@midscene/core 0.26.7-beta-20250818081955.0 → 0.26.7-beta-20250820105545.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (65) hide show
  1. package/dist/es/ai-model/common.mjs +58 -16
  2. package/dist/es/ai-model/common.mjs.map +1 -1
  3. package/dist/es/ai-model/index.mjs +3 -3
  4. package/dist/es/ai-model/inspect.mjs +28 -16
  5. package/dist/es/ai-model/inspect.mjs.map +1 -1
  6. package/dist/es/ai-model/llm-planning.mjs +26 -23
  7. package/dist/es/ai-model/llm-planning.mjs.map +1 -1
  8. package/dist/es/ai-model/prompt/llm-planning.mjs +50 -23
  9. package/dist/es/ai-model/prompt/llm-planning.mjs.map +1 -1
  10. package/dist/es/ai-model/prompt/playwright-generator.mjs +9 -3
  11. package/dist/es/ai-model/prompt/playwright-generator.mjs.map +1 -1
  12. package/dist/es/ai-model/prompt/util.mjs +2 -2
  13. package/dist/es/ai-model/prompt/util.mjs.map +1 -1
  14. package/dist/es/ai-model/prompt/yaml-generator.mjs +9 -3
  15. package/dist/es/ai-model/prompt/yaml-generator.mjs.map +1 -1
  16. package/dist/es/ai-model/service-caller/index.mjs +72 -118
  17. package/dist/es/ai-model/service-caller/index.mjs.map +1 -1
  18. package/dist/es/ai-model/ui-tars-planning.mjs +5 -5
  19. package/dist/es/ai-model/ui-tars-planning.mjs.map +1 -1
  20. package/dist/es/index.mjs +3 -2
  21. package/dist/es/index.mjs.map +1 -1
  22. package/dist/es/insight/index.mjs +13 -61
  23. package/dist/es/insight/index.mjs.map +1 -1
  24. package/dist/es/types.mjs.map +1 -1
  25. package/dist/es/utils.mjs +5 -6
  26. package/dist/es/utils.mjs.map +1 -1
  27. package/dist/lib/ai-model/common.js +80 -20
  28. package/dist/lib/ai-model/common.js.map +1 -1
  29. package/dist/lib/ai-model/index.js +14 -5
  30. package/dist/lib/ai-model/inspect.js +27 -15
  31. package/dist/lib/ai-model/inspect.js.map +1 -1
  32. package/dist/lib/ai-model/llm-planning.js +25 -22
  33. package/dist/lib/ai-model/llm-planning.js.map +1 -1
  34. package/dist/lib/ai-model/prompt/llm-planning.js +52 -25
  35. package/dist/lib/ai-model/prompt/llm-planning.js.map +1 -1
  36. package/dist/lib/ai-model/prompt/playwright-generator.js +9 -3
  37. package/dist/lib/ai-model/prompt/playwright-generator.js.map +1 -1
  38. package/dist/lib/ai-model/prompt/util.js +2 -2
  39. package/dist/lib/ai-model/prompt/util.js.map +1 -1
  40. package/dist/lib/ai-model/prompt/yaml-generator.js +9 -3
  41. package/dist/lib/ai-model/prompt/yaml-generator.js.map +1 -1
  42. package/dist/lib/ai-model/service-caller/index.js +75 -124
  43. package/dist/lib/ai-model/service-caller/index.js.map +1 -1
  44. package/dist/lib/ai-model/ui-tars-planning.js +5 -5
  45. package/dist/lib/ai-model/ui-tars-planning.js.map +1 -1
  46. package/dist/lib/index.js +20 -4
  47. package/dist/lib/index.js.map +1 -1
  48. package/dist/lib/insight/index.js +10 -58
  49. package/dist/lib/insight/index.js.map +1 -1
  50. package/dist/lib/types.js.map +1 -1
  51. package/dist/lib/utils.js +4 -5
  52. package/dist/lib/utils.js.map +1 -1
  53. package/dist/types/ai-model/common.d.ts +160 -7
  54. package/dist/types/ai-model/index.d.ts +2 -2
  55. package/dist/types/ai-model/inspect.d.ts +2 -0
  56. package/dist/types/ai-model/llm-planning.d.ts +1 -1
  57. package/dist/types/ai-model/prompt/llm-planning.d.ts +2 -2
  58. package/dist/types/ai-model/prompt/util.d.ts +2 -1
  59. package/dist/types/ai-model/service-caller/index.d.ts +6 -6
  60. package/dist/types/ai-model/ui-tars-planning.d.ts +3 -1
  61. package/dist/types/index.d.ts +3 -1
  62. package/dist/types/insight/index.d.ts +1 -4
  63. package/dist/types/types.d.ts +8 -11
  64. package/dist/types/yaml.d.ts +1 -0
  65. package/package.json +4 -3
@@ -1 +1 @@
1
- {"version":3,"file":"types.js","sources":["webpack://@midscene/core/webpack/runtime/compat_get_default_export","webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/types.ts"],"sourcesContent":["// getDefaultExport function for compatibility with non-ESM modules\n__webpack_require__.n = (module) => {\n\tvar getter = module && module.__esModule ?\n\t\t() => (module['default']) :\n\t\t() => (module);\n\t__webpack_require__.d(getter, { a: getter });\n\treturn getter;\n};\n","__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};","/* eslint-disable @typescript-eslint/no-explicit-any */\n\nimport type { NodeType } from '@midscene/shared/constants';\nimport type {\n BaseElement,\n ElementTreeNode,\n Rect,\n Size,\n} from '@midscene/shared/types';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { DetailedLocateParam, MidsceneYamlFlowItem } from './yaml';\n\nexport type {\n ElementTreeNode,\n BaseElement,\n Rect,\n Size,\n Point,\n} from '@midscene/shared/types';\nexport * from './yaml';\n\nexport type AIUsageInfo = Record<string, any> & {\n prompt_tokens: number | undefined;\n completion_tokens: number | undefined;\n total_tokens: number | undefined;\n time_cost: number | undefined;\n model_name: string | undefined;\n};\n\n/**\n * openai\n *\n */\nexport enum AIResponseFormat {\n JSON = 'json_object',\n TEXT = 'text',\n}\n\nexport type AISingleElementResponseById = {\n id: string;\n reason?: string;\n text?: string;\n xpaths?: string[];\n};\n\nexport type AISingleElementResponseByPosition = {\n position?: {\n x: number;\n y: number;\n };\n bbox?: [number, number, number, number];\n reason: string;\n text: string;\n};\n\nexport type AISingleElementResponse = AISingleElementResponseById;\nexport interface AIElementLocatorResponse {\n elements: {\n id: string;\n reason?: string;\n text?: string;\n xpaths?: string[];\n }[];\n bbox?: [number, number, number, number];\n isOrderSensitive?: boolean;\n errors?: string[];\n}\n\nexport interface AIElementCoordinatesResponse {\n bbox: [number, number, number, number];\n isOrderSensitive?: boolean;\n errors?: string[];\n}\n\nexport type AIElementResponse =\n | AIElementLocatorResponse\n | AIElementCoordinatesResponse;\n\nexport interface AIDataExtractionResponse<DataDemand> {\n data: DataDemand;\n errors?: string[];\n thought?: string;\n}\n\nexport interface AISectionLocatorResponse {\n bbox: [number, number, number, number];\n references_bbox?: [number, number, number, number][];\n error?: string;\n}\n\nexport interface AIAssertionResponse {\n pass: boolean;\n thought: string;\n}\n\nexport interface AIDescribeElementResponse {\n description: string;\n error?: string;\n}\n\nexport interface LocatorValidatorOption {\n centerDistanceThreshold?: number;\n}\n\nexport interface LocateValidatorResult {\n pass: boolean;\n rect: Rect;\n center: [number, number];\n centerDistance?: number;\n}\n\nexport interface AgentDescribeElementAtPointResult {\n prompt: string;\n deepThink: boolean;\n verifyResult?: LocateValidatorResult;\n}\n\n/**\n * context\n */\n\nexport abstract class UIContext<ElementType extends BaseElement = BaseElement> {\n abstract screenshotBase64: string;\n\n abstract tree: ElementTreeNode<ElementType>;\n\n abstract size: Size;\n}\n\n/**\n * insight\n */\n\nexport type CallAIFn = <T>(\n messages: ChatCompletionMessageParam[],\n) => Promise<T>;\n\nexport interface InsightOptions {\n taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n aiVendorFn?: CallAIFn;\n}\n\nexport type EnsureObject<T> = { [K in keyof T]: any };\n\nexport type InsightAction = 'locate' | 'extract' | 'assert' | 'describe';\n\nexport type InsightExtractParam = string | Record<string, string>;\n\nexport type LocateResultElement = {\n id: string;\n indexId?: number;\n center: [number, number];\n rect: Rect;\n xpaths: string[];\n attributes: {\n nodeType: NodeType;\n [key: string]: string;\n };\n isOrderSensitive?: boolean;\n};\n\nexport interface LocateResult {\n element: LocateResultElement | null;\n rect?: Rect;\n}\n\nexport interface InsightTaskInfo {\n durationMs: number;\n formatResponse?: string;\n rawResponse?: string;\n usage?: AIUsageInfo;\n searchArea?: Rect;\n searchAreaRawResponse?: string;\n searchAreaUsage?: AIUsageInfo;\n}\n\nexport interface DumpMeta {\n sdkVersion: string;\n logTime: number;\n}\n\nexport interface ReportDumpWithAttributes {\n dumpString: string;\n attributes?: Record<string, any>;\n}\n\nexport interface InsightDump extends DumpMeta {\n type: 'locate' | 'extract' | 'assert';\n logId: string;\n userQuery: {\n element?: TUserPrompt;\n dataDemand?: InsightExtractParam;\n assertion?: TUserPrompt;\n };\n matchedElement: BaseElement[];\n matchedRect?: Rect;\n deepThink?: boolean;\n data: any;\n assertionPass?: boolean;\n assertionThought?: string;\n taskInfo: InsightTaskInfo;\n error?: string;\n output?: any;\n}\n\nexport type PartialInsightDumpFromSDK = Omit<\n InsightDump,\n 'sdkVersion' | 'logTime' | 'logId' | 'model_name'\n>;\n\nexport type DumpSubscriber = (dump: InsightDump) => Promise<void> | void;\n\n// intermediate variables to optimize the return value by AI\nexport interface LiteUISection {\n name: string;\n description: string;\n sectionCharacteristics: string;\n textIds: string[];\n}\n\nexport type ElementById = (id: string) => BaseElement | null;\n\nexport type InsightAssertionResponse = AIAssertionResponse & {\n usage?: AIUsageInfo;\n};\n\n/**\n * agent\n */\n\nexport type OnTaskStartTip = (tip: string) => Promise<void> | void;\n\nexport interface AgentWaitForOpt {\n checkIntervalMs?: number;\n timeoutMs?: number;\n}\n\nexport interface AgentAssertOpt {\n keepRawResponse?: boolean;\n}\n\n/**\n * planning\n *\n */\n\nexport interface PlanningLocateParam extends DetailedLocateParam {\n id?: string;\n bbox?: [number, number, number, number];\n}\n\nexport interface PlanningAction<ParamType = any> {\n thought?: string;\n type:\n | 'Locate'\n | 'Tap'\n | 'RightClick'\n | 'Hover'\n | 'Drag'\n | 'Input'\n | 'KeyboardPress'\n | 'Scroll'\n | 'Error'\n | 'Assert'\n | 'AssertWithoutThrow'\n | 'Sleep'\n | 'Finished'\n | 'AndroidBackButton'\n | 'AndroidHomeButton'\n | 'AndroidRecentAppsButton'\n | 'AndroidLongPress'\n | 'AndroidPull';\n param: ParamType;\n locate?: PlanningLocateParam | null;\n}\n\nexport interface PlanningAIResponse {\n action?: PlanningAction; // this is the qwen mode\n actions?: PlanningAction[];\n more_actions_needed_by_instruction: boolean;\n log: string;\n sleep?: number;\n error?: string;\n usage?: AIUsageInfo;\n rawResponse?: string;\n yamlFlow?: MidsceneYamlFlowItem[];\n yamlString?: string;\n}\n\nexport type PlanningActionParamTap = null;\nexport type PlanningActionParamHover = null;\nexport type PlanningActionParamRightClick = null;\n\nexport interface PlanningActionParamInputOrKeyPress {\n value: string;\n autoDismissKeyboard?: boolean;\n}\n\nexport interface PlanningActionParamAssert {\n assertion: TUserPrompt;\n}\n\nexport interface PlanningActionParamSleep {\n timeMs: number;\n}\n\nexport interface PlanningActionParamError {\n thought: string;\n}\n\nexport type PlanningActionParamWaitFor = AgentWaitForOpt & {\n assertion: string;\n};\n\nexport interface AndroidLongPressParam {\n duration?: number;\n}\n\nexport interface AndroidPullParam {\n direction: 'up' | 'down';\n distance?: number;\n duration?: number;\n}\n/**\n * misc\n */\n\nexport interface Color {\n name: string;\n hex: string;\n}\n\nexport interface BaseAgentParserOpt {\n selector?: string;\n ignoreMarker?: boolean;\n}\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PuppeteerParserOpt extends BaseAgentParserOpt {}\n\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PlaywrightParserOpt extends BaseAgentParserOpt {}\n\n/*\naction\n*/\nexport interface ExecutionTaskProgressOptions {\n onTaskStart?: (task: ExecutionTask) => Promise<void> | void;\n}\n\nexport interface ExecutionRecorderItem {\n type: 'screenshot';\n ts: number;\n screenshot?: string;\n timing?: string;\n}\n\nexport type ExecutionTaskType =\n | 'Planning'\n | 'Insight'\n | 'Action'\n | 'Assertion'\n | 'Log';\n\nexport interface ExecutorContext {\n task: ExecutionTask;\n element?: LocateResultElement | null;\n}\n\nexport interface ExecutionTaskApply<\n Type extends ExecutionTaskType = any,\n TaskParam = any,\n TaskOutput = any,\n TaskLog = any,\n> {\n type: Type;\n subType?: string;\n param?: TaskParam;\n thought?: string;\n locate?: PlanningLocateParam | null;\n pageContext?: UIContext;\n executor: (\n param: TaskParam,\n context: ExecutorContext,\n ) => // biome-ignore lint/suspicious/noConfusingVoidType: <explanation>\n | Promise<ExecutionTaskReturn<TaskOutput, TaskLog> | undefined | void>\n | undefined\n | void;\n}\n\nexport interface ExecutionTaskHitBy {\n from: string;\n context: Record<string, any>;\n}\n\nexport interface ExecutionTaskReturn<TaskOutput = unknown, TaskLog = unknown> {\n output?: TaskOutput;\n log?: TaskLog;\n recorder?: ExecutionRecorderItem[];\n hitBy?: ExecutionTaskHitBy;\n}\n\nexport type ExecutionTask<\n E extends ExecutionTaskApply<any, any, any> = ExecutionTaskApply<\n any,\n any,\n any\n >,\n> = E &\n ExecutionTaskReturn<\n E extends ExecutionTaskApply<any, any, infer TaskOutput, any>\n ? TaskOutput\n : unknown,\n E extends ExecutionTaskApply<any, any, any, infer TaskLog>\n ? TaskLog\n : unknown\n > & {\n status: 'pending' | 'running' | 'finished' | 'failed' | 'cancelled';\n error?: Error;\n errorMessage?: string;\n errorStack?: string;\n timing?: {\n start: number;\n end?: number;\n cost?: number;\n };\n usage?: AIUsageInfo;\n };\n\nexport interface ExecutionDump extends DumpMeta {\n name: string;\n description?: string;\n tasks: ExecutionTask[];\n}\n\n/*\ntask - insight-locate\n*/\nexport type ExecutionTaskInsightLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskInsightLocateOutput {\n element: LocateResultElement | null;\n}\n\nexport interface ExecutionTaskInsightDumpLog {\n dump?: InsightDump;\n}\n\nexport type ExecutionTaskInsightLocateApply = ExecutionTaskApply<\n 'Insight',\n ExecutionTaskInsightLocateParam,\n ExecutionTaskInsightLocateOutput,\n ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightLocate =\n ExecutionTask<ExecutionTaskInsightLocateApply>;\n\n/*\ntask - insight-query\n*/\nexport interface ExecutionTaskInsightQueryParam {\n dataDemand: InsightExtractParam;\n}\n\nexport interface ExecutionTaskInsightQueryOutput {\n data: any;\n}\n\nexport type ExecutionTaskInsightQueryApply = ExecutionTaskApply<\n 'Insight',\n ExecutionTaskInsightQueryParam,\n any,\n ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightQuery =\n ExecutionTask<ExecutionTaskInsightQueryApply>;\n\n/*\ntask - assertion\n*/\nexport interface ExecutionTaskInsightAssertionParam {\n assertion: string;\n}\n\nexport type ExecutionTaskInsightAssertionApply = ExecutionTaskApply<\n 'Insight',\n ExecutionTaskInsightAssertionParam,\n InsightAssertionResponse,\n ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightAssertion =\n ExecutionTask<ExecutionTaskInsightAssertionApply>;\n\n/*\ntask - action (i.e. interact) \n*/\nexport type ExecutionTaskActionApply<ActionParam = any> = ExecutionTaskApply<\n 'Action',\n ActionParam,\n void,\n void\n>;\n\nexport type ExecutionTaskAction = ExecutionTask<ExecutionTaskActionApply>;\n\n/*\ntask - Log\n*/\n\nexport type ExecutionTaskLogApply<\n LogParam = {\n content: string;\n },\n> = ExecutionTaskApply<'Log', LogParam, void, void>;\n\nexport type ExecutionTaskLog = ExecutionTask<ExecutionTaskLogApply>;\n\n/*\ntask - planning\n*/\n\nexport type ExecutionTaskPlanningApply = ExecutionTaskApply<\n 'Planning',\n {\n userInstruction: string;\n log?: string;\n },\n PlanningAIResponse\n>;\n\nexport type ExecutionTaskPlanning = ExecutionTask<ExecutionTaskPlanningApply>;\n\n/*\nGrouped dump\n*/\nexport interface GroupedActionDump {\n groupName: string;\n groupDescription?: string;\n modelName: string;\n modelDescription: string;\n executions: ExecutionDump[];\n}\n\nexport type PageType =\n | 'puppeteer'\n | 'playwright'\n | 'static'\n | 'chrome-extension-proxy'\n | 'android';\n\nexport interface StreamingCodeGenerationOptions {\n /** Whether to enable streaming output */\n stream?: boolean;\n /** Callback function to handle streaming chunks */\n onChunk?: StreamingCallback;\n /** Callback function to handle streaming completion */\n onComplete?: (finalCode: string) => void;\n /** Callback function to handle streaming errors */\n onError?: (error: Error) => void;\n}\n\nexport type StreamingCallback = (chunk: CodeGenerationChunk) => void;\n\nexport interface CodeGenerationChunk {\n /** The incremental content chunk */\n content: string;\n /** The reasoning content */\n reasoning_content: string;\n /** The accumulated content so far */\n accumulated: string;\n /** Whether this is the final chunk */\n isComplete: boolean;\n /** Token usage information if available */\n usage?: AIUsageInfo;\n}\n\nexport interface StreamingAIResponse {\n /** The final accumulated content */\n content: string;\n /** Token usage information */\n usage?: AIUsageInfo;\n /** Whether the response was streamed */\n isStreamed: boolean;\n}\n\nexport type TMultimodalPrompt = {\n /**\n * Support use image to inspect elements.\n * The \"images\" field is an object that uses image name as key and image url as value.\n * The image url can be a local path, a http link , or a base64 string.\n */\n images?: {\n name: string;\n url: string;\n }[];\n /**\n * By default, the image url in the \"images\" filed starts with `https://` or `http://` will be directly sent to the LLM.\n * In case the images are not accessible to the LLM (One common case is that image url is internal network only.), you can enable this option.\n * Then image will be download and convert to base64 format.\n */\n convertHttpImage2Base64?: boolean;\n};\n\nexport type TUserPrompt =\n | string\n | ({\n prompt: string;\n } & Partial<TMultimodalPrompt>);\n\nexport interface DeviceAction<ParamType = any> {\n name: string;\n interfaceAlias?: string;\n description?: string;\n paramSchema?: string;\n paramDescription?: string;\n location?: 'required' | 'optional' | false;\n whatToLocate?: string; // what to locate if location is required or optional\n call: (context: ExecutorContext, param: ParamType) => Promise<void> | void;\n}\n"],"names":["__webpack_require__","module","getter","definition","key","Object","obj","prop","Symbol","AIResponseFormat","UIContext"],"mappings":";;;;;;;;;;;;;;;;;IACAA,oBAAoB,CAAC,GAAG,CAACC;QACxB,IAAIC,SAASD,UAAUA,OAAO,UAAU,GACvC,IAAOA,MAAM,CAAC,UAAU,GACxB,IAAOA;QACRD,oBAAoB,CAAC,CAACE,QAAQ;YAAE,GAAGA;QAAO;QAC1C,OAAOA;IACR;;;ICPAF,oBAAoB,CAAC,GAAG,CAAC,UAASG;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGH,oBAAoB,CAAC,CAACG,YAAYC,QAAQ,CAACJ,oBAAoB,CAAC,CAAC,UAASI,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAJ,oBAAoB,CAAC,GAAG,CAACM,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFP,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOQ,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;IC2BO,IAAKI,mBAAAA,WAAAA,GAAAA,SAAAA,gBAAgB;;;eAAhBA;;IAwFL,MAAeC;IAMtB"}
1
+ {"version":3,"file":"types.js","sources":["webpack://@midscene/core/webpack/runtime/compat_get_default_export","webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/types.ts"],"sourcesContent":["// getDefaultExport function for compatibility with non-ESM modules\n__webpack_require__.n = (module) => {\n\tvar getter = module && module.__esModule ?\n\t\t() => (module['default']) :\n\t\t() => (module);\n\t__webpack_require__.d(getter, { a: getter });\n\treturn getter;\n};\n","__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};","/* eslint-disable @typescript-eslint/no-explicit-any */\n\nimport type { NodeType } from '@midscene/shared/constants';\nimport type {\n BaseElement,\n ElementTreeNode,\n Rect,\n Size,\n} from '@midscene/shared/types';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { z } from 'zod';\nimport type { DetailedLocateParam, MidsceneYamlFlowItem } from './yaml';\n\nexport type {\n ElementTreeNode,\n BaseElement,\n Rect,\n Size,\n Point,\n} from '@midscene/shared/types';\nexport * from './yaml';\n\nexport type AIUsageInfo = Record<string, any> & {\n prompt_tokens: number | undefined;\n completion_tokens: number | undefined;\n total_tokens: number | undefined;\n time_cost: number | undefined;\n model_name: string | undefined;\n};\n\n/**\n * openai\n *\n */\nexport enum AIResponseFormat {\n JSON = 'json_object',\n TEXT = 'text',\n}\n\nexport type AISingleElementResponseById = {\n id: string;\n reason?: string;\n text?: string;\n xpaths?: string[];\n};\n\nexport type AISingleElementResponseByPosition = {\n position?: {\n x: number;\n y: number;\n };\n bbox?: [number, number, number, number];\n reason: string;\n text: string;\n};\n\nexport type AISingleElementResponse = AISingleElementResponseById;\nexport interface AIElementLocatorResponse {\n elements: {\n id: string;\n reason?: string;\n text?: string;\n xpaths?: string[];\n }[];\n bbox?: [number, number, number, number];\n isOrderSensitive?: boolean;\n errors?: string[];\n}\n\nexport interface AIElementCoordinatesResponse {\n bbox: [number, number, number, number];\n isOrderSensitive?: boolean;\n errors?: string[];\n}\n\nexport type AIElementResponse =\n | AIElementLocatorResponse\n | AIElementCoordinatesResponse;\n\nexport interface AIDataExtractionResponse<DataDemand> {\n data: DataDemand;\n errors?: string[];\n thought?: string;\n}\n\nexport interface AISectionLocatorResponse {\n bbox: [number, number, number, number];\n references_bbox?: [number, number, number, number][];\n error?: string;\n}\n\nexport interface AIAssertionResponse {\n pass: boolean;\n thought: string;\n}\n\nexport interface AIDescribeElementResponse {\n description: string;\n error?: string;\n}\n\nexport interface LocatorValidatorOption {\n centerDistanceThreshold?: number;\n}\n\nexport interface LocateValidatorResult {\n pass: boolean;\n rect: Rect;\n center: [number, number];\n centerDistance?: number;\n}\n\nexport interface AgentDescribeElementAtPointResult {\n prompt: string;\n deepThink: boolean;\n verifyResult?: LocateValidatorResult;\n}\n\n/**\n * context\n */\n\nexport abstract class UIContext<ElementType extends BaseElement = BaseElement> {\n abstract screenshotBase64: string;\n\n abstract tree: ElementTreeNode<ElementType>;\n\n abstract size: Size;\n}\n\n/**\n * insight\n */\n\nexport type CallAIFn = <T>(\n messages: ChatCompletionMessageParam[],\n) => Promise<T>;\n\nexport interface InsightOptions {\n taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n aiVendorFn?: CallAIFn;\n}\n\nexport type EnsureObject<T> = { [K in keyof T]: any };\n\nexport type InsightAction = 'locate' | 'extract' | 'assert' | 'describe';\n\nexport type InsightExtractParam = string | Record<string, string>;\n\nexport type LocateResultElement = {\n center: [number, number];\n rect: Rect;\n id: string;\n indexId?: number;\n xpaths: string[];\n attributes: {\n nodeType: NodeType;\n [key: string]: string;\n };\n isOrderSensitive?: boolean;\n};\n\nexport interface LocateResult {\n element: LocateResultElement | null;\n rect?: Rect;\n}\n\nexport interface InsightTaskInfo {\n durationMs: number;\n formatResponse?: string;\n rawResponse?: string;\n usage?: AIUsageInfo;\n searchArea?: Rect;\n searchAreaRawResponse?: string;\n searchAreaUsage?: AIUsageInfo;\n}\n\nexport interface DumpMeta {\n sdkVersion: string;\n logTime: number;\n}\n\nexport interface ReportDumpWithAttributes {\n dumpString: string;\n attributes?: Record<string, any>;\n}\n\nexport interface InsightDump extends DumpMeta {\n type: 'locate' | 'extract' | 'assert';\n logId: string;\n userQuery: {\n element?: TUserPrompt;\n dataDemand?: InsightExtractParam;\n assertion?: TUserPrompt;\n };\n matchedElement: BaseElement[];\n matchedRect?: Rect;\n deepThink?: boolean;\n data: any;\n assertionPass?: boolean;\n assertionThought?: string;\n taskInfo: InsightTaskInfo;\n error?: string;\n output?: any;\n}\n\nexport type PartialInsightDumpFromSDK = Omit<\n InsightDump,\n 'sdkVersion' | 'logTime' | 'logId' | 'model_name'\n>;\n\nexport type DumpSubscriber = (dump: InsightDump) => Promise<void> | void;\n\n// intermediate variables to optimize the return value by AI\nexport interface LiteUISection {\n name: string;\n description: string;\n sectionCharacteristics: string;\n textIds: string[];\n}\n\nexport type ElementById = (id: string) => BaseElement | null;\n\nexport type InsightAssertionResponse = AIAssertionResponse & {\n usage?: AIUsageInfo;\n};\n\n/**\n * agent\n */\n\nexport type OnTaskStartTip = (tip: string) => Promise<void> | void;\n\nexport interface AgentWaitForOpt {\n checkIntervalMs?: number;\n timeoutMs?: number;\n}\n\nexport interface AgentAssertOpt {\n keepRawResponse?: boolean;\n}\n\n/**\n * planning\n *\n */\n\nexport interface PlanningLocateParam extends DetailedLocateParam {\n id?: string;\n bbox?: [number, number, number, number];\n}\n\nexport interface PlanningAction<ParamType = any> {\n thought?: string;\n type: string;\n param: ParamType;\n locate?: PlanningLocateParam | null;\n}\n\nexport interface PlanningAIResponse {\n action?: PlanningAction; // this is the qwen mode\n actions?: PlanningAction[];\n more_actions_needed_by_instruction: boolean;\n log: string;\n sleep?: number;\n error?: string;\n usage?: AIUsageInfo;\n rawResponse?: string;\n yamlFlow?: MidsceneYamlFlowItem[];\n yamlString?: string;\n}\n\nexport type PlanningActionParamTap = null;\nexport type PlanningActionParamHover = null;\nexport type PlanningActionParamRightClick = null;\n\nexport interface PlanningActionParamInputOrKeyPress {\n value: string;\n autoDismissKeyboard?: boolean;\n}\n\nexport interface PlanningActionParamAssert {\n assertion: TUserPrompt;\n}\n\nexport interface PlanningActionParamSleep {\n timeMs: number;\n}\n\nexport interface PlanningActionParamError {\n thought: string;\n}\n\nexport type PlanningActionParamWaitFor = AgentWaitForOpt & {\n assertion: string;\n};\n\nexport interface AndroidLongPressParam {\n duration?: number;\n}\n\nexport interface AndroidPullParam {\n direction: 'up' | 'down';\n distance?: number;\n duration?: number;\n}\n/**\n * misc\n */\n\nexport interface Color {\n name: string;\n hex: string;\n}\n\nexport interface BaseAgentParserOpt {\n selector?: string;\n}\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PuppeteerParserOpt extends BaseAgentParserOpt {}\n\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PlaywrightParserOpt extends BaseAgentParserOpt {}\n\n/*\naction\n*/\nexport interface ExecutionTaskProgressOptions {\n onTaskStart?: (task: ExecutionTask) => Promise<void> | void;\n}\n\nexport interface ExecutionRecorderItem {\n type: 'screenshot';\n ts: number;\n screenshot?: string;\n timing?: string;\n}\n\nexport type ExecutionTaskType =\n | 'Planning'\n | 'Insight'\n | 'Action'\n | 'Assertion'\n | 'Log';\n\nexport interface ExecutorContext {\n task: ExecutionTask;\n element?: LocateResultElement | null;\n}\n\nexport interface ExecutionTaskApply<\n Type extends ExecutionTaskType = any,\n TaskParam = any,\n TaskOutput = any,\n TaskLog = any,\n> {\n type: Type;\n subType?: string;\n param?: TaskParam;\n thought?: string;\n locate?: PlanningLocateParam | null;\n pageContext?: UIContext;\n executor: (\n param: TaskParam,\n context: ExecutorContext,\n ) => // biome-ignore lint/suspicious/noConfusingVoidType: <explanation>\n | Promise<ExecutionTaskReturn<TaskOutput, TaskLog> | undefined | void>\n | undefined\n | void;\n}\n\nexport interface ExecutionTaskHitBy {\n from: string;\n context: Record<string, any>;\n}\n\nexport interface ExecutionTaskReturn<TaskOutput = unknown, TaskLog = unknown> {\n output?: TaskOutput;\n log?: TaskLog;\n recorder?: ExecutionRecorderItem[];\n hitBy?: ExecutionTaskHitBy;\n}\n\nexport type ExecutionTask<\n E extends ExecutionTaskApply<any, any, any> = ExecutionTaskApply<\n any,\n any,\n any\n >,\n> = E &\n ExecutionTaskReturn<\n E extends ExecutionTaskApply<any, any, infer TaskOutput, any>\n ? TaskOutput\n : unknown,\n E extends ExecutionTaskApply<any, any, any, infer TaskLog>\n ? TaskLog\n : unknown\n > & {\n status: 'pending' | 'running' | 'finished' | 'failed' | 'cancelled';\n error?: Error;\n errorMessage?: string;\n errorStack?: string;\n timing?: {\n start: number;\n end?: number;\n cost?: number;\n };\n usage?: AIUsageInfo;\n };\n\nexport interface ExecutionDump extends DumpMeta {\n name: string;\n description?: string;\n tasks: ExecutionTask[];\n}\n\n/*\ntask - insight-locate\n*/\nexport type ExecutionTaskInsightLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskInsightLocateOutput {\n element: LocateResultElement | null;\n}\n\nexport interface ExecutionTaskInsightDumpLog {\n dump?: InsightDump;\n}\n\nexport type ExecutionTaskInsightLocateApply = ExecutionTaskApply<\n 'Insight',\n ExecutionTaskInsightLocateParam,\n ExecutionTaskInsightLocateOutput,\n ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightLocate =\n ExecutionTask<ExecutionTaskInsightLocateApply>;\n\n/*\ntask - insight-query\n*/\nexport interface ExecutionTaskInsightQueryParam {\n dataDemand: InsightExtractParam;\n}\n\nexport interface ExecutionTaskInsightQueryOutput {\n data: any;\n}\n\nexport type ExecutionTaskInsightQueryApply = ExecutionTaskApply<\n 'Insight',\n ExecutionTaskInsightQueryParam,\n any,\n ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightQuery =\n ExecutionTask<ExecutionTaskInsightQueryApply>;\n\n/*\ntask - assertion\n*/\nexport interface ExecutionTaskInsightAssertionParam {\n assertion: string;\n}\n\nexport type ExecutionTaskInsightAssertionApply = ExecutionTaskApply<\n 'Insight',\n ExecutionTaskInsightAssertionParam,\n InsightAssertionResponse,\n ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightAssertion =\n ExecutionTask<ExecutionTaskInsightAssertionApply>;\n\n/*\ntask - action (i.e. interact) \n*/\nexport type ExecutionTaskActionApply<ActionParam = any> = ExecutionTaskApply<\n 'Action',\n ActionParam,\n void,\n void\n>;\n\nexport type ExecutionTaskAction = ExecutionTask<ExecutionTaskActionApply>;\n\n/*\ntask - Log\n*/\n\nexport type ExecutionTaskLogApply<\n LogParam = {\n content: string;\n },\n> = ExecutionTaskApply<'Log', LogParam, void, void>;\n\nexport type ExecutionTaskLog = ExecutionTask<ExecutionTaskLogApply>;\n\n/*\ntask - planning\n*/\n\nexport type ExecutionTaskPlanningApply = ExecutionTaskApply<\n 'Planning',\n {\n userInstruction: string;\n log?: string;\n },\n PlanningAIResponse\n>;\n\nexport type ExecutionTaskPlanning = ExecutionTask<ExecutionTaskPlanningApply>;\n\n/*\nGrouped dump\n*/\nexport interface GroupedActionDump {\n groupName: string;\n groupDescription?: string;\n modelName: string;\n modelDescription: string;\n executions: ExecutionDump[];\n}\n\nexport type PageType =\n | 'puppeteer'\n | 'playwright'\n | 'static'\n | 'chrome-extension-proxy'\n | 'android';\n\nexport interface StreamingCodeGenerationOptions {\n /** Whether to enable streaming output */\n stream?: boolean;\n /** Callback function to handle streaming chunks */\n onChunk?: StreamingCallback;\n /** Callback function to handle streaming completion */\n onComplete?: (finalCode: string) => void;\n /** Callback function to handle streaming errors */\n onError?: (error: Error) => void;\n}\n\nexport type StreamingCallback = (chunk: CodeGenerationChunk) => void;\n\nexport interface CodeGenerationChunk {\n /** The incremental content chunk */\n content: string;\n /** The reasoning content */\n reasoning_content: string;\n /** The accumulated content so far */\n accumulated: string;\n /** Whether this is the final chunk */\n isComplete: boolean;\n /** Token usage information if available */\n usage?: AIUsageInfo;\n}\n\nexport interface StreamingAIResponse {\n /** The final accumulated content */\n content: string;\n /** Token usage information */\n usage?: AIUsageInfo;\n /** Whether the response was streamed */\n isStreamed: boolean;\n}\n\nexport type TMultimodalPrompt = {\n /**\n * Support use image to inspect elements.\n * The \"images\" field is an object that uses image name as key and image url as value.\n * The image url can be a local path, a http link , or a base64 string.\n */\n images?: {\n name: string;\n url: string;\n }[];\n /**\n * By default, the image url in the \"images\" filed starts with `https://` or `http://` will be directly sent to the LLM.\n * In case the images are not accessible to the LLM (One common case is that image url is internal network only.), you can enable this option.\n * Then image will be download and convert to base64 format.\n */\n convertHttpImage2Base64?: boolean;\n};\n\nexport type TUserPrompt =\n | string\n | ({\n prompt: string;\n } & Partial<TMultimodalPrompt>);\n\n// biome-ignore lint/complexity/noBannedTypes: <explanation>\nexport interface DeviceAction<T = {}> {\n name: string;\n description?: string;\n interfaceAlias?: string;\n paramSchema?: z.ZodType<T>;\n call: (param: T, context: ExecutorContext) => Promise<void> | void;\n}\n"],"names":["__webpack_require__","module","getter","definition","key","Object","obj","prop","Symbol","AIResponseFormat","UIContext"],"mappings":";;;;;;;;;;;;;;;;;IACAA,oBAAoB,CAAC,GAAG,CAACC;QACxB,IAAIC,SAASD,UAAUA,OAAO,UAAU,GACvC,IAAOA,MAAM,CAAC,UAAU,GACxB,IAAOA;QACRD,oBAAoB,CAAC,CAACE,QAAQ;YAAE,GAAGA;QAAO;QAC1C,OAAOA;IACR;;;ICPAF,oBAAoB,CAAC,GAAG,CAAC,UAASG;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGH,oBAAoB,CAAC,CAACG,YAAYC,QAAQ,CAACJ,oBAAoB,CAAC,CAAC,UAASI,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAJ,oBAAoB,CAAC,GAAG,CAACM,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFP,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOQ,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;IC4BO,IAAKI,mBAAAA,WAAAA,GAAAA,SAAAA,gBAAgB;;;eAAhBA;;IAwFL,MAAeC;IAMtB"}