAEhD,OAAOC,MAAM,IAAI,CAAC;AACpB;AAEO,SAASC,aACdC,IAcY,EACZX,MAAc;IAEd,IAAIA,QACF,OAAO,GAAGW,KAAK,GAAG,EAAEX,QAAQ;IAE9B,OAAOW;AACT;AAEO,SAASC,SAASf,IAAmB;IAC1C,IAAIgB;IACJ,IAAIhB,AAAc,eAAdA,KAAK,IAAI,EAAiB;YACpBiB;QAARD,QAAShB,QAAAA,OAAAA,KAAAA,IAAAA,QAADiB,CAAAA,cAACjB,KAAgC,KAAK,AAAD,IAArCiB,KAAAA,IAAAA,YAAwC,eAAe;IACjE;IAEA,IAAIjB,AAAc,cAAdA,KAAK,IAAI,EAAgB;YAGzBkB,cACAC;QAHFH,QACEf,eAAgBD,QAAAA,OAAAA,KAAAA,IAAAA,KAAqC,KAAK,KACzDA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADkB,CAAAA,eAAClB,KAAoC,KAAK,AAAD,IAAzCkB,KAAAA,IAAAA,aAA4C,UAAU,AAAD,KACpDlB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADmB,CAAAA,eAACnB,KAAwC,KAAK,AAAD,IAA7CmB,KAAAA,IAAAA,aAAgD,SAAS,AAAD;IAC5D;IAEA,IAAInB,AAAc,aAAdA,KAAK,IAAI,EAAe;YAKfoB,cAGFC,cAIAC,cAKAC;QAhBT,MAAMrB,SAAUF,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,MAAM;QACpD,MAAMwB,YAAYtB,SAASD,eAAeC,UAAU;QAEpDc,QAAQhB,KAAK,OAAO,IAAI;QACxB,IAAI,AAAwD,YAAxD,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADoB,CAAAA,eAACpB,KAA8B,KAAK,AAAD,IAAnCoB,KAAAA,IAAAA,aAAsC,MAAM,AAAD,GAAgB;gBACzDK;YAAXT,QAAQ,GAAIhB,QAAAA,OAAAA,KAAAA,IAAAA,QAADyB,CAAAA,eAACzB,KAA8B,KAAK,AAAD,IAAnCyB,KAAAA,IAAAA,aAAsC,MAAM,CAAC,EAAE,CAAC;QAC7D,OAAO,IACL,AAA4D,YAA5D,OAAQzB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADqB,CAAAA,eAACrB,KAA8B,KAAK,AAAD,IAAnCqB,KAAAA,IAAAA,aAAsC,UAAU,AAAD,GAEtDL,QAAQR,eAAgBR,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;aACtD,IACL,AAA2D,YAA3D,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADsB,CAAAA,eAACtB,KAA8B,KAAK,AAAD,IAAnCsB,KAAAA,IAAAA,aAAsC,SAAS,AAAD,KACrD,AAACtB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,OAAO,AAAD,MAAM,eAE3CgB,QAAQN,aAAcV,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;aACpD,IACL,AAAuD,WAA/CA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADuB,CAAAA,eAACvB,KAA8B,KAAK,AAAD,IAAnCuB,KAAAA,IAAAA,aAAsC,KAAK,AAAD,GACjD;gBACQG;YAARV,QAAShB,QAAAA,OAAAA,KAAAA,IAAAA,QAAD0B,CAAAA,eAAC1B,KAA8B,KAAK,AAAD,IAAnC0B,KAAAA,IAAAA,aAAsC,KAAK;QACrD;QAEA,IAAIF,WAEAR,QADEA,QACM,GAAGQ,UAAU,GAAG,EAAER,OAAO,GAEzBQ;IAGd;IAEA,IAAI,AAAiB,WAAVR,OAAuB,OAAO;IAEzC,IAAI,AAAiB,YAAjB,OAAOA,OAAoB,OAAOA;IAEtC,IAAI,AAAiB,YAAjB,OAAOA,SAAsBf,eAAee,QAC9C,OAAOf,eAAee;IAGxB,OAAOW,KAAK,SAAS,CAACX,OAAOY,QAAW;AAC1C"}
|
package/dist/es/agent/utils.mjs
CHANGED
|
@@ -99,7 +99,7 @@ async function matchElementFromCache(context, cacheEntry, cachePrompt, cacheable
|
|
|
99
99
|
return;
|
|
100
100
|
}
|
|
101
101
|
}
|
|
102
|
-
const getMidsceneVersion = ()=>"1.0.1-beta-
|
|
102
|
+
const getMidsceneVersion = ()=>"1.0.1-beta-20251028121806.0";
|
|
103
103
|
const parsePrompt = (prompt)=>{
|
|
104
104
|
if ('string' == typeof prompt) return {
|
|
105
105
|
textPrompt: prompt,
|
|
@@ -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 InsightExtractOption,\n Rect,\n ReferenceImage,\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 { describeUserPage } from './prompt/util';\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 userInstructionPrompt = await findElementPrompt.format({\n pageDescription: await describeUserPage(context),\n targetElementDescription: extraTextFromUserPrompt(targetElementDescription),\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 === 'qwen-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 ('bbox' in res.content && Array.isArray(res.content.bbox)) {\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 =\n generateElementByPosition(rectCenter);\n errors = [];\n\n element.isOrderSensitive =\n typeof res.content === 'object' &&\n res.content !== null &&\n 'isOrderSensitive' in res.content\n ? (res.content as any).isOrderSensitive\n : undefined;\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 === 'qwen-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?: InsightExtractOption;\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"],"names":["debugInspect","getDebug","debugSection","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","_multimodalPrompt_images","msgs","item","base64","preProcessImageUrl","AiLocateElement","options","context","targetElementDescription","callAIFn","modelConfig","vlMode","screenshotBase64","assert","userInstructionPrompt","findElementPrompt","describeUserPage","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","undefined","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"],"mappings":";;;;;;;;;;AAmDA,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,wBAAwB,MAAMC,kBAAkB,MAAM,CAAC;QAC3D,iBAAiB,MAAMC,iBAAiBT;QACxC,0BAA0BX,wBAAwBY;IACpD;IACA,MAAMS,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,cAAXA,QAAsB;QAC/B,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,MAAML;gBACR;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAON,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,IAAI,UAAUA,IAAI,OAAO,IAAIO,MAAM,OAAO,CAACP,IAAI,OAAO,CAAC,IAAI,GAAG;gBAK1DQ,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,UACJC,0BAA0BF;YAC5BL,SAAS,EAAE;YAEXM,QAAQ,gBAAgB,GACtB,AAAuB,YAAvB,OAAOZ,IAAI,OAAO,IAClBA,AAAgB,SAAhBA,IAAI,OAAO,IACX,sBAAsBA,IAAI,OAAO,GAC5BA,IAAI,OAAO,CAAS,gBAAgB,GACrCc;YAEN,IAAIF,SACFP,kBAAkB;gBAACO;aAAQ;QAE/B;IACF,EAAE,OAAOG,GAAG;QACV,MAAMC,MACJD,aAAaE,QACT,CAAC,sBAAsB,EAAEF,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACT,UAAUA,AAAAA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM,AAAD,MAAM,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAEU,IAAI,CAAC,CAAC;aAFtBV,SAAS;YAACU;SAAI;IAIlB;IAEA,OAAO;QACL,MAAMZ;QACN,aAAa;YACX,UAAUC;YACV,QAAQC;QACV;QACAJ;QACA,OAAOF,IAAI,KAAK;IAClB;AACF;AAEO,eAAekB,gBAAgBzC,OAIrC;IAOC,MAAM,EAAEC,OAAO,EAAEyC,kBAAkB,EAAEtC,WAAW,EAAE,GAAGJ;IACrD,MAAM,EAAEK,MAAM,EAAE,GAAGD;IACnB,MAAM,EAAEE,gBAAgB,EAAE,GAAGL;IAE7B,MAAMU,eAAegC,4BAA4BtC;IACjD,MAAMuC,gCAAgC,MAAMC,0BAA0B,MAAM,CAAC;QAC3E,oBAAoBvD,wBAAwBoD;IAC9C;IACA,MAAM/C,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,MAAMsC;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAMpB,SAAS,MAAM9B,mBAAmB;YACtC,QAAQkD,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACA/C,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAMwB,SAAS,MAAMC,yBACnBpD,MACA6B,aAAa,YAAY,EACzBpB;IAGF,IAAI4C;IACJ,MAAMC,cAAcH,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIG,aAAa;QACf,MAAMC,aAAajB,gBACjBgB,aACAhD,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAEFhB,aAAa,0BAA0B6D;QAEvC,MAAMC,oBAAoBL,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9DzD,aAAa,wBAAwB8D;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAASvB,MAAM,OAAO,CAACuB,OAC/B,GAAG,CAAC,CAACA,OACGpB,gBACLoB,MACApD,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAGNhB,aAAa,qBAAqB+D;QAGlC,MAAME,aAAaC,WAAW;YAACL;eAAeE;SAAe;QAC7D/D,aAAa,iBAAiBiE;QAG9BN,cAAcQ,iBAAiBF,YAAYrD,QAAQ,IAAI,EAAEI;QACzDhB,aAAa,2BAA2B2D;IAC1C;IAEA,IAAIS,cAAcnD;IAClB,IAAI0C,aAAa;QACf,MAAMU,gBAAgB,MAAMC,WAC1BrD,kBACA0C,aACA3C,AAAW,cAAXA;QAEFoD,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,aAAapB,KAAK,SAAS,CAACoB,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAec,qBAAwB5D,OAO7C;IACC,MAAM,EAAE6D,SAAS,EAAE5D,OAAO,EAAE6D,aAAa,EAAErE,gBAAgB,EAAEW,WAAW,EAAE,GACxEJ;IACF,MAAMW,eAAeoD;IACrB,MAAM,EAAEzD,gBAAgB,EAAE,GAAGL;IAE7B,MAAM+D,wBAAwB,MAAMC,uBAClCjE,QAAQ,eAAe,IAAI,IAC3B6D;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,KAAK5D;YACL,QAAQ;QACV;IACF;IAGF4D,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAMrE,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAASuD;QACX;KACD;IAED,IAAIzE,kBAAkB;QACpB,MAAM6B,SAAS,MAAM9B,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAE,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAMwB,SAAS,MAAMC,yBACnBpD,MACA6B,aAAa,YAAY,EACzBpB;IAEF,OAAO;QACL,aAAa0C,OAAO,OAAO;QAC3B,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 { describeUserPage } from './prompt/util';\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 userInstructionPrompt = await findElementPrompt.format({\n pageDescription: await describeUserPage(context),\n targetElementDescription: extraTextFromUserPrompt(targetElementDescription),\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 === 'qwen-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 ('bbox' in res.content && Array.isArray(res.content.bbox)) {\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 =\n generateElementByPosition(rectCenter);\n errors = [];\n\n element.isOrderSensitive =\n typeof res.content === 'object' &&\n res.content !== null &&\n 'isOrderSensitive' in res.content\n ? (res.content as any).isOrderSensitive\n : undefined;\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 === 'qwen-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"],"names":["debugInspect","getDebug","debugSection","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","_multimodalPrompt_images","msgs","item","base64","preProcessImageUrl","AiLocateElement","options","context","targetElementDescription","callAIFn","modelConfig","vlMode","screenshotBase64","assert","userInstructionPrompt","findElementPrompt","describeUserPage","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","undefined","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"],"mappings":";;;;;;;;;;AAmDA,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,wBAAwB,MAAMC,kBAAkB,MAAM,CAAC;QAC3D,iBAAiB,MAAMC,iBAAiBT;QACxC,0BAA0BX,wBAAwBY;IACpD;IACA,MAAMS,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,cAAXA,QAAsB;QAC/B,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,MAAML;gBACR;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAON,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,IAAI,UAAUA,IAAI,OAAO,IAAIO,MAAM,OAAO,CAACP,IAAI,OAAO,CAAC,IAAI,GAAG;gBAK1DQ,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,UACJC,0BAA0BF;YAC5BL,SAAS,EAAE;YAEXM,QAAQ,gBAAgB,GACtB,AAAuB,YAAvB,OAAOZ,IAAI,OAAO,IAClBA,AAAgB,SAAhBA,IAAI,OAAO,IACX,sBAAsBA,IAAI,OAAO,GAC5BA,IAAI,OAAO,CAAS,gBAAgB,GACrCc;YAEN,IAAIF,SACFP,kBAAkB;gBAACO;aAAQ;QAE/B;IACF,EAAE,OAAOG,GAAG;QACV,MAAMC,MACJD,aAAaE,QACT,CAAC,sBAAsB,EAAEF,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACT,UAAUA,AAAAA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM,AAAD,MAAM,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAEU,IAAI,CAAC,CAAC;aAFtBV,SAAS;YAACU;SAAI;IAIlB;IAEA,OAAO;QACL,MAAMZ;QACN,aAAa;YACX,UAAUC;YACV,QAAQC;QACV;QACAJ;QACA,OAAOF,IAAI,KAAK;IAClB;AACF;AAEO,eAAekB,gBAAgBzC,OAIrC;IAOC,MAAM,EAAEC,OAAO,EAAEyC,kBAAkB,EAAEtC,WAAW,EAAE,GAAGJ;IACrD,MAAM,EAAEK,MAAM,EAAE,GAAGD;IACnB,MAAM,EAAEE,gBAAgB,EAAE,GAAGL;IAE7B,MAAMU,eAAegC,4BAA4BtC;IACjD,MAAMuC,gCAAgC,MAAMC,0BAA0B,MAAM,CAAC;QAC3E,oBAAoBvD,wBAAwBoD;IAC9C;IACA,MAAM/C,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,MAAMsC;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAMpB,SAAS,MAAM9B,mBAAmB;YACtC,QAAQkD,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACA/C,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAMwB,SAAS,MAAMC,yBACnBpD,MACA6B,aAAa,YAAY,EACzBpB;IAGF,IAAI4C;IACJ,MAAMC,cAAcH,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIG,aAAa;QACf,MAAMC,aAAajB,gBACjBgB,aACAhD,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAEFhB,aAAa,0BAA0B6D;QAEvC,MAAMC,oBAAoBL,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9DzD,aAAa,wBAAwB8D;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAASvB,MAAM,OAAO,CAACuB,OAC/B,GAAG,CAAC,CAACA,OACGpB,gBACLoB,MACApD,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAGNhB,aAAa,qBAAqB+D;QAGlC,MAAME,aAAaC,WAAW;YAACL;eAAeE;SAAe;QAC7D/D,aAAa,iBAAiBiE;QAG9BN,cAAcQ,iBAAiBF,YAAYrD,QAAQ,IAAI,EAAEI;QACzDhB,aAAa,2BAA2B2D;IAC1C;IAEA,IAAIS,cAAcnD;IAClB,IAAI0C,aAAa;QACf,MAAMU,gBAAgB,MAAMC,WAC1BrD,kBACA0C,aACA3C,AAAW,cAAXA;QAEFoD,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,aAAapB,KAAK,SAAS,CAACoB,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAec,qBAAwB5D,OAO7C;IACC,MAAM,EAAE6D,SAAS,EAAE5D,OAAO,EAAE6D,aAAa,EAAErE,gBAAgB,EAAEW,WAAW,EAAE,GACxEJ;IACF,MAAMW,eAAeoD;IACrB,MAAM,EAAEzD,gBAAgB,EAAE,GAAGL;IAE7B,MAAM+D,wBAAwB,MAAMC,uBAClCjE,QAAQ,eAAe,IAAI,IAC3B6D;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,KAAK5D;YACL,QAAQ;QACV;IACF;IAGF4D,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAMrE,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAASuD;QACX;KACD;IAED,IAAIzE,kBAAkB;QACpB,MAAM6B,SAAS,MAAM9B,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAE,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAMwB,SAAS,MAAMC,yBACnBpD,MACA6B,aAAa,YAAY,EACzBpB;IAEF,OAAO;QACL,aAAa0C,OAAO,OAAO;QAC3B,OAAOA,OAAO,KAAK;IACrB;AACF"}
|
package/dist/es/index.mjs
CHANGED
|
@@ -1,12 +1,12 @@
|
|
|
1
1
|
import { z } from "zod";
|
|
2
|
-
import
|
|
2
|
+
import service from "./service/index.mjs";
|
|
3
3
|
import { TaskRunner } from "./task-runner.mjs";
|
|
4
4
|
import { getVersion } from "./utils.mjs";
|
|
5
5
|
import { AiLocateElement, PointSchema, RectSchema, SizeSchema, TMultimodalPromptSchema, TUserPromptSchema, describeUserPage, getMidsceneLocationSchema, plan } from "./ai-model/index.mjs";
|
|
6
6
|
import { MIDSCENE_MODEL_NAME } from "@midscene/shared/env";
|
|
7
|
-
import {
|
|
7
|
+
import { ServiceError } from "./types.mjs";
|
|
8
8
|
import { Agent, createAgent } from "./agent/index.mjs";
|
|
9
|
-
const src =
|
|
10
|
-
export { Agent, AiLocateElement,
|
|
9
|
+
const src = service;
|
|
10
|
+
export { Agent, AiLocateElement, MIDSCENE_MODEL_NAME, PointSchema, RectSchema, service as Service, ServiceError, SizeSchema, TMultimodalPromptSchema, TUserPromptSchema, TaskRunner, createAgent, src as default, describeUserPage, getMidsceneLocationSchema, getVersion, plan, z };
|
|
11
11
|
|
|
12
12
|
//# sourceMappingURL=index.mjs.map
|
package/dist/es/index.mjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.mjs","sources":["webpack://@midscene/core/./src/index.ts"],"sourcesContent":["import { z } from 'zod';\nimport
|
|
1
|
+
{"version":3,"file":"index.mjs","sources":["webpack://@midscene/core/./src/index.ts"],"sourcesContent":["import { z } from 'zod';\nimport Service from './service/index';\nimport { TaskRunner } from './task-runner';\nimport { getVersion } from './utils';\n\nexport {\n plan,\n describeUserPage,\n AiLocateElement,\n getMidsceneLocationSchema,\n type MidsceneLocationResultType,\n PointSchema,\n SizeSchema,\n RectSchema,\n TMultimodalPromptSchema,\n TUserPromptSchema,\n type TMultimodalPrompt,\n type TUserPrompt,\n} from './ai-model/index';\n\nexport {\n MIDSCENE_MODEL_NAME,\n type CreateOpenAIClientFn,\n} from '@midscene/shared/env';\n\nexport type * from './types';\nexport { ServiceError } from './types';\n\nexport { z };\n\nexport default Service;\nexport { TaskRunner, Service, getVersion };\n\nexport type {\n MidsceneYamlScript,\n MidsceneYamlTask,\n MidsceneYamlFlowItem,\n MidsceneYamlConfigResult,\n MidsceneYamlConfig,\n MidsceneYamlScriptWebEnv,\n MidsceneYamlScriptAndroidEnv,\n MidsceneYamlScriptIOSEnv,\n MidsceneYamlScriptEnv,\n LocateOption,\n DetailedLocateParam,\n} from './yaml';\n\nexport { Agent, type AgentOpt, createAgent } from './agent';\n"],"names":["Service"],"mappings":";;;;;;;;AA8BA,YAAeA"}
|
|
@@ -2,12 +2,12 @@ import { AIActionType, expandSearchArea } from "../ai-model/common.mjs";
|
|
|
2
2
|
import { AiExtractElementInfo, AiLocateElement, callAIWithObjectResponse } from "../ai-model/index.mjs";
|
|
3
3
|
import { AiLocateSection } from "../ai-model/inspect.mjs";
|
|
4
4
|
import { elementDescriberInstruction } from "../ai-model/prompt/describe.mjs";
|
|
5
|
-
import {
|
|
5
|
+
import { ServiceError } from "../types.mjs";
|
|
6
6
|
import { MIDSCENE_FORCE_DEEP_THINK, globalConfigManager } from "@midscene/shared/env";
|
|
7
7
|
import { compositeElementInfoImg, cropByRect } from "@midscene/shared/img";
|
|
8
8
|
import { getDebug } from "@midscene/shared/logger";
|
|
9
9
|
import { assert } from "@midscene/shared/utils";
|
|
10
|
-
import {
|
|
10
|
+
import { createServiceDump } from "./utils.mjs";
|
|
11
11
|
function _define_property(obj, key, value) {
|
|
12
12
|
if (key in obj) Object.defineProperty(obj, key, {
|
|
13
13
|
value: value,
|
|
@@ -18,8 +18,8 @@ function _define_property(obj, key, value) {
|
|
|
18
18
|
else obj[key] = value;
|
|
19
19
|
return obj;
|
|
20
20
|
}
|
|
21
|
-
const debug = getDebug('ai:
|
|
22
|
-
class
|
|
21
|
+
const debug = getDebug('ai:service');
|
|
22
|
+
class Service {
|
|
23
23
|
async locate(query, opt, modelConfig) {
|
|
24
24
|
var _parseResult_errors;
|
|
25
25
|
const queryPrompt = 'string' == typeof query ? query : query.prompt;
|
|
@@ -84,12 +84,12 @@ class Insight {
|
|
|
84
84
|
error: errorLog
|
|
85
85
|
};
|
|
86
86
|
const elements = parseResult.elements || [];
|
|
87
|
-
const dump =
|
|
87
|
+
const dump = createServiceDump({
|
|
88
88
|
...dumpData,
|
|
89
89
|
matchedElement: elements
|
|
90
90
|
});
|
|
91
|
-
if (errorLog) throw new
|
|
92
|
-
if (elements.length > 1) throw new
|
|
91
|
+
if (errorLog) throw new ServiceError(errorLog, dump);
|
|
92
|
+
if (elements.length > 1) throw new ServiceError(`locate: multiple elements found, length = ${elements.length}`, dump);
|
|
93
93
|
if (1 === elements.length) return {
|
|
94
94
|
element: {
|
|
95
95
|
id: elements[0].id,
|
|
@@ -138,11 +138,11 @@ class Insight {
|
|
|
138
138
|
error: errorLog
|
|
139
139
|
};
|
|
140
140
|
const { data, thought } = parseResult || {};
|
|
141
|
-
const dump =
|
|
141
|
+
const dump = createServiceDump({
|
|
142
142
|
...dumpData,
|
|
143
143
|
data
|
|
144
144
|
});
|
|
145
|
-
if (errorLog && !data && !(null == opt ? void 0 : opt.doNotThrowError)) throw new
|
|
145
|
+
if (errorLog && !data && !(null == opt ? void 0 : opt.doNotThrowError)) throw new ServiceError(errorLog, dump);
|
|
146
146
|
return {
|
|
147
147
|
data,
|
|
148
148
|
thought,
|
|
@@ -151,10 +151,10 @@ class Insight {
|
|
|
151
151
|
};
|
|
152
152
|
}
|
|
153
153
|
async describe(target, modelConfig, opt) {
|
|
154
|
-
assert(target, 'target is required for
|
|
154
|
+
assert(target, 'target is required for service.describe');
|
|
155
155
|
const context = await this.contextRetrieverFn();
|
|
156
156
|
const { screenshotBase64, size } = context;
|
|
157
|
-
assert(screenshotBase64, 'screenshot is required for
|
|
157
|
+
assert(screenshotBase64, 'screenshot is required for service.describe');
|
|
158
158
|
const { vlMode } = modelConfig;
|
|
159
159
|
const systemPrompt = elementDescriberInstruction();
|
|
160
160
|
const defaultRectSize = 30;
|
|
@@ -209,13 +209,13 @@ class Insight {
|
|
|
209
209
|
_define_property(this, "contextRetrieverFn", void 0);
|
|
210
210
|
_define_property(this, "aiVendorFn", callAIWithObjectResponse);
|
|
211
211
|
_define_property(this, "taskInfo", void 0);
|
|
212
|
-
assert(context, 'context is required for
|
|
212
|
+
assert(context, 'context is required for Service');
|
|
213
213
|
if ('function' == typeof context) this.contextRetrieverFn = context;
|
|
214
214
|
else this.contextRetrieverFn = ()=>Promise.resolve(context);
|
|
215
215
|
if (void 0 !== (null == opt ? void 0 : opt.aiVendorFn)) this.aiVendorFn = opt.aiVendorFn;
|
|
216
216
|
if (void 0 !== (null == opt ? void 0 : opt.taskInfo)) this.taskInfo = opt.taskInfo;
|
|
217
217
|
}
|
|
218
218
|
}
|
|
219
|
-
export {
|
|
219
|
+
export { Service as default };
|
|
220
220
|
|
|
221
221
|
//# sourceMappingURL=index.mjs.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"insight/index.mjs","sources":["webpack://@midscene/core/./src/insight/index.ts"],"sourcesContent":["import { AIActionType, type AIArgs, expandSearchArea } from '@/ai-model/common';\nimport {\n AiExtractElementInfo,\n AiLocateElement,\n callAIWithObjectResponse,\n} from '@/ai-model/index';\nimport { AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n AIDescribeElementResponse,\n AIUsageInfo,\n DetailedLocateParam,\n InsightExtractOption,\n InsightExtractParam,\n InsightExtractResult,\n InsightTaskInfo,\n LocateResultWithDump,\n PartialInsightDumpFromSDK,\n Rect,\n UIContext,\n} from '@/types';\nimport { InsightError } from '@/types';\nimport {\n type IModelConfig,\n MIDSCENE_FORCE_DEEP_THINK,\n globalConfigManager,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TMultimodalPrompt } from '../ai-model/common';\nimport { createInsightDump } from './utils';\n\nexport interface LocateOpts {\n context?: UIContext;\n}\n\nexport type AnyValue<T> = {\n [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\ninterface InsightOptions {\n taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n aiVendorFn?: typeof callAIWithObjectResponse;\n}\n\nconst debug = getDebug('ai:insight');\nexport default class Insight {\n contextRetrieverFn: () => Promise<UIContext> | UIContext;\n\n aiVendorFn: Exclude<InsightOptions['aiVendorFn'], undefined> =\n callAIWithObjectResponse;\n\n taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n constructor(\n context: UIContext | (() => Promise<UIContext> | UIContext),\n opt?: InsightOptions,\n ) {\n assert(context, 'context is required for Insight');\n if (typeof context === 'function') {\n this.contextRetrieverFn = context;\n } else {\n this.contextRetrieverFn = () => Promise.resolve(context);\n }\n\n // just for unit test, aiVendorFn is callAIWithObjectResponse by default\n if (typeof opt?.aiVendorFn !== 'undefined') {\n this.aiVendorFn = opt.aiVendorFn;\n }\n if (typeof opt?.taskInfo !== 'undefined') {\n this.taskInfo = opt.taskInfo;\n }\n }\n\n async locate(\n query: DetailedLocateParam,\n opt: LocateOpts,\n modelConfig: IModelConfig,\n ): Promise<LocateResultWithDump> {\n const queryPrompt = typeof query === 'string' ? query : query.prompt;\n assert(queryPrompt, 'query is required for locate');\n\n assert(typeof query === 'object', 'query should be an object for locate');\n\n const globalDeepThinkSwitch = globalConfigManager.getEnvConfigInBoolean(\n MIDSCENE_FORCE_DEEP_THINK,\n );\n if (globalDeepThinkSwitch) {\n debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n }\n let searchAreaPrompt;\n if (query.deepThink || globalDeepThinkSwitch) {\n searchAreaPrompt = query.prompt;\n }\n\n const { vlMode } = modelConfig;\n\n if (searchAreaPrompt && !vlMode) {\n console.warn(\n 'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n );\n searchAreaPrompt = undefined;\n }\n\n const context = opt?.context || (await this.contextRetrieverFn());\n\n let searchArea: Rect | undefined = undefined;\n let searchAreaRawResponse: string | undefined = undefined;\n let searchAreaUsage: AIUsageInfo | undefined = undefined;\n let searchAreaResponse:\n | Awaited<ReturnType<typeof AiLocateSection>>\n | undefined = undefined;\n if (searchAreaPrompt) {\n searchAreaResponse = await AiLocateSection({\n context,\n sectionDescription: searchAreaPrompt,\n modelConfig,\n });\n assert(\n searchAreaResponse.rect,\n `cannot find search area for \"${searchAreaPrompt}\"${\n searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n }`,\n );\n searchAreaRawResponse = searchAreaResponse.rawResponse;\n searchAreaUsage = searchAreaResponse.usage;\n searchArea = searchAreaResponse.rect;\n }\n\n const startTime = Date.now();\n const { parseResult, rect, rawResponse, usage } = await AiLocateElement({\n callAIFn: this.aiVendorFn,\n context,\n targetElementDescription: queryPrompt,\n searchConfig: searchAreaResponse,\n modelConfig,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(rawResponse),\n formatResponse: JSON.stringify(parseResult),\n usage,\n searchArea,\n searchAreaRawResponse,\n searchAreaUsage,\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'locate',\n userQuery: {\n element: queryPrompt,\n },\n matchedElement: [],\n matchedRect: rect,\n data: null,\n taskInfo,\n deepThink: !!searchArea,\n error: errorLog,\n };\n\n const elements = parseResult.elements || [];\n\n const dump = createInsightDump({\n ...dumpData,\n matchedElement: elements,\n });\n\n if (errorLog) {\n throw new InsightError(errorLog, dump);\n }\n\n if (elements.length > 1) {\n throw new InsightError(\n `locate: multiple elements found, length = ${elements.length}`,\n dump,\n );\n }\n\n if (elements.length === 1) {\n return {\n element: {\n id: elements[0]!.id,\n center: elements[0]!.center,\n rect: elements[0]!.rect,\n isOrderSensitive: elements[0]!.isOrderSensitive,\n },\n rect,\n dump,\n };\n }\n\n return {\n element: null,\n rect,\n dump,\n };\n }\n\n async extract<T>(\n dataDemand: InsightExtractParam,\n modelConfig: IModelConfig,\n opt?: InsightExtractOption,\n pageDescription?: string,\n multimodalPrompt?: TMultimodalPrompt,\n ): Promise<InsightExtractResult<T>> {\n assert(\n typeof dataDemand === 'object' || typeof dataDemand === 'string',\n `dataDemand should be object or string, but get ${typeof dataDemand}`,\n );\n const context = await this.contextRetrieverFn();\n\n const startTime = Date.now();\n\n const { parseResult, usage } = await AiExtractElementInfo<T>({\n context,\n dataQuery: dataDemand,\n multimodalPrompt,\n extractOption: opt,\n modelConfig,\n pageDescription,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(parseResult),\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'extract',\n userQuery: {\n dataDemand,\n },\n matchedElement: [],\n data: null,\n taskInfo,\n error: errorLog,\n };\n\n const { data, thought } = parseResult || {};\n\n // 4\n const dump = createInsightDump({\n ...dumpData,\n data,\n });\n\n if (errorLog && !data && !opt?.doNotThrowError) {\n throw new InsightError(errorLog, dump);\n }\n\n return {\n data,\n thought,\n usage,\n dump,\n };\n }\n\n async describe(\n target: Rect | [number, number],\n modelConfig: IModelConfig,\n opt?: {\n deepThink?: boolean;\n },\n ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n assert(target, 'target is required for insight.describe');\n const context = await this.contextRetrieverFn();\n const { screenshotBase64, size } = context;\n assert(screenshotBase64, 'screenshot is required for insight.describe');\n // The result of the \"describe\" function will be used for positioning, so essentially it is a form of grounding.\n const { vlMode } = modelConfig;\n const systemPrompt = elementDescriberInstruction();\n\n // Convert [x,y] center point to Rect if needed\n const defaultRectSize = 30;\n const targetRect: Rect = Array.isArray(target)\n ? {\n left: Math.floor(target[0] - defaultRectSize / 2),\n top: Math.floor(target[1] - defaultRectSize / 2),\n width: defaultRectSize,\n height: defaultRectSize,\n }\n : target;\n\n let imagePayload = await compositeElementInfoImg({\n inputImgBase64: screenshotBase64,\n size,\n elementsPositionInfo: [\n {\n rect: targetRect,\n },\n ],\n borderThickness: 3,\n });\n\n if (opt?.deepThink) {\n const searchArea = expandSearchArea(targetRect, context.size, vlMode);\n debug('describe: set searchArea', searchArea);\n const croppedResult = await cropByRect(\n imagePayload,\n searchArea,\n vlMode === 'qwen-vl',\n );\n imagePayload = croppedResult.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 },\n ];\n\n const callAIFn = this\n .aiVendorFn as typeof callAIWithObjectResponse<AIDescribeElementResponse>;\n\n const res = await callAIFn(\n msgs,\n AIActionType.DESCRIBE_ELEMENT,\n modelConfig,\n );\n\n const { content } = res;\n assert(!content.error, `describe failed: ${content.error}`);\n assert(content.description, 'failed to describe the element');\n return content;\n }\n}\n"],"names":["debug","getDebug","Insight","query","opt","modelConfig","_parseResult_errors","queryPrompt","assert","globalDeepThinkSwitch","globalConfigManager","MIDSCENE_FORCE_DEEP_THINK","searchAreaPrompt","vlMode","console","undefined","context","searchArea","searchAreaRawResponse","searchAreaUsage","searchAreaResponse","AiLocateSection","startTime","Date","parseResult","rect","rawResponse","usage","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","elements","dump","createInsightDump","InsightError","dataDemand","pageDescription","multimodalPrompt","AiExtractElementInfo","data","thought","target","screenshotBase64","size","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","croppedResult","cropByRect","msgs","callAIFn","res","AIActionType","content","callAIWithObjectResponse","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;;AA8CA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IA4BnB,MAAM,OACJC,KAA0B,EAC1BC,GAAe,EACfC,WAAyB,EACM;YAyE3BC;QAxEJ,MAAMC,cAAc,AAAiB,YAAjB,OAAOJ,QAAqBA,QAAQA,MAAM,MAAM;QACpEK,OAAOD,aAAa;QAEpBC,OAAO,AAAiB,YAAjB,OAAOL,OAAoB;QAElC,MAAMM,wBAAwBC,oBAAoB,qBAAqB,CACrEC;QAEF,IAAIF,uBACFT,MAAM,yBAAyBS;QAEjC,IAAIG;QACJ,IAAIT,MAAM,SAAS,IAAIM,uBACrBG,mBAAmBT,MAAM,MAAM;QAGjC,MAAM,EAAEU,MAAM,EAAE,GAAGR;QAEnB,IAAIO,oBAAoB,CAACC,QAAQ;YAC/BC,QAAQ,IAAI,CACV;YAEFF,mBAAmBG;QACrB;QAEA,MAAMC,UAAUZ,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAM,MAAM,IAAI,CAAC,kBAAkB;QAE9D,IAAIa;QACJ,IAAIC;QACJ,IAAIC;QACJ,IAAIC;QAGJ,IAAIR,kBAAkB;YACpBQ,qBAAqB,MAAMC,gBAAgB;gBACzCL;gBACA,oBAAoBJ;gBACpBP;YACF;YACAG,OACEY,mBAAmB,IAAI,EACvB,CAAC,6BAA6B,EAAER,iBAAiB,CAAC,EAChDQ,mBAAmB,KAAK,GAAG,CAAC,EAAE,EAAEA,mBAAmB,KAAK,EAAE,GAAG,IAC7D;YAEJF,wBAAwBE,mBAAmB,WAAW;YACtDD,kBAAkBC,mBAAmB,KAAK;YAC1CH,aAAaG,mBAAmB,IAAI;QACtC;QAEA,MAAME,YAAYC,KAAK,GAAG;QAC1B,MAAM,EAAEC,WAAW,EAAEC,IAAI,EAAEC,WAAW,EAAEC,KAAK,EAAE,GAAG,MAAMC,gBAAgB;YACtE,UAAU,IAAI,CAAC,UAAU;YACzBZ;YACA,0BAA0BT;YAC1B,cAAca;YACdf;QACF;QAEA,MAAMwB,WAAWN,KAAK,GAAG,KAAKD;QAC9B,MAAMQ,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACL;YAC5B,gBAAgBK,KAAK,SAAS,CAACP;YAC/BG;YACAV;YACAC;YACAC;QACF;QAEA,IAAIa;QACJ,IAAI,QAAA1B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B0B,WAAW,CAAC,6BAA6B,EAAER,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG5E,MAAMS,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAAS1B;YACX;YACA,gBAAgB,EAAE;YAClB,aAAakB;YACb,MAAM;YACNK;YACA,WAAW,CAAC,CAACb;YACb,OAAOe;QACT;QAEA,MAAME,WAAWV,YAAY,QAAQ,IAAI,EAAE;QAE3C,MAAMW,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACX,gBAAgBC;QAClB;QAEA,IAAIF,UACF,MAAM,IAAIK,aAAaL,UAAUG;QAGnC,IAAID,SAAS,MAAM,GAAG,GACpB,MAAM,IAAIG,aACR,CAAC,0CAA0C,EAAEH,SAAS,MAAM,EAAE,EAC9DC;QAIJ,IAAID,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,SAAS;gBACP,IAAIA,QAAQ,CAAC,EAAE,CAAE,EAAE;gBACnB,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM;gBAC3B,MAAMA,QAAQ,CAAC,EAAE,CAAE,IAAI;gBACvB,kBAAkBA,QAAQ,CAAC,EAAE,CAAE,gBAAgB;YACjD;YACAT;YACAU;QACF;QAGF,OAAO;YACL,SAAS;YACTV;YACAU;QACF;IACF;IAEA,MAAM,QACJG,UAA+B,EAC/BjC,WAAyB,EACzBD,GAA0B,EAC1BmC,eAAwB,EACxBC,gBAAoC,EACF;YA0B9BlC;QAzBJE,OACE,AAAsB,YAAtB,OAAO8B,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAEvE,MAAMtB,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAE7C,MAAMM,YAAYC,KAAK,GAAG;QAE1B,MAAM,EAAEC,WAAW,EAAEG,KAAK,EAAE,GAAG,MAAMc,qBAAwB;YAC3DzB;YACA,WAAWsB;YACXE;YACA,eAAepC;YACfC;YACAkC;QACF;QAEA,MAAMV,WAAWN,KAAK,GAAG,KAAKD;QAC9B,MAAMQ,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACP;QAC9B;QAEA,IAAIQ;QACJ,IAAI,QAAA1B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B0B,WAAW,CAAC,qBAAqB,EAAER,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMS,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTK;YACF;YACA,gBAAgB,EAAE;YAClB,MAAM;YACNR;YACA,OAAOE;QACT;QAEA,MAAM,EAAEU,IAAI,EAAEC,OAAO,EAAE,GAAGnB,eAAe,CAAC;QAG1C,MAAMW,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACXS;QACF;QAEA,IAAIV,YAAY,CAACU,QAAQ,CAACtC,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,AAAD,GAC3C,MAAM,IAAIiC,aAAaL,UAAUG;QAGnC,OAAO;YACLO;YACAC;YACAhB;YACAQ;QACF;IACF;IAEA,MAAM,SACJS,MAA+B,EAC/BvC,WAAyB,EACzBD,GAEC,EACwD;QACzDI,OAAOoC,QAAQ;QACf,MAAM5B,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAC7C,MAAM,EAAE6B,gBAAgB,EAAEC,IAAI,EAAE,GAAG9B;QACnCR,OAAOqC,kBAAkB;QAEzB,MAAM,EAAEhC,MAAM,EAAE,GAAGR;QACnB,MAAM0C,eAAeC;QAGrB,MAAMC,kBAAkB;QACxB,MAAMC,aAAmBC,MAAM,OAAO,CAACP,UACnC;YACE,MAAMQ,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC/C,KAAKG,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC9C,OAAOA;YACP,QAAQA;QACV,IACAL;QAEJ,IAAIS,eAAe,MAAMC,wBAAwB;YAC/C,gBAAgBT;YAChBC;YACA,sBAAsB;gBACpB;oBACE,MAAMI;gBACR;aACD;YACD,iBAAiB;QACnB;QAEA,IAAI9C,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,EAAE;YAClB,MAAMa,aAAasC,iBAAiBL,YAAYlC,QAAQ,IAAI,EAAEH;YAC9Db,MAAM,4BAA4BiB;YAClC,MAAMuC,gBAAgB,MAAMC,WAC1BJ,cACApC,YACAJ,AAAW,cAAXA;YAEFwC,eAAeG,cAAc,WAAW;QAC1C;QAEA,MAAME,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,WAAW,IAAI,CAClB,UAAU;QAEb,MAAMC,MAAM,MAAMD,SAChBD,MACAG,aAAa,gBAAgB,EAC7BxD;QAGF,MAAM,EAAEyD,OAAO,EAAE,GAAGF;QACpBpD,OAAO,CAACsD,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1DtD,OAAOsD,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IAxSA,YACE9C,OAA2D,EAC3DZ,GAAoB,CACpB;QAVF;QAEA,qCACE2D;QAEF;QAMEvD,OAAOQ,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMgD,QAAQ,OAAO,CAAChD;QAIlD,IAAI,AAA2B,WAApBZ,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,UAAU,AAAD,GACvB,IAAI,CAAC,UAAU,GAAGA,IAAI,UAAU;QAElC,IAAI,AAAyB,WAAlBA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,QAAQ,AAAD,GACrB,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AAuRF"}
|
|
1
|
+
{"version":3,"file":"service/index.mjs","sources":["webpack://@midscene/core/./src/service/index.ts"],"sourcesContent":["import { AIActionType, type AIArgs, expandSearchArea } from '@/ai-model/common';\nimport {\n AiExtractElementInfo,\n AiLocateElement,\n callAIWithObjectResponse,\n} from '@/ai-model/index';\nimport { AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n AIDescribeElementResponse,\n AIUsageInfo,\n DetailedLocateParam,\n LocateResultWithDump,\n PartialServiceDumpFromSDK,\n Rect,\n ServiceExtractOption,\n ServiceExtractParam,\n ServiceExtractResult,\n ServiceTaskInfo,\n UIContext,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport {\n type IModelConfig,\n MIDSCENE_FORCE_DEEP_THINK,\n globalConfigManager,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TMultimodalPrompt } from '../ai-model/common';\nimport { createServiceDump } from './utils';\n\nexport interface LocateOpts {\n context?: UIContext;\n}\n\nexport type AnyValue<T> = {\n [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\ninterface ServiceOptions {\n taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n aiVendorFn?: typeof callAIWithObjectResponse;\n}\n\nconst debug = getDebug('ai:service');\nexport default class Service {\n contextRetrieverFn: () => Promise<UIContext> | UIContext;\n\n aiVendorFn: Exclude<ServiceOptions['aiVendorFn'], undefined> =\n callAIWithObjectResponse;\n\n taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n\n constructor(\n context: UIContext | (() => Promise<UIContext> | UIContext),\n opt?: ServiceOptions,\n ) {\n assert(context, 'context is required for Service');\n if (typeof context === 'function') {\n this.contextRetrieverFn = context;\n } else {\n this.contextRetrieverFn = () => Promise.resolve(context);\n }\n\n // just for unit test, aiVendorFn is callAIWithObjectResponse by default\n if (typeof opt?.aiVendorFn !== 'undefined') {\n this.aiVendorFn = opt.aiVendorFn;\n }\n if (typeof opt?.taskInfo !== 'undefined') {\n this.taskInfo = opt.taskInfo;\n }\n }\n\n async locate(\n query: DetailedLocateParam,\n opt: LocateOpts,\n modelConfig: IModelConfig,\n ): Promise<LocateResultWithDump> {\n const queryPrompt = typeof query === 'string' ? query : query.prompt;\n assert(queryPrompt, 'query is required for locate');\n\n assert(typeof query === 'object', 'query should be an object for locate');\n\n const globalDeepThinkSwitch = globalConfigManager.getEnvConfigInBoolean(\n MIDSCENE_FORCE_DEEP_THINK,\n );\n if (globalDeepThinkSwitch) {\n debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n }\n let searchAreaPrompt;\n if (query.deepThink || globalDeepThinkSwitch) {\n searchAreaPrompt = query.prompt;\n }\n\n const { vlMode } = modelConfig;\n\n if (searchAreaPrompt && !vlMode) {\n console.warn(\n 'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n );\n searchAreaPrompt = undefined;\n }\n\n const context = opt?.context || (await this.contextRetrieverFn());\n\n let searchArea: Rect | undefined = undefined;\n let searchAreaRawResponse: string | undefined = undefined;\n let searchAreaUsage: AIUsageInfo | undefined = undefined;\n let searchAreaResponse:\n | Awaited<ReturnType<typeof AiLocateSection>>\n | undefined = undefined;\n if (searchAreaPrompt) {\n searchAreaResponse = await AiLocateSection({\n context,\n sectionDescription: searchAreaPrompt,\n modelConfig,\n });\n assert(\n searchAreaResponse.rect,\n `cannot find search area for \"${searchAreaPrompt}\"${\n searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n }`,\n );\n searchAreaRawResponse = searchAreaResponse.rawResponse;\n searchAreaUsage = searchAreaResponse.usage;\n searchArea = searchAreaResponse.rect;\n }\n\n const startTime = Date.now();\n const { parseResult, rect, rawResponse, usage } = await AiLocateElement({\n callAIFn: this.aiVendorFn,\n context,\n targetElementDescription: queryPrompt,\n searchConfig: searchAreaResponse,\n modelConfig,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: ServiceTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(rawResponse),\n formatResponse: JSON.stringify(parseResult),\n usage,\n searchArea,\n searchAreaRawResponse,\n searchAreaUsage,\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialServiceDumpFromSDK = {\n type: 'locate',\n userQuery: {\n element: queryPrompt,\n },\n matchedElement: [],\n matchedRect: rect,\n data: null,\n taskInfo,\n deepThink: !!searchArea,\n error: errorLog,\n };\n\n const elements = parseResult.elements || [];\n\n const dump = createServiceDump({\n ...dumpData,\n matchedElement: elements,\n });\n\n if (errorLog) {\n throw new ServiceError(errorLog, dump);\n }\n\n if (elements.length > 1) {\n throw new ServiceError(\n `locate: multiple elements found, length = ${elements.length}`,\n dump,\n );\n }\n\n if (elements.length === 1) {\n return {\n element: {\n id: elements[0]!.id,\n center: elements[0]!.center,\n rect: elements[0]!.rect,\n isOrderSensitive: elements[0]!.isOrderSensitive,\n },\n rect,\n dump,\n };\n }\n\n return {\n element: null,\n rect,\n dump,\n };\n }\n\n async extract<T>(\n dataDemand: ServiceExtractParam,\n modelConfig: IModelConfig,\n opt?: ServiceExtractOption,\n pageDescription?: string,\n multimodalPrompt?: TMultimodalPrompt,\n ): Promise<ServiceExtractResult<T>> {\n assert(\n typeof dataDemand === 'object' || typeof dataDemand === 'string',\n `dataDemand should be object or string, but get ${typeof dataDemand}`,\n );\n const context = await this.contextRetrieverFn();\n\n const startTime = Date.now();\n\n const { parseResult, usage } = await AiExtractElementInfo<T>({\n context,\n dataQuery: dataDemand,\n multimodalPrompt,\n extractOption: opt,\n modelConfig,\n pageDescription,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: ServiceTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(parseResult),\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialServiceDumpFromSDK = {\n type: 'extract',\n userQuery: {\n dataDemand,\n },\n matchedElement: [],\n data: null,\n taskInfo,\n error: errorLog,\n };\n\n const { data, thought } = parseResult || {};\n\n // 4\n const dump = createServiceDump({\n ...dumpData,\n data,\n });\n\n if (errorLog && !data && !opt?.doNotThrowError) {\n throw new ServiceError(errorLog, dump);\n }\n\n return {\n data,\n thought,\n usage,\n dump,\n };\n }\n\n async describe(\n target: Rect | [number, number],\n modelConfig: IModelConfig,\n opt?: {\n deepThink?: boolean;\n },\n ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n assert(target, 'target is required for service.describe');\n const context = await this.contextRetrieverFn();\n const { screenshotBase64, size } = context;\n assert(screenshotBase64, 'screenshot is required for service.describe');\n // The result of the \"describe\" function will be used for positioning, so essentially it is a form of grounding.\n const { vlMode } = modelConfig;\n const systemPrompt = elementDescriberInstruction();\n\n // Convert [x,y] center point to Rect if needed\n const defaultRectSize = 30;\n const targetRect: Rect = Array.isArray(target)\n ? {\n left: Math.floor(target[0] - defaultRectSize / 2),\n top: Math.floor(target[1] - defaultRectSize / 2),\n width: defaultRectSize,\n height: defaultRectSize,\n }\n : target;\n\n let imagePayload = await compositeElementInfoImg({\n inputImgBase64: screenshotBase64,\n size,\n elementsPositionInfo: [\n {\n rect: targetRect,\n },\n ],\n borderThickness: 3,\n });\n\n if (opt?.deepThink) {\n const searchArea = expandSearchArea(targetRect, context.size, vlMode);\n debug('describe: set searchArea', searchArea);\n const croppedResult = await cropByRect(\n imagePayload,\n searchArea,\n vlMode === 'qwen-vl',\n );\n imagePayload = croppedResult.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 },\n ];\n\n const callAIFn = this\n .aiVendorFn as typeof callAIWithObjectResponse<AIDescribeElementResponse>;\n\n const res = await callAIFn(\n msgs,\n AIActionType.DESCRIBE_ELEMENT,\n modelConfig,\n );\n\n const { content } = res;\n assert(!content.error, `describe failed: ${content.error}`);\n assert(content.description, 'failed to describe the element');\n return content;\n }\n}\n"],"names":["debug","getDebug","Service","query","opt","modelConfig","_parseResult_errors","queryPrompt","assert","globalDeepThinkSwitch","globalConfigManager","MIDSCENE_FORCE_DEEP_THINK","searchAreaPrompt","vlMode","console","undefined","context","searchArea","searchAreaRawResponse","searchAreaUsage","searchAreaResponse","AiLocateSection","startTime","Date","parseResult","rect","rawResponse","usage","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","elements","dump","createServiceDump","ServiceError","dataDemand","pageDescription","multimodalPrompt","AiExtractElementInfo","data","thought","target","screenshotBase64","size","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","croppedResult","cropByRect","msgs","callAIFn","res","AIActionType","content","callAIWithObjectResponse","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;;AA8CA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IA4BnB,MAAM,OACJC,KAA0B,EAC1BC,GAAe,EACfC,WAAyB,EACM;YAyE3BC;QAxEJ,MAAMC,cAAc,AAAiB,YAAjB,OAAOJ,QAAqBA,QAAQA,MAAM,MAAM;QACpEK,OAAOD,aAAa;QAEpBC,OAAO,AAAiB,YAAjB,OAAOL,OAAoB;QAElC,MAAMM,wBAAwBC,oBAAoB,qBAAqB,CACrEC;QAEF,IAAIF,uBACFT,MAAM,yBAAyBS;QAEjC,IAAIG;QACJ,IAAIT,MAAM,SAAS,IAAIM,uBACrBG,mBAAmBT,MAAM,MAAM;QAGjC,MAAM,EAAEU,MAAM,EAAE,GAAGR;QAEnB,IAAIO,oBAAoB,CAACC,QAAQ;YAC/BC,QAAQ,IAAI,CACV;YAEFF,mBAAmBG;QACrB;QAEA,MAAMC,UAAUZ,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAM,MAAM,IAAI,CAAC,kBAAkB;QAE9D,IAAIa;QACJ,IAAIC;QACJ,IAAIC;QACJ,IAAIC;QAGJ,IAAIR,kBAAkB;YACpBQ,qBAAqB,MAAMC,gBAAgB;gBACzCL;gBACA,oBAAoBJ;gBACpBP;YACF;YACAG,OACEY,mBAAmB,IAAI,EACvB,CAAC,6BAA6B,EAAER,iBAAiB,CAAC,EAChDQ,mBAAmB,KAAK,GAAG,CAAC,EAAE,EAAEA,mBAAmB,KAAK,EAAE,GAAG,IAC7D;YAEJF,wBAAwBE,mBAAmB,WAAW;YACtDD,kBAAkBC,mBAAmB,KAAK;YAC1CH,aAAaG,mBAAmB,IAAI;QACtC;QAEA,MAAME,YAAYC,KAAK,GAAG;QAC1B,MAAM,EAAEC,WAAW,EAAEC,IAAI,EAAEC,WAAW,EAAEC,KAAK,EAAE,GAAG,MAAMC,gBAAgB;YACtE,UAAU,IAAI,CAAC,UAAU;YACzBZ;YACA,0BAA0BT;YAC1B,cAAca;YACdf;QACF;QAEA,MAAMwB,WAAWN,KAAK,GAAG,KAAKD;QAC9B,MAAMQ,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACL;YAC5B,gBAAgBK,KAAK,SAAS,CAACP;YAC/BG;YACAV;YACAC;YACAC;QACF;QAEA,IAAIa;QACJ,IAAI,QAAA1B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B0B,WAAW,CAAC,6BAA6B,EAAER,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG5E,MAAMS,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAAS1B;YACX;YACA,gBAAgB,EAAE;YAClB,aAAakB;YACb,MAAM;YACNK;YACA,WAAW,CAAC,CAACb;YACb,OAAOe;QACT;QAEA,MAAME,WAAWV,YAAY,QAAQ,IAAI,EAAE;QAE3C,MAAMW,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACX,gBAAgBC;QAClB;QAEA,IAAIF,UACF,MAAM,IAAIK,aAAaL,UAAUG;QAGnC,IAAID,SAAS,MAAM,GAAG,GACpB,MAAM,IAAIG,aACR,CAAC,0CAA0C,EAAEH,SAAS,MAAM,EAAE,EAC9DC;QAIJ,IAAID,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,SAAS;gBACP,IAAIA,QAAQ,CAAC,EAAE,CAAE,EAAE;gBACnB,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM;gBAC3B,MAAMA,QAAQ,CAAC,EAAE,CAAE,IAAI;gBACvB,kBAAkBA,QAAQ,CAAC,EAAE,CAAE,gBAAgB;YACjD;YACAT;YACAU;QACF;QAGF,OAAO;YACL,SAAS;YACTV;YACAU;QACF;IACF;IAEA,MAAM,QACJG,UAA+B,EAC/BjC,WAAyB,EACzBD,GAA0B,EAC1BmC,eAAwB,EACxBC,gBAAoC,EACF;YA0B9BlC;QAzBJE,OACE,AAAsB,YAAtB,OAAO8B,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAEvE,MAAMtB,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAE7C,MAAMM,YAAYC,KAAK,GAAG;QAE1B,MAAM,EAAEC,WAAW,EAAEG,KAAK,EAAE,GAAG,MAAMc,qBAAwB;YAC3DzB;YACA,WAAWsB;YACXE;YACA,eAAepC;YACfC;YACAkC;QACF;QAEA,MAAMV,WAAWN,KAAK,GAAG,KAAKD;QAC9B,MAAMQ,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACP;QAC9B;QAEA,IAAIQ;QACJ,IAAI,QAAA1B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B0B,WAAW,CAAC,qBAAqB,EAAER,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMS,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTK;YACF;YACA,gBAAgB,EAAE;YAClB,MAAM;YACNR;YACA,OAAOE;QACT;QAEA,MAAM,EAAEU,IAAI,EAAEC,OAAO,EAAE,GAAGnB,eAAe,CAAC;QAG1C,MAAMW,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACXS;QACF;QAEA,IAAIV,YAAY,CAACU,QAAQ,CAACtC,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,AAAD,GAC3C,MAAM,IAAIiC,aAAaL,UAAUG;QAGnC,OAAO;YACLO;YACAC;YACAhB;YACAQ;QACF;IACF;IAEA,MAAM,SACJS,MAA+B,EAC/BvC,WAAyB,EACzBD,GAEC,EACwD;QACzDI,OAAOoC,QAAQ;QACf,MAAM5B,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAC7C,MAAM,EAAE6B,gBAAgB,EAAEC,IAAI,EAAE,GAAG9B;QACnCR,OAAOqC,kBAAkB;QAEzB,MAAM,EAAEhC,MAAM,EAAE,GAAGR;QACnB,MAAM0C,eAAeC;QAGrB,MAAMC,kBAAkB;QACxB,MAAMC,aAAmBC,MAAM,OAAO,CAACP,UACnC;YACE,MAAMQ,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC/C,KAAKG,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC9C,OAAOA;YACP,QAAQA;QACV,IACAL;QAEJ,IAAIS,eAAe,MAAMC,wBAAwB;YAC/C,gBAAgBT;YAChBC;YACA,sBAAsB;gBACpB;oBACE,MAAMI;gBACR;aACD;YACD,iBAAiB;QACnB;QAEA,IAAI9C,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,EAAE;YAClB,MAAMa,aAAasC,iBAAiBL,YAAYlC,QAAQ,IAAI,EAAEH;YAC9Db,MAAM,4BAA4BiB;YAClC,MAAMuC,gBAAgB,MAAMC,WAC1BJ,cACApC,YACAJ,AAAW,cAAXA;YAEFwC,eAAeG,cAAc,WAAW;QAC1C;QAEA,MAAME,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,WAAW,IAAI,CAClB,UAAU;QAEb,MAAMC,MAAM,MAAMD,SAChBD,MACAG,aAAa,gBAAgB,EAC7BxD;QAGF,MAAM,EAAEyD,OAAO,EAAE,GAAGF;QACpBpD,OAAO,CAACsD,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1DtD,OAAOsD,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IAxSA,YACE9C,OAA2D,EAC3DZ,GAAoB,CACpB;QAVF;QAEA,qCACE2D;QAEF;QAMEvD,OAAOQ,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMgD,QAAQ,OAAO,CAAChD;QAIlD,IAAI,AAA2B,WAApBZ,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,UAAU,AAAD,GACvB,IAAI,CAAC,UAAU,GAAGA,IAAI,UAAU;QAElC,IAAI,AAAyB,WAAlBA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,QAAQ,AAAD,GACrB,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AAuRF"}
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
import { uuid } from "@midscene/shared/utils";
|
|
2
|
-
function
|
|
2
|
+
function createServiceDump(data) {
|
|
3
3
|
const baseData = {
|
|
4
4
|
logTime: Date.now()
|
|
5
5
|
};
|
|
@@ -10,6 +10,6 @@ function createInsightDump(data) {
|
|
|
10
10
|
};
|
|
11
11
|
return finalData;
|
|
12
12
|
}
|
|
13
|
-
export {
|
|
13
|
+
export { createServiceDump };
|
|
14
14
|
|
|
15
15
|
//# sourceMappingURL=utils.mjs.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"service/utils.mjs","sources":["webpack://@midscene/core/./src/service/utils.ts"],"sourcesContent":["import type { DumpMeta, PartialServiceDumpFromSDK, ServiceDump } from '@/types';\nimport { uuid } from '@midscene/shared/utils';\n\nexport function createServiceDump(\n data: PartialServiceDumpFromSDK,\n): ServiceDump {\n const baseData: DumpMeta = {\n logTime: Date.now(),\n };\n const finalData: ServiceDump = {\n logId: uuid(),\n ...baseData,\n ...data,\n };\n\n return finalData;\n}\n"],"names":["createServiceDump","data","baseData","Date","finalData","uuid"],"mappings":";AAGO,SAASA,kBACdC,IAA+B;IAE/B,MAAMC,WAAqB;QACzB,SAASC,KAAK,GAAG;IACnB;IACA,MAAMC,YAAyB;QAC7B,OAAOC;QACP,GAAGH,QAAQ;QACX,GAAGD,IAAI;IACT;IAEA,OAAOG;AACT"}
|
package/dist/es/task-runner.mjs
CHANGED
|
@@ -105,7 +105,7 @@ class TaskRunner {
|
|
|
105
105
|
uiContext
|
|
106
106
|
};
|
|
107
107
|
if ('Insight' === task.type) {
|
|
108
|
-
assert('Locate' === task.subType || 'Query' === task.subType || 'Assert' === task.subType || 'WaitFor' === task.subType || 'Boolean' === task.subType || 'Number' === task.subType || 'String' === task.subType, `unsupported
|
|
108
|
+
assert('Locate' === task.subType || 'Query' === task.subType || 'Assert' === task.subType || 'WaitFor' === task.subType || 'Boolean' === task.subType || 'Number' === task.subType || 'String' === task.subType, `unsupported service subType: ${task.subType}`);
|
|
109
109
|
returnValue = await task.executor(param, executorContext);
|
|
110
110
|
if ('Locate' === task.subType) previousFindOutput = null == returnValue ? void 0 : returnValue.output;
|
|
111
111
|
} else if ('Action' === task.type || 'Planning' === task.type) returnValue = await task.executor(param, executorContext);
|