@donggui/core 1.5.14 → 1.6.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (36) hide show
  1. package/dist/es/agent/agent.mjs +187 -10
  2. package/dist/es/agent/agent.mjs.map +1 -1
  3. package/dist/es/agent/task-builder.mjs +46 -2
  4. package/dist/es/agent/task-builder.mjs.map +1 -1
  5. package/dist/es/agent/utils.mjs +1 -1
  6. package/dist/es/ai-model/assert.mjs +415 -0
  7. package/dist/es/ai-model/assert.mjs.map +1 -0
  8. package/dist/es/ai-model/index.mjs +2 -1
  9. package/dist/es/ai-model/prompt/cache-verify.mjs +34 -0
  10. package/dist/es/ai-model/prompt/cache-verify.mjs.map +1 -0
  11. package/dist/es/service/index.mjs +81 -6
  12. package/dist/es/service/index.mjs.map +1 -1
  13. package/dist/es/types.mjs.map +1 -1
  14. package/dist/es/utils.mjs +2 -2
  15. package/dist/lib/agent/agent.js +185 -8
  16. package/dist/lib/agent/agent.js.map +1 -1
  17. package/dist/lib/agent/task-builder.js +46 -2
  18. package/dist/lib/agent/task-builder.js.map +1 -1
  19. package/dist/lib/agent/utils.js +1 -1
  20. package/dist/lib/ai-model/assert.js +455 -0
  21. package/dist/lib/ai-model/assert.js.map +1 -0
  22. package/dist/lib/ai-model/index.js +18 -11
  23. package/dist/lib/ai-model/prompt/cache-verify.js +68 -0
  24. package/dist/lib/ai-model/prompt/cache-verify.js.map +1 -0
  25. package/dist/lib/service/index.js +81 -6
  26. package/dist/lib/service/index.js.map +1 -1
  27. package/dist/lib/types.js +3 -3
  28. package/dist/lib/types.js.map +1 -1
  29. package/dist/lib/utils.js +2 -2
  30. package/dist/types/agent/agent.d.ts +38 -2
  31. package/dist/types/ai-model/assert.d.ts +66 -0
  32. package/dist/types/ai-model/index.d.ts +2 -0
  33. package/dist/types/ai-model/prompt/cache-verify.d.ts +1 -0
  34. package/dist/types/service/index.d.ts +3 -2
  35. package/dist/types/types.d.ts +166 -0
  36. package/package.json +25 -44
@@ -1 +1 @@
1
- {"version":3,"file":"agent/task-builder.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../src/agent/task-builder.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import { findAllMidsceneLocatorField, parseActionParam } from '@/ai-model';\nimport type { AbstractInterface } from '@/device';\nimport type Service from '@/service';\nimport { setTimingFieldOnce } from '@/task-timing';\nimport type {\n DetailedLocateParam,\n DeviceAction,\n ElementCacheFeature,\n ExecutionTaskActionApply,\n ExecutionTaskApply,\n ExecutionTaskHitBy,\n ExecutionTaskPlanningLocateApply,\n LocateResultElement,\n LocateResultWithDump,\n PlanningAction,\n PlanningLocateParam,\n Rect,\n ServiceDump,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport { sleep } from '@/utils';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { generateElementByRect } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TaskCache } from './task-cache';\nimport {\n ifPlanLocateParamIsBbox,\n matchElementFromCache,\n matchElementFromPlan,\n transformLogicalElementToScreenshot,\n transformLogicalRectToScreenshotRect,\n} from './utils';\n\nconst debug = getDebug('agent:task-builder');\n\n/**\n * Check if a cache object is non-empty\n */\nfunction hasNonEmptyCache(cache: unknown): boolean {\n return (\n cache !== null &&\n cache !== undefined &&\n typeof cache === 'object' &&\n Object.keys(cache).length > 0\n );\n}\n\nexport function locatePlanForLocate(param: string | DetailedLocateParam) {\n const locate = typeof param === 'string' ? { prompt: param } : param;\n const locatePlan: PlanningAction<PlanningLocateParam> = {\n type: 'Locate',\n param: locate,\n thought: '',\n };\n return locatePlan;\n}\n\ninterface TaskBuilderDeps {\n interfaceInstance: AbstractInterface;\n service: Service;\n taskCache?: TaskCache;\n actionSpace: DeviceAction[];\n waitAfterAction?: number;\n}\n\ninterface BuildOptions {\n cacheable?: boolean;\n deepLocate?: boolean;\n abortSignal?: AbortSignal;\n}\n\ninterface PlanBuildContext {\n tasks: ExecutionTaskApply[];\n modelConfigForPlanning: IModelConfig;\n modelConfigForDefaultIntent: IModelConfig;\n cacheable?: boolean;\n deepLocate?: boolean;\n abortSignal?: AbortSignal;\n}\n\nexport class TaskBuilder {\n private readonly interface: AbstractInterface;\n\n private readonly service: Service;\n\n private readonly taskCache?: TaskCache;\n\n private readonly actionSpace: DeviceAction[];\n\n private readonly waitAfterAction?: number;\n\n constructor({\n interfaceInstance,\n service,\n taskCache,\n actionSpace,\n waitAfterAction,\n }: TaskBuilderDeps) {\n this.interface = interfaceInstance;\n this.service = service;\n this.taskCache = taskCache;\n this.actionSpace = actionSpace;\n this.waitAfterAction = waitAfterAction;\n }\n\n public async build(\n plans: PlanningAction[],\n modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n options?: BuildOptions,\n ): Promise<{ tasks: ExecutionTaskApply[] }> {\n const tasks: ExecutionTaskApply[] = [];\n const cacheable = options?.cacheable;\n\n const context: PlanBuildContext = {\n tasks,\n modelConfigForPlanning,\n modelConfigForDefaultIntent,\n cacheable,\n deepLocate: options?.deepLocate,\n abortSignal: options?.abortSignal,\n };\n\n type PlanHandler = (plan: PlanningAction) => Promise<void> | void;\n\n const planHandlers = new Map<string, PlanHandler>([\n [\n 'Locate',\n (plan) =>\n this.handleLocatePlan(\n plan as PlanningAction<PlanningLocateParam>,\n context,\n ),\n ],\n ['Finished', (plan) => this.handleFinishedPlan(plan, context)],\n ]);\n\n const defaultHandler: PlanHandler = (plan) =>\n this.handleActionPlan(plan, context);\n\n for (const plan of plans) {\n const handler = planHandlers.get(plan.type) ?? defaultHandler;\n await handler(plan);\n }\n\n return {\n tasks,\n };\n }\n\n private handleFinishedPlan(\n plan: PlanningAction,\n context: PlanBuildContext,\n ): void {\n const taskActionFinished: ExecutionTaskActionApply<null> = {\n type: 'Action Space',\n subType: 'Finished',\n param: null,\n thought: plan.thought,\n executor: async () => {},\n };\n context.tasks.push(taskActionFinished);\n }\n\n private async handleLocatePlan(\n plan: PlanningAction<PlanningLocateParam>,\n context: PlanBuildContext,\n ): Promise<void> {\n const taskLocate = this.createLocateTask(plan, plan.param, context);\n context.tasks.push(taskLocate);\n }\n\n private async handleActionPlan(\n plan: PlanningAction,\n context: PlanBuildContext,\n ): Promise<void> {\n const planType = plan.type;\n const actionSpace = this.actionSpace;\n const action = actionSpace.find((item) => item.name === planType);\n const param = plan.param;\n\n if (!action) {\n throw new Error(`Action type '${planType}' not found`);\n }\n\n const locateFields = action\n ? findAllMidsceneLocatorField(action.paramSchema)\n : [];\n\n const requiredLocateFields = action\n ? findAllMidsceneLocatorField(action.paramSchema, true)\n : [];\n\n locateFields.forEach((field) => {\n if (param[field]) {\n // Always use createLocateTask for all locate params (including bbox)\n // This ensures cache writing happens even when bbox is available\n const locatePlan = locatePlanForLocate(param[field]);\n debug(\n 'will prepend locate param for field',\n `action.type=${planType}`,\n `param=${JSON.stringify(param[field])}`,\n `locatePlan=${JSON.stringify(locatePlan)}`,\n `hasBbox=${ifPlanLocateParamIsBbox(param[field])}`,\n );\n const locateTask = this.createLocateTask(\n locatePlan,\n param[field],\n context,\n (result) => {\n param[field] = result;\n },\n );\n context.tasks.push(locateTask);\n } else {\n assert(\n !requiredLocateFields.includes(field),\n `Required locate field '${field}' is not provided for action ${planType}`,\n );\n debug(`field '${field}' is not provided for action ${planType}`);\n }\n });\n\n const task: ExecutionTaskApply<\n 'Action Space',\n any,\n { success: boolean; action: string; param: any },\n void\n > = {\n type: 'Action Space',\n subType: planType,\n thought: plan.thought,\n param: plan.param,\n executor: async (param, taskContext) => {\n const timing = taskContext.task.timing;\n\n debug(\n 'executing action',\n planType,\n param,\n `taskContext.element.center: ${taskContext.element?.center}`,\n );\n\n const uiContext = taskContext.uiContext;\n assert(uiContext, 'uiContext is required for Action task');\n\n requiredLocateFields.forEach((field) => {\n assert(\n param[field],\n `field '${field}' is required for action ${planType} but not provided. Cannot execute action ${planType}.`,\n );\n });\n\n setTimingFieldOnce(timing, 'beforeInvokeActionHookStart');\n try {\n await Promise.all([\n (async () => {\n if (this.interface.beforeInvokeAction) {\n debug(\n `will call \"beforeInvokeAction\" for interface with action name ${action.name}`,\n );\n await this.interface.beforeInvokeAction(action.name, param);\n debug(\n `called \"beforeInvokeAction\" for interface with action name ${action.name}`,\n );\n }\n })(),\n sleep(200),\n ]);\n } catch (originalError: any) {\n const originalMessage =\n originalError?.message || String(originalError);\n throw new Error(\n `error in running beforeInvokeAction for ${action.name}: ${originalMessage}`,\n { cause: originalError },\n );\n }\n setTimingFieldOnce(timing, 'beforeInvokeActionHookEnd');\n\n const { shrunkShotToLogicalRatio } = uiContext;\n if (shrunkShotToLogicalRatio === undefined) {\n throw new Error(\n 'shrunkShotToLogicalRatio is not defined in Action task',\n );\n }\n\n if (action.paramSchema) {\n try {\n param = parseActionParam(param, action.paramSchema, {\n shrunkShotToLogicalRatio,\n });\n } catch (error: any) {\n throw new Error(\n `Invalid parameters for action ${action.name}: ${error.message}\\nParameters: ${JSON.stringify(param)}`,\n { cause: error },\n );\n }\n }\n\n setTimingFieldOnce(timing, 'callActionStart');\n\n debug('calling action', action.name);\n const actionFn = action.call.bind(this.interface);\n const actionResult = await actionFn(param, taskContext);\n setTimingFieldOnce(timing, 'callActionEnd');\n debug('called action', action.name, 'result:', actionResult);\n\n setTimingFieldOnce(timing, 'afterInvokeActionHookStart');\n\n const delayAfterRunner =\n action.delayAfterRunner ?? this.waitAfterAction ?? 300;\n if (delayAfterRunner > 0) {\n await sleep(delayAfterRunner);\n }\n\n try {\n if (this.interface.afterInvokeAction) {\n debug(\n `will call \"afterInvokeAction\" for interface with action name ${action.name}`,\n );\n await this.interface.afterInvokeAction(action.name, param);\n debug(\n `called \"afterInvokeAction\" for interface with action name ${action.name}`,\n );\n }\n } catch (originalError: any) {\n const originalMessage =\n originalError?.message || String(originalError);\n throw new Error(\n `error in running afterInvokeAction for ${action.name}: ${originalMessage}`,\n { cause: originalError },\n );\n }\n\n setTimingFieldOnce(timing, 'afterInvokeActionHookEnd');\n\n return {\n output: actionResult,\n };\n },\n };\n\n context.tasks.push(task);\n }\n\n private createLocateTask(\n plan: PlanningAction<PlanningLocateParam>,\n detailedLocateParam: DetailedLocateParam | string,\n context: PlanBuildContext,\n onResult?: (result: LocateResultElement) => void,\n ): ExecutionTaskPlanningLocateApply {\n const { cacheable, modelConfigForDefaultIntent, deepLocate, abortSignal } =\n context;\n\n let locateParam = detailedLocateParam;\n\n if (typeof locateParam === 'string') {\n locateParam = {\n prompt: locateParam,\n };\n }\n\n if (cacheable !== undefined) {\n locateParam = {\n ...locateParam,\n cacheable,\n };\n }\n\n if (deepLocate && !locateParam.deepLocate) {\n locateParam = {\n ...locateParam,\n deepLocate: true,\n };\n }\n\n const taskLocator: ExecutionTaskPlanningLocateApply = {\n type: 'Planning',\n subType: 'Locate',\n param: locateParam,\n thought: plan.thought,\n executor: async (param, taskContext) => {\n const { task } = taskContext;\n let { uiContext } = taskContext;\n\n assert(\n param?.prompt || param?.bbox,\n `No prompt or id or position or bbox to locate, param=${JSON.stringify(\n param,\n )}`,\n );\n\n if (!uiContext) {\n uiContext = await this.service.contextRetrieverFn();\n }\n\n assert(uiContext, 'uiContext is required for Service task');\n\n const { shrunkShotToLogicalRatio } = uiContext;\n\n if (shrunkShotToLogicalRatio === undefined) {\n throw new Error(\n 'shrunkShotToLogicalRatio is not defined in locate task',\n );\n }\n\n let locateDump: ServiceDump | undefined;\n let locateResult: LocateResultWithDump | undefined;\n\n const applyDump = (dump?: ServiceDump) => {\n if (!dump) {\n return;\n }\n locateDump = dump;\n task.log = {\n dump,\n rawResponse: dump.taskInfo?.rawResponse,\n };\n task.usage = dump.taskInfo?.usage;\n if (dump.taskInfo?.searchAreaUsage) {\n task.searchAreaUsage = dump.taskInfo.searchAreaUsage;\n }\n if (dump.taskInfo?.reasoning_content) {\n task.reasoning_content = dump.taskInfo.reasoning_content;\n }\n };\n\n // from bbox (plan hit)\n const elementFromBbox = ifPlanLocateParamIsBbox(param)\n ? matchElementFromPlan(param)\n : undefined;\n const isPlanHit = !!elementFromBbox;\n\n // from xpath\n let rectFromXpath: Rect | undefined;\n if (\n !isPlanHit &&\n param.xpath &&\n this.interface.rectMatchesCacheFeature\n ) {\n try {\n rectFromXpath = await this.interface.rectMatchesCacheFeature({\n xpaths: [param.xpath],\n });\n } catch {\n // xpath locate failed, allow fallback to cache or AI locate\n }\n }\n\n const elementFromXpath = rectFromXpath\n ? generateElementByRect(\n // rectFromXpath is in logical coordinates, which should be transformed to screenshot coordinates;\n transformLogicalRectToScreenshotRect(\n rectFromXpath,\n shrunkShotToLogicalRatio,\n ),\n typeof param.prompt === 'string'\n ? param.prompt\n : param.prompt?.prompt || '',\n )\n : undefined;\n\n const isXpathHit = !!elementFromXpath;\n\n const cachePrompt = param.prompt;\n const locateCacheRecord =\n await this.taskCache?.matchLocateCache(cachePrompt);\n const cacheEntry = locateCacheRecord?.cacheContent?.cache;\n\n const elementFromCacheResult =\n isPlanHit || isXpathHit\n ? null\n : await matchElementFromCache(\n {\n taskCache: this.taskCache,\n interfaceInstance: this.interface,\n },\n cacheEntry,\n cachePrompt,\n param.cacheable,\n );\n\n // elementFromCacheResult is in logical coordinates, which should be transformed to screenshot coordinates;\n const elementFromCache = elementFromCacheResult\n ? transformLogicalElementToScreenshot(\n elementFromCacheResult,\n shrunkShotToLogicalRatio,\n )\n : undefined;\n\n const isCacheHit = !!elementFromCache;\n\n let elementFromAiLocate: LocateResultElement | null | undefined;\n const timing = taskContext.task.timing;\n if (!isXpathHit && !isCacheHit && !isPlanHit) {\n try {\n setTimingFieldOnce(timing, 'callAiStart');\n locateResult = await this.service.locate(\n param,\n {\n context: uiContext,\n },\n modelConfigForDefaultIntent,\n abortSignal,\n );\n applyDump(locateResult.dump);\n elementFromAiLocate = locateResult.element;\n } catch (error) {\n if (error instanceof ServiceError) {\n applyDump(error.dump);\n }\n throw error;\n } finally {\n setTimingFieldOnce(timing, 'callAiEnd');\n }\n }\n\n const element =\n elementFromBbox ||\n elementFromXpath ||\n elementFromCache ||\n elementFromAiLocate;\n\n // Check if locate cache already exists (for planHitFlag case)\n const locateCacheAlreadyExists = hasNonEmptyCache(\n locateCacheRecord?.cacheContent?.cache,\n );\n\n let currentCacheEntry: ElementCacheFeature | undefined;\n // Write cache if:\n // 1. element found\n // 2. taskCache enabled\n // 3. not a cache hit (otherwise we'd be writing what we just read)\n // 4. not already cached for plan hit case (avoid redundant writes), OR allow update if cache validation failed\n // 5. cacheable is not explicitly false\n if (\n element &&\n this.taskCache &&\n !isCacheHit &&\n (!isPlanHit || !locateCacheAlreadyExists) &&\n param?.cacheable !== false\n ) {\n if (this.interface.cacheFeatureForPoint) {\n try {\n // Transform coordinates to logical space for cacheFeatureForPoint\n // cacheFeatureForPoint needs logical coordinates to locate elements in DOM\n let pointForCache: [number, number] = element.center;\n if (shrunkShotToLogicalRatio !== 1) {\n pointForCache = [\n Math.round(element.center[0] / shrunkShotToLogicalRatio),\n Math.round(element.center[1] / shrunkShotToLogicalRatio),\n ];\n debug(\n 'Transformed coordinates for cacheFeatureForPoint: %o -> %o',\n element.center,\n pointForCache,\n );\n }\n\n const feature = await this.interface.cacheFeatureForPoint(\n pointForCache,\n {\n targetDescription:\n typeof param.prompt === 'string'\n ? param.prompt\n : param.prompt?.prompt,\n modelConfig: modelConfigForDefaultIntent,\n },\n );\n if (hasNonEmptyCache(feature)) {\n debug(\n 'update cache, prompt: %s, cache: %o',\n cachePrompt,\n feature,\n );\n currentCacheEntry = feature;\n await this.taskCache.updateOrAppendCacheRecord(\n {\n type: 'locate',\n prompt: cachePrompt,\n cache: feature,\n },\n locateCacheRecord,\n );\n } else {\n debug(\n 'no cache data returned, skip cache update, prompt: %s',\n cachePrompt,\n );\n }\n } catch (error) {\n debug('cacheFeatureForPoint failed: %s', error);\n }\n } else {\n debug('cacheFeatureForPoint is not supported, skip cache update');\n }\n }\n\n if (!element) {\n if (locateDump) {\n throw new ServiceError(\n `Element not found : ${param.prompt}`,\n locateDump,\n );\n }\n throw new Error(`Element not found: ${param.prompt}`);\n }\n\n let hitBy: ExecutionTaskHitBy | undefined;\n\n if (isPlanHit) {\n hitBy = {\n from: 'Plan',\n context: {\n bbox: param.bbox,\n },\n };\n } else if (isXpathHit) {\n hitBy = {\n from: 'User expected path',\n context: {\n xpath: param.xpath,\n },\n };\n } else if (isCacheHit) {\n hitBy = {\n from: 'Cache',\n context: {\n cacheEntry,\n cacheToSave: currentCacheEntry,\n },\n };\n }\n\n onResult?.(element);\n\n return {\n output: {\n element: {\n ...element,\n // backward compatibility for aiLocate, which return value needs a dpr field\n dpr: uiContext.deprecatedDpr,\n },\n },\n hitBy,\n };\n },\n };\n\n return taskLocator;\n }\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","hasNonEmptyCache","cache","locatePlanForLocate","param","locate","locatePlan","TaskBuilder","plans","modelConfigForPlanning","modelConfigForDefaultIntent","options","tasks","cacheable","context","planHandlers","Map","plan","defaultHandler","handler","taskActionFinished","taskLocate","planType","actionSpace","action","item","Error","locateFields","findAllMidsceneLocatorField","requiredLocateFields","field","JSON","ifPlanLocateParamIsBbox","locateTask","result","assert","task","taskContext","timing","uiContext","setTimingFieldOnce","Promise","sleep","originalError","originalMessage","String","shrunkShotToLogicalRatio","undefined","parseActionParam","error","actionFn","actionResult","delayAfterRunner","detailedLocateParam","onResult","deepLocate","abortSignal","locateParam","taskLocator","locateDump","locateResult","applyDump","dump","elementFromBbox","matchElementFromPlan","isPlanHit","rectFromXpath","elementFromXpath","generateElementByRect","transformLogicalRectToScreenshotRect","isXpathHit","cachePrompt","locateCacheRecord","cacheEntry","elementFromCacheResult","matchElementFromCache","elementFromCache","transformLogicalElementToScreenshot","isCacheHit","elementFromAiLocate","ServiceError","element","locateCacheAlreadyExists","currentCacheEntry","pointForCache","Math","feature","hitBy","interfaceInstance","service","taskCache","waitAfterAction"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;;;;;;;AC4BA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAKvB,SAASC,iBAAiBC,KAAc;IACtC,OACEA,QAAAA,SAEA,AAAiB,YAAjB,OAAOA,SACPP,OAAO,IAAI,CAACO,OAAO,MAAM,GAAG;AAEhC;AAEO,SAASC,oBAAoBC,KAAmC;IACrE,MAAMC,SAAS,AAAiB,YAAjB,OAAOD,QAAqB;QAAE,QAAQA;IAAM,IAAIA;IAC/D,MAAME,aAAkD;QACtD,MAAM;QACN,OAAOD;QACP,SAAS;IACX;IACA,OAAOC;AACT;AAyBO,MAAMC;IAyBX,MAAa,MACXC,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACzCC,OAAsB,EACoB;QAC1C,MAAMC,QAA8B,EAAE;QACtC,MAAMC,YAAYF,SAAS;QAE3B,MAAMG,UAA4B;YAChCF;YACAH;YACAC;YACAG;YACA,YAAYF,SAAS;YACrB,aAAaA,SAAS;QACxB;QAIA,MAAMI,eAAe,IAAIC,IAAyB;YAChD;gBACE;gBACA,CAACC,OACC,IAAI,CAAC,gBAAgB,CACnBA,MACAH;aAEL;YACD;gBAAC;gBAAY,CAACG,OAAS,IAAI,CAAC,kBAAkB,CAACA,MAAMH;aAAS;SAC/D;QAED,MAAMI,iBAA8B,CAACD,OACnC,IAAI,CAAC,gBAAgB,CAACA,MAAMH;QAE9B,KAAK,MAAMG,QAAQT,MAAO;YACxB,MAAMW,UAAUJ,aAAa,GAAG,CAACE,KAAK,IAAI,KAAKC;YAC/C,MAAMC,QAAQF;QAChB;QAEA,OAAO;YACLL;QACF;IACF;IAEQ,mBACNK,IAAoB,EACpBH,OAAyB,EACnB;QACN,MAAMM,qBAAqD;YACzD,MAAM;YACN,SAAS;YACT,OAAO;YACP,SAASH,KAAK,OAAO;YACrB,UAAU,WAAa;QACzB;QACAH,QAAQ,KAAK,CAAC,IAAI,CAACM;IACrB;IAEA,MAAc,iBACZH,IAAyC,EACzCH,OAAyB,EACV;QACf,MAAMO,aAAa,IAAI,CAAC,gBAAgB,CAACJ,MAAMA,KAAK,KAAK,EAAEH;QAC3DA,QAAQ,KAAK,CAAC,IAAI,CAACO;IACrB;IAEA,MAAc,iBACZJ,IAAoB,EACpBH,OAAyB,EACV;QACf,MAAMQ,WAAWL,KAAK,IAAI;QAC1B,MAAMM,cAAc,IAAI,CAAC,WAAW;QACpC,MAAMC,SAASD,YAAY,IAAI,CAAC,CAACE,OAASA,KAAK,IAAI,KAAKH;QACxD,MAAMlB,QAAQa,KAAK,KAAK;QAExB,IAAI,CAACO,QACH,MAAM,IAAIE,MAAM,CAAC,aAAa,EAAEJ,SAAS,WAAW,CAAC;QAGvD,MAAMK,eAAeH,SACjBI,AAAAA,IAAAA,yBAAAA,2BAAAA,AAAAA,EAA4BJ,OAAO,WAAW,IAC9C,EAAE;QAEN,MAAMK,uBAAuBL,SACzBI,AAAAA,IAAAA,yBAAAA,2BAAAA,AAAAA,EAA4BJ,OAAO,WAAW,EAAE,QAChD,EAAE;QAENG,aAAa,OAAO,CAAC,CAACG;YACpB,IAAI1B,KAAK,CAAC0B,MAAM,EAAE;gBAGhB,MAAMxB,aAAaH,oBAAoBC,KAAK,CAAC0B,MAAM;gBACnD/B,MACE,uCACA,CAAC,YAAY,EAAEuB,UAAU,EACzB,CAAC,MAAM,EAAES,KAAK,SAAS,CAAC3B,KAAK,CAAC0B,MAAM,GAAG,EACvC,CAAC,WAAW,EAAEC,KAAK,SAAS,CAACzB,aAAa,EAC1C,CAAC,QAAQ,EAAE0B,AAAAA,IAAAA,oCAAAA,uBAAAA,AAAAA,EAAwB5B,KAAK,CAAC0B,MAAM,GAAG;gBAEpD,MAAMG,aAAa,IAAI,CAAC,gBAAgB,CACtC3B,YACAF,KAAK,CAAC0B,MAAM,EACZhB,SACA,CAACoB;oBACC9B,KAAK,CAAC0B,MAAM,GAAGI;gBACjB;gBAEFpB,QAAQ,KAAK,CAAC,IAAI,CAACmB;YACrB,OAAO;gBACLE,IAAAA,sBAAAA,MAAAA,AAAAA,EACE,CAACN,qBAAqB,QAAQ,CAACC,QAC/B,CAAC,uBAAuB,EAAEA,MAAM,6BAA6B,EAAER,UAAU;gBAE3EvB,MAAM,CAAC,OAAO,EAAE+B,MAAM,6BAA6B,EAAER,UAAU;YACjE;QACF;QAEA,MAAMc,OAKF;YACF,MAAM;YACN,SAASd;YACT,SAASL,KAAK,OAAO;YACrB,OAAOA,KAAK,KAAK;YACjB,UAAU,OAAOb,OAAOiC;gBACtB,MAAMC,SAASD,YAAY,IAAI,CAAC,MAAM;gBAEtCtC,MACE,oBACAuB,UACAlB,OACA,CAAC,4BAA4B,EAAEiC,YAAY,OAAO,EAAE,QAAQ;gBAG9D,MAAME,YAAYF,YAAY,SAAS;gBACvCF,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOI,WAAW;gBAElBV,qBAAqB,OAAO,CAAC,CAACC;oBAC5BK,IAAAA,sBAAAA,MAAAA,AAAAA,EACE/B,KAAK,CAAC0B,MAAM,EACZ,CAAC,OAAO,EAAEA,MAAM,yBAAyB,EAAER,SAAS,yCAAyC,EAAEA,SAAS,CAAC,CAAC;gBAE9G;gBAEAkB,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;gBAC3B,IAAI;oBACF,MAAMG,QAAQ,GAAG,CAAC;wBACf;4BACC,IAAI,IAAI,CAAC,SAAS,CAAC,kBAAkB,EAAE;gCACrC1C,MACE,CAAC,8DAA8D,EAAEyB,OAAO,IAAI,EAAE;gCAEhF,MAAM,IAAI,CAAC,SAAS,CAAC,kBAAkB,CAACA,OAAO,IAAI,EAAEpB;gCACrDL,MACE,CAAC,2DAA2D,EAAEyB,OAAO,IAAI,EAAE;4BAE/E;wBACF;wBACAkB,IAAAA,kCAAAA,KAAAA,AAAAA,EAAM;qBACP;gBACH,EAAE,OAAOC,eAAoB;oBAC3B,MAAMC,kBACJD,eAAe,WAAWE,OAAOF;oBACnC,MAAM,IAAIjB,MACR,CAAC,wCAAwC,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEoB,iBAAiB,EAC5E;wBAAE,OAAOD;oBAAc;gBAE3B;gBACAH,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;gBAE3B,MAAM,EAAEQ,wBAAwB,EAAE,GAAGP;gBACrC,IAAIO,AAA6BC,WAA7BD,0BACF,MAAM,IAAIpB,MACR;gBAIJ,IAAIF,OAAO,WAAW,EACpB,IAAI;oBACFpB,QAAQ4C,AAAAA,IAAAA,yBAAAA,gBAAAA,AAAAA,EAAiB5C,OAAOoB,OAAO,WAAW,EAAE;wBAClDsB;oBACF;gBACF,EAAE,OAAOG,OAAY;oBACnB,MAAM,IAAIvB,MACR,CAAC,8BAA8B,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEyB,MAAM,OAAO,CAAC,cAAc,EAAElB,KAAK,SAAS,CAAC3B,QAAQ,EACtG;wBAAE,OAAO6C;oBAAM;gBAEnB;gBAGFT,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;gBAE3BvC,MAAM,kBAAkByB,OAAO,IAAI;gBACnC,MAAM0B,WAAW1B,OAAO,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS;gBAChD,MAAM2B,eAAe,MAAMD,SAAS9C,OAAOiC;gBAC3CG,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;gBAC3BvC,MAAM,iBAAiByB,OAAO,IAAI,EAAE,WAAW2B;gBAE/CX,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;gBAE3B,MAAMc,mBACJ5B,OAAO,gBAAgB,IAAI,IAAI,CAAC,eAAe,IAAI;gBACrD,IAAI4B,mBAAmB,GACrB,MAAMV,AAAAA,IAAAA,kCAAAA,KAAAA,AAAAA,EAAMU;gBAGd,IAAI;oBACF,IAAI,IAAI,CAAC,SAAS,CAAC,iBAAiB,EAAE;wBACpCrD,MACE,CAAC,6DAA6D,EAAEyB,OAAO,IAAI,EAAE;wBAE/E,MAAM,IAAI,CAAC,SAAS,CAAC,iBAAiB,CAACA,OAAO,IAAI,EAAEpB;wBACpDL,MACE,CAAC,0DAA0D,EAAEyB,OAAO,IAAI,EAAE;oBAE9E;gBACF,EAAE,OAAOmB,eAAoB;oBAC3B,MAAMC,kBACJD,eAAe,WAAWE,OAAOF;oBACnC,MAAM,IAAIjB,MACR,CAAC,uCAAuC,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEoB,iBAAiB,EAC3E;wBAAE,OAAOD;oBAAc;gBAE3B;gBAEAH,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;gBAE3B,OAAO;oBACL,QAAQa;gBACV;YACF;QACF;QAEArC,QAAQ,KAAK,CAAC,IAAI,CAACsB;IACrB;IAEQ,iBACNnB,IAAyC,EACzCoC,mBAAiD,EACjDvC,OAAyB,EACzBwC,QAAgD,EACd;QAClC,MAAM,EAAEzC,SAAS,EAAEH,2BAA2B,EAAE6C,UAAU,EAAEC,WAAW,EAAE,GACvE1C;QAEF,IAAI2C,cAAcJ;QAElB,IAAI,AAAuB,YAAvB,OAAOI,aACTA,cAAc;YACZ,QAAQA;QACV;QAGF,IAAI5C,AAAckC,WAAdlC,WACF4C,cAAc;YACZ,GAAGA,WAAW;YACd5C;QACF;QAGF,IAAI0C,cAAc,CAACE,YAAY,UAAU,EACvCA,cAAc;YACZ,GAAGA,WAAW;YACd,YAAY;QACd;QAGF,MAAMC,cAAgD;YACpD,MAAM;YACN,SAAS;YACT,OAAOD;YACP,SAASxC,KAAK,OAAO;YACrB,UAAU,OAAOb,OAAOiC;gBACtB,MAAM,EAAED,IAAI,EAAE,GAAGC;gBACjB,IAAI,EAAEE,SAAS,EAAE,GAAGF;gBAEpBF,IAAAA,sBAAAA,MAAAA,AAAAA,EACE/B,OAAO,UAAUA,OAAO,MACxB,CAAC,qDAAqD,EAAE2B,KAAK,SAAS,CACpE3B,QACC;gBAGL,IAAI,CAACmC,WACHA,YAAY,MAAM,IAAI,CAAC,OAAO,CAAC,kBAAkB;gBAGnDJ,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOI,WAAW;gBAElB,MAAM,EAAEO,wBAAwB,EAAE,GAAGP;gBAErC,IAAIO,AAA6BC,WAA7BD,0BACF,MAAM,IAAIpB,MACR;gBAIJ,IAAIiC;gBACJ,IAAIC;gBAEJ,MAAMC,YAAY,CAACC;oBACjB,IAAI,CAACA,MACH;oBAEFH,aAAaG;oBACb1B,KAAK,GAAG,GAAG;wBACT0B;wBACA,aAAaA,KAAK,QAAQ,EAAE;oBAC9B;oBACA1B,KAAK,KAAK,GAAG0B,KAAK,QAAQ,EAAE;oBAC5B,IAAIA,KAAK,QAAQ,EAAE,iBACjB1B,KAAK,eAAe,GAAG0B,KAAK,QAAQ,CAAC,eAAe;oBAEtD,IAAIA,KAAK,QAAQ,EAAE,mBACjB1B,KAAK,iBAAiB,GAAG0B,KAAK,QAAQ,CAAC,iBAAiB;gBAE5D;gBAGA,MAAMC,kBAAkB/B,AAAAA,IAAAA,oCAAAA,uBAAAA,AAAAA,EAAwB5B,SAC5C4D,AAAAA,IAAAA,oCAAAA,oBAAAA,AAAAA,EAAqB5D,SACrB2C;gBACJ,MAAMkB,YAAY,CAAC,CAACF;gBAGpB,IAAIG;gBACJ,IACE,CAACD,aACD7D,MAAM,KAAK,IACX,IAAI,CAAC,SAAS,CAAC,uBAAuB,EAEtC,IAAI;oBACF8D,gBAAgB,MAAM,IAAI,CAAC,SAAS,CAAC,uBAAuB,CAAC;wBAC3D,QAAQ;4BAAC9D,MAAM,KAAK;yBAAC;oBACvB;gBACF,EAAE,OAAM,CAER;gBAGF,MAAM+D,mBAAmBD,gBACrBE,AAAAA,IAAAA,0BAAAA,qBAAAA,AAAAA,EAEEC,AAAAA,IAAAA,oCAAAA,oCAAAA,AAAAA,EACEH,eACApB,2BAEF,AAAwB,YAAxB,OAAO1C,MAAM,MAAM,GACfA,MAAM,MAAM,GACZA,MAAM,MAAM,EAAE,UAAU,MAE9B2C;gBAEJ,MAAMuB,aAAa,CAAC,CAACH;gBAErB,MAAMI,cAAcnE,MAAM,MAAM;gBAChC,MAAMoE,oBACJ,MAAM,IAAI,CAAC,SAAS,EAAE,iBAAiBD;gBACzC,MAAME,aAAaD,mBAAmB,cAAc;gBAEpD,MAAME,yBACJT,aAAaK,aACT,OACA,MAAMK,AAAAA,IAAAA,oCAAAA,qBAAAA,AAAAA,EACJ;oBACE,WAAW,IAAI,CAAC,SAAS;oBACzB,mBAAmB,IAAI,CAAC,SAAS;gBACnC,GACAF,YACAF,aACAnE,MAAM,SAAS;gBAIvB,MAAMwE,mBAAmBF,yBACrBG,AAAAA,IAAAA,oCAAAA,mCAAAA,AAAAA,EACEH,wBACA5B,4BAEFC;gBAEJ,MAAM+B,aAAa,CAAC,CAACF;gBAErB,IAAIG;gBACJ,MAAMzC,SAASD,YAAY,IAAI,CAAC,MAAM;gBACtC,IAAI,CAACiC,cAAc,CAACQ,cAAc,CAACb,WACjC,IAAI;oBACFzB,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;oBAC3BsB,eAAe,MAAM,IAAI,CAAC,OAAO,CAAC,MAAM,CACtCxD,OACA;wBACE,SAASmC;oBACX,GACA7B,6BACA8C;oBAEFK,UAAUD,aAAa,IAAI;oBAC3BmB,sBAAsBnB,aAAa,OAAO;gBAC5C,EAAE,OAAOX,OAAO;oBACd,IAAIA,iBAAiB+B,kCAAAA,YAAYA,EAC/BnB,UAAUZ,MAAM,IAAI;oBAEtB,MAAMA;gBACR,SAAU;oBACRT,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;gBAC7B;gBAGF,MAAM2C,UACJlB,mBACAI,oBACAS,oBACAG;gBAGF,MAAMG,2BAA2BjF,iBAC/BuE,mBAAmB,cAAc;gBAGnC,IAAIW;gBAOJ,IACEF,WACA,IAAI,CAAC,SAAS,IACd,CAACH,cACA,EAACb,aAAa,CAACiB,wBAAuB,KACvC9E,OAAO,cAAc,OAErB,IAAI,IAAI,CAAC,SAAS,CAAC,oBAAoB,EACrC,IAAI;oBAGF,IAAIgF,gBAAkCH,QAAQ,MAAM;oBACpD,IAAInC,AAA6B,MAA7BA,0BAAgC;wBAClCsC,gBAAgB;4BACdC,KAAK,KAAK,CAACJ,QAAQ,MAAM,CAAC,EAAE,GAAGnC;4BAC/BuC,KAAK,KAAK,CAACJ,QAAQ,MAAM,CAAC,EAAE,GAAGnC;yBAChC;wBACD/C,MACE,8DACAkF,QAAQ,MAAM,EACdG;oBAEJ;oBAEA,MAAME,UAAU,MAAM,IAAI,CAAC,SAAS,CAAC,oBAAoB,CACvDF,eACA;wBACE,mBACE,AAAwB,YAAxB,OAAOhF,MAAM,MAAM,GACfA,MAAM,MAAM,GACZA,MAAM,MAAM,EAAE;wBACpB,aAAaM;oBACf;oBAEF,IAAIT,iBAAiBqF,UAAU;wBAC7BvF,MACE,uCACAwE,aACAe;wBAEFH,oBAAoBG;wBACpB,MAAM,IAAI,CAAC,SAAS,CAAC,yBAAyB,CAC5C;4BACE,MAAM;4BACN,QAAQf;4BACR,OAAOe;wBACT,GACAd;oBAEJ,OACEzE,MACE,yDACAwE;gBAGN,EAAE,OAAOtB,OAAO;oBACdlD,MAAM,mCAAmCkD;gBAC3C;qBAEAlD,MAAM;gBAIV,IAAI,CAACkF,SAAS;oBACZ,IAAItB,YACF,MAAM,IAAIqB,kCAAAA,YAAYA,CACpB,CAAC,oBAAoB,EAAE5E,MAAM,MAAM,EAAE,EACrCuD;oBAGJ,MAAM,IAAIjC,MAAM,CAAC,mBAAmB,EAAEtB,MAAM,MAAM,EAAE;gBACtD;gBAEA,IAAImF;gBAEJ,IAAItB,WACFsB,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACP,MAAMnF,MAAM,IAAI;oBAClB;gBACF;qBACK,IAAIkE,YACTiB,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACP,OAAOnF,MAAM,KAAK;oBACpB;gBACF;qBACK,IAAI0E,YACTS,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACPd;wBACA,aAAaU;oBACf;gBACF;gBAGF7B,WAAW2B;gBAEX,OAAO;oBACL,QAAQ;wBACN,SAAS;4BACP,GAAGA,OAAO;4BAEV,KAAK1C,UAAU,aAAa;wBAC9B;oBACF;oBACAgD;gBACF;YACF;QACF;QAEA,OAAO7B;IACT;IA/iBA,YAAY,EACV8B,iBAAiB,EACjBC,OAAO,EACPC,SAAS,EACTnE,WAAW,EACXoE,eAAe,EACC,CAAE;QAhBpB,uBAAiB,aAAjB;QAEA,uBAAiB,WAAjB;QAEA,uBAAiB,aAAjB;QAEA,uBAAiB,eAAjB;QAEA,uBAAiB,mBAAjB;QASE,IAAI,CAAC,SAAS,GAAGH;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC;QACjB,IAAI,CAAC,WAAW,GAAGnE;QACnB,IAAI,CAAC,eAAe,GAAGoE;IACzB;AAoiBF"}
1
+ {"version":3,"file":"agent/task-builder.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../src/agent/task-builder.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import { findAllMidsceneLocatorField, parseActionParam } from '@/ai-model';\nimport type { AbstractInterface } from '@/device';\nimport type Service from '@/service';\nimport { setTimingFieldOnce } from '@/task-timing';\nimport type {\n CacheValidationOptions,\n DetailedLocateParam,\n DeviceAction,\n ElementCacheFeature,\n ExecutionTaskActionApply,\n ExecutionTaskApply,\n ExecutionTaskHitBy,\n ExecutionTaskPlanningLocateApply,\n LocateResultElement,\n LocateResultWithDump,\n PlanningAction,\n PlanningLocateParam,\n Rect,\n ServiceDump,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport { sleep } from '@/utils';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport {\n MIDSCENE_CACHE_COORD_OFFSET_THRESHOLD,\n MIDSCENE_CACHE_ENABLE_COORD_CHECK,\n MIDSCENE_CACHE_ENABLE_VISUAL_VERIFY,\n} from '@midscene/shared/env/constants';\nimport { generateElementByRect } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TaskCache } from './task-cache';\nimport {\n ifPlanLocateParamIsBbox,\n matchElementFromCache,\n matchElementFromPlan,\n transformLogicalElementToScreenshot,\n transformLogicalRectToScreenshotRect,\n} from './utils';\n\nconst debug = getDebug('agent:task-builder');\n\n/**\n * Check if a cache object is non-empty\n */\nfunction hasNonEmptyCache(cache: unknown): boolean {\n return (\n cache !== null &&\n cache !== undefined &&\n typeof cache === 'object' &&\n Object.keys(cache).length > 0\n );\n}\n\nexport function locatePlanForLocate(param: string | DetailedLocateParam) {\n const locate = typeof param === 'string' ? { prompt: param } : param;\n const locatePlan: PlanningAction<PlanningLocateParam> = {\n type: 'Locate',\n param: locate,\n thought: '',\n };\n return locatePlan;\n}\n\ninterface TaskBuilderDeps {\n interfaceInstance: AbstractInterface;\n service: Service;\n taskCache?: TaskCache;\n actionSpace: DeviceAction[];\n waitAfterAction?: number;\n}\n\ninterface BuildOptions {\n cacheable?: boolean;\n deepLocate?: boolean;\n abortSignal?: AbortSignal;\n}\n\ninterface PlanBuildContext {\n tasks: ExecutionTaskApply[];\n modelConfigForPlanning: IModelConfig;\n modelConfigForDefaultIntent: IModelConfig;\n cacheable?: boolean;\n deepLocate?: boolean;\n abortSignal?: AbortSignal;\n}\n\nexport class TaskBuilder {\n private readonly interface: AbstractInterface;\n\n private readonly service: Service;\n\n private readonly taskCache?: TaskCache;\n\n private readonly actionSpace: DeviceAction[];\n\n private readonly waitAfterAction?: number;\n\n constructor({\n interfaceInstance,\n service,\n taskCache,\n actionSpace,\n waitAfterAction,\n }: TaskBuilderDeps) {\n this.interface = interfaceInstance;\n this.service = service;\n this.taskCache = taskCache;\n this.actionSpace = actionSpace;\n this.waitAfterAction = waitAfterAction;\n }\n\n public async build(\n plans: PlanningAction[],\n modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n options?: BuildOptions,\n ): Promise<{ tasks: ExecutionTaskApply[] }> {\n const tasks: ExecutionTaskApply[] = [];\n const cacheable = options?.cacheable;\n\n const context: PlanBuildContext = {\n tasks,\n modelConfigForPlanning,\n modelConfigForDefaultIntent,\n cacheable,\n deepLocate: options?.deepLocate,\n abortSignal: options?.abortSignal,\n };\n\n type PlanHandler = (plan: PlanningAction) => Promise<void> | void;\n\n const planHandlers = new Map<string, PlanHandler>([\n [\n 'Locate',\n (plan) =>\n this.handleLocatePlan(\n plan as PlanningAction<PlanningLocateParam>,\n context,\n ),\n ],\n ['Finished', (plan) => this.handleFinishedPlan(plan, context)],\n ]);\n\n const defaultHandler: PlanHandler = (plan) =>\n this.handleActionPlan(plan, context);\n\n for (const plan of plans) {\n const handler = planHandlers.get(plan.type) ?? defaultHandler;\n await handler(plan);\n }\n\n return {\n tasks,\n };\n }\n\n private handleFinishedPlan(\n plan: PlanningAction,\n context: PlanBuildContext,\n ): void {\n const taskActionFinished: ExecutionTaskActionApply<null> = {\n type: 'Action Space',\n subType: 'Finished',\n param: null,\n thought: plan.thought,\n executor: async () => {},\n };\n context.tasks.push(taskActionFinished);\n }\n\n private async handleLocatePlan(\n plan: PlanningAction<PlanningLocateParam>,\n context: PlanBuildContext,\n ): Promise<void> {\n const taskLocate = this.createLocateTask(plan, plan.param, context);\n context.tasks.push(taskLocate);\n }\n\n private async handleActionPlan(\n plan: PlanningAction,\n context: PlanBuildContext,\n ): Promise<void> {\n const planType = plan.type;\n const actionSpace = this.actionSpace;\n const action = actionSpace.find((item) => item.name === planType);\n const param = plan.param;\n\n if (!action) {\n throw new Error(`Action type '${planType}' not found`);\n }\n\n const locateFields = action\n ? findAllMidsceneLocatorField(action.paramSchema)\n : [];\n\n const requiredLocateFields = action\n ? findAllMidsceneLocatorField(action.paramSchema, true)\n : [];\n\n locateFields.forEach((field) => {\n if (param[field]) {\n // Always use createLocateTask for all locate params (including bbox)\n // This ensures cache writing happens even when bbox is available\n const locatePlan = locatePlanForLocate(param[field]);\n debug(\n 'will prepend locate param for field',\n `action.type=${planType}`,\n `param=${JSON.stringify(param[field])}`,\n `locatePlan=${JSON.stringify(locatePlan)}`,\n `hasBbox=${ifPlanLocateParamIsBbox(param[field])}`,\n );\n const locateTask = this.createLocateTask(\n locatePlan,\n param[field],\n context,\n (result) => {\n param[field] = result;\n },\n );\n context.tasks.push(locateTask);\n } else {\n assert(\n !requiredLocateFields.includes(field),\n `Required locate field '${field}' is not provided for action ${planType}`,\n );\n debug(`field '${field}' is not provided for action ${planType}`);\n }\n });\n\n const task: ExecutionTaskApply<\n 'Action Space',\n any,\n { success: boolean; action: string; param: any },\n void\n > = {\n type: 'Action Space',\n subType: planType,\n thought: plan.thought,\n param: plan.param,\n executor: async (param, taskContext) => {\n const timing = taskContext.task.timing;\n\n debug(\n 'executing action',\n planType,\n param,\n `taskContext.element.center: ${taskContext.element?.center}`,\n );\n\n const uiContext = taskContext.uiContext;\n assert(uiContext, 'uiContext is required for Action task');\n\n requiredLocateFields.forEach((field) => {\n assert(\n param[field],\n `field '${field}' is required for action ${planType} but not provided. Cannot execute action ${planType}.`,\n );\n });\n\n setTimingFieldOnce(timing, 'beforeInvokeActionHookStart');\n try {\n await Promise.all([\n (async () => {\n if (this.interface.beforeInvokeAction) {\n debug(\n `will call \"beforeInvokeAction\" for interface with action name ${action.name}`,\n );\n await this.interface.beforeInvokeAction(action.name, param);\n debug(\n `called \"beforeInvokeAction\" for interface with action name ${action.name}`,\n );\n }\n })(),\n sleep(200),\n ]);\n } catch (originalError: any) {\n const originalMessage =\n originalError?.message || String(originalError);\n throw new Error(\n `error in running beforeInvokeAction for ${action.name}: ${originalMessage}`,\n { cause: originalError },\n );\n }\n setTimingFieldOnce(timing, 'beforeInvokeActionHookEnd');\n\n const { shrunkShotToLogicalRatio } = uiContext;\n if (shrunkShotToLogicalRatio === undefined) {\n throw new Error(\n 'shrunkShotToLogicalRatio is not defined in Action task',\n );\n }\n\n if (action.paramSchema) {\n try {\n param = parseActionParam(param, action.paramSchema, {\n shrunkShotToLogicalRatio,\n });\n } catch (error: any) {\n throw new Error(\n `Invalid parameters for action ${action.name}: ${error.message}\\nParameters: ${JSON.stringify(param)}`,\n { cause: error },\n );\n }\n }\n\n setTimingFieldOnce(timing, 'callActionStart');\n\n debug('calling action', action.name);\n const actionFn = action.call.bind(this.interface);\n const actionResult = await actionFn(param, taskContext);\n setTimingFieldOnce(timing, 'callActionEnd');\n debug('called action', action.name, 'result:', actionResult);\n\n setTimingFieldOnce(timing, 'afterInvokeActionHookStart');\n\n const delayAfterRunner =\n action.delayAfterRunner ?? this.waitAfterAction ?? 300;\n if (delayAfterRunner > 0) {\n await sleep(delayAfterRunner);\n }\n\n try {\n if (this.interface.afterInvokeAction) {\n debug(\n `will call \"afterInvokeAction\" for interface with action name ${action.name}`,\n );\n await this.interface.afterInvokeAction(action.name, param);\n debug(\n `called \"afterInvokeAction\" for interface with action name ${action.name}`,\n );\n }\n } catch (originalError: any) {\n const originalMessage =\n originalError?.message || String(originalError);\n throw new Error(\n `error in running afterInvokeAction for ${action.name}: ${originalMessage}`,\n { cause: originalError },\n );\n }\n\n setTimingFieldOnce(timing, 'afterInvokeActionHookEnd');\n\n return {\n output: actionResult,\n };\n },\n };\n\n context.tasks.push(task);\n }\n\n private createLocateTask(\n plan: PlanningAction<PlanningLocateParam>,\n detailedLocateParam: DetailedLocateParam | string,\n context: PlanBuildContext,\n onResult?: (result: LocateResultElement) => void,\n ): ExecutionTaskPlanningLocateApply {\n const { cacheable, modelConfigForDefaultIntent, deepLocate, abortSignal } =\n context;\n\n let locateParam = detailedLocateParam;\n\n if (typeof locateParam === 'string') {\n locateParam = {\n prompt: locateParam,\n };\n }\n\n if (cacheable !== undefined) {\n locateParam = {\n ...locateParam,\n cacheable,\n };\n }\n\n if (deepLocate && !locateParam.deepLocate) {\n locateParam = {\n ...locateParam,\n deepLocate: true,\n };\n }\n\n const taskLocator: ExecutionTaskPlanningLocateApply = {\n type: 'Planning',\n subType: 'Locate',\n param: locateParam,\n thought: plan.thought,\n executor: async (param, taskContext) => {\n const { task } = taskContext;\n let { uiContext } = taskContext;\n\n assert(\n param?.prompt || param?.bbox,\n `No prompt or id or position or bbox to locate, param=${JSON.stringify(\n param,\n )}`,\n );\n\n if (!uiContext) {\n uiContext = await this.service.contextRetrieverFn();\n }\n\n assert(uiContext, 'uiContext is required for Service task');\n\n const { shrunkShotToLogicalRatio } = uiContext;\n\n if (shrunkShotToLogicalRatio === undefined) {\n throw new Error(\n 'shrunkShotToLogicalRatio is not defined in locate task',\n );\n }\n\n let locateDump: ServiceDump | undefined;\n let locateResult: LocateResultWithDump | undefined;\n\n const applyDump = (dump?: ServiceDump) => {\n if (!dump) {\n return;\n }\n locateDump = dump;\n task.log = {\n dump,\n rawResponse: dump.taskInfo?.rawResponse,\n };\n task.usage = dump.taskInfo?.usage;\n if (dump.taskInfo?.searchAreaUsage) {\n task.searchAreaUsage = dump.taskInfo.searchAreaUsage;\n }\n if (dump.taskInfo?.reasoning_content) {\n task.reasoning_content = dump.taskInfo.reasoning_content;\n }\n };\n\n // from bbox (plan hit)\n const elementFromBbox = ifPlanLocateParamIsBbox(param)\n ? matchElementFromPlan(param)\n : undefined;\n const isPlanHit = !!elementFromBbox;\n\n // from xpath\n let rectFromXpath: Rect | undefined;\n if (\n !isPlanHit &&\n param.xpath &&\n this.interface.rectMatchesCacheFeature\n ) {\n try {\n rectFromXpath = await this.interface.rectMatchesCacheFeature({\n xpaths: [param.xpath],\n });\n } catch {\n // xpath locate failed, allow fallback to cache or AI locate\n }\n }\n\n const elementFromXpath = rectFromXpath\n ? generateElementByRect(\n // rectFromXpath is in logical coordinates, which should be transformed to screenshot coordinates;\n transformLogicalRectToScreenshotRect(\n rectFromXpath,\n shrunkShotToLogicalRatio,\n ),\n typeof param.prompt === 'string'\n ? param.prompt\n : param.prompt?.prompt || '',\n )\n : undefined;\n\n const isXpathHit = !!elementFromXpath;\n\n const cachePrompt = param.prompt;\n const locateCacheRecord =\n await this.taskCache?.matchLocateCache(cachePrompt);\n const cacheEntry = locateCacheRecord?.cacheContent?.cache;\n\n const elementFromCacheResult =\n isPlanHit || isXpathHit\n ? null\n : await matchElementFromCache(\n {\n taskCache: this.taskCache,\n interfaceInstance: this.interface,\n },\n cacheEntry,\n cachePrompt,\n param.cacheable,\n );\n\n // elementFromCacheResult is in logical coordinates, which should be transformed to screenshot coordinates;\n const elementFromCache = elementFromCacheResult\n ? transformLogicalElementToScreenshot(\n elementFromCacheResult,\n shrunkShotToLogicalRatio,\n )\n : undefined;\n\n let isCacheHit = !!elementFromCache;\n const timing = taskContext.task.timing;\n let elementFromAiLocate: LocateResultElement | null | undefined;\n\n if (isCacheHit) {\n const cacheFeature = cacheEntry as ElementCacheFeature;\n const cachedCenter = cacheFeature?.cachedCenter as\n | [number, number]\n | undefined;\n\n const coordOffsetThreshold =\n Number.parseInt(\n process.env[MIDSCENE_CACHE_COORD_OFFSET_THRESHOLD] || '16',\n 10,\n ) || 16;\n const enableCoordCheck =\n process.env[MIDSCENE_CACHE_ENABLE_COORD_CHECK] !== 'false';\n const enableVisualVerify =\n process.env[MIDSCENE_CACHE_ENABLE_VISUAL_VERIFY] !== 'false';\n\n try {\n if (enableCoordCheck && cachedCenter) {\n const offset = Math.sqrt(\n (elementFromCache!.center[0] - cachedCenter[0]) ** 2 +\n (elementFromCache!.center[1] - cachedCenter[1]) ** 2,\n );\n\n debug('cache coord offset check', {\n cachedCenter,\n currentCenter: elementFromCache!.center,\n offset: Math.round(offset),\n threshold: coordOffsetThreshold,\n });\n\n if (offset > coordOffsetThreshold) {\n debug(\n 'cache coord offset exceeded threshold, fallback to AI locate',\n { offset, threshold: coordOffsetThreshold },\n );\n isCacheHit = false;\n }\n }\n\n if (isCacheHit && enableVisualVerify) {\n const verification = await this.service.verifyCachedElement(\n elementFromCache!.center,\n cachePrompt,\n modelConfigForDefaultIntent,\n uiContext,\n );\n if (!verification.pass) {\n debug(\n 'cache hit but visual verification failed, fallback to AI locate',\n {\n reason: verification.reason,\n description: verification.description,\n prompt: cachePrompt,\n },\n );\n isCacheHit = false;\n } else {\n debug('cache hit and visual verification passed', {\n description: verification.description,\n });\n }\n }\n } catch (verifyError) {\n debug(\n 'cache verification error, fallback to AI locate',\n verifyError,\n );\n isCacheHit = false;\n }\n }\n\n if (!isXpathHit && !isCacheHit && !isPlanHit) {\n try {\n setTimingFieldOnce(timing, 'callAiStart');\n locateResult = await this.service.locate(\n param,\n {\n context: uiContext,\n },\n modelConfigForDefaultIntent,\n abortSignal,\n );\n applyDump(locateResult.dump);\n elementFromAiLocate = locateResult.element;\n } catch (error) {\n if (error instanceof ServiceError) {\n applyDump(error.dump);\n }\n throw error;\n } finally {\n setTimingFieldOnce(timing, 'callAiEnd');\n }\n }\n\n const element =\n elementFromBbox ||\n elementFromXpath ||\n elementFromCache ||\n elementFromAiLocate;\n\n // Check if locate cache already exists (for planHitFlag case)\n const locateCacheAlreadyExists = hasNonEmptyCache(\n locateCacheRecord?.cacheContent?.cache,\n );\n\n let currentCacheEntry: ElementCacheFeature | undefined;\n // Write cache if:\n // 1. element found\n // 2. taskCache enabled\n // 3. not a cache hit (otherwise we'd be writing what we just read)\n // 4. not already cached for plan hit case (avoid redundant writes), OR allow update if cache validation failed\n // 5. cacheable is not explicitly false\n if (\n element &&\n this.taskCache &&\n !isCacheHit &&\n (!isPlanHit || !locateCacheAlreadyExists) &&\n param?.cacheable !== false\n ) {\n if (this.interface.cacheFeatureForPoint) {\n try {\n // Transform coordinates to logical space for cacheFeatureForPoint\n // cacheFeatureForPoint needs logical coordinates to locate elements in DOM\n let pointForCache: [number, number] = element.center;\n if (shrunkShotToLogicalRatio !== 1) {\n pointForCache = [\n Math.round(element.center[0] / shrunkShotToLogicalRatio),\n Math.round(element.center[1] / shrunkShotToLogicalRatio),\n ];\n debug(\n 'Transformed coordinates for cacheFeatureForPoint: %o -> %o',\n element.center,\n pointForCache,\n );\n }\n\n const feature = await this.interface.cacheFeatureForPoint(\n pointForCache,\n {\n targetDescription:\n typeof param.prompt === 'string'\n ? param.prompt\n : param.prompt?.prompt,\n modelConfig: modelConfigForDefaultIntent,\n },\n );\n if (hasNonEmptyCache(feature)) {\n feature.cachedCenter = pointForCache;\n debug(\n 'update cache, prompt: %s, cache: %o',\n cachePrompt,\n feature,\n );\n currentCacheEntry = feature;\n await this.taskCache.updateOrAppendCacheRecord(\n {\n type: 'locate',\n prompt: cachePrompt,\n cache: feature,\n },\n locateCacheRecord,\n );\n } else {\n debug(\n 'no cache data returned, skip cache update, prompt: %s',\n cachePrompt,\n );\n }\n } catch (error) {\n debug('cacheFeatureForPoint failed: %s', error);\n }\n } else {\n debug('cacheFeatureForPoint is not supported, skip cache update');\n }\n }\n\n if (!element) {\n if (locateDump) {\n throw new ServiceError(\n `Element not found : ${param.prompt}`,\n locateDump,\n );\n }\n throw new Error(`Element not found: ${param.prompt}`);\n }\n\n let hitBy: ExecutionTaskHitBy | undefined;\n\n if (isPlanHit) {\n hitBy = {\n from: 'Plan',\n context: {\n bbox: param.bbox,\n },\n };\n } else if (isXpathHit) {\n hitBy = {\n from: 'User expected path',\n context: {\n xpath: param.xpath,\n },\n };\n } else if (isCacheHit) {\n hitBy = {\n from: 'Cache',\n context: {\n cacheEntry,\n cacheToSave: currentCacheEntry,\n },\n };\n }\n\n onResult?.(element);\n\n return {\n output: {\n element: {\n ...element,\n // backward compatibility for aiLocate, which return value needs a dpr field\n dpr: uiContext.deprecatedDpr,\n },\n },\n hitBy,\n };\n },\n };\n\n return taskLocator;\n }\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","hasNonEmptyCache","cache","locatePlanForLocate","param","locate","locatePlan","TaskBuilder","plans","modelConfigForPlanning","modelConfigForDefaultIntent","options","tasks","cacheable","context","planHandlers","Map","plan","defaultHandler","handler","taskActionFinished","taskLocate","planType","actionSpace","action","item","Error","locateFields","findAllMidsceneLocatorField","requiredLocateFields","field","JSON","ifPlanLocateParamIsBbox","locateTask","result","assert","task","taskContext","timing","uiContext","setTimingFieldOnce","Promise","sleep","originalError","originalMessage","String","shrunkShotToLogicalRatio","undefined","parseActionParam","error","actionFn","actionResult","delayAfterRunner","detailedLocateParam","onResult","deepLocate","abortSignal","locateParam","taskLocator","locateDump","locateResult","applyDump","dump","elementFromBbox","matchElementFromPlan","isPlanHit","rectFromXpath","elementFromXpath","generateElementByRect","transformLogicalRectToScreenshotRect","isXpathHit","cachePrompt","locateCacheRecord","cacheEntry","elementFromCacheResult","matchElementFromCache","elementFromCache","transformLogicalElementToScreenshot","isCacheHit","elementFromAiLocate","cacheFeature","cachedCenter","coordOffsetThreshold","Number","process","MIDSCENE_CACHE_COORD_OFFSET_THRESHOLD","enableCoordCheck","MIDSCENE_CACHE_ENABLE_COORD_CHECK","enableVisualVerify","MIDSCENE_CACHE_ENABLE_VISUAL_VERIFY","offset","Math","verification","verifyError","ServiceError","element","locateCacheAlreadyExists","currentCacheEntry","pointForCache","feature","hitBy","interfaceInstance","service","taskCache","waitAfterAction"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;;;;;;;;ACkCA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAKvB,SAASC,iBAAiBC,KAAc;IACtC,OACEA,QAAAA,SAEA,AAAiB,YAAjB,OAAOA,SACPP,OAAO,IAAI,CAACO,OAAO,MAAM,GAAG;AAEhC;AAEO,SAASC,oBAAoBC,KAAmC;IACrE,MAAMC,SAAS,AAAiB,YAAjB,OAAOD,QAAqB;QAAE,QAAQA;IAAM,IAAIA;IAC/D,MAAME,aAAkD;QACtD,MAAM;QACN,OAAOD;QACP,SAAS;IACX;IACA,OAAOC;AACT;AAyBO,MAAMC;IAyBX,MAAa,MACXC,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACzCC,OAAsB,EACoB;QAC1C,MAAMC,QAA8B,EAAE;QACtC,MAAMC,YAAYF,SAAS;QAE3B,MAAMG,UAA4B;YAChCF;YACAH;YACAC;YACAG;YACA,YAAYF,SAAS;YACrB,aAAaA,SAAS;QACxB;QAIA,MAAMI,eAAe,IAAIC,IAAyB;YAChD;gBACE;gBACA,CAACC,OACC,IAAI,CAAC,gBAAgB,CACnBA,MACAH;aAEL;YACD;gBAAC;gBAAY,CAACG,OAAS,IAAI,CAAC,kBAAkB,CAACA,MAAMH;aAAS;SAC/D;QAED,MAAMI,iBAA8B,CAACD,OACnC,IAAI,CAAC,gBAAgB,CAACA,MAAMH;QAE9B,KAAK,MAAMG,QAAQT,MAAO;YACxB,MAAMW,UAAUJ,aAAa,GAAG,CAACE,KAAK,IAAI,KAAKC;YAC/C,MAAMC,QAAQF;QAChB;QAEA,OAAO;YACLL;QACF;IACF;IAEQ,mBACNK,IAAoB,EACpBH,OAAyB,EACnB;QACN,MAAMM,qBAAqD;YACzD,MAAM;YACN,SAAS;YACT,OAAO;YACP,SAASH,KAAK,OAAO;YACrB,UAAU,WAAa;QACzB;QACAH,QAAQ,KAAK,CAAC,IAAI,CAACM;IACrB;IAEA,MAAc,iBACZH,IAAyC,EACzCH,OAAyB,EACV;QACf,MAAMO,aAAa,IAAI,CAAC,gBAAgB,CAACJ,MAAMA,KAAK,KAAK,EAAEH;QAC3DA,QAAQ,KAAK,CAAC,IAAI,CAACO;IACrB;IAEA,MAAc,iBACZJ,IAAoB,EACpBH,OAAyB,EACV;QACf,MAAMQ,WAAWL,KAAK,IAAI;QAC1B,MAAMM,cAAc,IAAI,CAAC,WAAW;QACpC,MAAMC,SAASD,YAAY,IAAI,CAAC,CAACE,OAASA,KAAK,IAAI,KAAKH;QACxD,MAAMlB,QAAQa,KAAK,KAAK;QAExB,IAAI,CAACO,QACH,MAAM,IAAIE,MAAM,CAAC,aAAa,EAAEJ,SAAS,WAAW,CAAC;QAGvD,MAAMK,eAAeH,SACjBI,AAAAA,IAAAA,yBAAAA,2BAAAA,AAAAA,EAA4BJ,OAAO,WAAW,IAC9C,EAAE;QAEN,MAAMK,uBAAuBL,SACzBI,AAAAA,IAAAA,yBAAAA,2BAAAA,AAAAA,EAA4BJ,OAAO,WAAW,EAAE,QAChD,EAAE;QAENG,aAAa,OAAO,CAAC,CAACG;YACpB,IAAI1B,KAAK,CAAC0B,MAAM,EAAE;gBAGhB,MAAMxB,aAAaH,oBAAoBC,KAAK,CAAC0B,MAAM;gBACnD/B,MACE,uCACA,CAAC,YAAY,EAAEuB,UAAU,EACzB,CAAC,MAAM,EAAES,KAAK,SAAS,CAAC3B,KAAK,CAAC0B,MAAM,GAAG,EACvC,CAAC,WAAW,EAAEC,KAAK,SAAS,CAACzB,aAAa,EAC1C,CAAC,QAAQ,EAAE0B,AAAAA,IAAAA,oCAAAA,uBAAAA,AAAAA,EAAwB5B,KAAK,CAAC0B,MAAM,GAAG;gBAEpD,MAAMG,aAAa,IAAI,CAAC,gBAAgB,CACtC3B,YACAF,KAAK,CAAC0B,MAAM,EACZhB,SACA,CAACoB;oBACC9B,KAAK,CAAC0B,MAAM,GAAGI;gBACjB;gBAEFpB,QAAQ,KAAK,CAAC,IAAI,CAACmB;YACrB,OAAO;gBACLE,IAAAA,sBAAAA,MAAAA,AAAAA,EACE,CAACN,qBAAqB,QAAQ,CAACC,QAC/B,CAAC,uBAAuB,EAAEA,MAAM,6BAA6B,EAAER,UAAU;gBAE3EvB,MAAM,CAAC,OAAO,EAAE+B,MAAM,6BAA6B,EAAER,UAAU;YACjE;QACF;QAEA,MAAMc,OAKF;YACF,MAAM;YACN,SAASd;YACT,SAASL,KAAK,OAAO;YACrB,OAAOA,KAAK,KAAK;YACjB,UAAU,OAAOb,OAAOiC;gBACtB,MAAMC,SAASD,YAAY,IAAI,CAAC,MAAM;gBAEtCtC,MACE,oBACAuB,UACAlB,OACA,CAAC,4BAA4B,EAAEiC,YAAY,OAAO,EAAE,QAAQ;gBAG9D,MAAME,YAAYF,YAAY,SAAS;gBACvCF,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOI,WAAW;gBAElBV,qBAAqB,OAAO,CAAC,CAACC;oBAC5BK,IAAAA,sBAAAA,MAAAA,AAAAA,EACE/B,KAAK,CAAC0B,MAAM,EACZ,CAAC,OAAO,EAAEA,MAAM,yBAAyB,EAAER,SAAS,yCAAyC,EAAEA,SAAS,CAAC,CAAC;gBAE9G;gBAEAkB,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;gBAC3B,IAAI;oBACF,MAAMG,QAAQ,GAAG,CAAC;wBACf;4BACC,IAAI,IAAI,CAAC,SAAS,CAAC,kBAAkB,EAAE;gCACrC1C,MACE,CAAC,8DAA8D,EAAEyB,OAAO,IAAI,EAAE;gCAEhF,MAAM,IAAI,CAAC,SAAS,CAAC,kBAAkB,CAACA,OAAO,IAAI,EAAEpB;gCACrDL,MACE,CAAC,2DAA2D,EAAEyB,OAAO,IAAI,EAAE;4BAE/E;wBACF;wBACAkB,IAAAA,kCAAAA,KAAAA,AAAAA,EAAM;qBACP;gBACH,EAAE,OAAOC,eAAoB;oBAC3B,MAAMC,kBACJD,eAAe,WAAWE,OAAOF;oBACnC,MAAM,IAAIjB,MACR,CAAC,wCAAwC,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEoB,iBAAiB,EAC5E;wBAAE,OAAOD;oBAAc;gBAE3B;gBACAH,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;gBAE3B,MAAM,EAAEQ,wBAAwB,EAAE,GAAGP;gBACrC,IAAIO,AAA6BC,WAA7BD,0BACF,MAAM,IAAIpB,MACR;gBAIJ,IAAIF,OAAO,WAAW,EACpB,IAAI;oBACFpB,QAAQ4C,AAAAA,IAAAA,yBAAAA,gBAAAA,AAAAA,EAAiB5C,OAAOoB,OAAO,WAAW,EAAE;wBAClDsB;oBACF;gBACF,EAAE,OAAOG,OAAY;oBACnB,MAAM,IAAIvB,MACR,CAAC,8BAA8B,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEyB,MAAM,OAAO,CAAC,cAAc,EAAElB,KAAK,SAAS,CAAC3B,QAAQ,EACtG;wBAAE,OAAO6C;oBAAM;gBAEnB;gBAGFT,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;gBAE3BvC,MAAM,kBAAkByB,OAAO,IAAI;gBACnC,MAAM0B,WAAW1B,OAAO,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS;gBAChD,MAAM2B,eAAe,MAAMD,SAAS9C,OAAOiC;gBAC3CG,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;gBAC3BvC,MAAM,iBAAiByB,OAAO,IAAI,EAAE,WAAW2B;gBAE/CX,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;gBAE3B,MAAMc,mBACJ5B,OAAO,gBAAgB,IAAI,IAAI,CAAC,eAAe,IAAI;gBACrD,IAAI4B,mBAAmB,GACrB,MAAMV,AAAAA,IAAAA,kCAAAA,KAAAA,AAAAA,EAAMU;gBAGd,IAAI;oBACF,IAAI,IAAI,CAAC,SAAS,CAAC,iBAAiB,EAAE;wBACpCrD,MACE,CAAC,6DAA6D,EAAEyB,OAAO,IAAI,EAAE;wBAE/E,MAAM,IAAI,CAAC,SAAS,CAAC,iBAAiB,CAACA,OAAO,IAAI,EAAEpB;wBACpDL,MACE,CAAC,0DAA0D,EAAEyB,OAAO,IAAI,EAAE;oBAE9E;gBACF,EAAE,OAAOmB,eAAoB;oBAC3B,MAAMC,kBACJD,eAAe,WAAWE,OAAOF;oBACnC,MAAM,IAAIjB,MACR,CAAC,uCAAuC,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEoB,iBAAiB,EAC3E;wBAAE,OAAOD;oBAAc;gBAE3B;gBAEAH,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;gBAE3B,OAAO;oBACL,QAAQa;gBACV;YACF;QACF;QAEArC,QAAQ,KAAK,CAAC,IAAI,CAACsB;IACrB;IAEQ,iBACNnB,IAAyC,EACzCoC,mBAAiD,EACjDvC,OAAyB,EACzBwC,QAAgD,EACd;QAClC,MAAM,EAAEzC,SAAS,EAAEH,2BAA2B,EAAE6C,UAAU,EAAEC,WAAW,EAAE,GACvE1C;QAEF,IAAI2C,cAAcJ;QAElB,IAAI,AAAuB,YAAvB,OAAOI,aACTA,cAAc;YACZ,QAAQA;QACV;QAGF,IAAI5C,AAAckC,WAAdlC,WACF4C,cAAc;YACZ,GAAGA,WAAW;YACd5C;QACF;QAGF,IAAI0C,cAAc,CAACE,YAAY,UAAU,EACvCA,cAAc;YACZ,GAAGA,WAAW;YACd,YAAY;QACd;QAGF,MAAMC,cAAgD;YACpD,MAAM;YACN,SAAS;YACT,OAAOD;YACP,SAASxC,KAAK,OAAO;YACrB,UAAU,OAAOb,OAAOiC;gBACtB,MAAM,EAAED,IAAI,EAAE,GAAGC;gBACjB,IAAI,EAAEE,SAAS,EAAE,GAAGF;gBAEpBF,IAAAA,sBAAAA,MAAAA,AAAAA,EACE/B,OAAO,UAAUA,OAAO,MACxB,CAAC,qDAAqD,EAAE2B,KAAK,SAAS,CACpE3B,QACC;gBAGL,IAAI,CAACmC,WACHA,YAAY,MAAM,IAAI,CAAC,OAAO,CAAC,kBAAkB;gBAGnDJ,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOI,WAAW;gBAElB,MAAM,EAAEO,wBAAwB,EAAE,GAAGP;gBAErC,IAAIO,AAA6BC,WAA7BD,0BACF,MAAM,IAAIpB,MACR;gBAIJ,IAAIiC;gBACJ,IAAIC;gBAEJ,MAAMC,YAAY,CAACC;oBACjB,IAAI,CAACA,MACH;oBAEFH,aAAaG;oBACb1B,KAAK,GAAG,GAAG;wBACT0B;wBACA,aAAaA,KAAK,QAAQ,EAAE;oBAC9B;oBACA1B,KAAK,KAAK,GAAG0B,KAAK,QAAQ,EAAE;oBAC5B,IAAIA,KAAK,QAAQ,EAAE,iBACjB1B,KAAK,eAAe,GAAG0B,KAAK,QAAQ,CAAC,eAAe;oBAEtD,IAAIA,KAAK,QAAQ,EAAE,mBACjB1B,KAAK,iBAAiB,GAAG0B,KAAK,QAAQ,CAAC,iBAAiB;gBAE5D;gBAGA,MAAMC,kBAAkB/B,AAAAA,IAAAA,oCAAAA,uBAAAA,AAAAA,EAAwB5B,SAC5C4D,AAAAA,IAAAA,oCAAAA,oBAAAA,AAAAA,EAAqB5D,SACrB2C;gBACJ,MAAMkB,YAAY,CAAC,CAACF;gBAGpB,IAAIG;gBACJ,IACE,CAACD,aACD7D,MAAM,KAAK,IACX,IAAI,CAAC,SAAS,CAAC,uBAAuB,EAEtC,IAAI;oBACF8D,gBAAgB,MAAM,IAAI,CAAC,SAAS,CAAC,uBAAuB,CAAC;wBAC3D,QAAQ;4BAAC9D,MAAM,KAAK;yBAAC;oBACvB;gBACF,EAAE,OAAM,CAER;gBAGF,MAAM+D,mBAAmBD,gBACrBE,AAAAA,IAAAA,0BAAAA,qBAAAA,AAAAA,EAEEC,AAAAA,IAAAA,oCAAAA,oCAAAA,AAAAA,EACEH,eACApB,2BAEF,AAAwB,YAAxB,OAAO1C,MAAM,MAAM,GACfA,MAAM,MAAM,GACZA,MAAM,MAAM,EAAE,UAAU,MAE9B2C;gBAEJ,MAAMuB,aAAa,CAAC,CAACH;gBAErB,MAAMI,cAAcnE,MAAM,MAAM;gBAChC,MAAMoE,oBACJ,MAAM,IAAI,CAAC,SAAS,EAAE,iBAAiBD;gBACzC,MAAME,aAAaD,mBAAmB,cAAc;gBAEpD,MAAME,yBACJT,aAAaK,aACT,OACA,MAAMK,AAAAA,IAAAA,oCAAAA,qBAAAA,AAAAA,EACJ;oBACE,WAAW,IAAI,CAAC,SAAS;oBACzB,mBAAmB,IAAI,CAAC,SAAS;gBACnC,GACAF,YACAF,aACAnE,MAAM,SAAS;gBAIvB,MAAMwE,mBAAmBF,yBACrBG,AAAAA,IAAAA,oCAAAA,mCAAAA,AAAAA,EACEH,wBACA5B,4BAEFC;gBAEJ,IAAI+B,aAAa,CAAC,CAACF;gBACnB,MAAMtC,SAASD,YAAY,IAAI,CAAC,MAAM;gBACtC,IAAI0C;gBAEJ,IAAID,YAAY;oBACd,MAAME,eAAeP;oBACrB,MAAMQ,eAAeD,cAAc;oBAInC,MAAME,uBACJC,OAAO,QAAQ,CACbC,QAAQ,GAAG,CAACC,0BAAAA,qCAAqCA,CAAC,IAAI,MACtD,OACG;oBACP,MAAMC,mBACJF,AAAmD,YAAnDA,QAAQ,GAAG,CAACG,0BAAAA,iCAAiCA,CAAC;oBAChD,MAAMC,qBACJJ,AAAqD,YAArDA,QAAQ,GAAG,CAACK,0BAAAA,mCAAmCA,CAAC;oBAElD,IAAI;wBACF,IAAIH,oBAAoBL,cAAc;4BACpC,MAAMS,SAASC,KAAK,IAAI,CACrBf,AAAAA,CAAAA,iBAAkB,MAAM,CAAC,EAAE,GAAGK,YAAY,CAAC,EAAC,KAAM,IAChDL,AAAAA,CAAAA,iBAAkB,MAAM,CAAC,EAAE,GAAGK,YAAY,CAAC,EAAC,KAAM;4BAGvDlF,MAAM,4BAA4B;gCAChCkF;gCACA,eAAeL,iBAAkB,MAAM;gCACvC,QAAQe,KAAK,KAAK,CAACD;gCACnB,WAAWR;4BACb;4BAEA,IAAIQ,SAASR,sBAAsB;gCACjCnF,MACE,gEACA;oCAAE2F;oCAAQ,WAAWR;gCAAqB;gCAE5CJ,aAAa;4BACf;wBACF;wBAEA,IAAIA,cAAcU,oBAAoB;4BACpC,MAAMI,eAAe,MAAM,IAAI,CAAC,OAAO,CAAC,mBAAmB,CACzDhB,iBAAkB,MAAM,EACxBL,aACA7D,6BACA6B;4BAEF,IAAKqD,aAAa,IAAI,EAWpB7F,MAAM,4CAA4C;gCAChD,aAAa6F,aAAa,WAAW;4BACvC;iCAbsB;gCACtB7F,MACE,mEACA;oCACE,QAAQ6F,aAAa,MAAM;oCAC3B,aAAaA,aAAa,WAAW;oCACrC,QAAQrB;gCACV;gCAEFO,aAAa;4BACf;wBAKF;oBACF,EAAE,OAAOe,aAAa;wBACpB9F,MACE,mDACA8F;wBAEFf,aAAa;oBACf;gBACF;gBAEA,IAAI,CAACR,cAAc,CAACQ,cAAc,CAACb,WACjC,IAAI;oBACFzB,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;oBAC3BsB,eAAe,MAAM,IAAI,CAAC,OAAO,CAAC,MAAM,CACtCxD,OACA;wBACE,SAASmC;oBACX,GACA7B,6BACA8C;oBAEFK,UAAUD,aAAa,IAAI;oBAC3BmB,sBAAsBnB,aAAa,OAAO;gBAC5C,EAAE,OAAOX,OAAO;oBACd,IAAIA,iBAAiB6C,kCAAAA,YAAYA,EAC/BjC,UAAUZ,MAAM,IAAI;oBAEtB,MAAMA;gBACR,SAAU;oBACRT,IAAAA,wCAAAA,kBAAAA,AAAAA,EAAmBF,QAAQ;gBAC7B;gBAGF,MAAMyD,UACJhC,mBACAI,oBACAS,oBACAG;gBAGF,MAAMiB,2BAA2B/F,iBAC/BuE,mBAAmB,cAAc;gBAGnC,IAAIyB;gBAOJ,IACEF,WACA,IAAI,CAAC,SAAS,IACd,CAACjB,cACA,EAACb,aAAa,CAAC+B,wBAAuB,KACvC5F,OAAO,cAAc,OAErB,IAAI,IAAI,CAAC,SAAS,CAAC,oBAAoB,EACrC,IAAI;oBAGF,IAAI8F,gBAAkCH,QAAQ,MAAM;oBACpD,IAAIjD,AAA6B,MAA7BA,0BAAgC;wBAClCoD,gBAAgB;4BACdP,KAAK,KAAK,CAACI,QAAQ,MAAM,CAAC,EAAE,GAAGjD;4BAC/B6C,KAAK,KAAK,CAACI,QAAQ,MAAM,CAAC,EAAE,GAAGjD;yBAChC;wBACD/C,MACE,8DACAgG,QAAQ,MAAM,EACdG;oBAEJ;oBAEA,MAAMC,UAAU,MAAM,IAAI,CAAC,SAAS,CAAC,oBAAoB,CACvDD,eACA;wBACE,mBACE,AAAwB,YAAxB,OAAO9F,MAAM,MAAM,GACfA,MAAM,MAAM,GACZA,MAAM,MAAM,EAAE;wBACpB,aAAaM;oBACf;oBAEF,IAAIT,iBAAiBkG,UAAU;wBAC7BA,QAAQ,YAAY,GAAGD;wBACvBnG,MACE,uCACAwE,aACA4B;wBAEFF,oBAAoBE;wBACpB,MAAM,IAAI,CAAC,SAAS,CAAC,yBAAyB,CAC5C;4BACE,MAAM;4BACN,QAAQ5B;4BACR,OAAO4B;wBACT,GACA3B;oBAEJ,OACEzE,MACE,yDACAwE;gBAGN,EAAE,OAAOtB,OAAO;oBACdlD,MAAM,mCAAmCkD;gBAC3C;qBAEAlD,MAAM;gBAIV,IAAI,CAACgG,SAAS;oBACZ,IAAIpC,YACF,MAAM,IAAImC,kCAAAA,YAAYA,CACpB,CAAC,oBAAoB,EAAE1F,MAAM,MAAM,EAAE,EACrCuD;oBAGJ,MAAM,IAAIjC,MAAM,CAAC,mBAAmB,EAAEtB,MAAM,MAAM,EAAE;gBACtD;gBAEA,IAAIgG;gBAEJ,IAAInC,WACFmC,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACP,MAAMhG,MAAM,IAAI;oBAClB;gBACF;qBACK,IAAIkE,YACT8B,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACP,OAAOhG,MAAM,KAAK;oBACpB;gBACF;qBACK,IAAI0E,YACTsB,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACP3B;wBACA,aAAawB;oBACf;gBACF;gBAGF3C,WAAWyC;gBAEX,OAAO;oBACL,QAAQ;wBACN,SAAS;4BACP,GAAGA,OAAO;4BAEV,KAAKxD,UAAU,aAAa;wBAC9B;oBACF;oBACA6D;gBACF;YACF;QACF;QAEA,OAAO1C;IACT;IAvnBA,YAAY,EACV2C,iBAAiB,EACjBC,OAAO,EACPC,SAAS,EACThF,WAAW,EACXiF,eAAe,EACC,CAAE;QAhBpB,uBAAiB,aAAjB;QAEA,uBAAiB,WAAjB;QAEA,uBAAiB,aAAjB;QAEA,uBAAiB,eAAjB;QAEA,uBAAiB,mBAAjB;QASE,IAAI,CAAC,SAAS,GAAGH;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC;QACjB,IAAI,CAAC,WAAW,GAAGhF;QACnB,IAAI,CAAC,eAAe,GAAGiF;IACzB;AA4mBF"}
@@ -201,7 +201,7 @@ async function matchElementFromCache(context, cacheEntry, cachePrompt, cacheable
201
201
  return;
202
202
  }
203
203
  }
204
- const getMidsceneVersion = ()=>"1.5.14";
204
+ const getMidsceneVersion = ()=>"1.6.0";
205
205
  const parsePrompt = (prompt)=>{
206
206
  if ('string' == typeof prompt) return {
207
207
  textPrompt: prompt,
@@ -0,0 +1,455 @@
1
+ "use strict";
2
+ var __webpack_require__ = {};
3
+ (()=>{
4
+ __webpack_require__.d = (exports1, definition)=>{
5
+ for(var key in definition)if (__webpack_require__.o(definition, key) && !__webpack_require__.o(exports1, key)) Object.defineProperty(exports1, key, {
6
+ enumerable: true,
7
+ get: definition[key]
8
+ });
9
+ };
10
+ })();
11
+ (()=>{
12
+ __webpack_require__.o = (obj, prop)=>Object.prototype.hasOwnProperty.call(obj, prop);
13
+ })();
14
+ (()=>{
15
+ __webpack_require__.r = (exports1)=>{
16
+ if ('undefined' != typeof Symbol && Symbol.toStringTag) Object.defineProperty(exports1, Symbol.toStringTag, {
17
+ value: 'Module'
18
+ });
19
+ Object.defineProperty(exports1, '__esModule', {
20
+ value: true
21
+ });
22
+ };
23
+ })();
24
+ var __webpack_exports__ = {};
25
+ __webpack_require__.r(__webpack_exports__);
26
+ __webpack_require__.d(__webpack_exports__, {
27
+ AiAssertDiff: ()=>AiAssertDiff,
28
+ AiAssertVideo: ()=>AiAssertVideo,
29
+ AiAssertElement: ()=>AiAssertElement
30
+ });
31
+ const logger_namespaceObject = require("@midscene/shared/logger");
32
+ const index_js_namespaceObject = require("./service-caller/index.js");
33
+ const debug = (0, logger_namespaceObject.getDebug)('ai:assert');
34
+ const DEFAULT_SYSTEM_CHECK_PROMPT = `请检查以下系统级问题:
35
+
36
+ 1. **白屏检测**:页面是否完全白屏或大面积空白
37
+ 2. **布局遮挡**:重要元素是否存在组件间重叠遮挡
38
+ 3. **未加载内容**:页面是否有加载中的占位符、骨架屏
39
+ 4. **错误提示**:页面是否显示错误信息、异常提示
40
+ 5. **后端错误**:是否有后端请求失败的提示,如"活动太火爆了"、报错 Toast、错误码 等
41
+
42
+ 如果发现以上问题,请在 systemCheckResults 中标注对应字段为 true。`;
43
+ function buildAssertionPrompt(options) {
44
+ const { assertion, businessContext, systemCheckPrompt } = options;
45
+ let prompt = `## 用户断言描述
46
+ ${assertion}`;
47
+ if (businessContext) prompt += `
48
+
49
+ ## 业务知识上下文
50
+ ${businessContext}`;
51
+ if (systemCheckPrompt) prompt += `
52
+
53
+ ## 系统校验规则
54
+ ${systemCheckPrompt}`;
55
+ prompt += `
56
+
57
+ ## 输出要求
58
+ 请以 JSON 格式输出断言结果:
59
+ {
60
+ "pass": boolean, // 断言是否通过
61
+ "thought": string, // 思考过程
62
+ "reason": string, // 失败原因(如果失败)
63
+ "systemCheckResults": { // 系统校验结果(可选)
64
+ "whiteScreen": boolean,
65
+ "layoutBlocked": boolean,
66
+ "loadingContent": boolean,
67
+ "errorPrompt": boolean,
68
+ "backendError": boolean
69
+ }
70
+ }`;
71
+ return prompt;
72
+ }
73
+ async function AiAssertElement(options) {
74
+ const { beforeScreenshot, afterScreenshot, assertion, businessContext, enableSystemCheck = true, customSystemCheckRules, modelConfig, abortSignal } = options;
75
+ const systemPrompt = `你是一个自动化测试断言专家。请根据以下信息判断断言是否通过。
76
+
77
+ 你需要:
78
+ 1. 分析操作前后的截图变化(如果提供了操作前截图)
79
+ 2. 验证用户的断言描述是否成立
80
+ 3. 检查是否存在系统级问题(如果启用)
81
+ 4. 给出详细的思考过程和判断结果`;
82
+ const systemCheckPrompt = enableSystemCheck ? customSystemCheckRules || DEFAULT_SYSTEM_CHECK_PROMPT : '';
83
+ const userContent = [];
84
+ if (beforeScreenshot) {
85
+ userContent.push({
86
+ type: 'text',
87
+ text: '## 操作前截图(执行操作前的页面状态)'
88
+ });
89
+ userContent.push({
90
+ type: 'image_url',
91
+ image_url: {
92
+ url: beforeScreenshot,
93
+ detail: 'high'
94
+ }
95
+ });
96
+ userContent.push({
97
+ type: 'text',
98
+ text: '## 操作后截图(执行操作后的页面状态)'
99
+ });
100
+ } else userContent.push({
101
+ type: 'text',
102
+ text: '## 当前页面截图'
103
+ });
104
+ userContent.push({
105
+ type: 'image_url',
106
+ image_url: {
107
+ url: afterScreenshot,
108
+ detail: 'high'
109
+ }
110
+ });
111
+ userContent.push({
112
+ type: 'text',
113
+ text: buildAssertionPrompt({
114
+ assertion,
115
+ businessContext,
116
+ systemCheckPrompt
117
+ })
118
+ });
119
+ const msgs = [
120
+ {
121
+ role: 'system',
122
+ content: systemPrompt
123
+ },
124
+ {
125
+ role: 'user',
126
+ content: userContent
127
+ }
128
+ ];
129
+ debug('calling AI for assertion:', assertion);
130
+ try {
131
+ const result = await (0, index_js_namespaceObject.callAIWithObjectResponse)(msgs, modelConfig, {
132
+ abortSignal
133
+ });
134
+ debug('assertion result:', result.content);
135
+ return {
136
+ pass: result.content.pass,
137
+ thought: result.content.thought || '',
138
+ reason: result.content.reason,
139
+ usage: result.usage,
140
+ systemCheckResults: result.content.systemCheckResults,
141
+ rawResponse: JSON.stringify(result.content)
142
+ };
143
+ } catch (error) {
144
+ debug('assertion error:', error);
145
+ throw error;
146
+ }
147
+ }
148
+ const DIFF_SYSTEM_PROMPT = `你是一个自动化测试的图像对比专家。你的任务是对比当前页面截图与基准图片,判断页面样式是否符合预期。
149
+
150
+ 你需要:
151
+ 1. 仔细对比两张图片的布局、颜色、元素位置
152
+ 2. 分析差异是否在可接受范围内
153
+ 3. 识别动态内容(如时间、日期)导致的差异
154
+ 4. 给出详细的对比分析和判断结果`;
155
+ function buildDiffPrompt(options) {
156
+ const { assertion, businessContext, diffThreshold = 0.1, ignoreRegions, ignoreDynamicContent, strictMode } = options;
157
+ let prompt = `## 图像对比任务
158
+
159
+ 请对比基准图片(预期样式)和当前截图(实际样式),判断页面是否符合预期。
160
+
161
+ ## 用户断言描述
162
+ ${assertion}`;
163
+ if (businessContext) prompt += `
164
+
165
+ ## 业务知识上下文
166
+ ${businessContext}`;
167
+ prompt += `
168
+
169
+ ## 对比要求
170
+
171
+ 请从以下维度进行对比分析:
172
+
173
+ 1. **布局一致性**:页面布局是否与基准图片一致
174
+ 2. **颜色一致性**:主要颜色是否与基准图片一致
175
+ 3. **元素位置**:关键元素的位置是否与基准图片一致
176
+ 4. **文字内容**:文字内容是否与基准图片一致
177
+ 5. **图片资源**:图片、图标等资源是否正确加载
178
+
179
+ ## 差异阈值
180
+ 允许的差异阈值:${100 * diffThreshold}%`;
181
+ if (ignoreRegions && ignoreRegions.length > 0) {
182
+ prompt += `
183
+
184
+ ## 忽略对比的区域
185
+ 以下区域不参与对比:`;
186
+ ignoreRegions.forEach((region, index)=>{
187
+ prompt += `
188
+ ${index + 1}. 位置 (${region.x}, ${region.y}),尺寸 ${region.width}x${region.height}`;
189
+ });
190
+ }
191
+ if (ignoreDynamicContent) prompt += `
192
+
193
+ ## 动态内容处理
194
+ 请忽略以下动态内容导致的差异:
195
+ - 时间显示
196
+ - 日期显示
197
+ - 随机验证码
198
+ - 动画效果`;
199
+ if (strictMode) prompt += `
200
+
201
+ ## 严格模式
202
+ 当前为严格模式,任何差异都视为失败。`;
203
+ prompt += `
204
+
205
+ ## 输出要求
206
+ 请以 JSON 格式输出对比结果:
207
+ {
208
+ "pass": boolean, // 断言是否通过
209
+ "thought": string, // 对比分析过程
210
+ "reason": string, // 失败原因(如果失败)
211
+ "diffDetails": { // 差异详情
212
+ "layoutMatch": boolean, // 布局是否匹配
213
+ "colorMatch": boolean, // 颜色是否匹配
214
+ "elementPositionMatch": boolean, // 元素位置是否匹配
215
+ "textContentMatch": boolean, // 文字内容是否匹配
216
+ "resourceMatch": boolean, // 资源是否匹配
217
+ "acceptableDifferences": string[], // 可接受的差异列表
218
+ "unacceptableDifferences": string[] // 不可接受的差异列表
219
+ }
220
+ }`;
221
+ return prompt;
222
+ }
223
+ async function AiAssertDiff(options) {
224
+ const { currentScreenshot, referenceImages, assertion, businessContext, diffThreshold, ignoreRegions, ignoreDynamicContent, strictMode, modelConfig, abortSignal } = options;
225
+ const userContent = [];
226
+ for (const refImage of referenceImages){
227
+ userContent.push({
228
+ type: 'text',
229
+ text: `## 基准图片:${refImage.name}`
230
+ });
231
+ userContent.push({
232
+ type: 'image_url',
233
+ image_url: {
234
+ url: refImage.url,
235
+ detail: 'high'
236
+ }
237
+ });
238
+ }
239
+ userContent.push({
240
+ type: 'text',
241
+ text: '## 当前截图(实际样式)'
242
+ });
243
+ userContent.push({
244
+ type: 'image_url',
245
+ image_url: {
246
+ url: currentScreenshot,
247
+ detail: 'high'
248
+ }
249
+ });
250
+ userContent.push({
251
+ type: 'text',
252
+ text: buildDiffPrompt({
253
+ assertion,
254
+ businessContext,
255
+ diffThreshold,
256
+ ignoreRegions,
257
+ ignoreDynamicContent,
258
+ strictMode
259
+ })
260
+ });
261
+ const msgs = [
262
+ {
263
+ role: 'system',
264
+ content: DIFF_SYSTEM_PROMPT
265
+ },
266
+ {
267
+ role: 'user',
268
+ content: userContent
269
+ }
270
+ ];
271
+ debug('calling AI for diff assertion:', assertion);
272
+ try {
273
+ const result = await (0, index_js_namespaceObject.callAIWithObjectResponse)(msgs, modelConfig, {
274
+ abortSignal
275
+ });
276
+ debug('diff assertion result:', result.content);
277
+ return {
278
+ pass: result.content.pass,
279
+ thought: result.content.thought || '',
280
+ reason: result.content.reason,
281
+ diffDetails: result.content.diffDetails,
282
+ usage: result.usage,
283
+ rawResponse: JSON.stringify(result.content)
284
+ };
285
+ } catch (error) {
286
+ debug('diff assertion error:', error);
287
+ throw error;
288
+ }
289
+ }
290
+ const VIDEO_SYSTEM_PROMPT = `你是一个自动化测试的视频/动画分析专家。你的任务是对比当前录制的视频与基准视频,判断动画效果是否符合预期。
291
+
292
+ 你需要:
293
+ 1. 分析视频中的动画流畅度、时长、完整性
294
+ 2. 对比基准视频与当前视频的差异
295
+ 3. 识别动画类型(过渡动画、加载动画、交互动画等)
296
+ 4. 检测动画质量问题(卡顿、跳帧等)
297
+ 5. 给出详细的分析过程和判断结果`;
298
+ function buildVideoPrompt(options) {
299
+ const { assertion, businessContext, videoOptions } = options;
300
+ let prompt = `## 视频/动画对比任务
301
+
302
+ 请对比基准视频(预期动画效果)和当前录制的视频(实际动画效果),判断动画是否符合预期。
303
+
304
+ ## 用户断言描述
305
+ ${assertion}`;
306
+ if (businessContext) prompt += `
307
+
308
+ ## 业务知识上下文
309
+ ${businessContext}`;
310
+ prompt += `
311
+
312
+ ## 分析要求
313
+
314
+ 请从以下维度进行动画分析:
315
+
316
+ 1. **动画流畅度**:动画是否流畅,有无卡顿、跳帧
317
+ 2. **动画时长**:动画时长是否在预期范围内
318
+ 3. **动画完整性**:动画是否完整执行,有无中断
319
+ 4. **动画类型**:识别动画类型(过渡、加载、交互等)
320
+ 5. **关键帧匹配**:关键时间点的画面是否符合预期`;
321
+ if (videoOptions) {
322
+ if (videoOptions.checkSmoothness) prompt += `
323
+
324
+ ## 流畅度检测
325
+ 流畅度阈值:${videoOptions.smoothnessThreshold || 60}/100
326
+ 请评估动画流畅度并给出评分。`;
327
+ if (videoOptions.checkDuration && videoOptions.expectedDuration) prompt += `
328
+
329
+ ## 时长检测
330
+ 预期时长范围:${videoOptions.expectedDuration.min || 0}秒 - ${videoOptions.expectedDuration.max || '无限制'}秒`;
331
+ if (videoOptions.keyframes && videoOptions.keyframes.timestamps.length > 0) {
332
+ prompt += `
333
+
334
+ ## 关键帧验证
335
+ 需要在以下时间点验证画面:`;
336
+ videoOptions.keyframes.timestamps.forEach((ts, index)=>{
337
+ const desc = videoOptions.keyframes?.descriptions?.[index] || '未描述';
338
+ prompt += `
339
+ - ${ts}秒:${desc}`;
340
+ });
341
+ }
342
+ if (videoOptions.animationType && 'auto' !== videoOptions.animationType) prompt += `
343
+
344
+ ## 动画类型
345
+ 预期动画类型:${videoOptions.animationType}`;
346
+ }
347
+ prompt += `
348
+
349
+ ## 输出要求
350
+ 请以 JSON 格式输出对比结果:
351
+ {
352
+ "pass": boolean, // 断言是否通过
353
+ "thought": string, // 分析过程
354
+ "reason": string, // 失败原因(如果失败)
355
+ "videoDetails": { // 视频详情
356
+ "smoothnessScore": number, // 流畅度评分(0-100)
357
+ "duration": number, // 动画时长(秒)
358
+ "isComplete": boolean, // 动画是否完整
359
+ "keyframeMatches": [ // 关键帧匹配结果
360
+ {
361
+ "timestamp": number,
362
+ "matched": boolean,
363
+ "description": string
364
+ }
365
+ ],
366
+ "detectedAnimationType": string, // 检测到的动画类型
367
+ "qualityAssessment": { // 质量评估
368
+ "hasStuttering": boolean, // 是否有卡顿
369
+ "hasFrameDropping": boolean, // 是否有跳帧
370
+ "averageFrameInterval": number, // 平均帧间隔(ms)
371
+ "frameIntervalStdDev": number // 帧间隔标准差
372
+ },
373
+ "acceptableDifferences": string[], // 可接受的差异列表
374
+ "unacceptableDifferences": string[] // 不可接受的差异列表
375
+ }
376
+ }`;
377
+ return prompt;
378
+ }
379
+ async function AiAssertVideo(options) {
380
+ const { currentVideoFrames, assertion, businessContext, videoOptions, modelConfig, abortSignal } = options;
381
+ const MAX_DURATION = 5;
382
+ const DEFAULT_FPS = 30;
383
+ const MAX_FRAMES = MAX_DURATION * DEFAULT_FPS;
384
+ if (currentVideoFrames.length > MAX_FRAMES) throw new Error(`Video frames exceed maximum limit. Maximum allowed: ${MAX_FRAMES} frames (${MAX_DURATION}s at ${DEFAULT_FPS}fps), got: ${currentVideoFrames.length} frames`);
385
+ const userContent = [];
386
+ userContent.push({
387
+ type: 'text',
388
+ text: `## 当前视频(共 ${currentVideoFrames.length} 帧,最大限制 ${MAX_FRAMES} 帧)
389
+
390
+ 以下为当前录制视频的关键帧截图:`
391
+ });
392
+ const frameInterval = Math.max(1, Math.floor(currentVideoFrames.length / 10));
393
+ for(let i = 0; i < currentVideoFrames.length; i += frameInterval){
394
+ userContent.push({
395
+ type: 'text',
396
+ text: `帧 ${i + 1}/${currentVideoFrames.length}`
397
+ });
398
+ userContent.push({
399
+ type: 'image_url',
400
+ image_url: {
401
+ url: currentVideoFrames[i],
402
+ detail: 'low'
403
+ }
404
+ });
405
+ }
406
+ userContent.push({
407
+ type: 'text',
408
+ text: buildVideoPrompt({
409
+ assertion,
410
+ businessContext,
411
+ videoOptions
412
+ })
413
+ });
414
+ const msgs = [
415
+ {
416
+ role: 'system',
417
+ content: VIDEO_SYSTEM_PROMPT
418
+ },
419
+ {
420
+ role: 'user',
421
+ content: userContent
422
+ }
423
+ ];
424
+ debug('calling AI for video assertion:', assertion);
425
+ try {
426
+ const result = await (0, index_js_namespaceObject.callAIWithObjectResponse)(msgs, modelConfig, {
427
+ abortSignal
428
+ });
429
+ debug('video assertion result:', result.content);
430
+ return {
431
+ pass: result.content.pass,
432
+ thought: result.content.thought || '',
433
+ reason: result.content.reason,
434
+ videoDetails: result.content.videoDetails,
435
+ usage: result.usage,
436
+ rawResponse: JSON.stringify(result.content)
437
+ };
438
+ } catch (error) {
439
+ debug('video assertion error:', error);
440
+ throw error;
441
+ }
442
+ }
443
+ exports.AiAssertDiff = __webpack_exports__.AiAssertDiff;
444
+ exports.AiAssertElement = __webpack_exports__.AiAssertElement;
445
+ exports.AiAssertVideo = __webpack_exports__.AiAssertVideo;
446
+ for(var __rspack_i in __webpack_exports__)if (-1 === [
447
+ "AiAssertDiff",
448
+ "AiAssertElement",
449
+ "AiAssertVideo"
450
+ ].indexOf(__rspack_i)) exports[__rspack_i] = __webpack_exports__[__rspack_i];
451
+ Object.defineProperty(exports, '__esModule', {
452
+ value: true
453
+ });
454
+
455
+ //# sourceMappingURL=assert.js.map