@midscene/core 0.29.7-beta-20250930035234.0 → 0.29.7-beta-20250930083551.0

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.
@@ -515,7 +515,8 @@ class Agent {
515
515
  const config = opts.cache;
516
516
  if (!config.id) throw new Error('cache configuration requires an explicit id. Please provide:\nExample: cache: { id: "my-cache-id" }');
517
517
  const id = config.id;
518
- const isReadOnly = 'read-only' === config.strategy;
518
+ const strategy = config.strategy ?? 'read-write';
519
+ const isReadOnly = 'read-only' === strategy;
519
520
  return {
520
521
  id,
521
522
  enabled: true,
@@ -1 +1 @@
1
- {"version":3,"file":"agent/agent.mjs","sources":["webpack://@midscene/core/./src/agent/agent.ts"],"sourcesContent":["import {\n type AgentAssertOpt,\n type AgentDescribeElementAtPointResult,\n type AgentWaitForOpt,\n type DeviceAction,\n type ExecutionDump,\n type ExecutionRecorderItem,\n type ExecutionTask,\n type ExecutionTaskLog,\n type Executor,\n type GroupedActionDump,\n Insight,\n type InsightAction,\n type InsightExtractOption,\n type InsightExtractParam,\n type LocateOption,\n type LocateResultElement,\n type LocateValidatorResult,\n type LocatorValidatorOption,\n type MidsceneYamlScript,\n type OnTaskStartTip,\n type PlanningAction,\n type Rect,\n type ScrollParam,\n type TUserPrompt,\n type UIContext,\n} from '../index';\n\nimport yaml from 'js-yaml';\n\nimport {\n groupedActionDumpFileExt,\n reportHTMLContent,\n stringifyDumpData,\n writeLogFile,\n} from '@/utils';\nimport {\n ScriptPlayer,\n buildDetailedLocateParam,\n parseYamlScript,\n} from '../yaml/index';\n\nimport type { AbstractInterface } from '@/device';\nimport type { AgentOpt } from '@/types';\nimport {\n MIDSCENE_CACHE,\n ModelConfigManager,\n globalConfigManager,\n globalModelConfigManager,\n} from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\n// import type { AndroidDeviceInputOpt } from '../device';\nimport { TaskCache } from './task-cache';\nimport { TaskExecutor, locatePlanForLocate } from './tasks';\nimport { locateParamStr, paramStr, taskTitleStr, typeStr } from './ui-utils';\nimport {\n commonContextParser,\n getReportFileName,\n parsePrompt,\n printReportMsg,\n} from './utils';\nimport { trimContextByViewport } from './utils';\n\nconst debug = getDebug('agent');\n\nconst distanceOfTwoPoints = (p1: [number, number], p2: [number, number]) => {\n const [x1, y1] = p1;\n const [x2, y2] = p2;\n return Math.round(Math.sqrt((x1 - x2) ** 2 + (y1 - y2) ** 2));\n};\n\nconst includedInRect = (point: [number, number], rect: Rect) => {\n const [x, y] = point;\n const { left, top, width, height } = rect;\n return x >= left && x <= left + width && y >= top && y <= top + height;\n};\n\nconst defaultInsightExtractOption: InsightExtractOption = {\n domIncluded: false,\n screenshotIncluded: true,\n};\n\nexport class Agent<\n InterfaceType extends AbstractInterface = AbstractInterface,\n> {\n interface: InterfaceType;\n\n insight: Insight;\n\n dump: GroupedActionDump;\n\n reportFile?: string | null;\n\n reportFileName?: string;\n\n taskExecutor: TaskExecutor;\n\n opts: AgentOpt;\n\n /**\n * If true, the agent will not perform any actions\n */\n dryMode = false;\n\n onTaskStartTip?: OnTaskStartTip;\n\n taskCache?: TaskCache;\n\n onDumpUpdate?: (dump: string) => void;\n\n destroyed = false;\n\n modelConfigManager: ModelConfigManager;\n\n /**\n * Frozen page context for consistent AI operations\n */\n private frozenUIContext?: UIContext;\n\n /**\n * Flag to track if VL model warning has been shown\n */\n private hasWarnedNonVLModel = false;\n\n // @deprecated use .interface instead\n get page() {\n return this.interface;\n }\n\n /**\n * Ensures VL model warning is shown once when needed\n */\n private ensureVLModelWarning() {\n if (\n !this.hasWarnedNonVLModel &&\n this.interface.interfaceType !== 'puppeteer' &&\n this.interface.interfaceType !== 'playwright' &&\n this.interface.interfaceType !== 'static' &&\n this.interface.interfaceType !== 'chrome-extension-proxy'\n ) {\n this.modelConfigManager.throwErrorIfNonVLModel();\n this.hasWarnedNonVLModel = true;\n }\n }\n\n constructor(interfaceInstance: InterfaceType, opts?: AgentOpt) {\n this.interface = interfaceInstance;\n this.opts = Object.assign(\n {\n generateReport: true,\n autoPrintReportMsg: true,\n groupName: 'Midscene Report',\n groupDescription: '',\n },\n opts || {},\n );\n\n if (opts?.modelConfig && typeof opts?.modelConfig !== 'function') {\n throw new Error(\n `opts.modelConfig must be one of function or undefined, but got ${typeof opts?.modelConfig}`,\n );\n }\n this.modelConfigManager = opts?.modelConfig\n ? new ModelConfigManager(opts.modelConfig)\n : globalModelConfigManager;\n\n this.onTaskStartTip = this.opts.onTaskStartTip;\n\n this.insight = new Insight(async (action: InsightAction) => {\n return this.getUIContext(action);\n });\n\n // Process cache configuration\n const cacheConfig = this.processCacheConfig(opts || {});\n if (cacheConfig) {\n this.taskCache = new TaskCache(\n cacheConfig.id,\n cacheConfig.enabled,\n undefined, // cacheFilePath\n cacheConfig.readOnly,\n );\n }\n\n this.taskExecutor = new TaskExecutor(this.interface, this.insight, {\n taskCache: this.taskCache,\n onTaskStart: this.callbackOnTaskStartTip.bind(this),\n replanningCycleLimit: this.opts.replanningCycleLimit,\n });\n this.dump = this.resetDump();\n this.reportFileName =\n opts?.reportFileName ||\n getReportFileName(opts?.testId || this.interface.interfaceType || 'web');\n }\n\n async getActionSpace(): Promise<DeviceAction[]> {\n return this.interface.actionSpace();\n }\n\n async getUIContext(action?: InsightAction): Promise<UIContext> {\n // Check VL model configuration when UI context is first needed\n this.ensureVLModelWarning();\n\n // If page context is frozen, return the frozen context for all actions\n if (this.frozenUIContext) {\n debug('Using frozen page context for action:', action);\n return this.frozenUIContext;\n }\n\n if (this.interface.getContext) {\n debug('Using page.getContext for action:', action);\n return await this.interface.getContext();\n } else {\n debug('Using commonContextParser for action:', action);\n return await commonContextParser(this.interface, {\n uploadServerUrl: this.modelConfigManager.getUploadTestServerUrl(),\n });\n }\n }\n\n async _snapshotContext(): Promise<UIContext> {\n return await this.getUIContext('locate');\n }\n\n async setAIActionContext(prompt: string) {\n if (this.opts.aiActionContext) {\n console.warn(\n 'aiActionContext is already set, and it is called again, will override the previous setting',\n );\n }\n this.opts.aiActionContext = prompt;\n }\n\n resetDump() {\n this.dump = {\n groupName: this.opts.groupName!,\n groupDescription: this.opts.groupDescription,\n executions: [],\n modelBriefs: [],\n };\n\n return this.dump;\n }\n\n appendExecutionDump(execution: ExecutionDump) {\n // use trimContextByViewport to process execution\n const trimmedExecution = trimContextByViewport(execution);\n const currentDump = this.dump;\n currentDump.executions.push(trimmedExecution);\n }\n\n dumpDataString() {\n // update dump info\n this.dump.groupName = this.opts.groupName!;\n this.dump.groupDescription = this.opts.groupDescription;\n return stringifyDumpData(this.dump);\n }\n\n reportHTMLString() {\n return reportHTMLContent(this.dumpDataString());\n }\n\n writeOutActionDumps() {\n if (this.destroyed) {\n throw new Error(\n 'PageAgent has been destroyed. Cannot update report file.',\n );\n }\n const { generateReport, autoPrintReportMsg } = this.opts;\n this.reportFile = writeLogFile({\n fileName: this.reportFileName!,\n fileExt: groupedActionDumpFileExt,\n fileContent: this.dumpDataString(),\n type: 'dump',\n generateReport,\n });\n debug('writeOutActionDumps', this.reportFile);\n if (generateReport && autoPrintReportMsg && this.reportFile) {\n printReportMsg(this.reportFile);\n }\n }\n\n private async callbackOnTaskStartTip(task: ExecutionTask) {\n const param = paramStr(task);\n const tip = param ? `${typeStr(task)} - ${param}` : typeStr(task);\n\n if (this.onTaskStartTip) {\n await this.onTaskStartTip(tip);\n }\n }\n\n private async afterTaskRunning(executor: Executor, doNotThrowError = false) {\n this.appendExecutionDump(executor.dump());\n\n try {\n if (this.onDumpUpdate) {\n this.onDumpUpdate(this.dumpDataString());\n }\n } catch (error) {\n console.error('Error in onDumpUpdate', error);\n }\n\n this.writeOutActionDumps();\n\n if (executor.isInErrorState() && !doNotThrowError) {\n const errorTask = executor.latestErrorTask();\n throw new Error(`${errorTask?.errorMessage}\\n${errorTask?.errorStack}`, {\n cause: errorTask?.error,\n });\n }\n }\n\n async callActionInActionSpace<T = any>(\n type: string,\n opt?: T, // and all other action params\n ) {\n debug('callActionInActionSpace', type, ',', opt);\n\n const actionPlan: PlanningAction<T> = {\n type: type as any,\n param: (opt as any) || {},\n thought: '',\n };\n debug('actionPlan', actionPlan); // , ', in which the locateParam is', locateParam);\n\n const plans: PlanningAction[] = [actionPlan].filter(\n Boolean,\n ) as PlanningAction[];\n\n const title = taskTitleStr(\n type as any,\n locateParamStr((opt as any)?.locate || {}),\n );\n\n // assume all operation in action space is related to locating\n const modelConfig = this.modelConfigManager.getModelConfig('grounding');\n\n const { output, executor } = await this.taskExecutor.runPlans(\n title,\n plans,\n modelConfig,\n );\n await this.afterTaskRunning(executor);\n return output;\n }\n\n async aiTap(locatePrompt: TUserPrompt, opt?: LocateOption) {\n assert(locatePrompt, 'missing locate prompt for tap');\n\n const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n return this.callActionInActionSpace('Tap', {\n locate: detailedLocateParam,\n });\n }\n\n async aiRightClick(locatePrompt: TUserPrompt, opt?: LocateOption) {\n assert(locatePrompt, 'missing locate prompt for right click');\n\n const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n return this.callActionInActionSpace('RightClick', {\n locate: detailedLocateParam,\n });\n }\n\n async aiDoubleClick(locatePrompt: TUserPrompt, opt?: LocateOption) {\n assert(locatePrompt, 'missing locate prompt for double click');\n\n const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n return this.callActionInActionSpace('DoubleClick', {\n locate: detailedLocateParam,\n });\n }\n\n async aiHover(locatePrompt: TUserPrompt, opt?: LocateOption) {\n assert(locatePrompt, 'missing locate prompt for hover');\n\n const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n return this.callActionInActionSpace('Hover', {\n locate: detailedLocateParam,\n });\n }\n\n // New signature, always use locatePrompt as the first param\n async aiInput(\n locatePrompt: TUserPrompt,\n opt: LocateOption & { value: string } & { autoDismissKeyboard?: boolean },\n ): Promise<any>;\n\n // Legacy signature - deprecated\n /**\n * @deprecated Use aiInput(locatePrompt, opt) instead where opt contains the value\n */\n async aiInput(\n value: string,\n locatePrompt: TUserPrompt,\n opt?: LocateOption & { autoDismissKeyboard?: boolean }, // AndroidDeviceInputOpt &\n ): Promise<any>;\n\n // Implementation\n async aiInput(\n locatePromptOrValue: TUserPrompt | string,\n locatePromptOrOpt:\n | TUserPrompt\n | (LocateOption & { value: string } & { autoDismissKeyboard?: boolean }) // AndroidDeviceInputOpt &\n | undefined,\n optOrUndefined?: LocateOption, // AndroidDeviceInputOpt &\n ) {\n let value: string;\n let locatePrompt: TUserPrompt;\n let opt:\n | (LocateOption & { value: string } & { autoDismissKeyboard?: boolean }) // AndroidDeviceInputOpt &\n | undefined;\n\n // Check if using new signature (first param is locatePrompt, second has value)\n if (\n typeof locatePromptOrOpt === 'object' &&\n locatePromptOrOpt !== null &&\n 'value' in locatePromptOrOpt\n ) {\n // New signature: aiInput(locatePrompt, opt)\n locatePrompt = locatePromptOrValue as TUserPrompt;\n const optWithValue = locatePromptOrOpt as LocateOption & {\n // AndroidDeviceInputOpt &\n value: string;\n autoDismissKeyboard?: boolean;\n };\n value = optWithValue.value;\n opt = optWithValue;\n } else {\n // Legacy signature: aiInput(value, locatePrompt, opt)\n value = locatePromptOrValue as string;\n locatePrompt = locatePromptOrOpt as TUserPrompt;\n opt = {\n ...optOrUndefined,\n value,\n };\n }\n\n assert(\n typeof value === 'string',\n 'input value must be a string, use empty string if you want to clear the input',\n );\n assert(locatePrompt, 'missing locate prompt for input');\n\n const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n return this.callActionInActionSpace('Input', {\n ...(opt || {}),\n locate: detailedLocateParam,\n });\n }\n\n // New signature\n async aiKeyboardPress(\n locatePrompt: TUserPrompt,\n opt: LocateOption & { keyName: string },\n ): Promise<any>;\n\n // Legacy signature - deprecated\n /**\n * @deprecated Use aiKeyboardPress(locatePrompt, opt) instead where opt contains the keyName\n */\n async aiKeyboardPress(\n keyName: string,\n locatePrompt?: TUserPrompt,\n opt?: LocateOption,\n ): Promise<any>;\n\n // Implementation\n async aiKeyboardPress(\n locatePromptOrKeyName: TUserPrompt | string,\n locatePromptOrOpt:\n | TUserPrompt\n | (LocateOption & { keyName: string })\n | undefined,\n optOrUndefined?: LocateOption,\n ) {\n let keyName: string;\n let locatePrompt: TUserPrompt | undefined;\n let opt: (LocateOption & { keyName: string }) | undefined;\n\n // Check if using new signature (first param is locatePrompt, second has keyName)\n if (\n typeof locatePromptOrOpt === 'object' &&\n locatePromptOrOpt !== null &&\n 'keyName' in locatePromptOrOpt\n ) {\n // New signature: aiKeyboardPress(locatePrompt, opt)\n locatePrompt = locatePromptOrKeyName as TUserPrompt;\n opt = locatePromptOrOpt as LocateOption & {\n keyName: string;\n };\n } else {\n // Legacy signature: aiKeyboardPress(keyName, locatePrompt, opt)\n keyName = locatePromptOrKeyName as string;\n locatePrompt = locatePromptOrOpt as TUserPrompt | undefined;\n opt = {\n ...(optOrUndefined || {}),\n keyName,\n };\n }\n\n assert(opt?.keyName, 'missing keyName for keyboard press');\n\n const detailedLocateParam = locatePrompt\n ? buildDetailedLocateParam(locatePrompt, opt)\n : undefined;\n\n return this.callActionInActionSpace('KeyboardPress', {\n ...(opt || {}),\n locate: detailedLocateParam,\n });\n }\n\n // New signature\n async aiScroll(\n locatePrompt: TUserPrompt | undefined,\n opt: LocateOption & ScrollParam,\n ): Promise<any>;\n\n // Legacy signature - deprecated\n /**\n * @deprecated Use aiScroll(locatePrompt, opt) instead where opt contains the scroll parameters\n */\n async aiScroll(\n scrollParam: ScrollParam,\n locatePrompt?: TUserPrompt,\n opt?: LocateOption,\n ): Promise<any>;\n\n // Implementation\n async aiScroll(\n locatePromptOrScrollParam: TUserPrompt | ScrollParam | undefined,\n locatePromptOrOpt: TUserPrompt | (LocateOption & ScrollParam) | undefined,\n optOrUndefined?: LocateOption,\n ) {\n let scrollParam: ScrollParam;\n let locatePrompt: TUserPrompt | undefined;\n let opt: LocateOption | undefined;\n\n // Check if using new signature (first param is locatePrompt, second has scroll params)\n if (\n typeof locatePromptOrOpt === 'object' &&\n ('direction' in locatePromptOrOpt ||\n 'scrollType' in locatePromptOrOpt ||\n 'distance' in locatePromptOrOpt)\n ) {\n // New signature: aiScroll(locatePrompt, opt)\n locatePrompt = locatePromptOrScrollParam as TUserPrompt;\n opt = locatePromptOrOpt as LocateOption & ScrollParam;\n } else {\n // Legacy signature: aiScroll(scrollParam, locatePrompt, opt)\n scrollParam = locatePromptOrScrollParam as ScrollParam;\n locatePrompt = locatePromptOrOpt as TUserPrompt | undefined;\n opt = {\n ...(optOrUndefined || {}),\n ...(scrollParam || {}),\n };\n }\n\n const detailedLocateParam = buildDetailedLocateParam(\n locatePrompt || '',\n opt,\n );\n\n return this.callActionInActionSpace('Scroll', {\n ...(opt || {}),\n locate: detailedLocateParam,\n });\n }\n\n async aiAction(\n taskPrompt: string,\n opt?: {\n cacheable?: boolean;\n },\n ) {\n const modelConfig = this.modelConfigManager.getModelConfig('planning');\n\n const cacheable = opt?.cacheable;\n // if vlm-ui-tars, plan cache is not used\n const isVlmUiTars = modelConfig.vlMode === 'vlm-ui-tars';\n const matchedCache =\n isVlmUiTars || cacheable === false\n ? undefined\n : this.taskCache?.matchPlanCache(taskPrompt);\n if (matchedCache && this.taskCache?.isCacheResultUsed) {\n // log into report file\n const { executor } = await this.taskExecutor.loadYamlFlowAsPlanning(\n taskPrompt,\n matchedCache.cacheContent?.yamlWorkflow,\n );\n\n await this.afterTaskRunning(executor);\n\n debug('matched cache, will call .runYaml to run the action');\n const yaml = matchedCache.cacheContent?.yamlWorkflow;\n return this.runYaml(yaml);\n }\n\n const { output, executor } = await this.taskExecutor.action(\n taskPrompt,\n modelConfig,\n this.opts.aiActionContext,\n );\n\n // update cache\n if (this.taskCache && output?.yamlFlow && cacheable !== false) {\n const yamlContent: MidsceneYamlScript = {\n tasks: [\n {\n name: taskPrompt,\n flow: output.yamlFlow,\n },\n ],\n };\n const yamlFlowStr = yaml.dump(yamlContent);\n this.taskCache.updateOrAppendCacheRecord(\n {\n type: 'plan',\n prompt: taskPrompt,\n yamlWorkflow: yamlFlowStr,\n },\n matchedCache,\n );\n }\n\n await this.afterTaskRunning(executor);\n return output;\n }\n\n async aiQuery<ReturnType = any>(\n demand: InsightExtractParam,\n opt: InsightExtractOption = defaultInsightExtractOption,\n ): Promise<ReturnType> {\n const modelConfig = this.modelConfigManager.getModelConfig('VQA');\n const { output, executor } =\n await this.taskExecutor.createTypeQueryExecution(\n 'Query',\n demand,\n modelConfig,\n opt,\n );\n await this.afterTaskRunning(executor);\n return output as ReturnType;\n }\n\n async aiBoolean(\n prompt: TUserPrompt,\n opt: InsightExtractOption = defaultInsightExtractOption,\n ): Promise<boolean> {\n const modelConfig = this.modelConfigManager.getModelConfig('VQA');\n\n const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n const { output, executor } =\n await this.taskExecutor.createTypeQueryExecution(\n 'Boolean',\n textPrompt,\n modelConfig,\n opt,\n multimodalPrompt,\n );\n await this.afterTaskRunning(executor);\n return output as boolean;\n }\n\n async aiNumber(\n prompt: TUserPrompt,\n opt: InsightExtractOption = defaultInsightExtractOption,\n ): Promise<number> {\n const modelConfig = this.modelConfigManager.getModelConfig('VQA');\n\n const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n const { output, executor } =\n await this.taskExecutor.createTypeQueryExecution(\n 'Number',\n textPrompt,\n modelConfig,\n opt,\n multimodalPrompt,\n );\n await this.afterTaskRunning(executor);\n return output as number;\n }\n\n async aiString(\n prompt: TUserPrompt,\n opt: InsightExtractOption = defaultInsightExtractOption,\n ): Promise<string> {\n const modelConfig = this.modelConfigManager.getModelConfig('VQA');\n\n const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n const { output, executor } =\n await this.taskExecutor.createTypeQueryExecution(\n 'String',\n textPrompt,\n modelConfig,\n opt,\n multimodalPrompt,\n );\n await this.afterTaskRunning(executor);\n return output as string;\n }\n\n async aiAsk(\n prompt: TUserPrompt,\n opt: InsightExtractOption = defaultInsightExtractOption,\n ): Promise<string> {\n return this.aiString(prompt, opt);\n }\n\n async describeElementAtPoint(\n center: [number, number],\n opt?: {\n verifyPrompt?: boolean;\n retryLimit?: number;\n deepThink?: boolean;\n } & LocatorValidatorOption,\n ): Promise<AgentDescribeElementAtPointResult> {\n const { verifyPrompt = true, retryLimit = 3 } = opt || {};\n\n let success = false;\n let retryCount = 0;\n let resultPrompt = '';\n let deepThink = opt?.deepThink || false;\n let verifyResult: LocateValidatorResult | undefined;\n\n while (!success && retryCount < retryLimit) {\n if (retryCount >= 2) {\n deepThink = true;\n }\n debug(\n 'aiDescribe',\n center,\n 'verifyPrompt',\n verifyPrompt,\n 'retryCount',\n retryCount,\n 'deepThink',\n deepThink,\n );\n // use same intent as aiLocate\n const modelConfig = this.modelConfigManager.getModelConfig('grounding');\n\n const text = await this.insight.describe(center, modelConfig, {\n deepThink,\n });\n debug('aiDescribe text', text);\n assert(text.description, `failed to describe element at [${center}]`);\n resultPrompt = text.description;\n\n verifyResult = await this.verifyLocator(\n resultPrompt,\n deepThink ? { deepThink: true } : undefined,\n center,\n opt,\n );\n if (verifyResult.pass) {\n success = true;\n } else {\n retryCount++;\n }\n }\n\n return {\n prompt: resultPrompt,\n deepThink,\n verifyResult,\n };\n }\n\n async verifyLocator(\n prompt: string,\n locateOpt: LocateOption | undefined,\n expectCenter: [number, number],\n verifyLocateOption?: LocatorValidatorOption,\n ): Promise<LocateValidatorResult> {\n debug('verifyLocator', prompt, locateOpt, expectCenter, verifyLocateOption);\n\n const { center: verifyCenter, rect: verifyRect } = await this.aiLocate(\n prompt,\n locateOpt,\n );\n const distance = distanceOfTwoPoints(expectCenter, verifyCenter);\n const included = includedInRect(expectCenter, verifyRect);\n const pass =\n distance <= (verifyLocateOption?.centerDistanceThreshold || 20) ||\n included;\n const verifyResult = {\n pass,\n rect: verifyRect,\n center: verifyCenter,\n centerDistance: distance,\n };\n debug('aiDescribe verifyResult', verifyResult);\n return verifyResult;\n }\n\n async aiLocate(prompt: TUserPrompt, opt?: LocateOption) {\n const locateParam = buildDetailedLocateParam(prompt, opt);\n assert(locateParam, 'cannot get locate param for aiLocate');\n const locatePlan = locatePlanForLocate(locateParam);\n const plans = [locatePlan];\n const modelConfig = this.modelConfigManager.getModelConfig('grounding');\n\n const { executor, output } = await this.taskExecutor.runPlans(\n taskTitleStr('Locate', locateParamStr(locateParam)),\n plans,\n modelConfig,\n );\n await this.afterTaskRunning(executor);\n\n const { element } = output;\n\n return {\n rect: element?.rect,\n center: element?.center,\n scale: (await this.interface.size()).dpr,\n } as Pick<LocateResultElement, 'rect' | 'center'> & {\n scale: number;\n };\n }\n\n async aiAssert(\n assertion: TUserPrompt,\n msg?: string,\n opt?: AgentAssertOpt & InsightExtractOption,\n ) {\n const modelConfig = this.modelConfigManager.getModelConfig('VQA');\n\n const insightOpt: InsightExtractOption = {\n domIncluded: opt?.domIncluded ?? defaultInsightExtractOption.domIncluded,\n screenshotIncluded:\n opt?.screenshotIncluded ??\n defaultInsightExtractOption.screenshotIncluded,\n isWaitForAssert: opt?.isWaitForAssert,\n doNotThrowError: opt?.doNotThrowError,\n };\n\n const { output, executor, thought } = await this.taskExecutor.assert(\n assertion,\n modelConfig,\n insightOpt,\n );\n await this.afterTaskRunning(executor, true);\n\n const message = output\n ? undefined\n : `Assertion failed: ${msg || (typeof assertion === 'string' ? assertion : assertion.prompt)}\\nReason: ${\n thought || executor.latestErrorTask()?.error || '(no_reason)'\n }`;\n\n if (opt?.keepRawResponse) {\n return {\n pass: output,\n thought,\n message,\n };\n }\n\n if (!output) {\n throw new Error(message);\n }\n }\n\n async aiWaitFor(assertion: TUserPrompt, opt?: AgentWaitForOpt) {\n const modelConfig = this.modelConfigManager.getModelConfig('VQA');\n const { executor } = await this.taskExecutor.waitFor(\n assertion,\n {\n timeoutMs: opt?.timeoutMs || 15 * 1000,\n checkIntervalMs: opt?.checkIntervalMs || 3 * 1000,\n },\n modelConfig,\n );\n await this.afterTaskRunning(executor, true);\n\n if (executor.isInErrorState()) {\n const errorTask = executor.latestErrorTask();\n throw new Error(`${errorTask?.error}\\n${errorTask?.errorStack}`);\n }\n }\n\n async ai(taskPrompt: string, type = 'action') {\n if (type === 'action') {\n return this.aiAction(taskPrompt);\n }\n if (type === 'query') {\n return this.aiQuery(taskPrompt);\n }\n\n if (type === 'assert') {\n return this.aiAssert(taskPrompt);\n }\n\n if (type === 'tap') {\n return this.aiTap(taskPrompt);\n }\n\n if (type === 'rightClick') {\n return this.aiRightClick(taskPrompt);\n }\n\n if (type === 'doubleClick') {\n return this.aiDoubleClick(taskPrompt);\n }\n\n throw new Error(\n `Unknown type: ${type}, only support 'action', 'query', 'assert', 'tap', 'rightClick', 'doubleClick'`,\n );\n }\n\n async runYaml(yamlScriptContent: string): Promise<{\n result: Record<string, any>;\n }> {\n const script = parseYamlScript(yamlScriptContent, 'yaml');\n const player = new ScriptPlayer(script, async () => {\n return { agent: this, freeFn: [] };\n });\n await player.run();\n\n if (player.status === 'error') {\n const errors = player.taskStatusList\n .filter((task) => task.status === 'error')\n .map((task) => {\n return `task - ${task.name}: ${task.error?.message}`;\n })\n .join('\\n');\n throw new Error(`Error(s) occurred in running yaml script:\\n${errors}`);\n }\n\n return {\n result: player.result,\n };\n }\n\n async evaluateJavaScript(script: string) {\n assert(\n this.interface.evaluateJavaScript,\n 'evaluateJavaScript is not supported in current agent',\n );\n return this.interface.evaluateJavaScript(script);\n }\n\n async destroy() {\n await this.interface.destroy?.();\n this.resetDump(); // reset dump to release memory\n this.destroyed = true;\n }\n\n async logScreenshot(\n title?: string,\n opt?: {\n content: string;\n },\n ) {\n // 1. screenshot\n const base64 = await this.interface.screenshotBase64();\n const now = Date.now();\n // 2. build recorder\n const recorder: ExecutionRecorderItem[] = [\n {\n type: 'screenshot',\n ts: now,\n screenshot: base64,\n },\n ];\n // 3. build ExecutionTaskLog\n const task: ExecutionTaskLog = {\n type: 'Log',\n subType: 'Screenshot',\n status: 'finished',\n recorder,\n timing: {\n start: now,\n end: now,\n cost: 0,\n },\n param: {\n content: opt?.content || '',\n },\n executor: async () => {},\n };\n // 4. build ExecutionDump\n const executionDump: ExecutionDump = {\n sdkVersion: '',\n logTime: now,\n name: `Log - ${title || 'untitled'}`,\n description: opt?.content || '',\n tasks: [task],\n };\n // 5. append to execution dump\n this.appendExecutionDump(executionDump);\n\n try {\n this.onDumpUpdate?.(this.dumpDataString());\n } catch (error) {\n console.error('Failed to update dump', error);\n }\n\n this.writeOutActionDumps();\n }\n\n _unstableLogContent() {\n const { groupName, groupDescription, executions } = this.dump;\n const newExecutions = Array.isArray(executions)\n ? executions.map((execution: any) => {\n const { tasks, ...restExecution } = execution;\n let newTasks = tasks;\n if (Array.isArray(tasks)) {\n newTasks = tasks.map((task: any) => {\n // only remove uiContext and log from task\n const { uiContext, log, ...restTask } = task;\n return restTask;\n });\n }\n return { ...restExecution, ...(newTasks ? { tasks: newTasks } : {}) };\n })\n : [];\n return {\n groupName,\n groupDescription,\n executions: newExecutions,\n };\n }\n\n /**\n * Freezes the current page context to be reused in subsequent AI operations\n * This avoids recalculating page context for each operation\n */\n async freezePageContext(): Promise<void> {\n debug('Freezing page context');\n const context = await this._snapshotContext();\n // Mark the context as frozen\n context._isFrozen = true;\n this.frozenUIContext = context;\n debug('Page context frozen successfully');\n }\n\n /**\n * Unfreezes the page context, allowing AI operations to calculate context dynamically\n */\n async unfreezePageContext(): Promise<void> {\n debug('Unfreezing page context');\n this.frozenUIContext = undefined;\n debug('Page context unfrozen successfully');\n }\n\n /**\n * Process cache configuration and return normalized cache settings\n */\n private processCacheConfig(\n opts: AgentOpt,\n ): { id: string; enabled: boolean; readOnly: boolean } | null {\n // 1. New cache object configuration (highest priority)\n if (opts.cache !== undefined) {\n if (opts.cache === false) {\n return null; // Completely disable cache\n }\n\n if (opts.cache === true) {\n throw new Error(\n 'cache: true requires an explicit cache ID. Please provide:\\n' +\n 'Example: cache: { id: \"my-cache-id\" }',\n );\n }\n\n // cache is object configuration\n if (typeof opts.cache === 'object') {\n const config = opts.cache;\n if (!config.id) {\n throw new Error(\n 'cache configuration requires an explicit id. Please provide:\\n' +\n 'Example: cache: { id: \"my-cache-id\" }',\n );\n }\n const id = config.id;\n const isReadOnly = config.strategy === 'read-only';\n\n return {\n id,\n enabled: true,\n readOnly: isReadOnly,\n };\n }\n }\n\n // 2. Backward compatibility: support old cacheId (requires environment variable)\n if (opts.cacheId) {\n const envEnabled =\n globalConfigManager.getEnvConfigInBoolean(MIDSCENE_CACHE);\n if (envEnabled) {\n return {\n id: opts.cacheId,\n enabled: true,\n readOnly: false,\n };\n }\n }\n\n // 3. No cache configuration\n return null;\n }\n\n /**\n * Manually flush cache to file\n * Only meaningful in read-only mode, other modes will throw error\n */\n async flushCache(): Promise<void> {\n if (!this.taskCache) {\n throw new Error('Cache is not configured');\n }\n\n this.taskCache.flushCacheToFile();\n }\n}\n\nexport const createAgent = (\n interfaceInstance: AbstractInterface,\n opts?: AgentOpt,\n) => {\n return new Agent(interfaceInstance, opts);\n};\n"],"names":["debug","getDebug","distanceOfTwoPoints","p1","p2","x1","y1","x2","y2","Math","includedInRect","point","rect","x","y","left","top","width","height","defaultInsightExtractOption","Agent","action","commonContextParser","prompt","console","execution","trimmedExecution","trimContextByViewport","currentDump","stringifyDumpData","reportHTMLContent","Error","generateReport","autoPrintReportMsg","writeLogFile","groupedActionDumpFileExt","printReportMsg","task","param","paramStr","tip","typeStr","executor","doNotThrowError","error","errorTask","type","opt","actionPlan","plans","Boolean","title","taskTitleStr","locateParamStr","modelConfig","output","locatePrompt","assert","detailedLocateParam","buildDetailedLocateParam","locatePromptOrValue","locatePromptOrOpt","optOrUndefined","value","optWithValue","locatePromptOrKeyName","keyName","undefined","locatePromptOrScrollParam","scrollParam","taskPrompt","_this_taskCache","_this_taskCache1","cacheable","isVlmUiTars","matchedCache","_matchedCache_cacheContent","_matchedCache_cacheContent1","yaml","yamlContent","yamlFlowStr","demand","textPrompt","multimodalPrompt","parsePrompt","center","verifyPrompt","retryLimit","success","retryCount","resultPrompt","deepThink","verifyResult","text","locateOpt","expectCenter","verifyLocateOption","verifyCenter","verifyRect","distance","included","pass","locateParam","locatePlan","locatePlanForLocate","element","assertion","msg","_executor_latestErrorTask","insightOpt","thought","message","yamlScriptContent","script","parseYamlScript","player","ScriptPlayer","errors","_task_error","_this_interface","base64","now","Date","recorder","executionDump","_this","groupName","groupDescription","executions","newExecutions","Array","tasks","restExecution","newTasks","uiContext","log","restTask","context","opts","config","id","isReadOnly","envEnabled","globalConfigManager","MIDSCENE_CACHE","interfaceInstance","Object","ModelConfigManager","globalModelConfigManager","Insight","cacheConfig","TaskCache","TaskExecutor","getReportFileName","createAgent"],"mappings":";;;;;;;;;;;;;;;;;;;;;AAgEA,MAAMA,QAAQC,SAAS;AAEvB,MAAMC,sBAAsB,CAACC,IAAsBC;IACjD,MAAM,CAACC,IAAIC,GAAG,GAAGH;IACjB,MAAM,CAACI,IAAIC,GAAG,GAAGJ;IACjB,OAAOK,KAAK,KAAK,CAACA,KAAK,IAAI,CAAEJ,AAAAA,CAAAA,KAAKE,EAAC,KAAM,IAAKD,AAAAA,CAAAA,KAAKE,EAAC,KAAM;AAC5D;AAEA,MAAME,iBAAiB,CAACC,OAAyBC;IAC/C,MAAM,CAACC,GAAGC,EAAE,GAAGH;IACf,MAAM,EAAEI,IAAI,EAAEC,GAAG,EAAEC,KAAK,EAAEC,MAAM,EAAE,GAAGN;IACrC,OAAOC,KAAKE,QAAQF,KAAKE,OAAOE,SAASH,KAAKE,OAAOF,KAAKE,MAAME;AAClE;AAEA,MAAMC,8BAAoD;IACxD,aAAa;IACb,oBAAoB;AACtB;AAEO,MAAMC;IA2CX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IAKQ,uBAAuB;QAC7B,IACE,CAAC,IAAI,CAAC,mBAAmB,IACzB,AAAiC,gBAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,IAC5B,AAAiC,iBAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,IAC5B,AAAiC,aAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,IAC5B,AAAiC,6BAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,EAC5B;YACA,IAAI,CAAC,kBAAkB,CAAC,sBAAsB;YAC9C,IAAI,CAAC,mBAAmB,GAAG;QAC7B;IACF;IAmDA,MAAM,iBAA0C;QAC9C,OAAO,IAAI,CAAC,SAAS,CAAC,WAAW;IACnC;IAEA,MAAM,aAAaC,MAAsB,EAAsB;QAE7D,IAAI,CAAC,oBAAoB;QAGzB,IAAI,IAAI,CAAC,eAAe,EAAE;YACxBrB,MAAM,yCAAyCqB;YAC/C,OAAO,IAAI,CAAC,eAAe;QAC7B;QAEA,IAAI,IAAI,CAAC,SAAS,CAAC,UAAU,EAAE;YAC7BrB,MAAM,qCAAqCqB;YAC3C,OAAO,MAAM,IAAI,CAAC,SAAS,CAAC,UAAU;QACxC;QACErB,MAAM,yCAAyCqB;QAC/C,OAAO,MAAMC,oBAAoB,IAAI,CAAC,SAAS,EAAE;YAC/C,iBAAiB,IAAI,CAAC,kBAAkB,CAAC,sBAAsB;QACjE;IAEJ;IAEA,MAAM,mBAAuC;QAC3C,OAAO,MAAM,IAAI,CAAC,YAAY,CAAC;IACjC;IAEA,MAAM,mBAAmBC,MAAc,EAAE;QACvC,IAAI,IAAI,CAAC,IAAI,CAAC,eAAe,EAC3BC,QAAQ,IAAI,CACV;QAGJ,IAAI,CAAC,IAAI,CAAC,eAAe,GAAGD;IAC9B;IAEA,YAAY;QACV,IAAI,CAAC,IAAI,GAAG;YACV,WAAW,IAAI,CAAC,IAAI,CAAC,SAAS;YAC9B,kBAAkB,IAAI,CAAC,IAAI,CAAC,gBAAgB;YAC5C,YAAY,EAAE;YACd,aAAa,EAAE;QACjB;QAEA,OAAO,IAAI,CAAC,IAAI;IAClB;IAEA,oBAAoBE,SAAwB,EAAE;QAE5C,MAAMC,mBAAmBC,sBAAsBF;QAC/C,MAAMG,cAAc,IAAI,CAAC,IAAI;QAC7BA,YAAY,UAAU,CAAC,IAAI,CAACF;IAC9B;IAEA,iBAAiB;QAEf,IAAI,CAAC,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,IAAI,CAAC,SAAS;QACzC,IAAI,CAAC,IAAI,CAAC,gBAAgB,GAAG,IAAI,CAAC,IAAI,CAAC,gBAAgB;QACvD,OAAOG,kBAAkB,IAAI,CAAC,IAAI;IACpC;IAEA,mBAAmB;QACjB,OAAOC,kBAAkB,IAAI,CAAC,cAAc;IAC9C;IAEA,sBAAsB;QACpB,IAAI,IAAI,CAAC,SAAS,EAChB,MAAM,IAAIC,MACR;QAGJ,MAAM,EAAEC,cAAc,EAAEC,kBAAkB,EAAE,GAAG,IAAI,CAAC,IAAI;QACxD,IAAI,CAAC,UAAU,GAAGC,aAAa;YAC7B,UAAU,IAAI,CAAC,cAAc;YAC7B,SAASC;YACT,aAAa,IAAI,CAAC,cAAc;YAChC,MAAM;YACNH;QACF;QACAhC,MAAM,uBAAuB,IAAI,CAAC,UAAU;QAC5C,IAAIgC,kBAAkBC,sBAAsB,IAAI,CAAC,UAAU,EACzDG,eAAe,IAAI,CAAC,UAAU;IAElC;IAEA,MAAc,uBAAuBC,IAAmB,EAAE;QACxD,MAAMC,QAAQC,SAASF;QACvB,MAAMG,MAAMF,QAAQ,GAAGG,QAAQJ,MAAM,GAAG,EAAEC,OAAO,GAAGG,QAAQJ;QAE5D,IAAI,IAAI,CAAC,cAAc,EACrB,MAAM,IAAI,CAAC,cAAc,CAACG;IAE9B;IAEA,MAAc,iBAAiBE,QAAkB,EAAEC,kBAAkB,KAAK,EAAE;QAC1E,IAAI,CAAC,mBAAmB,CAACD,SAAS,IAAI;QAEtC,IAAI;YACF,IAAI,IAAI,CAAC,YAAY,EACnB,IAAI,CAAC,YAAY,CAAC,IAAI,CAAC,cAAc;QAEzC,EAAE,OAAOE,OAAO;YACdpB,QAAQ,KAAK,CAAC,yBAAyBoB;QACzC;QAEA,IAAI,CAAC,mBAAmB;QAExB,IAAIF,SAAS,cAAc,MAAM,CAACC,iBAAiB;YACjD,MAAME,YAAYH,SAAS,eAAe;YAC1C,MAAM,IAAIX,MAAM,GAAGc,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,YAAY,CAAC,EAAE,EAAEA,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,UAAU,EAAE,EAAE;gBACtE,OAAOA,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,KAAK;YACzB;QACF;IACF;IAEA,MAAM,wBACJC,IAAY,EACZC,GAAO,EACP;QACA/C,MAAM,2BAA2B8C,MAAM,KAAKC;QAE5C,MAAMC,aAAgC;YACpC,MAAMF;YACN,OAAQC,OAAe,CAAC;YACxB,SAAS;QACX;QACA/C,MAAM,cAAcgD;QAEpB,MAAMC,QAA0B;YAACD;SAAW,CAAC,MAAM,CACjDE;QAGF,MAAMC,QAAQC,aACZN,MACAO,eAAe,AAACN,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAa,MAAM,AAAD,KAAK,CAAC;QAI1C,MAAMO,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAEC,MAAM,EAAEb,QAAQ,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,CAC3DS,OACAF,OACAK;QAEF,MAAM,IAAI,CAAC,gBAAgB,CAACZ;QAC5B,OAAOa;IACT;IAEA,MAAM,MAAMC,YAAyB,EAAET,GAAkB,EAAE;QACzDU,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,OAAO;YACzC,QAAQW;QACV;IACF;IAEA,MAAM,aAAaF,YAAyB,EAAET,GAAkB,EAAE;QAChEU,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,cAAc;YAChD,QAAQW;QACV;IACF;IAEA,MAAM,cAAcF,YAAyB,EAAET,GAAkB,EAAE;QACjEU,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,eAAe;YACjD,QAAQW;QACV;IACF;IAEA,MAAM,QAAQF,YAAyB,EAAET,GAAkB,EAAE;QAC3DU,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,SAAS;YAC3C,QAAQW;QACV;IACF;IAmBA,MAAM,QACJE,mBAAyC,EACzCC,iBAGa,EACbC,cAA6B,EAC7B;QACA,IAAIC;QACJ,IAAIP;QACJ,IAAIT;QAKJ,IACE,AAA6B,YAA7B,OAAOc,qBACPA,AAAsB,SAAtBA,qBACA,WAAWA,mBACX;YAEAL,eAAeI;YACf,MAAMI,eAAeH;YAKrBE,QAAQC,aAAa,KAAK;YAC1BjB,MAAMiB;QACR,OAAO;YAELD,QAAQH;YACRJ,eAAeK;YACfd,MAAM;gBACJ,GAAGe,cAAc;gBACjBC;YACF;QACF;QAEAN,OACE,AAAiB,YAAjB,OAAOM,OACP;QAEFN,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,SAAS;YAC3C,GAAIA,OAAO,CAAC,CAAC;YACb,QAAQW;QACV;IACF;IAmBA,MAAM,gBACJO,qBAA2C,EAC3CJ,iBAGa,EACbC,cAA6B,EAC7B;QACA,IAAII;QACJ,IAAIV;QACJ,IAAIT;QAGJ,IACE,AAA6B,YAA7B,OAAOc,qBACPA,AAAsB,SAAtBA,qBACA,aAAaA,mBACb;YAEAL,eAAeS;YACflB,MAAMc;QAGR,OAAO;YAELK,UAAUD;YACVT,eAAeK;YACfd,MAAM;gBACJ,GAAIe,kBAAkB,CAAC,CAAC;gBACxBI;YACF;QACF;QAEAT,OAAOV,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,EAAE;QAErB,MAAMW,sBAAsBF,eACxBG,yBAAyBH,cAAcT,OACvCoB;QAEJ,OAAO,IAAI,CAAC,uBAAuB,CAAC,iBAAiB;YACnD,GAAIpB,OAAO,CAAC,CAAC;YACb,QAAQW;QACV;IACF;IAmBA,MAAM,SACJU,yBAAgE,EAChEP,iBAAyE,EACzEC,cAA6B,EAC7B;QACA,IAAIO;QACJ,IAAIb;QACJ,IAAIT;QAGJ,IACE,AAA6B,YAA7B,OAAOc,qBACN,gBAAeA,qBACd,gBAAgBA,qBAChB,cAAcA,iBAAgB,GAChC;YAEAL,eAAeY;YACfrB,MAAMc;QACR,OAAO;YAELQ,cAAcD;YACdZ,eAAeK;YACfd,MAAM;gBACJ,GAAIe,kBAAkB,CAAC,CAAC;gBACxB,GAAIO,eAAe,CAAC,CAAC;YACvB;QACF;QAEA,MAAMX,sBAAsBC,yBAC1BH,gBAAgB,IAChBT;QAGF,OAAO,IAAI,CAAC,uBAAuB,CAAC,UAAU;YAC5C,GAAIA,OAAO,CAAC,CAAC;YACb,QAAQW;QACV;IACF;IAEA,MAAM,SACJY,UAAkB,EAClBvB,GAEC,EACD;YASMwB,iBACcC;QATpB,MAAMlB,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAMmB,YAAY1B,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS;QAEhC,MAAM2B,cAAcpB,AAAuB,kBAAvBA,YAAY,MAAM;QACtC,MAAMqB,eACJD,eAAeD,AAAc,UAAdA,YACXN,SAAAA,QACAI,CAAAA,kBAAAA,IAAI,CAAC,SAAS,AAAD,IAAbA,KAAAA,IAAAA,gBAAgB,cAAc,CAACD;QACrC,IAAIK,gBAAAA,SAAgBH,CAAAA,mBAAAA,IAAI,CAAC,SAAS,AAAD,IAAbA,KAAAA,IAAAA,iBAAgB,iBAAiB,AAAD,GAAG;gBAInDI,4BAMWC;YARb,MAAM,EAAEnC,QAAQ,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,sBAAsB,CACjE4B,YAAAA,QACAM,CAAAA,6BAAAA,aAAa,YAAY,AAAD,IAAxBA,KAAAA,IAAAA,2BAA2B,YAAY;YAGzC,MAAM,IAAI,CAAC,gBAAgB,CAAClC;YAE5B1C,MAAM;YACN,MAAM8E,OAAO,QAAAD,CAAAA,8BAAAA,aAAa,YAAY,AAAD,IAAxBA,KAAAA,IAAAA,4BAA2B,YAAY;YACpD,OAAO,IAAI,CAAC,OAAO,CAACC;QACtB;QAEA,MAAM,EAAEvB,MAAM,EAAEb,QAAQ,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,MAAM,CACzD4B,YACAhB,aACA,IAAI,CAAC,IAAI,CAAC,eAAe;QAI3B,IAAI,IAAI,CAAC,SAAS,IAAIC,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,QAAQ,AAAD,KAAKkB,AAAc,UAAdA,WAAqB;YAC7D,MAAMM,cAAkC;gBACtC,OAAO;oBACL;wBACE,MAAMT;wBACN,MAAMf,OAAO,QAAQ;oBACvB;iBACD;YACH;YACA,MAAMyB,cAAcF,QAAAA,IAAS,CAACC;YAC9B,IAAI,CAAC,SAAS,CAAC,yBAAyB,CACtC;gBACE,MAAM;gBACN,QAAQT;gBACR,cAAcU;YAChB,GACAL;QAEJ;QAEA,MAAM,IAAI,CAAC,gBAAgB,CAACjC;QAC5B,OAAOa;IACT;IAEA,MAAM,QACJ0B,MAA2B,EAC3BlC,MAA4B5B,2BAA2B,EAClC;QACrB,MAAMmC,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAC3D,MAAM,EAAEC,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,SACAuC,QACA3B,aACAP;QAEJ,MAAM,IAAI,CAAC,gBAAgB,CAACL;QAC5B,OAAOa;IACT;IAEA,MAAM,UACJhC,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACrC;QAClB,MAAMmC,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAE4B,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAY7D;QACrD,MAAM,EAAEgC,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,WACAwC,YACA5B,aACAP,KACAoC;QAEJ,MAAM,IAAI,CAAC,gBAAgB,CAACzC;QAC5B,OAAOa;IACT;IAEA,MAAM,SACJhC,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACtC;QACjB,MAAMmC,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAE4B,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAY7D;QACrD,MAAM,EAAEgC,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,UACAwC,YACA5B,aACAP,KACAoC;QAEJ,MAAM,IAAI,CAAC,gBAAgB,CAACzC;QAC5B,OAAOa;IACT;IAEA,MAAM,SACJhC,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACtC;QACjB,MAAMmC,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAE4B,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAY7D;QACrD,MAAM,EAAEgC,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,UACAwC,YACA5B,aACAP,KACAoC;QAEJ,MAAM,IAAI,CAAC,gBAAgB,CAACzC;QAC5B,OAAOa;IACT;IAEA,MAAM,MACJhC,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACtC;QACjB,OAAO,IAAI,CAAC,QAAQ,CAACI,QAAQwB;IAC/B;IAEA,MAAM,uBACJsC,MAAwB,EACxBtC,GAI0B,EACkB;QAC5C,MAAM,EAAEuC,eAAe,IAAI,EAAEC,aAAa,CAAC,EAAE,GAAGxC,OAAO,CAAC;QAExD,IAAIyC,UAAU;QACd,IAAIC,aAAa;QACjB,IAAIC,eAAe;QACnB,IAAIC,YAAY5C,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,AAAD,KAAK;QAClC,IAAI6C;QAEJ,MAAO,CAACJ,WAAWC,aAAaF,WAAY;YAC1C,IAAIE,cAAc,GAChBE,YAAY;YAEd3F,MACE,cACAqF,QACA,gBACAC,cACA,cACAG,YACA,aACAE;YAGF,MAAMrC,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;YAE3D,MAAMuC,OAAO,MAAM,IAAI,CAAC,OAAO,CAAC,QAAQ,CAACR,QAAQ/B,aAAa;gBAC5DqC;YACF;YACA3F,MAAM,mBAAmB6F;YACzBpC,OAAOoC,KAAK,WAAW,EAAE,CAAC,+BAA+B,EAAER,OAAO,CAAC,CAAC;YACpEK,eAAeG,KAAK,WAAW;YAE/BD,eAAe,MAAM,IAAI,CAAC,aAAa,CACrCF,cACAC,YAAY;gBAAE,WAAW;YAAK,IAAIxB,QAClCkB,QACAtC;YAEF,IAAI6C,aAAa,IAAI,EACnBJ,UAAU;iBAEVC;QAEJ;QAEA,OAAO;YACL,QAAQC;YACRC;YACAC;QACF;IACF;IAEA,MAAM,cACJrE,MAAc,EACduE,SAAmC,EACnCC,YAA8B,EAC9BC,kBAA2C,EACX;QAChChG,MAAM,iBAAiBuB,QAAQuE,WAAWC,cAAcC;QAExD,MAAM,EAAE,QAAQC,YAAY,EAAE,MAAMC,UAAU,EAAE,GAAG,MAAM,IAAI,CAAC,QAAQ,CACpE3E,QACAuE;QAEF,MAAMK,WAAWjG,oBAAoB6F,cAAcE;QACnD,MAAMG,WAAW1F,eAAeqF,cAAcG;QAC9C,MAAMG,OACJF,YAAaH,CAAAA,CAAAA,QAAAA,qBAAAA,KAAAA,IAAAA,mBAAoB,uBAAuB,AAAD,KAAK,EAAC,KAC7DI;QACF,MAAMR,eAAe;YACnBS;YACA,MAAMH;YACN,QAAQD;YACR,gBAAgBE;QAClB;QACAnG,MAAM,2BAA2B4F;QACjC,OAAOA;IACT;IAEA,MAAM,SAASrE,MAAmB,EAAEwB,GAAkB,EAAE;QACtD,MAAMuD,cAAc3C,yBAAyBpC,QAAQwB;QACrDU,OAAO6C,aAAa;QACpB,MAAMC,aAAaC,oBAAoBF;QACvC,MAAMrD,QAAQ;YAACsD;SAAW;QAC1B,MAAMjD,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAEZ,QAAQ,EAAEa,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,CAC3DH,aAAa,UAAUC,eAAeiD,eACtCrD,OACAK;QAEF,MAAM,IAAI,CAAC,gBAAgB,CAACZ;QAE5B,MAAM,EAAE+D,OAAO,EAAE,GAAGlD;QAEpB,OAAO;YACL,MAAMkD,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,IAAI;YACnB,QAAQA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM;YACvB,OAAQ,OAAM,IAAI,CAAC,SAAS,CAAC,IAAI,EAAC,EAAG,GAAG;QAC1C;IAGF;IAEA,MAAM,SACJC,SAAsB,EACtBC,GAAY,EACZ5D,GAA2C,EAC3C;YAsBiB6D;QArBjB,MAAMtD,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAMuD,aAAmC;YACvC,aAAa9D,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,KAAK5B,4BAA4B,WAAW;YACxE,oBACE4B,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,kBAAkB,AAAD,KACtB5B,4BAA4B,kBAAkB;YAChD,iBAAiB4B,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe;YACrC,iBAAiBA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe;QACvC;QAEA,MAAM,EAAEQ,MAAM,EAAEb,QAAQ,EAAEoE,OAAO,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,MAAM,CAClEJ,WACApD,aACAuD;QAEF,MAAM,IAAI,CAAC,gBAAgB,CAACnE,UAAU;QAEtC,MAAMqE,UAAUxD,SACZY,SACA,CAAC,kBAAkB,EAAEwC,OAAQ,CAAqB,YAArB,OAAOD,YAAyBA,YAAYA,UAAU,MAAK,EAAG,UAAU,EACnGI,WAAAA,SAAWF,CAAAA,4BAAAA,SAAS,eAAe,EAAC,IAAzBA,KAAAA,IAAAA,0BAA4B,KAAK,AAAD,KAAK,eAChD;QAEN,IAAI7D,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,EACtB,OAAO;YACL,MAAMQ;YACNuD;YACAC;QACF;QAGF,IAAI,CAACxD,QACH,MAAM,IAAIxB,MAAMgF;IAEpB;IAEA,MAAM,UAAUL,SAAsB,EAAE3D,GAAqB,EAAE;QAC7D,MAAMO,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAC3D,MAAM,EAAEZ,QAAQ,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,OAAO,CAClDgE,WACA;YACE,WAAW3D,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,AAAD,KAAK;YAC7B,iBAAiBA,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,AAAD,KAAK;QAC3C,GACAO;QAEF,MAAM,IAAI,CAAC,gBAAgB,CAACZ,UAAU;QAEtC,IAAIA,SAAS,cAAc,IAAI;YAC7B,MAAMG,YAAYH,SAAS,eAAe;YAC1C,MAAM,IAAIX,MAAM,GAAGc,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,KAAK,CAAC,EAAE,EAAEA,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,UAAU,EAAE;QACjE;IACF;IAEA,MAAM,GAAGyB,UAAkB,EAAExB,OAAO,QAAQ,EAAE;QAC5C,IAAIA,AAAS,aAATA,MACF,OAAO,IAAI,CAAC,QAAQ,CAACwB;QAEvB,IAAIxB,AAAS,YAATA,MACF,OAAO,IAAI,CAAC,OAAO,CAACwB;QAGtB,IAAIxB,AAAS,aAATA,MACF,OAAO,IAAI,CAAC,QAAQ,CAACwB;QAGvB,IAAIxB,AAAS,UAATA,MACF,OAAO,IAAI,CAAC,KAAK,CAACwB;QAGpB,IAAIxB,AAAS,iBAATA,MACF,OAAO,IAAI,CAAC,YAAY,CAACwB;QAG3B,IAAIxB,AAAS,kBAATA,MACF,OAAO,IAAI,CAAC,aAAa,CAACwB;QAG5B,MAAM,IAAIvC,MACR,CAAC,cAAc,EAAEe,KAAK,8EAA8E,CAAC;IAEzG;IAEA,MAAM,QAAQkE,iBAAyB,EAEpC;QACD,MAAMC,SAASC,gBAAgBF,mBAAmB;QAClD,MAAMG,SAAS,IAAIC,aAAaH,QAAQ,UAC/B;gBAAE,OAAO,IAAI;gBAAE,QAAQ,EAAE;YAAC;QAEnC,MAAME,OAAO,GAAG;QAEhB,IAAIA,AAAkB,YAAlBA,OAAO,MAAM,EAAc;YAC7B,MAAME,SAASF,OAAO,cAAc,CACjC,MAAM,CAAC,CAAC9E,OAASA,AAAgB,YAAhBA,KAAK,MAAM,EAC5B,GAAG,CAAC,CAACA;oBAC2BiF;gBAA/B,OAAO,CAAC,OAAO,EAAEjF,KAAK,IAAI,CAAC,EAAE,EAAE,QAAAiF,CAAAA,cAAAA,KAAK,KAAK,AAAD,IAATA,KAAAA,IAAAA,YAAY,OAAO,EAAE;YACtD,GACC,IAAI,CAAC;YACR,MAAM,IAAIvF,MAAM,CAAC,2CAA2C,EAAEsF,QAAQ;QACxE;QAEA,OAAO;YACL,QAAQF,OAAO,MAAM;QACvB;IACF;IAEA,MAAM,mBAAmBF,MAAc,EAAE;QACvCxD,OACE,IAAI,CAAC,SAAS,CAAC,kBAAkB,EACjC;QAEF,OAAO,IAAI,CAAC,SAAS,CAAC,kBAAkB,CAACwD;IAC3C;IAEA,MAAM,UAAU;YACRM,yBAAAA;QAAN,eAAMA,CAAAA,0BAAAA,AAAAA,CAAAA,kBAAAA,IAAI,CAAC,SAAS,AAAD,EAAE,OAAO,AAAD,IAArBA,KAAAA,IAAAA,wBAAAA,IAAAA,CAAAA,gBAAAA;QACN,IAAI,CAAC,SAAS;QACd,IAAI,CAAC,SAAS,GAAG;IACnB;IAEA,MAAM,cACJpE,KAAc,EACdJ,GAEC,EACD;QAEA,MAAMyE,SAAS,MAAM,IAAI,CAAC,SAAS,CAAC,gBAAgB;QACpD,MAAMC,MAAMC,KAAK,GAAG;QAEpB,MAAMC,WAAoC;YACxC;gBACE,MAAM;gBACN,IAAIF;gBACJ,YAAYD;YACd;SACD;QAED,MAAMnF,OAAyB;YAC7B,MAAM;YACN,SAAS;YACT,QAAQ;YACRsF;YACA,QAAQ;gBACN,OAAOF;gBACP,KAAKA;gBACL,MAAM;YACR;YACA,OAAO;gBACL,SAAS1E,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAK;YAC3B;YACA,UAAU,WAAa;QACzB;QAEA,MAAM6E,gBAA+B;YACnC,YAAY;YACZ,SAASH;YACT,MAAM,CAAC,MAAM,EAAEtE,SAAS,YAAY;YACpC,aAAaJ,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAK;YAC7B,OAAO;gBAACV;aAAK;QACf;QAEA,IAAI,CAAC,mBAAmB,CAACuF;QAEzB,IAAI;gBACFC,oBAAAA;oBAAAA,CAAAA,qBAAAA,AAAAA,CAAAA,QAAAA,IAAI,AAAD,EAAE,YAAY,AAAD,KAAhBA,mBAAAA,IAAAA,CAAAA,OAAoB,IAAI,CAAC,cAAc;QACzC,EAAE,OAAOjF,OAAO;YACdpB,QAAQ,KAAK,CAAC,yBAAyBoB;QACzC;QAEA,IAAI,CAAC,mBAAmB;IAC1B;IAEA,sBAAsB;QACpB,MAAM,EAAEkF,SAAS,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAG,IAAI,CAAC,IAAI;QAC7D,MAAMC,gBAAgBC,MAAM,OAAO,CAACF,cAChCA,WAAW,GAAG,CAAC,CAACvG;YACd,MAAM,EAAE0G,KAAK,EAAE,GAAGC,eAAe,GAAG3G;YACpC,IAAI4G,WAAWF;YACf,IAAID,MAAM,OAAO,CAACC,QAChBE,WAAWF,MAAM,GAAG,CAAC,CAAC9F;gBAEpB,MAAM,EAAEiG,SAAS,EAAEC,GAAG,EAAE,GAAGC,UAAU,GAAGnG;gBACxC,OAAOmG;YACT;YAEF,OAAO;gBAAE,GAAGJ,aAAa;gBAAE,GAAIC,WAAW;oBAAE,OAAOA;gBAAS,IAAI,CAAC,CAAC;YAAE;QACtE,KACA,EAAE;QACN,OAAO;YACLP;YACAC;YACA,YAAYE;QACd;IACF;IAMA,MAAM,oBAAmC;QACvCjI,MAAM;QACN,MAAMyI,UAAU,MAAM,IAAI,CAAC,gBAAgB;QAE3CA,QAAQ,SAAS,GAAG;QACpB,IAAI,CAAC,eAAe,GAAGA;QACvBzI,MAAM;IACR;IAKA,MAAM,sBAAqC;QACzCA,MAAM;QACN,IAAI,CAAC,eAAe,GAAGmE;QACvBnE,MAAM;IACR;IAKQ,mBACN0I,IAAc,EAC8C;QAE5D,IAAIA,AAAevE,WAAfuE,KAAK,KAAK,EAAgB;YAC5B,IAAIA,AAAe,UAAfA,KAAK,KAAK,EACZ,OAAO;YAGT,IAAIA,AAAe,SAAfA,KAAK,KAAK,EACZ,MAAM,IAAI3G,MACR;YAMJ,IAAI,AAAsB,YAAtB,OAAO2G,KAAK,KAAK,EAAe;gBAClC,MAAMC,SAASD,KAAK,KAAK;gBACzB,IAAI,CAACC,OAAO,EAAE,EACZ,MAAM,IAAI5G,MACR;gBAIJ,MAAM6G,KAAKD,OAAO,EAAE;gBACpB,MAAME,aAAaF,AAAoB,gBAApBA,OAAO,QAAQ;gBAElC,OAAO;oBACLC;oBACA,SAAS;oBACT,UAAUC;gBACZ;YACF;QACF;QAGA,IAAIH,KAAK,OAAO,EAAE;YAChB,MAAMI,aACJC,oBAAoB,qBAAqB,CAACC;YAC5C,IAAIF,YACF,OAAO;gBACL,IAAIJ,KAAK,OAAO;gBAChB,SAAS;gBACT,UAAU;YACZ;QAEJ;QAGA,OAAO;IACT;IAMA,MAAM,aAA4B;QAChC,IAAI,CAAC,IAAI,CAAC,SAAS,EACjB,MAAM,IAAI3G,MAAM;QAGlB,IAAI,CAAC,SAAS,CAAC,gBAAgB;IACjC;IA58BA,YAAYkH,iBAAgC,EAAEP,IAAe,CAAE;QA5D/D;QAEA;QAEA;QAEA;QAEA;QAEA;QAEA;QAKA,kCAAU;QAEV;QAEA;QAEA;QAEA,oCAAY;QAEZ;QAKA,uBAAQ,mBAAR;QAKA,uBAAQ,uBAAsB;QAwB5B,IAAI,CAAC,SAAS,GAAGO;QACjB,IAAI,CAAC,IAAI,GAAGC,OAAO,MAAM,CACvB;YACE,gBAAgB;YAChB,oBAAoB;YACpB,WAAW;YACX,kBAAkB;QACpB,GACAR,QAAQ,CAAC;QAGX,IAAIA,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW,AAAD,KAAK,AAA6B,cAA7B,OAAOA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW,AAAD,GAC9C,MAAM,IAAI3G,MACR,CAAC,+DAA+D,EAAE,OAAO2G,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW,AAAD,GAAG;QAGhG,IAAI,CAAC,kBAAkB,GAAGA,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW,AAAD,IACtC,IAAIS,mBAAmBT,KAAK,WAAW,IACvCU;QAEJ,IAAI,CAAC,cAAc,GAAG,IAAI,CAAC,IAAI,CAAC,cAAc;QAE9C,IAAI,CAAC,OAAO,GAAG,IAAIC,QAAQ,OAAOhI,SACzB,IAAI,CAAC,YAAY,CAACA;QAI3B,MAAMiI,cAAc,IAAI,CAAC,kBAAkB,CAACZ,QAAQ,CAAC;QACrD,IAAIY,aACF,IAAI,CAAC,SAAS,GAAG,IAAIC,UACnBD,YAAY,EAAE,EACdA,YAAY,OAAO,EACnBnF,QACAmF,YAAY,QAAQ;QAIxB,IAAI,CAAC,YAAY,GAAG,IAAIE,aAAa,IAAI,CAAC,SAAS,EAAE,IAAI,CAAC,OAAO,EAAE;YACjE,WAAW,IAAI,CAAC,SAAS;YACzB,aAAa,IAAI,CAAC,sBAAsB,CAAC,IAAI,CAAC,IAAI;YAClD,sBAAsB,IAAI,CAAC,IAAI,CAAC,oBAAoB;QACtD;QACA,IAAI,CAAC,IAAI,GAAG,IAAI,CAAC,SAAS;QAC1B,IAAI,CAAC,cAAc,GACjBd,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,cAAc,AAAD,KACnBe,kBAAkBf,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,MAAM,AAAD,KAAK,IAAI,CAAC,SAAS,CAAC,aAAa,IAAI;IACtE;AA85BF;AAEO,MAAMgB,cAAc,CACzBT,mBACAP,OAEO,IAAItH,MAAM6H,mBAAmBP"}
1
+ {"version":3,"file":"agent/agent.mjs","sources":["webpack://@midscene/core/./src/agent/agent.ts"],"sourcesContent":["import {\n type AgentAssertOpt,\n type AgentDescribeElementAtPointResult,\n type AgentWaitForOpt,\n type DeviceAction,\n type ExecutionDump,\n type ExecutionRecorderItem,\n type ExecutionTask,\n type ExecutionTaskLog,\n type Executor,\n type GroupedActionDump,\n Insight,\n type InsightAction,\n type InsightExtractOption,\n type InsightExtractParam,\n type LocateOption,\n type LocateResultElement,\n type LocateValidatorResult,\n type LocatorValidatorOption,\n type MidsceneYamlScript,\n type OnTaskStartTip,\n type PlanningAction,\n type Rect,\n type ScrollParam,\n type TUserPrompt,\n type UIContext,\n} from '../index';\n\nimport yaml from 'js-yaml';\n\nimport {\n groupedActionDumpFileExt,\n reportHTMLContent,\n stringifyDumpData,\n writeLogFile,\n} from '@/utils';\nimport {\n ScriptPlayer,\n buildDetailedLocateParam,\n parseYamlScript,\n} from '../yaml/index';\n\nimport type { AbstractInterface } from '@/device';\nimport type { AgentOpt } from '@/types';\nimport {\n MIDSCENE_CACHE,\n ModelConfigManager,\n globalConfigManager,\n globalModelConfigManager,\n} from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\n// import type { AndroidDeviceInputOpt } from '../device';\nimport { TaskCache } from './task-cache';\nimport { TaskExecutor, locatePlanForLocate } from './tasks';\nimport { locateParamStr, paramStr, taskTitleStr, typeStr } from './ui-utils';\nimport {\n commonContextParser,\n getReportFileName,\n parsePrompt,\n printReportMsg,\n} from './utils';\nimport { trimContextByViewport } from './utils';\n\nconst debug = getDebug('agent');\n\nconst distanceOfTwoPoints = (p1: [number, number], p2: [number, number]) => {\n const [x1, y1] = p1;\n const [x2, y2] = p2;\n return Math.round(Math.sqrt((x1 - x2) ** 2 + (y1 - y2) ** 2));\n};\n\nconst includedInRect = (point: [number, number], rect: Rect) => {\n const [x, y] = point;\n const { left, top, width, height } = rect;\n return x >= left && x <= left + width && y >= top && y <= top + height;\n};\n\nconst defaultInsightExtractOption: InsightExtractOption = {\n domIncluded: false,\n screenshotIncluded: true,\n};\n\nexport class Agent<\n InterfaceType extends AbstractInterface = AbstractInterface,\n> {\n interface: InterfaceType;\n\n insight: Insight;\n\n dump: GroupedActionDump;\n\n reportFile?: string | null;\n\n reportFileName?: string;\n\n taskExecutor: TaskExecutor;\n\n opts: AgentOpt;\n\n /**\n * If true, the agent will not perform any actions\n */\n dryMode = false;\n\n onTaskStartTip?: OnTaskStartTip;\n\n taskCache?: TaskCache;\n\n onDumpUpdate?: (dump: string) => void;\n\n destroyed = false;\n\n modelConfigManager: ModelConfigManager;\n\n /**\n * Frozen page context for consistent AI operations\n */\n private frozenUIContext?: UIContext;\n\n /**\n * Flag to track if VL model warning has been shown\n */\n private hasWarnedNonVLModel = false;\n\n // @deprecated use .interface instead\n get page() {\n return this.interface;\n }\n\n /**\n * Ensures VL model warning is shown once when needed\n */\n private ensureVLModelWarning() {\n if (\n !this.hasWarnedNonVLModel &&\n this.interface.interfaceType !== 'puppeteer' &&\n this.interface.interfaceType !== 'playwright' &&\n this.interface.interfaceType !== 'static' &&\n this.interface.interfaceType !== 'chrome-extension-proxy'\n ) {\n this.modelConfigManager.throwErrorIfNonVLModel();\n this.hasWarnedNonVLModel = true;\n }\n }\n\n constructor(interfaceInstance: InterfaceType, opts?: AgentOpt) {\n this.interface = interfaceInstance;\n this.opts = Object.assign(\n {\n generateReport: true,\n autoPrintReportMsg: true,\n groupName: 'Midscene Report',\n groupDescription: '',\n },\n opts || {},\n );\n\n if (opts?.modelConfig && typeof opts?.modelConfig !== 'function') {\n throw new Error(\n `opts.modelConfig must be one of function or undefined, but got ${typeof opts?.modelConfig}`,\n );\n }\n this.modelConfigManager = opts?.modelConfig\n ? new ModelConfigManager(opts.modelConfig)\n : globalModelConfigManager;\n\n this.onTaskStartTip = this.opts.onTaskStartTip;\n\n this.insight = new Insight(async (action: InsightAction) => {\n return this.getUIContext(action);\n });\n\n // Process cache configuration\n const cacheConfig = this.processCacheConfig(opts || {});\n if (cacheConfig) {\n this.taskCache = new TaskCache(\n cacheConfig.id,\n cacheConfig.enabled,\n undefined, // cacheFilePath\n cacheConfig.readOnly,\n );\n }\n\n this.taskExecutor = new TaskExecutor(this.interface, this.insight, {\n taskCache: this.taskCache,\n onTaskStart: this.callbackOnTaskStartTip.bind(this),\n replanningCycleLimit: this.opts.replanningCycleLimit,\n });\n this.dump = this.resetDump();\n this.reportFileName =\n opts?.reportFileName ||\n getReportFileName(opts?.testId || this.interface.interfaceType || 'web');\n }\n\n async getActionSpace(): Promise<DeviceAction[]> {\n return this.interface.actionSpace();\n }\n\n async getUIContext(action?: InsightAction): Promise<UIContext> {\n // Check VL model configuration when UI context is first needed\n this.ensureVLModelWarning();\n\n // If page context is frozen, return the frozen context for all actions\n if (this.frozenUIContext) {\n debug('Using frozen page context for action:', action);\n return this.frozenUIContext;\n }\n\n if (this.interface.getContext) {\n debug('Using page.getContext for action:', action);\n return await this.interface.getContext();\n } else {\n debug('Using commonContextParser for action:', action);\n return await commonContextParser(this.interface, {\n uploadServerUrl: this.modelConfigManager.getUploadTestServerUrl(),\n });\n }\n }\n\n async _snapshotContext(): Promise<UIContext> {\n return await this.getUIContext('locate');\n }\n\n async setAIActionContext(prompt: string) {\n if (this.opts.aiActionContext) {\n console.warn(\n 'aiActionContext is already set, and it is called again, will override the previous setting',\n );\n }\n this.opts.aiActionContext = prompt;\n }\n\n resetDump() {\n this.dump = {\n groupName: this.opts.groupName!,\n groupDescription: this.opts.groupDescription,\n executions: [],\n modelBriefs: [],\n };\n\n return this.dump;\n }\n\n appendExecutionDump(execution: ExecutionDump) {\n // use trimContextByViewport to process execution\n const trimmedExecution = trimContextByViewport(execution);\n const currentDump = this.dump;\n currentDump.executions.push(trimmedExecution);\n }\n\n dumpDataString() {\n // update dump info\n this.dump.groupName = this.opts.groupName!;\n this.dump.groupDescription = this.opts.groupDescription;\n return stringifyDumpData(this.dump);\n }\n\n reportHTMLString() {\n return reportHTMLContent(this.dumpDataString());\n }\n\n writeOutActionDumps() {\n if (this.destroyed) {\n throw new Error(\n 'PageAgent has been destroyed. Cannot update report file.',\n );\n }\n const { generateReport, autoPrintReportMsg } = this.opts;\n this.reportFile = writeLogFile({\n fileName: this.reportFileName!,\n fileExt: groupedActionDumpFileExt,\n fileContent: this.dumpDataString(),\n type: 'dump',\n generateReport,\n });\n debug('writeOutActionDumps', this.reportFile);\n if (generateReport && autoPrintReportMsg && this.reportFile) {\n printReportMsg(this.reportFile);\n }\n }\n\n private async callbackOnTaskStartTip(task: ExecutionTask) {\n const param = paramStr(task);\n const tip = param ? `${typeStr(task)} - ${param}` : typeStr(task);\n\n if (this.onTaskStartTip) {\n await this.onTaskStartTip(tip);\n }\n }\n\n private async afterTaskRunning(executor: Executor, doNotThrowError = false) {\n this.appendExecutionDump(executor.dump());\n\n try {\n if (this.onDumpUpdate) {\n this.onDumpUpdate(this.dumpDataString());\n }\n } catch (error) {\n console.error('Error in onDumpUpdate', error);\n }\n\n this.writeOutActionDumps();\n\n if (executor.isInErrorState() && !doNotThrowError) {\n const errorTask = executor.latestErrorTask();\n throw new Error(`${errorTask?.errorMessage}\\n${errorTask?.errorStack}`, {\n cause: errorTask?.error,\n });\n }\n }\n\n async callActionInActionSpace<T = any>(\n type: string,\n opt?: T, // and all other action params\n ) {\n debug('callActionInActionSpace', type, ',', opt);\n\n const actionPlan: PlanningAction<T> = {\n type: type as any,\n param: (opt as any) || {},\n thought: '',\n };\n debug('actionPlan', actionPlan); // , ', in which the locateParam is', locateParam);\n\n const plans: PlanningAction[] = [actionPlan].filter(\n Boolean,\n ) as PlanningAction[];\n\n const title = taskTitleStr(\n type as any,\n locateParamStr((opt as any)?.locate || {}),\n );\n\n // assume all operation in action space is related to locating\n const modelConfig = this.modelConfigManager.getModelConfig('grounding');\n\n const { output, executor } = await this.taskExecutor.runPlans(\n title,\n plans,\n modelConfig,\n );\n await this.afterTaskRunning(executor);\n return output;\n }\n\n async aiTap(locatePrompt: TUserPrompt, opt?: LocateOption) {\n assert(locatePrompt, 'missing locate prompt for tap');\n\n const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n return this.callActionInActionSpace('Tap', {\n locate: detailedLocateParam,\n });\n }\n\n async aiRightClick(locatePrompt: TUserPrompt, opt?: LocateOption) {\n assert(locatePrompt, 'missing locate prompt for right click');\n\n const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n return this.callActionInActionSpace('RightClick', {\n locate: detailedLocateParam,\n });\n }\n\n async aiDoubleClick(locatePrompt: TUserPrompt, opt?: LocateOption) {\n assert(locatePrompt, 'missing locate prompt for double click');\n\n const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n return this.callActionInActionSpace('DoubleClick', {\n locate: detailedLocateParam,\n });\n }\n\n async aiHover(locatePrompt: TUserPrompt, opt?: LocateOption) {\n assert(locatePrompt, 'missing locate prompt for hover');\n\n const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n return this.callActionInActionSpace('Hover', {\n locate: detailedLocateParam,\n });\n }\n\n // New signature, always use locatePrompt as the first param\n async aiInput(\n locatePrompt: TUserPrompt,\n opt: LocateOption & { value: string } & { autoDismissKeyboard?: boolean },\n ): Promise<any>;\n\n // Legacy signature - deprecated\n /**\n * @deprecated Use aiInput(locatePrompt, opt) instead where opt contains the value\n */\n async aiInput(\n value: string,\n locatePrompt: TUserPrompt,\n opt?: LocateOption & { autoDismissKeyboard?: boolean }, // AndroidDeviceInputOpt &\n ): Promise<any>;\n\n // Implementation\n async aiInput(\n locatePromptOrValue: TUserPrompt | string,\n locatePromptOrOpt:\n | TUserPrompt\n | (LocateOption & { value: string } & { autoDismissKeyboard?: boolean }) // AndroidDeviceInputOpt &\n | undefined,\n optOrUndefined?: LocateOption, // AndroidDeviceInputOpt &\n ) {\n let value: string;\n let locatePrompt: TUserPrompt;\n let opt:\n | (LocateOption & { value: string } & { autoDismissKeyboard?: boolean }) // AndroidDeviceInputOpt &\n | undefined;\n\n // Check if using new signature (first param is locatePrompt, second has value)\n if (\n typeof locatePromptOrOpt === 'object' &&\n locatePromptOrOpt !== null &&\n 'value' in locatePromptOrOpt\n ) {\n // New signature: aiInput(locatePrompt, opt)\n locatePrompt = locatePromptOrValue as TUserPrompt;\n const optWithValue = locatePromptOrOpt as LocateOption & {\n // AndroidDeviceInputOpt &\n value: string;\n autoDismissKeyboard?: boolean;\n };\n value = optWithValue.value;\n opt = optWithValue;\n } else {\n // Legacy signature: aiInput(value, locatePrompt, opt)\n value = locatePromptOrValue as string;\n locatePrompt = locatePromptOrOpt as TUserPrompt;\n opt = {\n ...optOrUndefined,\n value,\n };\n }\n\n assert(\n typeof value === 'string',\n 'input value must be a string, use empty string if you want to clear the input',\n );\n assert(locatePrompt, 'missing locate prompt for input');\n\n const detailedLocateParam = buildDetailedLocateParam(locatePrompt, opt);\n\n return this.callActionInActionSpace('Input', {\n ...(opt || {}),\n locate: detailedLocateParam,\n });\n }\n\n // New signature\n async aiKeyboardPress(\n locatePrompt: TUserPrompt,\n opt: LocateOption & { keyName: string },\n ): Promise<any>;\n\n // Legacy signature - deprecated\n /**\n * @deprecated Use aiKeyboardPress(locatePrompt, opt) instead where opt contains the keyName\n */\n async aiKeyboardPress(\n keyName: string,\n locatePrompt?: TUserPrompt,\n opt?: LocateOption,\n ): Promise<any>;\n\n // Implementation\n async aiKeyboardPress(\n locatePromptOrKeyName: TUserPrompt | string,\n locatePromptOrOpt:\n | TUserPrompt\n | (LocateOption & { keyName: string })\n | undefined,\n optOrUndefined?: LocateOption,\n ) {\n let keyName: string;\n let locatePrompt: TUserPrompt | undefined;\n let opt: (LocateOption & { keyName: string }) | undefined;\n\n // Check if using new signature (first param is locatePrompt, second has keyName)\n if (\n typeof locatePromptOrOpt === 'object' &&\n locatePromptOrOpt !== null &&\n 'keyName' in locatePromptOrOpt\n ) {\n // New signature: aiKeyboardPress(locatePrompt, opt)\n locatePrompt = locatePromptOrKeyName as TUserPrompt;\n opt = locatePromptOrOpt as LocateOption & {\n keyName: string;\n };\n } else {\n // Legacy signature: aiKeyboardPress(keyName, locatePrompt, opt)\n keyName = locatePromptOrKeyName as string;\n locatePrompt = locatePromptOrOpt as TUserPrompt | undefined;\n opt = {\n ...(optOrUndefined || {}),\n keyName,\n };\n }\n\n assert(opt?.keyName, 'missing keyName for keyboard press');\n\n const detailedLocateParam = locatePrompt\n ? buildDetailedLocateParam(locatePrompt, opt)\n : undefined;\n\n return this.callActionInActionSpace('KeyboardPress', {\n ...(opt || {}),\n locate: detailedLocateParam,\n });\n }\n\n // New signature\n async aiScroll(\n locatePrompt: TUserPrompt | undefined,\n opt: LocateOption & ScrollParam,\n ): Promise<any>;\n\n // Legacy signature - deprecated\n /**\n * @deprecated Use aiScroll(locatePrompt, opt) instead where opt contains the scroll parameters\n */\n async aiScroll(\n scrollParam: ScrollParam,\n locatePrompt?: TUserPrompt,\n opt?: LocateOption,\n ): Promise<any>;\n\n // Implementation\n async aiScroll(\n locatePromptOrScrollParam: TUserPrompt | ScrollParam | undefined,\n locatePromptOrOpt: TUserPrompt | (LocateOption & ScrollParam) | undefined,\n optOrUndefined?: LocateOption,\n ) {\n let scrollParam: ScrollParam;\n let locatePrompt: TUserPrompt | undefined;\n let opt: LocateOption | undefined;\n\n // Check if using new signature (first param is locatePrompt, second has scroll params)\n if (\n typeof locatePromptOrOpt === 'object' &&\n ('direction' in locatePromptOrOpt ||\n 'scrollType' in locatePromptOrOpt ||\n 'distance' in locatePromptOrOpt)\n ) {\n // New signature: aiScroll(locatePrompt, opt)\n locatePrompt = locatePromptOrScrollParam as TUserPrompt;\n opt = locatePromptOrOpt as LocateOption & ScrollParam;\n } else {\n // Legacy signature: aiScroll(scrollParam, locatePrompt, opt)\n scrollParam = locatePromptOrScrollParam as ScrollParam;\n locatePrompt = locatePromptOrOpt as TUserPrompt | undefined;\n opt = {\n ...(optOrUndefined || {}),\n ...(scrollParam || {}),\n };\n }\n\n const detailedLocateParam = buildDetailedLocateParam(\n locatePrompt || '',\n opt,\n );\n\n return this.callActionInActionSpace('Scroll', {\n ...(opt || {}),\n locate: detailedLocateParam,\n });\n }\n\n async aiAction(\n taskPrompt: string,\n opt?: {\n cacheable?: boolean;\n },\n ) {\n const modelConfig = this.modelConfigManager.getModelConfig('planning');\n\n const cacheable = opt?.cacheable;\n // if vlm-ui-tars, plan cache is not used\n const isVlmUiTars = modelConfig.vlMode === 'vlm-ui-tars';\n const matchedCache =\n isVlmUiTars || cacheable === false\n ? undefined\n : this.taskCache?.matchPlanCache(taskPrompt);\n if (matchedCache && this.taskCache?.isCacheResultUsed) {\n // log into report file\n const { executor } = await this.taskExecutor.loadYamlFlowAsPlanning(\n taskPrompt,\n matchedCache.cacheContent?.yamlWorkflow,\n );\n\n await this.afterTaskRunning(executor);\n\n debug('matched cache, will call .runYaml to run the action');\n const yaml = matchedCache.cacheContent?.yamlWorkflow;\n return this.runYaml(yaml);\n }\n\n const { output, executor } = await this.taskExecutor.action(\n taskPrompt,\n modelConfig,\n this.opts.aiActionContext,\n );\n\n // update cache\n if (this.taskCache && output?.yamlFlow && cacheable !== false) {\n const yamlContent: MidsceneYamlScript = {\n tasks: [\n {\n name: taskPrompt,\n flow: output.yamlFlow,\n },\n ],\n };\n const yamlFlowStr = yaml.dump(yamlContent);\n this.taskCache.updateOrAppendCacheRecord(\n {\n type: 'plan',\n prompt: taskPrompt,\n yamlWorkflow: yamlFlowStr,\n },\n matchedCache,\n );\n }\n\n await this.afterTaskRunning(executor);\n return output;\n }\n\n async aiQuery<ReturnType = any>(\n demand: InsightExtractParam,\n opt: InsightExtractOption = defaultInsightExtractOption,\n ): Promise<ReturnType> {\n const modelConfig = this.modelConfigManager.getModelConfig('VQA');\n const { output, executor } =\n await this.taskExecutor.createTypeQueryExecution(\n 'Query',\n demand,\n modelConfig,\n opt,\n );\n await this.afterTaskRunning(executor);\n return output as ReturnType;\n }\n\n async aiBoolean(\n prompt: TUserPrompt,\n opt: InsightExtractOption = defaultInsightExtractOption,\n ): Promise<boolean> {\n const modelConfig = this.modelConfigManager.getModelConfig('VQA');\n\n const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n const { output, executor } =\n await this.taskExecutor.createTypeQueryExecution(\n 'Boolean',\n textPrompt,\n modelConfig,\n opt,\n multimodalPrompt,\n );\n await this.afterTaskRunning(executor);\n return output as boolean;\n }\n\n async aiNumber(\n prompt: TUserPrompt,\n opt: InsightExtractOption = defaultInsightExtractOption,\n ): Promise<number> {\n const modelConfig = this.modelConfigManager.getModelConfig('VQA');\n\n const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n const { output, executor } =\n await this.taskExecutor.createTypeQueryExecution(\n 'Number',\n textPrompt,\n modelConfig,\n opt,\n multimodalPrompt,\n );\n await this.afterTaskRunning(executor);\n return output as number;\n }\n\n async aiString(\n prompt: TUserPrompt,\n opt: InsightExtractOption = defaultInsightExtractOption,\n ): Promise<string> {\n const modelConfig = this.modelConfigManager.getModelConfig('VQA');\n\n const { textPrompt, multimodalPrompt } = parsePrompt(prompt);\n const { output, executor } =\n await this.taskExecutor.createTypeQueryExecution(\n 'String',\n textPrompt,\n modelConfig,\n opt,\n multimodalPrompt,\n );\n await this.afterTaskRunning(executor);\n return output as string;\n }\n\n async aiAsk(\n prompt: TUserPrompt,\n opt: InsightExtractOption = defaultInsightExtractOption,\n ): Promise<string> {\n return this.aiString(prompt, opt);\n }\n\n async describeElementAtPoint(\n center: [number, number],\n opt?: {\n verifyPrompt?: boolean;\n retryLimit?: number;\n deepThink?: boolean;\n } & LocatorValidatorOption,\n ): Promise<AgentDescribeElementAtPointResult> {\n const { verifyPrompt = true, retryLimit = 3 } = opt || {};\n\n let success = false;\n let retryCount = 0;\n let resultPrompt = '';\n let deepThink = opt?.deepThink || false;\n let verifyResult: LocateValidatorResult | undefined;\n\n while (!success && retryCount < retryLimit) {\n if (retryCount >= 2) {\n deepThink = true;\n }\n debug(\n 'aiDescribe',\n center,\n 'verifyPrompt',\n verifyPrompt,\n 'retryCount',\n retryCount,\n 'deepThink',\n deepThink,\n );\n // use same intent as aiLocate\n const modelConfig = this.modelConfigManager.getModelConfig('grounding');\n\n const text = await this.insight.describe(center, modelConfig, {\n deepThink,\n });\n debug('aiDescribe text', text);\n assert(text.description, `failed to describe element at [${center}]`);\n resultPrompt = text.description;\n\n verifyResult = await this.verifyLocator(\n resultPrompt,\n deepThink ? { deepThink: true } : undefined,\n center,\n opt,\n );\n if (verifyResult.pass) {\n success = true;\n } else {\n retryCount++;\n }\n }\n\n return {\n prompt: resultPrompt,\n deepThink,\n verifyResult,\n };\n }\n\n async verifyLocator(\n prompt: string,\n locateOpt: LocateOption | undefined,\n expectCenter: [number, number],\n verifyLocateOption?: LocatorValidatorOption,\n ): Promise<LocateValidatorResult> {\n debug('verifyLocator', prompt, locateOpt, expectCenter, verifyLocateOption);\n\n const { center: verifyCenter, rect: verifyRect } = await this.aiLocate(\n prompt,\n locateOpt,\n );\n const distance = distanceOfTwoPoints(expectCenter, verifyCenter);\n const included = includedInRect(expectCenter, verifyRect);\n const pass =\n distance <= (verifyLocateOption?.centerDistanceThreshold || 20) ||\n included;\n const verifyResult = {\n pass,\n rect: verifyRect,\n center: verifyCenter,\n centerDistance: distance,\n };\n debug('aiDescribe verifyResult', verifyResult);\n return verifyResult;\n }\n\n async aiLocate(prompt: TUserPrompt, opt?: LocateOption) {\n const locateParam = buildDetailedLocateParam(prompt, opt);\n assert(locateParam, 'cannot get locate param for aiLocate');\n const locatePlan = locatePlanForLocate(locateParam);\n const plans = [locatePlan];\n const modelConfig = this.modelConfigManager.getModelConfig('grounding');\n\n const { executor, output } = await this.taskExecutor.runPlans(\n taskTitleStr('Locate', locateParamStr(locateParam)),\n plans,\n modelConfig,\n );\n await this.afterTaskRunning(executor);\n\n const { element } = output;\n\n return {\n rect: element?.rect,\n center: element?.center,\n scale: (await this.interface.size()).dpr,\n } as Pick<LocateResultElement, 'rect' | 'center'> & {\n scale: number;\n };\n }\n\n async aiAssert(\n assertion: TUserPrompt,\n msg?: string,\n opt?: AgentAssertOpt & InsightExtractOption,\n ) {\n const modelConfig = this.modelConfigManager.getModelConfig('VQA');\n\n const insightOpt: InsightExtractOption = {\n domIncluded: opt?.domIncluded ?? defaultInsightExtractOption.domIncluded,\n screenshotIncluded:\n opt?.screenshotIncluded ??\n defaultInsightExtractOption.screenshotIncluded,\n isWaitForAssert: opt?.isWaitForAssert,\n doNotThrowError: opt?.doNotThrowError,\n };\n\n const { output, executor, thought } = await this.taskExecutor.assert(\n assertion,\n modelConfig,\n insightOpt,\n );\n await this.afterTaskRunning(executor, true);\n\n const message = output\n ? undefined\n : `Assertion failed: ${msg || (typeof assertion === 'string' ? assertion : assertion.prompt)}\\nReason: ${\n thought || executor.latestErrorTask()?.error || '(no_reason)'\n }`;\n\n if (opt?.keepRawResponse) {\n return {\n pass: output,\n thought,\n message,\n };\n }\n\n if (!output) {\n throw new Error(message);\n }\n }\n\n async aiWaitFor(assertion: TUserPrompt, opt?: AgentWaitForOpt) {\n const modelConfig = this.modelConfigManager.getModelConfig('VQA');\n const { executor } = await this.taskExecutor.waitFor(\n assertion,\n {\n timeoutMs: opt?.timeoutMs || 15 * 1000,\n checkIntervalMs: opt?.checkIntervalMs || 3 * 1000,\n },\n modelConfig,\n );\n await this.afterTaskRunning(executor, true);\n\n if (executor.isInErrorState()) {\n const errorTask = executor.latestErrorTask();\n throw new Error(`${errorTask?.error}\\n${errorTask?.errorStack}`);\n }\n }\n\n async ai(taskPrompt: string, type = 'action') {\n if (type === 'action') {\n return this.aiAction(taskPrompt);\n }\n if (type === 'query') {\n return this.aiQuery(taskPrompt);\n }\n\n if (type === 'assert') {\n return this.aiAssert(taskPrompt);\n }\n\n if (type === 'tap') {\n return this.aiTap(taskPrompt);\n }\n\n if (type === 'rightClick') {\n return this.aiRightClick(taskPrompt);\n }\n\n if (type === 'doubleClick') {\n return this.aiDoubleClick(taskPrompt);\n }\n\n throw new Error(\n `Unknown type: ${type}, only support 'action', 'query', 'assert', 'tap', 'rightClick', 'doubleClick'`,\n );\n }\n\n async runYaml(yamlScriptContent: string): Promise<{\n result: Record<string, any>;\n }> {\n const script = parseYamlScript(yamlScriptContent, 'yaml');\n const player = new ScriptPlayer(script, async () => {\n return { agent: this, freeFn: [] };\n });\n await player.run();\n\n if (player.status === 'error') {\n const errors = player.taskStatusList\n .filter((task) => task.status === 'error')\n .map((task) => {\n return `task - ${task.name}: ${task.error?.message}`;\n })\n .join('\\n');\n throw new Error(`Error(s) occurred in running yaml script:\\n${errors}`);\n }\n\n return {\n result: player.result,\n };\n }\n\n async evaluateJavaScript(script: string) {\n assert(\n this.interface.evaluateJavaScript,\n 'evaluateJavaScript is not supported in current agent',\n );\n return this.interface.evaluateJavaScript(script);\n }\n\n async destroy() {\n await this.interface.destroy?.();\n this.resetDump(); // reset dump to release memory\n this.destroyed = true;\n }\n\n async logScreenshot(\n title?: string,\n opt?: {\n content: string;\n },\n ) {\n // 1. screenshot\n const base64 = await this.interface.screenshotBase64();\n const now = Date.now();\n // 2. build recorder\n const recorder: ExecutionRecorderItem[] = [\n {\n type: 'screenshot',\n ts: now,\n screenshot: base64,\n },\n ];\n // 3. build ExecutionTaskLog\n const task: ExecutionTaskLog = {\n type: 'Log',\n subType: 'Screenshot',\n status: 'finished',\n recorder,\n timing: {\n start: now,\n end: now,\n cost: 0,\n },\n param: {\n content: opt?.content || '',\n },\n executor: async () => {},\n };\n // 4. build ExecutionDump\n const executionDump: ExecutionDump = {\n sdkVersion: '',\n logTime: now,\n name: `Log - ${title || 'untitled'}`,\n description: opt?.content || '',\n tasks: [task],\n };\n // 5. append to execution dump\n this.appendExecutionDump(executionDump);\n\n try {\n this.onDumpUpdate?.(this.dumpDataString());\n } catch (error) {\n console.error('Failed to update dump', error);\n }\n\n this.writeOutActionDumps();\n }\n\n _unstableLogContent() {\n const { groupName, groupDescription, executions } = this.dump;\n const newExecutions = Array.isArray(executions)\n ? executions.map((execution: any) => {\n const { tasks, ...restExecution } = execution;\n let newTasks = tasks;\n if (Array.isArray(tasks)) {\n newTasks = tasks.map((task: any) => {\n // only remove uiContext and log from task\n const { uiContext, log, ...restTask } = task;\n return restTask;\n });\n }\n return { ...restExecution, ...(newTasks ? { tasks: newTasks } : {}) };\n })\n : [];\n return {\n groupName,\n groupDescription,\n executions: newExecutions,\n };\n }\n\n /**\n * Freezes the current page context to be reused in subsequent AI operations\n * This avoids recalculating page context for each operation\n */\n async freezePageContext(): Promise<void> {\n debug('Freezing page context');\n const context = await this._snapshotContext();\n // Mark the context as frozen\n context._isFrozen = true;\n this.frozenUIContext = context;\n debug('Page context frozen successfully');\n }\n\n /**\n * Unfreezes the page context, allowing AI operations to calculate context dynamically\n */\n async unfreezePageContext(): Promise<void> {\n debug('Unfreezing page context');\n this.frozenUIContext = undefined;\n debug('Page context unfrozen successfully');\n }\n\n /**\n * Process cache configuration and return normalized cache settings\n */\n private processCacheConfig(\n opts: AgentOpt,\n ): { id: string; enabled: boolean; readOnly: boolean } | null {\n // 1. New cache object configuration (highest priority)\n if (opts.cache !== undefined) {\n if (opts.cache === false) {\n return null; // Completely disable cache\n }\n\n if (opts.cache === true) {\n throw new Error(\n 'cache: true requires an explicit cache ID. Please provide:\\n' +\n 'Example: cache: { id: \"my-cache-id\" }',\n );\n }\n\n // cache is object configuration\n if (typeof opts.cache === 'object') {\n const config = opts.cache;\n if (!config.id) {\n throw new Error(\n 'cache configuration requires an explicit id. Please provide:\\n' +\n 'Example: cache: { id: \"my-cache-id\" }',\n );\n }\n const id = config.id;\n // Default strategy is 'read-write'\n const strategy = config.strategy ?? 'read-write';\n const isReadOnly = strategy === 'read-only';\n\n return {\n id,\n enabled: true,\n readOnly: isReadOnly,\n };\n }\n }\n\n // 2. Backward compatibility: support old cacheId (requires environment variable)\n if (opts.cacheId) {\n const envEnabled =\n globalConfigManager.getEnvConfigInBoolean(MIDSCENE_CACHE);\n if (envEnabled) {\n return {\n id: opts.cacheId,\n enabled: true,\n readOnly: false,\n };\n }\n }\n\n // 3. No cache configuration\n return null;\n }\n\n /**\n * Manually flush cache to file\n * Only meaningful in read-only mode, other modes will throw error\n */\n async flushCache(): Promise<void> {\n if (!this.taskCache) {\n throw new Error('Cache is not configured');\n }\n\n this.taskCache.flushCacheToFile();\n }\n}\n\nexport const createAgent = (\n interfaceInstance: AbstractInterface,\n opts?: AgentOpt,\n) => {\n return new Agent(interfaceInstance, opts);\n};\n"],"names":["debug","getDebug","distanceOfTwoPoints","p1","p2","x1","y1","x2","y2","Math","includedInRect","point","rect","x","y","left","top","width","height","defaultInsightExtractOption","Agent","action","commonContextParser","prompt","console","execution","trimmedExecution","trimContextByViewport","currentDump","stringifyDumpData","reportHTMLContent","Error","generateReport","autoPrintReportMsg","writeLogFile","groupedActionDumpFileExt","printReportMsg","task","param","paramStr","tip","typeStr","executor","doNotThrowError","error","errorTask","type","opt","actionPlan","plans","Boolean","title","taskTitleStr","locateParamStr","modelConfig","output","locatePrompt","assert","detailedLocateParam","buildDetailedLocateParam","locatePromptOrValue","locatePromptOrOpt","optOrUndefined","value","optWithValue","locatePromptOrKeyName","keyName","undefined","locatePromptOrScrollParam","scrollParam","taskPrompt","_this_taskCache","_this_taskCache1","cacheable","isVlmUiTars","matchedCache","_matchedCache_cacheContent","_matchedCache_cacheContent1","yaml","yamlContent","yamlFlowStr","demand","textPrompt","multimodalPrompt","parsePrompt","center","verifyPrompt","retryLimit","success","retryCount","resultPrompt","deepThink","verifyResult","text","locateOpt","expectCenter","verifyLocateOption","verifyCenter","verifyRect","distance","included","pass","locateParam","locatePlan","locatePlanForLocate","element","assertion","msg","_executor_latestErrorTask","insightOpt","thought","message","yamlScriptContent","script","parseYamlScript","player","ScriptPlayer","errors","_task_error","_this_interface","base64","now","Date","recorder","executionDump","_this","groupName","groupDescription","executions","newExecutions","Array","tasks","restExecution","newTasks","uiContext","log","restTask","context","opts","config","id","strategy","isReadOnly","envEnabled","globalConfigManager","MIDSCENE_CACHE","interfaceInstance","Object","ModelConfigManager","globalModelConfigManager","Insight","cacheConfig","TaskCache","TaskExecutor","getReportFileName","createAgent"],"mappings":";;;;;;;;;;;;;;;;;;;;;AAgEA,MAAMA,QAAQC,SAAS;AAEvB,MAAMC,sBAAsB,CAACC,IAAsBC;IACjD,MAAM,CAACC,IAAIC,GAAG,GAAGH;IACjB,MAAM,CAACI,IAAIC,GAAG,GAAGJ;IACjB,OAAOK,KAAK,KAAK,CAACA,KAAK,IAAI,CAAEJ,AAAAA,CAAAA,KAAKE,EAAC,KAAM,IAAKD,AAAAA,CAAAA,KAAKE,EAAC,KAAM;AAC5D;AAEA,MAAME,iBAAiB,CAACC,OAAyBC;IAC/C,MAAM,CAACC,GAAGC,EAAE,GAAGH;IACf,MAAM,EAAEI,IAAI,EAAEC,GAAG,EAAEC,KAAK,EAAEC,MAAM,EAAE,GAAGN;IACrC,OAAOC,KAAKE,QAAQF,KAAKE,OAAOE,SAASH,KAAKE,OAAOF,KAAKE,MAAME;AAClE;AAEA,MAAMC,8BAAoD;IACxD,aAAa;IACb,oBAAoB;AACtB;AAEO,MAAMC;IA2CX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IAKQ,uBAAuB;QAC7B,IACE,CAAC,IAAI,CAAC,mBAAmB,IACzB,AAAiC,gBAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,IAC5B,AAAiC,iBAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,IAC5B,AAAiC,aAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,IAC5B,AAAiC,6BAAjC,IAAI,CAAC,SAAS,CAAC,aAAa,EAC5B;YACA,IAAI,CAAC,kBAAkB,CAAC,sBAAsB;YAC9C,IAAI,CAAC,mBAAmB,GAAG;QAC7B;IACF;IAmDA,MAAM,iBAA0C;QAC9C,OAAO,IAAI,CAAC,SAAS,CAAC,WAAW;IACnC;IAEA,MAAM,aAAaC,MAAsB,EAAsB;QAE7D,IAAI,CAAC,oBAAoB;QAGzB,IAAI,IAAI,CAAC,eAAe,EAAE;YACxBrB,MAAM,yCAAyCqB;YAC/C,OAAO,IAAI,CAAC,eAAe;QAC7B;QAEA,IAAI,IAAI,CAAC,SAAS,CAAC,UAAU,EAAE;YAC7BrB,MAAM,qCAAqCqB;YAC3C,OAAO,MAAM,IAAI,CAAC,SAAS,CAAC,UAAU;QACxC;QACErB,MAAM,yCAAyCqB;QAC/C,OAAO,MAAMC,oBAAoB,IAAI,CAAC,SAAS,EAAE;YAC/C,iBAAiB,IAAI,CAAC,kBAAkB,CAAC,sBAAsB;QACjE;IAEJ;IAEA,MAAM,mBAAuC;QAC3C,OAAO,MAAM,IAAI,CAAC,YAAY,CAAC;IACjC;IAEA,MAAM,mBAAmBC,MAAc,EAAE;QACvC,IAAI,IAAI,CAAC,IAAI,CAAC,eAAe,EAC3BC,QAAQ,IAAI,CACV;QAGJ,IAAI,CAAC,IAAI,CAAC,eAAe,GAAGD;IAC9B;IAEA,YAAY;QACV,IAAI,CAAC,IAAI,GAAG;YACV,WAAW,IAAI,CAAC,IAAI,CAAC,SAAS;YAC9B,kBAAkB,IAAI,CAAC,IAAI,CAAC,gBAAgB;YAC5C,YAAY,EAAE;YACd,aAAa,EAAE;QACjB;QAEA,OAAO,IAAI,CAAC,IAAI;IAClB;IAEA,oBAAoBE,SAAwB,EAAE;QAE5C,MAAMC,mBAAmBC,sBAAsBF;QAC/C,MAAMG,cAAc,IAAI,CAAC,IAAI;QAC7BA,YAAY,UAAU,CAAC,IAAI,CAACF;IAC9B;IAEA,iBAAiB;QAEf,IAAI,CAAC,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,IAAI,CAAC,SAAS;QACzC,IAAI,CAAC,IAAI,CAAC,gBAAgB,GAAG,IAAI,CAAC,IAAI,CAAC,gBAAgB;QACvD,OAAOG,kBAAkB,IAAI,CAAC,IAAI;IACpC;IAEA,mBAAmB;QACjB,OAAOC,kBAAkB,IAAI,CAAC,cAAc;IAC9C;IAEA,sBAAsB;QACpB,IAAI,IAAI,CAAC,SAAS,EAChB,MAAM,IAAIC,MACR;QAGJ,MAAM,EAAEC,cAAc,EAAEC,kBAAkB,EAAE,GAAG,IAAI,CAAC,IAAI;QACxD,IAAI,CAAC,UAAU,GAAGC,aAAa;YAC7B,UAAU,IAAI,CAAC,cAAc;YAC7B,SAASC;YACT,aAAa,IAAI,CAAC,cAAc;YAChC,MAAM;YACNH;QACF;QACAhC,MAAM,uBAAuB,IAAI,CAAC,UAAU;QAC5C,IAAIgC,kBAAkBC,sBAAsB,IAAI,CAAC,UAAU,EACzDG,eAAe,IAAI,CAAC,UAAU;IAElC;IAEA,MAAc,uBAAuBC,IAAmB,EAAE;QACxD,MAAMC,QAAQC,SAASF;QACvB,MAAMG,MAAMF,QAAQ,GAAGG,QAAQJ,MAAM,GAAG,EAAEC,OAAO,GAAGG,QAAQJ;QAE5D,IAAI,IAAI,CAAC,cAAc,EACrB,MAAM,IAAI,CAAC,cAAc,CAACG;IAE9B;IAEA,MAAc,iBAAiBE,QAAkB,EAAEC,kBAAkB,KAAK,EAAE;QAC1E,IAAI,CAAC,mBAAmB,CAACD,SAAS,IAAI;QAEtC,IAAI;YACF,IAAI,IAAI,CAAC,YAAY,EACnB,IAAI,CAAC,YAAY,CAAC,IAAI,CAAC,cAAc;QAEzC,EAAE,OAAOE,OAAO;YACdpB,QAAQ,KAAK,CAAC,yBAAyBoB;QACzC;QAEA,IAAI,CAAC,mBAAmB;QAExB,IAAIF,SAAS,cAAc,MAAM,CAACC,iBAAiB;YACjD,MAAME,YAAYH,SAAS,eAAe;YAC1C,MAAM,IAAIX,MAAM,GAAGc,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,YAAY,CAAC,EAAE,EAAEA,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,UAAU,EAAE,EAAE;gBACtE,OAAOA,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,KAAK;YACzB;QACF;IACF;IAEA,MAAM,wBACJC,IAAY,EACZC,GAAO,EACP;QACA/C,MAAM,2BAA2B8C,MAAM,KAAKC;QAE5C,MAAMC,aAAgC;YACpC,MAAMF;YACN,OAAQC,OAAe,CAAC;YACxB,SAAS;QACX;QACA/C,MAAM,cAAcgD;QAEpB,MAAMC,QAA0B;YAACD;SAAW,CAAC,MAAM,CACjDE;QAGF,MAAMC,QAAQC,aACZN,MACAO,eAAe,AAACN,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAa,MAAM,AAAD,KAAK,CAAC;QAI1C,MAAMO,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAEC,MAAM,EAAEb,QAAQ,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,CAC3DS,OACAF,OACAK;QAEF,MAAM,IAAI,CAAC,gBAAgB,CAACZ;QAC5B,OAAOa;IACT;IAEA,MAAM,MAAMC,YAAyB,EAAET,GAAkB,EAAE;QACzDU,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,OAAO;YACzC,QAAQW;QACV;IACF;IAEA,MAAM,aAAaF,YAAyB,EAAET,GAAkB,EAAE;QAChEU,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,cAAc;YAChD,QAAQW;QACV;IACF;IAEA,MAAM,cAAcF,YAAyB,EAAET,GAAkB,EAAE;QACjEU,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,eAAe;YACjD,QAAQW;QACV;IACF;IAEA,MAAM,QAAQF,YAAyB,EAAET,GAAkB,EAAE;QAC3DU,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,SAAS;YAC3C,QAAQW;QACV;IACF;IAmBA,MAAM,QACJE,mBAAyC,EACzCC,iBAGa,EACbC,cAA6B,EAC7B;QACA,IAAIC;QACJ,IAAIP;QACJ,IAAIT;QAKJ,IACE,AAA6B,YAA7B,OAAOc,qBACPA,AAAsB,SAAtBA,qBACA,WAAWA,mBACX;YAEAL,eAAeI;YACf,MAAMI,eAAeH;YAKrBE,QAAQC,aAAa,KAAK;YAC1BjB,MAAMiB;QACR,OAAO;YAELD,QAAQH;YACRJ,eAAeK;YACfd,MAAM;gBACJ,GAAGe,cAAc;gBACjBC;YACF;QACF;QAEAN,OACE,AAAiB,YAAjB,OAAOM,OACP;QAEFN,OAAOD,cAAc;QAErB,MAAME,sBAAsBC,yBAAyBH,cAAcT;QAEnE,OAAO,IAAI,CAAC,uBAAuB,CAAC,SAAS;YAC3C,GAAIA,OAAO,CAAC,CAAC;YACb,QAAQW;QACV;IACF;IAmBA,MAAM,gBACJO,qBAA2C,EAC3CJ,iBAGa,EACbC,cAA6B,EAC7B;QACA,IAAII;QACJ,IAAIV;QACJ,IAAIT;QAGJ,IACE,AAA6B,YAA7B,OAAOc,qBACPA,AAAsB,SAAtBA,qBACA,aAAaA,mBACb;YAEAL,eAAeS;YACflB,MAAMc;QAGR,OAAO;YAELK,UAAUD;YACVT,eAAeK;YACfd,MAAM;gBACJ,GAAIe,kBAAkB,CAAC,CAAC;gBACxBI;YACF;QACF;QAEAT,OAAOV,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,EAAE;QAErB,MAAMW,sBAAsBF,eACxBG,yBAAyBH,cAAcT,OACvCoB;QAEJ,OAAO,IAAI,CAAC,uBAAuB,CAAC,iBAAiB;YACnD,GAAIpB,OAAO,CAAC,CAAC;YACb,QAAQW;QACV;IACF;IAmBA,MAAM,SACJU,yBAAgE,EAChEP,iBAAyE,EACzEC,cAA6B,EAC7B;QACA,IAAIO;QACJ,IAAIb;QACJ,IAAIT;QAGJ,IACE,AAA6B,YAA7B,OAAOc,qBACN,gBAAeA,qBACd,gBAAgBA,qBAChB,cAAcA,iBAAgB,GAChC;YAEAL,eAAeY;YACfrB,MAAMc;QACR,OAAO;YAELQ,cAAcD;YACdZ,eAAeK;YACfd,MAAM;gBACJ,GAAIe,kBAAkB,CAAC,CAAC;gBACxB,GAAIO,eAAe,CAAC,CAAC;YACvB;QACF;QAEA,MAAMX,sBAAsBC,yBAC1BH,gBAAgB,IAChBT;QAGF,OAAO,IAAI,CAAC,uBAAuB,CAAC,UAAU;YAC5C,GAAIA,OAAO,CAAC,CAAC;YACb,QAAQW;QACV;IACF;IAEA,MAAM,SACJY,UAAkB,EAClBvB,GAEC,EACD;YASMwB,iBACcC;QATpB,MAAMlB,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAMmB,YAAY1B,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS;QAEhC,MAAM2B,cAAcpB,AAAuB,kBAAvBA,YAAY,MAAM;QACtC,MAAMqB,eACJD,eAAeD,AAAc,UAAdA,YACXN,SAAAA,QACAI,CAAAA,kBAAAA,IAAI,CAAC,SAAS,AAAD,IAAbA,KAAAA,IAAAA,gBAAgB,cAAc,CAACD;QACrC,IAAIK,gBAAAA,SAAgBH,CAAAA,mBAAAA,IAAI,CAAC,SAAS,AAAD,IAAbA,KAAAA,IAAAA,iBAAgB,iBAAiB,AAAD,GAAG;gBAInDI,4BAMWC;YARb,MAAM,EAAEnC,QAAQ,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,sBAAsB,CACjE4B,YAAAA,QACAM,CAAAA,6BAAAA,aAAa,YAAY,AAAD,IAAxBA,KAAAA,IAAAA,2BAA2B,YAAY;YAGzC,MAAM,IAAI,CAAC,gBAAgB,CAAClC;YAE5B1C,MAAM;YACN,MAAM8E,OAAO,QAAAD,CAAAA,8BAAAA,aAAa,YAAY,AAAD,IAAxBA,KAAAA,IAAAA,4BAA2B,YAAY;YACpD,OAAO,IAAI,CAAC,OAAO,CAACC;QACtB;QAEA,MAAM,EAAEvB,MAAM,EAAEb,QAAQ,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,MAAM,CACzD4B,YACAhB,aACA,IAAI,CAAC,IAAI,CAAC,eAAe;QAI3B,IAAI,IAAI,CAAC,SAAS,IAAIC,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,QAAQ,AAAD,KAAKkB,AAAc,UAAdA,WAAqB;YAC7D,MAAMM,cAAkC;gBACtC,OAAO;oBACL;wBACE,MAAMT;wBACN,MAAMf,OAAO,QAAQ;oBACvB;iBACD;YACH;YACA,MAAMyB,cAAcF,QAAAA,IAAS,CAACC;YAC9B,IAAI,CAAC,SAAS,CAAC,yBAAyB,CACtC;gBACE,MAAM;gBACN,QAAQT;gBACR,cAAcU;YAChB,GACAL;QAEJ;QAEA,MAAM,IAAI,CAAC,gBAAgB,CAACjC;QAC5B,OAAOa;IACT;IAEA,MAAM,QACJ0B,MAA2B,EAC3BlC,MAA4B5B,2BAA2B,EAClC;QACrB,MAAMmC,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAC3D,MAAM,EAAEC,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,SACAuC,QACA3B,aACAP;QAEJ,MAAM,IAAI,CAAC,gBAAgB,CAACL;QAC5B,OAAOa;IACT;IAEA,MAAM,UACJhC,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACrC;QAClB,MAAMmC,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAE4B,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAY7D;QACrD,MAAM,EAAEgC,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,WACAwC,YACA5B,aACAP,KACAoC;QAEJ,MAAM,IAAI,CAAC,gBAAgB,CAACzC;QAC5B,OAAOa;IACT;IAEA,MAAM,SACJhC,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACtC;QACjB,MAAMmC,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAE4B,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAY7D;QACrD,MAAM,EAAEgC,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,UACAwC,YACA5B,aACAP,KACAoC;QAEJ,MAAM,IAAI,CAAC,gBAAgB,CAACzC;QAC5B,OAAOa;IACT;IAEA,MAAM,SACJhC,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACtC;QACjB,MAAMmC,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAE4B,UAAU,EAAEC,gBAAgB,EAAE,GAAGC,YAAY7D;QACrD,MAAM,EAAEgC,MAAM,EAAEb,QAAQ,EAAE,GACxB,MAAM,IAAI,CAAC,YAAY,CAAC,wBAAwB,CAC9C,UACAwC,YACA5B,aACAP,KACAoC;QAEJ,MAAM,IAAI,CAAC,gBAAgB,CAACzC;QAC5B,OAAOa;IACT;IAEA,MAAM,MACJhC,MAAmB,EACnBwB,MAA4B5B,2BAA2B,EACtC;QACjB,OAAO,IAAI,CAAC,QAAQ,CAACI,QAAQwB;IAC/B;IAEA,MAAM,uBACJsC,MAAwB,EACxBtC,GAI0B,EACkB;QAC5C,MAAM,EAAEuC,eAAe,IAAI,EAAEC,aAAa,CAAC,EAAE,GAAGxC,OAAO,CAAC;QAExD,IAAIyC,UAAU;QACd,IAAIC,aAAa;QACjB,IAAIC,eAAe;QACnB,IAAIC,YAAY5C,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,AAAD,KAAK;QAClC,IAAI6C;QAEJ,MAAO,CAACJ,WAAWC,aAAaF,WAAY;YAC1C,IAAIE,cAAc,GAChBE,YAAY;YAEd3F,MACE,cACAqF,QACA,gBACAC,cACA,cACAG,YACA,aACAE;YAGF,MAAMrC,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;YAE3D,MAAMuC,OAAO,MAAM,IAAI,CAAC,OAAO,CAAC,QAAQ,CAACR,QAAQ/B,aAAa;gBAC5DqC;YACF;YACA3F,MAAM,mBAAmB6F;YACzBpC,OAAOoC,KAAK,WAAW,EAAE,CAAC,+BAA+B,EAAER,OAAO,CAAC,CAAC;YACpEK,eAAeG,KAAK,WAAW;YAE/BD,eAAe,MAAM,IAAI,CAAC,aAAa,CACrCF,cACAC,YAAY;gBAAE,WAAW;YAAK,IAAIxB,QAClCkB,QACAtC;YAEF,IAAI6C,aAAa,IAAI,EACnBJ,UAAU;iBAEVC;QAEJ;QAEA,OAAO;YACL,QAAQC;YACRC;YACAC;QACF;IACF;IAEA,MAAM,cACJrE,MAAc,EACduE,SAAmC,EACnCC,YAA8B,EAC9BC,kBAA2C,EACX;QAChChG,MAAM,iBAAiBuB,QAAQuE,WAAWC,cAAcC;QAExD,MAAM,EAAE,QAAQC,YAAY,EAAE,MAAMC,UAAU,EAAE,GAAG,MAAM,IAAI,CAAC,QAAQ,CACpE3E,QACAuE;QAEF,MAAMK,WAAWjG,oBAAoB6F,cAAcE;QACnD,MAAMG,WAAW1F,eAAeqF,cAAcG;QAC9C,MAAMG,OACJF,YAAaH,CAAAA,CAAAA,QAAAA,qBAAAA,KAAAA,IAAAA,mBAAoB,uBAAuB,AAAD,KAAK,EAAC,KAC7DI;QACF,MAAMR,eAAe;YACnBS;YACA,MAAMH;YACN,QAAQD;YACR,gBAAgBE;QAClB;QACAnG,MAAM,2BAA2B4F;QACjC,OAAOA;IACT;IAEA,MAAM,SAASrE,MAAmB,EAAEwB,GAAkB,EAAE;QACtD,MAAMuD,cAAc3C,yBAAyBpC,QAAQwB;QACrDU,OAAO6C,aAAa;QACpB,MAAMC,aAAaC,oBAAoBF;QACvC,MAAMrD,QAAQ;YAACsD;SAAW;QAC1B,MAAMjD,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAM,EAAEZ,QAAQ,EAAEa,MAAM,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,QAAQ,CAC3DH,aAAa,UAAUC,eAAeiD,eACtCrD,OACAK;QAEF,MAAM,IAAI,CAAC,gBAAgB,CAACZ;QAE5B,MAAM,EAAE+D,OAAO,EAAE,GAAGlD;QAEpB,OAAO;YACL,MAAMkD,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,IAAI;YACnB,QAAQA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM;YACvB,OAAQ,OAAM,IAAI,CAAC,SAAS,CAAC,IAAI,EAAC,EAAG,GAAG;QAC1C;IAGF;IAEA,MAAM,SACJC,SAAsB,EACtBC,GAAY,EACZ5D,GAA2C,EAC3C;YAsBiB6D;QArBjB,MAAMtD,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAE3D,MAAMuD,aAAmC;YACvC,aAAa9D,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,KAAK5B,4BAA4B,WAAW;YACxE,oBACE4B,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,kBAAkB,AAAD,KACtB5B,4BAA4B,kBAAkB;YAChD,iBAAiB4B,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe;YACrC,iBAAiBA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe;QACvC;QAEA,MAAM,EAAEQ,MAAM,EAAEb,QAAQ,EAAEoE,OAAO,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,MAAM,CAClEJ,WACApD,aACAuD;QAEF,MAAM,IAAI,CAAC,gBAAgB,CAACnE,UAAU;QAEtC,MAAMqE,UAAUxD,SACZY,SACA,CAAC,kBAAkB,EAAEwC,OAAQ,CAAqB,YAArB,OAAOD,YAAyBA,YAAYA,UAAU,MAAK,EAAG,UAAU,EACnGI,WAAAA,SAAWF,CAAAA,4BAAAA,SAAS,eAAe,EAAC,IAAzBA,KAAAA,IAAAA,0BAA4B,KAAK,AAAD,KAAK,eAChD;QAEN,IAAI7D,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,EACtB,OAAO;YACL,MAAMQ;YACNuD;YACAC;QACF;QAGF,IAAI,CAACxD,QACH,MAAM,IAAIxB,MAAMgF;IAEpB;IAEA,MAAM,UAAUL,SAAsB,EAAE3D,GAAqB,EAAE;QAC7D,MAAMO,cAAc,IAAI,CAAC,kBAAkB,CAAC,cAAc,CAAC;QAC3D,MAAM,EAAEZ,QAAQ,EAAE,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,OAAO,CAClDgE,WACA;YACE,WAAW3D,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,AAAD,KAAK;YAC7B,iBAAiBA,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,AAAD,KAAK;QAC3C,GACAO;QAEF,MAAM,IAAI,CAAC,gBAAgB,CAACZ,UAAU;QAEtC,IAAIA,SAAS,cAAc,IAAI;YAC7B,MAAMG,YAAYH,SAAS,eAAe;YAC1C,MAAM,IAAIX,MAAM,GAAGc,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,KAAK,CAAC,EAAE,EAAEA,QAAAA,YAAAA,KAAAA,IAAAA,UAAW,UAAU,EAAE;QACjE;IACF;IAEA,MAAM,GAAGyB,UAAkB,EAAExB,OAAO,QAAQ,EAAE;QAC5C,IAAIA,AAAS,aAATA,MACF,OAAO,IAAI,CAAC,QAAQ,CAACwB;QAEvB,IAAIxB,AAAS,YAATA,MACF,OAAO,IAAI,CAAC,OAAO,CAACwB;QAGtB,IAAIxB,AAAS,aAATA,MACF,OAAO,IAAI,CAAC,QAAQ,CAACwB;QAGvB,IAAIxB,AAAS,UAATA,MACF,OAAO,IAAI,CAAC,KAAK,CAACwB;QAGpB,IAAIxB,AAAS,iBAATA,MACF,OAAO,IAAI,CAAC,YAAY,CAACwB;QAG3B,IAAIxB,AAAS,kBAATA,MACF,OAAO,IAAI,CAAC,aAAa,CAACwB;QAG5B,MAAM,IAAIvC,MACR,CAAC,cAAc,EAAEe,KAAK,8EAA8E,CAAC;IAEzG;IAEA,MAAM,QAAQkE,iBAAyB,EAEpC;QACD,MAAMC,SAASC,gBAAgBF,mBAAmB;QAClD,MAAMG,SAAS,IAAIC,aAAaH,QAAQ,UAC/B;gBAAE,OAAO,IAAI;gBAAE,QAAQ,EAAE;YAAC;QAEnC,MAAME,OAAO,GAAG;QAEhB,IAAIA,AAAkB,YAAlBA,OAAO,MAAM,EAAc;YAC7B,MAAME,SAASF,OAAO,cAAc,CACjC,MAAM,CAAC,CAAC9E,OAASA,AAAgB,YAAhBA,KAAK,MAAM,EAC5B,GAAG,CAAC,CAACA;oBAC2BiF;gBAA/B,OAAO,CAAC,OAAO,EAAEjF,KAAK,IAAI,CAAC,EAAE,EAAE,QAAAiF,CAAAA,cAAAA,KAAK,KAAK,AAAD,IAATA,KAAAA,IAAAA,YAAY,OAAO,EAAE;YACtD,GACC,IAAI,CAAC;YACR,MAAM,IAAIvF,MAAM,CAAC,2CAA2C,EAAEsF,QAAQ;QACxE;QAEA,OAAO;YACL,QAAQF,OAAO,MAAM;QACvB;IACF;IAEA,MAAM,mBAAmBF,MAAc,EAAE;QACvCxD,OACE,IAAI,CAAC,SAAS,CAAC,kBAAkB,EACjC;QAEF,OAAO,IAAI,CAAC,SAAS,CAAC,kBAAkB,CAACwD;IAC3C;IAEA,MAAM,UAAU;YACRM,yBAAAA;QAAN,eAAMA,CAAAA,0BAAAA,AAAAA,CAAAA,kBAAAA,IAAI,CAAC,SAAS,AAAD,EAAE,OAAO,AAAD,IAArBA,KAAAA,IAAAA,wBAAAA,IAAAA,CAAAA,gBAAAA;QACN,IAAI,CAAC,SAAS;QACd,IAAI,CAAC,SAAS,GAAG;IACnB;IAEA,MAAM,cACJpE,KAAc,EACdJ,GAEC,EACD;QAEA,MAAMyE,SAAS,MAAM,IAAI,CAAC,SAAS,CAAC,gBAAgB;QACpD,MAAMC,MAAMC,KAAK,GAAG;QAEpB,MAAMC,WAAoC;YACxC;gBACE,MAAM;gBACN,IAAIF;gBACJ,YAAYD;YACd;SACD;QAED,MAAMnF,OAAyB;YAC7B,MAAM;YACN,SAAS;YACT,QAAQ;YACRsF;YACA,QAAQ;gBACN,OAAOF;gBACP,KAAKA;gBACL,MAAM;YACR;YACA,OAAO;gBACL,SAAS1E,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAK;YAC3B;YACA,UAAU,WAAa;QACzB;QAEA,MAAM6E,gBAA+B;YACnC,YAAY;YACZ,SAASH;YACT,MAAM,CAAC,MAAM,EAAEtE,SAAS,YAAY;YACpC,aAAaJ,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAK;YAC7B,OAAO;gBAACV;aAAK;QACf;QAEA,IAAI,CAAC,mBAAmB,CAACuF;QAEzB,IAAI;gBACFC,oBAAAA;oBAAAA,CAAAA,qBAAAA,AAAAA,CAAAA,QAAAA,IAAI,AAAD,EAAE,YAAY,AAAD,KAAhBA,mBAAAA,IAAAA,CAAAA,OAAoB,IAAI,CAAC,cAAc;QACzC,EAAE,OAAOjF,OAAO;YACdpB,QAAQ,KAAK,CAAC,yBAAyBoB;QACzC;QAEA,IAAI,CAAC,mBAAmB;IAC1B;IAEA,sBAAsB;QACpB,MAAM,EAAEkF,SAAS,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAG,IAAI,CAAC,IAAI;QAC7D,MAAMC,gBAAgBC,MAAM,OAAO,CAACF,cAChCA,WAAW,GAAG,CAAC,CAACvG;YACd,MAAM,EAAE0G,KAAK,EAAE,GAAGC,eAAe,GAAG3G;YACpC,IAAI4G,WAAWF;YACf,IAAID,MAAM,OAAO,CAACC,QAChBE,WAAWF,MAAM,GAAG,CAAC,CAAC9F;gBAEpB,MAAM,EAAEiG,SAAS,EAAEC,GAAG,EAAE,GAAGC,UAAU,GAAGnG;gBACxC,OAAOmG;YACT;YAEF,OAAO;gBAAE,GAAGJ,aAAa;gBAAE,GAAIC,WAAW;oBAAE,OAAOA;gBAAS,IAAI,CAAC,CAAC;YAAE;QACtE,KACA,EAAE;QACN,OAAO;YACLP;YACAC;YACA,YAAYE;QACd;IACF;IAMA,MAAM,oBAAmC;QACvCjI,MAAM;QACN,MAAMyI,UAAU,MAAM,IAAI,CAAC,gBAAgB;QAE3CA,QAAQ,SAAS,GAAG;QACpB,IAAI,CAAC,eAAe,GAAGA;QACvBzI,MAAM;IACR;IAKA,MAAM,sBAAqC;QACzCA,MAAM;QACN,IAAI,CAAC,eAAe,GAAGmE;QACvBnE,MAAM;IACR;IAKQ,mBACN0I,IAAc,EAC8C;QAE5D,IAAIA,AAAevE,WAAfuE,KAAK,KAAK,EAAgB;YAC5B,IAAIA,AAAe,UAAfA,KAAK,KAAK,EACZ,OAAO;YAGT,IAAIA,AAAe,SAAfA,KAAK,KAAK,EACZ,MAAM,IAAI3G,MACR;YAMJ,IAAI,AAAsB,YAAtB,OAAO2G,KAAK,KAAK,EAAe;gBAClC,MAAMC,SAASD,KAAK,KAAK;gBACzB,IAAI,CAACC,OAAO,EAAE,EACZ,MAAM,IAAI5G,MACR;gBAIJ,MAAM6G,KAAKD,OAAO,EAAE;gBAEpB,MAAME,WAAWF,OAAO,QAAQ,IAAI;gBACpC,MAAMG,aAAaD,AAAa,gBAAbA;gBAEnB,OAAO;oBACLD;oBACA,SAAS;oBACT,UAAUE;gBACZ;YACF;QACF;QAGA,IAAIJ,KAAK,OAAO,EAAE;YAChB,MAAMK,aACJC,oBAAoB,qBAAqB,CAACC;YAC5C,IAAIF,YACF,OAAO;gBACL,IAAIL,KAAK,OAAO;gBAChB,SAAS;gBACT,UAAU;YACZ;QAEJ;QAGA,OAAO;IACT;IAMA,MAAM,aAA4B;QAChC,IAAI,CAAC,IAAI,CAAC,SAAS,EACjB,MAAM,IAAI3G,MAAM;QAGlB,IAAI,CAAC,SAAS,CAAC,gBAAgB;IACjC;IA98BA,YAAYmH,iBAAgC,EAAER,IAAe,CAAE;QA5D/D;QAEA;QAEA;QAEA;QAEA;QAEA;QAEA;QAKA,kCAAU;QAEV;QAEA;QAEA;QAEA,oCAAY;QAEZ;QAKA,uBAAQ,mBAAR;QAKA,uBAAQ,uBAAsB;QAwB5B,IAAI,CAAC,SAAS,GAAGQ;QACjB,IAAI,CAAC,IAAI,GAAGC,OAAO,MAAM,CACvB;YACE,gBAAgB;YAChB,oBAAoB;YACpB,WAAW;YACX,kBAAkB;QACpB,GACAT,QAAQ,CAAC;QAGX,IAAIA,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW,AAAD,KAAK,AAA6B,cAA7B,OAAOA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW,AAAD,GAC9C,MAAM,IAAI3G,MACR,CAAC,+DAA+D,EAAE,OAAO2G,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW,AAAD,GAAG;QAGhG,IAAI,CAAC,kBAAkB,GAAGA,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,WAAW,AAAD,IACtC,IAAIU,mBAAmBV,KAAK,WAAW,IACvCW;QAEJ,IAAI,CAAC,cAAc,GAAG,IAAI,CAAC,IAAI,CAAC,cAAc;QAE9C,IAAI,CAAC,OAAO,GAAG,IAAIC,QAAQ,OAAOjI,SACzB,IAAI,CAAC,YAAY,CAACA;QAI3B,MAAMkI,cAAc,IAAI,CAAC,kBAAkB,CAACb,QAAQ,CAAC;QACrD,IAAIa,aACF,IAAI,CAAC,SAAS,GAAG,IAAIC,UACnBD,YAAY,EAAE,EACdA,YAAY,OAAO,EACnBpF,QACAoF,YAAY,QAAQ;QAIxB,IAAI,CAAC,YAAY,GAAG,IAAIE,aAAa,IAAI,CAAC,SAAS,EAAE,IAAI,CAAC,OAAO,EAAE;YACjE,WAAW,IAAI,CAAC,SAAS;YACzB,aAAa,IAAI,CAAC,sBAAsB,CAAC,IAAI,CAAC,IAAI;YAClD,sBAAsB,IAAI,CAAC,IAAI,CAAC,oBAAoB;QACtD;QACA,IAAI,CAAC,IAAI,GAAG,IAAI,CAAC,SAAS;QAC1B,IAAI,CAAC,cAAc,GACjBf,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,cAAc,AAAD,KACnBgB,kBAAkBhB,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,MAAM,AAAD,KAAK,IAAI,CAAC,SAAS,CAAC,aAAa,IAAI;IACtE;AAg6BF;AAEO,MAAMiB,cAAc,CACzBT,mBACAR,OAEO,IAAItH,MAAM8H,mBAAmBR"}
@@ -152,7 +152,7 @@ function trimContextByViewport(execution) {
152
152
  }) : execution.tasks
153
153
  };
154
154
  }
155
- const getMidsceneVersion = ()=>"0.29.7-beta-20250930035234.0";
155
+ const getMidsceneVersion = ()=>"0.29.7-beta-20250930083551.0";
156
156
  const parsePrompt = (prompt)=>{
157
157
  if ('string' == typeof prompt) return {
158
158
  textPrompt: prompt,
@@ -64,7 +64,20 @@ async function createChatClient({ AIActionTypeValue, modelConfig }) {
64
64
  ...(null == openaiExtraConfig ? void 0 : openaiExtraConfig.defaultHeaders) || {},
65
65
  [MIDSCENE_API_TYPE]: AIActionTypeValue.toString()
66
66
  },
67
- dangerouslyAllowBrowser: true
67
+ dangerouslyAllowBrowser: true,
68
+ fetch: async (...args)=>{
69
+ const result = await fetch(...args);
70
+ if (!result.ok) {
71
+ const clone = result.clone();
72
+ try {
73
+ const text = await clone.text();
74
+ console.log(`call AI model service error with status code ${result.status} and response text: ${text}`);
75
+ } catch (e) {
76
+ console.log(`call AI model service error with status code ${result.status} but get response text failed.`, e);
77
+ }
78
+ }
79
+ return result;
80
+ }
68
81
  });
69
82
  if (openai && globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGSMITH_DEBUG)) {
70
83
  if (ifInBrowser) throw new Error('langsmith is not supported in browser');
@@ -299,7 +312,7 @@ async function callAI(messages, AIActionTypeValue, modelConfig, options) {
299
312
  isStreamed: !!isStreaming
300
313
  };
301
314
  } catch (e) {
302
- console.error(' call AI error', e);
315
+ console.error('call AI model service error', e);
303
316
  const newError = new Error(`failed to call ${isStreaming ? 'streaming ' : ''}AI model service: ${e.message}. Trouble shooting: https://midscenejs.com/model-provider.html`, {
304
317
  cause: e
305
318
  });
@@ -1 +1 @@
1
- {"version":3,"file":"ai-model/service-caller/index.mjs","sources":["webpack://@midscene/core/./src/ai-model/service-caller/index.ts"],"sourcesContent":["import { AIResponseFormat, type AIUsageInfo } from '@/types';\nimport type { CodeGenerationChunk, StreamingCallback } from '@/types';\nimport { Anthropic } from '@anthropic-ai/sdk';\nimport {\n DefaultAzureCredential,\n getBearerTokenProvider,\n} from '@azure/identity';\nimport {\n type IModelConfig,\n MIDSCENE_API_TYPE,\n MIDSCENE_LANGSMITH_DEBUG,\n OPENAI_MAX_TOKENS,\n type TVlModeTypes,\n type UITarsModelVersion,\n globalConfigManager,\n} from '@midscene/shared/env';\n\nimport { parseBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { ifInBrowser } from '@midscene/shared/utils';\nimport { HttpsProxyAgent } from 'https-proxy-agent';\nimport { jsonrepair } from 'jsonrepair';\nimport OpenAI, { AzureOpenAI } from 'openai';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { Stream } from 'openai/streaming';\nimport { SocksProxyAgent } from 'socks-proxy-agent';\nimport { AIActionType, type AIArgs } from '../common';\nimport { assertSchema } from '../prompt/assertion';\nimport { locatorSchema } from '../prompt/llm-locator';\nimport { planSchema } from '../prompt/llm-planning';\n\nasync function createChatClient({\n AIActionTypeValue,\n modelConfig,\n}: {\n AIActionTypeValue: AIActionType;\n modelConfig: IModelConfig;\n}): Promise<{\n completion: OpenAI.Chat.Completions;\n style: 'openai' | 'anthropic';\n modelName: string;\n modelDescription: string;\n uiTarsVersion?: UITarsModelVersion;\n vlMode: TVlModeTypes | undefined;\n}> {\n const {\n socksProxy,\n httpProxy,\n modelName,\n openaiBaseURL,\n openaiApiKey,\n openaiExtraConfig,\n openaiUseAzureDeprecated,\n useAzureOpenai,\n azureOpenaiScope,\n azureOpenaiKey,\n azureOpenaiEndpoint,\n azureOpenaiApiVersion,\n azureOpenaiDeployment,\n azureExtraConfig,\n useAnthropicSdk,\n anthropicApiKey,\n modelDescription,\n uiTarsModelVersion: uiTarsVersion,\n vlMode,\n } = modelConfig;\n\n let openai: OpenAI | AzureOpenAI | undefined;\n\n let proxyAgent = undefined;\n const debugProxy = getDebug('ai:call:proxy');\n if (httpProxy) {\n debugProxy('using http proxy', httpProxy);\n proxyAgent = new HttpsProxyAgent(httpProxy);\n } else if (socksProxy) {\n debugProxy('using socks proxy', socksProxy);\n proxyAgent = new SocksProxyAgent(socksProxy);\n }\n\n if (openaiUseAzureDeprecated) {\n // this is deprecated\n openai = new AzureOpenAI({\n baseURL: openaiBaseURL,\n apiKey: openaiApiKey,\n httpAgent: proxyAgent,\n ...openaiExtraConfig,\n dangerouslyAllowBrowser: true,\n }) as OpenAI;\n } else if (useAzureOpenai) {\n // https://learn.microsoft.com/en-us/azure/ai-services/openai/chatgpt-quickstart?tabs=bash%2Cjavascript-key%2Ctypescript-keyless%2Cpython&pivots=programming-language-javascript#rest-api\n // keyless authentication\n let tokenProvider: any = undefined;\n if (azureOpenaiScope) {\n assert(\n !ifInBrowser,\n 'Azure OpenAI is not supported in browser with Midscene.',\n );\n const credential = new DefaultAzureCredential();\n\n tokenProvider = getBearerTokenProvider(credential, azureOpenaiScope);\n\n openai = new AzureOpenAI({\n azureADTokenProvider: tokenProvider,\n endpoint: azureOpenaiEndpoint,\n apiVersion: azureOpenaiApiVersion,\n deployment: azureOpenaiDeployment,\n ...openaiExtraConfig,\n ...azureExtraConfig,\n });\n } else {\n // endpoint, apiKey, apiVersion, deployment\n openai = new AzureOpenAI({\n apiKey: azureOpenaiKey,\n endpoint: azureOpenaiEndpoint,\n apiVersion: azureOpenaiApiVersion,\n deployment: azureOpenaiDeployment,\n dangerouslyAllowBrowser: true,\n ...openaiExtraConfig,\n ...azureExtraConfig,\n });\n }\n } else if (!useAnthropicSdk) {\n openai = new OpenAI({\n baseURL: openaiBaseURL,\n apiKey: openaiApiKey,\n httpAgent: proxyAgent,\n ...openaiExtraConfig,\n defaultHeaders: {\n ...(openaiExtraConfig?.defaultHeaders || {}),\n [MIDSCENE_API_TYPE]: AIActionTypeValue.toString(),\n },\n dangerouslyAllowBrowser: true,\n });\n }\n\n if (\n openai &&\n globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGSMITH_DEBUG)\n ) {\n if (ifInBrowser) {\n throw new Error('langsmith is not supported in browser');\n }\n console.log('DEBUGGING MODE: langsmith wrapper enabled');\n const { wrapOpenAI } = await import('langsmith/wrappers');\n openai = wrapOpenAI(openai);\n }\n\n if (typeof openai !== 'undefined') {\n return {\n completion: openai.chat.completions,\n style: 'openai',\n modelName,\n modelDescription,\n uiTarsVersion,\n vlMode,\n };\n }\n\n // Anthropic\n if (useAnthropicSdk) {\n openai = new Anthropic({\n apiKey: anthropicApiKey,\n httpAgent: proxyAgent,\n dangerouslyAllowBrowser: true,\n }) as any;\n }\n\n if (typeof openai !== 'undefined' && (openai as any).messages) {\n return {\n completion: (openai as any).messages,\n style: 'anthropic',\n modelName,\n modelDescription,\n uiTarsVersion,\n vlMode,\n };\n }\n\n throw new Error('Openai SDK or Anthropic SDK is not initialized');\n}\n\nexport async function callAI(\n messages: ChatCompletionMessageParam[],\n AIActionTypeValue: AIActionType,\n modelConfig: IModelConfig,\n options?: {\n stream?: boolean;\n onChunk?: StreamingCallback;\n },\n): Promise<{ content: string; usage?: AIUsageInfo; isStreamed: boolean }> {\n const {\n completion,\n style,\n modelName,\n modelDescription,\n uiTarsVersion,\n vlMode,\n } = await createChatClient({\n AIActionTypeValue,\n modelConfig,\n });\n\n const responseFormat = getResponseFormat(modelName, AIActionTypeValue);\n\n const maxTokens = globalConfigManager.getEnvConfigValue(OPENAI_MAX_TOKENS);\n const debugCall = getDebug('ai:call');\n const debugProfileStats = getDebug('ai:profile:stats');\n const debugProfileDetail = getDebug('ai:profile:detail');\n\n const startTime = Date.now();\n\n const isStreaming = options?.stream && options?.onChunk;\n let content: string | undefined;\n let accumulated = '';\n let usage: OpenAI.CompletionUsage | undefined;\n let timeCost: number | undefined;\n\n const commonConfig = {\n temperature: vlMode === 'vlm-ui-tars' ? 0.0 : 0.1,\n stream: !!isStreaming,\n max_tokens:\n typeof maxTokens === 'number'\n ? maxTokens\n : Number.parseInt(maxTokens || '2048', 10),\n ...(vlMode === 'qwen-vl' || vlMode === 'qwen3-vl' // qwen specific config\n ? {\n vl_high_resolution_images: true,\n }\n : {}),\n };\n\n try {\n if (style === 'openai') {\n debugCall(\n `sending ${isStreaming ? 'streaming ' : ''}request to ${modelName}`,\n );\n\n if (isStreaming) {\n const stream = (await completion.create(\n {\n model: modelName,\n messages,\n response_format: responseFormat,\n ...commonConfig,\n },\n {\n stream: true,\n },\n )) as Stream<OpenAI.Chat.Completions.ChatCompletionChunk> & {\n _request_id?: string | null;\n };\n\n for await (const chunk of stream) {\n const content = chunk.choices?.[0]?.delta?.content || '';\n const reasoning_content =\n (chunk.choices?.[0]?.delta as any)?.reasoning_content || '';\n\n // Check for usage info in any chunk (OpenAI provides usage in separate chunks)\n if (chunk.usage) {\n usage = chunk.usage;\n }\n\n if (content || reasoning_content) {\n accumulated += content;\n const chunkData: CodeGenerationChunk = {\n content,\n reasoning_content,\n accumulated,\n isComplete: false,\n usage: undefined,\n };\n options.onChunk!(chunkData);\n }\n\n // Check if stream is complete\n if (chunk.choices?.[0]?.finish_reason) {\n timeCost = Date.now() - startTime;\n\n // If usage is not available from the stream, provide a basic usage info\n if (!usage) {\n // Estimate token counts based on content length (rough approximation)\n const estimatedTokens = Math.max(\n 1,\n Math.floor(accumulated.length / 4),\n );\n usage = {\n prompt_tokens: estimatedTokens,\n completion_tokens: estimatedTokens,\n total_tokens: estimatedTokens * 2,\n };\n }\n\n // Send final chunk\n const finalChunk: CodeGenerationChunk = {\n content: '',\n accumulated,\n reasoning_content: '',\n isComplete: true,\n usage: {\n prompt_tokens: usage.prompt_tokens ?? 0,\n completion_tokens: usage.completion_tokens ?? 0,\n total_tokens: usage.total_tokens ?? 0,\n time_cost: timeCost ?? 0,\n model_name: modelName,\n model_description: modelDescription,\n intent: modelConfig.intent,\n },\n };\n options.onChunk!(finalChunk);\n break;\n }\n }\n content = accumulated;\n debugProfileStats(\n `streaming model, ${modelName}, mode, ${vlMode || 'default'}, cost-ms, ${timeCost}`,\n );\n } else {\n const result = await completion.create({\n model: modelName,\n messages,\n response_format: responseFormat,\n ...commonConfig,\n } as any);\n timeCost = Date.now() - startTime;\n\n debugProfileStats(\n `model, ${modelName}, mode, ${vlMode || 'default'}, ui-tars-version, ${uiTarsVersion}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}`,\n );\n\n debugProfileDetail(\n `model usage detail: ${JSON.stringify(result.usage)}`,\n );\n\n assert(\n result.choices,\n `invalid response from LLM service: ${JSON.stringify(result)}`,\n );\n content = result.choices[0].message.content!;\n usage = result.usage;\n }\n\n debugCall(`response: ${content}`);\n assert(content, 'empty content');\n } else if (style === 'anthropic') {\n const convertImageContent = (content: any) => {\n if (content.type === 'image_url') {\n const imgBase64 = content.image_url.url;\n assert(imgBase64, 'image_url is required');\n const { mimeType, body } = parseBase64(content.image_url.url);\n return {\n source: {\n type: 'base64',\n media_type: mimeType,\n data: body,\n },\n type: 'image',\n };\n }\n return content;\n };\n\n if (isStreaming) {\n const stream = (await completion.create({\n model: modelName,\n system: 'You are a versatile professional in software UI automation',\n messages: messages.map((m) => ({\n role: 'user',\n content: Array.isArray(m.content)\n ? (m.content as any).map(convertImageContent)\n : m.content,\n })),\n response_format: responseFormat,\n ...commonConfig,\n } as any)) as any;\n\n for await (const chunk of stream) {\n const content = chunk.delta?.text || '';\n if (content) {\n accumulated += content;\n const chunkData: CodeGenerationChunk = {\n content,\n accumulated,\n reasoning_content: '',\n isComplete: false,\n usage: undefined,\n };\n options.onChunk!(chunkData);\n }\n\n // Check if stream is complete\n if (chunk.type === 'message_stop') {\n timeCost = Date.now() - startTime;\n const anthropicUsage = chunk.usage;\n\n // Send final chunk\n const finalChunk: CodeGenerationChunk = {\n content: '',\n accumulated,\n reasoning_content: '',\n isComplete: true,\n usage: anthropicUsage\n ? {\n prompt_tokens: anthropicUsage.input_tokens ?? 0,\n completion_tokens: anthropicUsage.output_tokens ?? 0,\n total_tokens:\n (anthropicUsage.input_tokens ?? 0) +\n (anthropicUsage.output_tokens ?? 0),\n time_cost: timeCost ?? 0,\n model_name: modelName,\n model_description: modelDescription,\n intent: modelConfig.intent,\n }\n : undefined,\n };\n options.onChunk!(finalChunk);\n break;\n }\n }\n content = accumulated;\n } else {\n const result = await completion.create({\n model: modelName,\n system: 'You are a versatile professional in software UI automation',\n messages: messages.map((m) => ({\n role: 'user',\n content: Array.isArray(m.content)\n ? (m.content as any).map(convertImageContent)\n : m.content,\n })),\n response_format: responseFormat,\n ...commonConfig,\n } as any);\n timeCost = Date.now() - startTime;\n content = (result as any).content[0].text as string;\n usage = result.usage;\n }\n\n assert(content, 'empty content');\n }\n // Ensure we always have usage info for streaming responses\n if (isStreaming && !usage) {\n // Estimate token counts based on content length (rough approximation)\n const estimatedTokens = Math.max(\n 1,\n Math.floor((content || '').length / 4),\n );\n usage = {\n prompt_tokens: estimatedTokens,\n completion_tokens: estimatedTokens,\n total_tokens: estimatedTokens * 2,\n };\n }\n\n return {\n content: content || '',\n usage: usage\n ? {\n prompt_tokens: usage.prompt_tokens ?? 0,\n completion_tokens: usage.completion_tokens ?? 0,\n total_tokens: usage.total_tokens ?? 0,\n time_cost: timeCost ?? 0,\n model_name: modelName,\n model_description: modelDescription,\n intent: modelConfig.intent,\n }\n : undefined,\n isStreamed: !!isStreaming,\n };\n } catch (e: any) {\n console.error(' call AI error', e);\n const newError = new Error(\n `failed to call ${isStreaming ? 'streaming ' : ''}AI model service: ${e.message}. Trouble shooting: https://midscenejs.com/model-provider.html`,\n {\n cause: e,\n },\n );\n throw newError;\n }\n}\n\nexport const getResponseFormat = (\n modelName: string,\n AIActionTypeValue: AIActionType,\n):\n | OpenAI.ChatCompletionCreateParams['response_format']\n | OpenAI.ResponseFormatJSONObject => {\n let responseFormat:\n | OpenAI.ChatCompletionCreateParams['response_format']\n | OpenAI.ResponseFormatJSONObject\n | undefined;\n\n if (modelName.includes('gpt-4')) {\n switch (AIActionTypeValue) {\n case AIActionType.ASSERT:\n responseFormat = assertSchema;\n break;\n case AIActionType.INSPECT_ELEMENT:\n responseFormat = locatorSchema;\n break;\n case AIActionType.PLAN:\n responseFormat = planSchema;\n break;\n case AIActionType.EXTRACT_DATA:\n case AIActionType.DESCRIBE_ELEMENT:\n responseFormat = { type: AIResponseFormat.JSON };\n break;\n case AIActionType.TEXT:\n // No response format for plain text - return as-is\n responseFormat = undefined;\n break;\n }\n }\n\n // gpt-4o-2024-05-13 only supports json_object response format\n // Skip for plain text to allow string output\n if (\n modelName === 'gpt-4o-2024-05-13' &&\n AIActionTypeValue !== AIActionType.TEXT\n ) {\n responseFormat = { type: AIResponseFormat.JSON };\n }\n\n return responseFormat;\n};\n\nexport async function callAIWithObjectResponse<T>(\n messages: ChatCompletionMessageParam[],\n AIActionTypeValue: AIActionType,\n modelConfig: IModelConfig,\n): Promise<{ content: T; usage?: AIUsageInfo }> {\n const response = await callAI(messages, AIActionTypeValue, modelConfig);\n assert(response, 'empty response');\n const vlMode = modelConfig.vlMode;\n const jsonContent = safeParseJson(response.content, vlMode);\n return { content: jsonContent, usage: response.usage };\n}\n\nexport async function callAIWithStringResponse(\n msgs: AIArgs,\n AIActionTypeValue: AIActionType,\n modelConfig: IModelConfig,\n): Promise<{ content: string; usage?: AIUsageInfo }> {\n const { content, usage } = await callAI(msgs, AIActionTypeValue, modelConfig);\n return { content, usage };\n}\n\nexport function extractJSONFromCodeBlock(response: string) {\n try {\n // First, try to match a JSON object directly in the response\n const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\n if (jsonMatch) {\n return jsonMatch[1];\n }\n\n // If no direct JSON object is found, try to extract JSON from a code block\n const codeBlockMatch = response.match(\n /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\n );\n if (codeBlockMatch) {\n return codeBlockMatch[1];\n }\n\n // If no code block is found, try to find a JSON-like structure in the text\n const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\n if (jsonLikeMatch) {\n return jsonLikeMatch[0];\n }\n } catch {}\n // If no JSON-like structure is found, return the original response\n return response;\n}\n\nexport function preprocessDoubaoBboxJson(input: string) {\n if (input.includes('bbox')) {\n // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\n while (/\\d+\\s+\\d+/.test(input)) {\n input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n }\n }\n return input;\n}\n\nexport function safeParseJson(input: string, vlMode: TVlModeTypes | undefined) {\n const cleanJsonString = extractJSONFromCodeBlock(input);\n // match the point\n if (cleanJsonString?.match(/\\((\\d+),(\\d+)\\)/)) {\n return cleanJsonString\n .match(/\\((\\d+),(\\d+)\\)/)\n ?.slice(1)\n .map(Number);\n }\n try {\n return JSON.parse(cleanJsonString);\n } catch {}\n try {\n return JSON.parse(jsonrepair(cleanJsonString));\n } catch (e) {}\n\n if (vlMode === 'doubao-vision' || vlMode === 'vlm-ui-tars') {\n const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\n return JSON.parse(jsonrepair(jsonString));\n }\n throw Error(`failed to parse json response: ${input}`);\n}\n"],"names":["createChatClient","AIActionTypeValue","modelConfig","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","openaiUseAzureDeprecated","useAzureOpenai","azureOpenaiScope","azureOpenaiKey","azureOpenaiEndpoint","azureOpenaiApiVersion","azureOpenaiDeployment","azureExtraConfig","useAnthropicSdk","anthropicApiKey","modelDescription","uiTarsVersion","vlMode","openai","proxyAgent","debugProxy","getDebug","HttpsProxyAgent","SocksProxyAgent","AzureOpenAI","tokenProvider","assert","ifInBrowser","credential","DefaultAzureCredential","getBearerTokenProvider","OpenAI","MIDSCENE_API_TYPE","globalConfigManager","MIDSCENE_LANGSMITH_DEBUG","Error","console","wrapOpenAI","Anthropic","callAI","messages","options","completion","style","responseFormat","getResponseFormat","maxTokens","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","isStreaming","content","accumulated","usage","timeCost","commonConfig","Number","stream","chunk","_chunk_choices__delta","_chunk_choices__delta1","_chunk_choices_2","reasoning_content","chunkData","undefined","estimatedTokens","Math","finalChunk","_result_usage","_result_usage1","_result_usage2","result","JSON","convertImageContent","imgBase64","mimeType","body","parseBase64","m","Array","_chunk_delta","anthropicUsage","e","newError","AIActionType","assertSchema","locatorSchema","planSchema","AIResponseFormat","callAIWithObjectResponse","response","jsonContent","safeParseJson","callAIWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","cleanJsonString","_cleanJsonString_match","jsonrepair","jsonString"],"mappings":";;;;;;;;;;;;;;;AAgCA,eAAeA,iBAAiB,EAC9BC,iBAAiB,EACjBC,WAAW,EAIZ;IAQC,MAAM,EACJC,UAAU,EACVC,SAAS,EACTC,SAAS,EACTC,aAAa,EACbC,YAAY,EACZC,iBAAiB,EACjBC,wBAAwB,EACxBC,cAAc,EACdC,gBAAgB,EAChBC,cAAc,EACdC,mBAAmB,EACnBC,qBAAqB,EACrBC,qBAAqB,EACrBC,gBAAgB,EAChBC,eAAe,EACfC,eAAe,EACfC,gBAAgB,EAChB,oBAAoBC,aAAa,EACjCC,MAAM,EACP,GAAGnB;IAEJ,IAAIoB;IAEJ,IAAIC;IACJ,MAAMC,aAAaC,SAAS;IAC5B,IAAIrB,WAAW;QACboB,WAAW,oBAAoBpB;QAC/BmB,aAAa,IAAIG,gBAAgBtB;IACnC,OAAO,IAAID,YAAY;QACrBqB,WAAW,qBAAqBrB;QAChCoB,aAAa,IAAII,gBAAgBxB;IACnC;IAEA,IAAIM,0BAEFa,SAAS,IAAIM,YAAY;QACvB,SAAStB;QACT,QAAQC;QACR,WAAWgB;QACX,GAAGf,iBAAiB;QACpB,yBAAyB;IAC3B;SACK,IAAIE,gBAAgB;QAGzB,IAAImB;QACJ,IAAIlB,kBAAkB;YACpBmB,OACE,CAACC,aACD;YAEF,MAAMC,aAAa,IAAIC;YAEvBJ,gBAAgBK,uBAAuBF,YAAYrB;YAEnDW,SAAS,IAAIM,YAAY;gBACvB,sBAAsBC;gBACtB,UAAUhB;gBACV,YAAYC;gBACZ,YAAYC;gBACZ,GAAGP,iBAAiB;gBACpB,GAAGQ,gBAAgB;YACrB;QACF,OAEEM,SAAS,IAAIM,YAAY;YACvB,QAAQhB;YACR,UAAUC;YACV,YAAYC;YACZ,YAAYC;YACZ,yBAAyB;YACzB,GAAGP,iBAAiB;YACpB,GAAGQ,gBAAgB;QACrB;IAEJ,OAAO,IAAI,CAACC,iBACVK,SAAS,IAAIa,SAAO;QAClB,SAAS7B;QACT,QAAQC;QACR,WAAWgB;QACX,GAAGf,iBAAiB;QACpB,gBAAgB;YACd,GAAIA,AAAAA,CAAAA,QAAAA,oBAAAA,KAAAA,IAAAA,kBAAmB,cAAc,AAAD,KAAK,CAAC,CAAC;YAC3C,CAAC4B,kBAAkB,EAAEnC,kBAAkB,QAAQ;QACjD;QACA,yBAAyB;IAC3B;IAGF,IACEqB,UACAe,oBAAoB,qBAAqB,CAACC,2BAC1C;QACA,IAAIP,aACF,MAAM,IAAIQ,MAAM;QAElBC,QAAQ,GAAG,CAAC;QACZ,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAAC;QACpCnB,SAASmB,WAAWnB;IACtB;IAEA,IAAI,AAAkB,WAAXA,QACT,OAAO;QACL,YAAYA,OAAO,IAAI,CAAC,WAAW;QACnC,OAAO;QACPjB;QACAc;QACAC;QACAC;IACF;IAIF,IAAIJ,iBACFK,SAAS,IAAIoB,UAAU;QACrB,QAAQxB;QACR,WAAWK;QACX,yBAAyB;IAC3B;IAGF,IAAI,AAAkB,WAAXD,UAA2BA,OAAe,QAAQ,EAC3D,OAAO;QACL,YAAaA,OAAe,QAAQ;QACpC,OAAO;QACPjB;QACAc;QACAC;QACAC;IACF;IAGF,MAAM,IAAIkB,MAAM;AAClB;AAEO,eAAeI,OACpBC,QAAsC,EACtC3C,iBAA+B,EAC/BC,WAAyB,EACzB2C,OAGC;IAED,MAAM,EACJC,UAAU,EACVC,KAAK,EACL1C,SAAS,EACTc,gBAAgB,EAChBC,aAAa,EACbC,MAAM,EACP,GAAG,MAAMrB,iBAAiB;QACzBC;QACAC;IACF;IAEA,MAAM8C,iBAAiBC,kBAAkB5C,WAAWJ;IAEpD,MAAMiD,YAAYb,oBAAoB,iBAAiB,CAACc;IACxD,MAAMC,YAAY3B,SAAS;IAC3B,MAAM4B,oBAAoB5B,SAAS;IACnC,MAAM6B,qBAAqB7B,SAAS;IAEpC,MAAM8B,YAAYC,KAAK,GAAG;IAE1B,MAAMC,cAAcZ,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM,AAAD,KAAKA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO,AAAD;IACtD,IAAIa;IACJ,IAAIC,cAAc;IAClB,IAAIC;IACJ,IAAIC;IAEJ,MAAMC,eAAe;QACnB,aAAazC,AAAW,kBAAXA,SAA2B,MAAM;QAC9C,QAAQ,CAAC,CAACoC;QACV,YACE,AAAqB,YAArB,OAAOP,YACHA,YACAa,OAAO,QAAQ,CAACb,aAAa,QAAQ;QAC3C,GAAI7B,AAAW,cAAXA,UAAwBA,AAAW,eAAXA,SACxB;YACE,2BAA2B;QAC7B,IACA,CAAC,CAAC;IACR;IAEA,IAAI;QACF,IAAI0B,AAAU,aAAVA,OAAoB;YACtBK,UACE,CAAC,QAAQ,EAAEK,cAAc,eAAe,GAAG,WAAW,EAAEpD,WAAW;YAGrE,IAAIoD,aAAa;gBACf,MAAMO,SAAU,MAAMlB,WAAW,MAAM,CACrC;oBACE,OAAOzC;oBACPuC;oBACA,iBAAiBI;oBACjB,GAAGc,YAAY;gBACjB,GACA;oBACE,QAAQ;gBACV;gBAKF,WAAW,MAAMG,SAASD,OAAQ;wBAChBE,uBAAAA,iBAAAA,gBAEbC,wBAAAA,kBAAAA,iBAoBCC,kBAAAA;oBAtBJ,MAAMV,UAAUQ,AAAAA,SAAAA,CAAAA,iBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,kBAAAA,cAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,QAAAA,CAAAA,wBAAAA,gBAAoB,KAAK,AAAD,IAAxBA,KAAAA,IAAAA,sBAA2B,OAAO,AAAD,KAAK;oBACtD,MAAMG,oBACJ,AAAC,SAAAF,CAAAA,kBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,mBAAAA,eAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,QAAAA,CAAAA,yBAAAA,iBAAoB,KAAK,AAAD,IAAxBA,KAAAA,IAAAA,uBAAmC,iBAAiB,AAAD,KAAK;oBAG3D,IAAIF,MAAM,KAAK,EACbL,QAAQK,MAAM,KAAK;oBAGrB,IAAIP,WAAWW,mBAAmB;wBAChCV,eAAeD;wBACf,MAAMY,YAAiC;4BACrCZ;4BACAW;4BACAV;4BACA,YAAY;4BACZ,OAAOY;wBACT;wBACA1B,QAAQ,OAAO,CAAEyB;oBACnB;oBAGA,IAAI,QAAAF,CAAAA,kBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,mBAAAA,eAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,iBAAoB,aAAa,EAAE;wBACrCP,WAAWL,KAAK,GAAG,KAAKD;wBAGxB,IAAI,CAACK,OAAO;4BAEV,MAAMY,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAACd,YAAY,MAAM,GAAG;4BAElCC,QAAQ;gCACN,eAAeY;gCACf,mBAAmBA;gCACnB,cAAcA,AAAkB,IAAlBA;4BAChB;wBACF;wBAGA,MAAME,aAAkC;4BACtC,SAAS;4BACTf;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAO;gCACL,eAAeC,MAAM,aAAa,IAAI;gCACtC,mBAAmBA,MAAM,iBAAiB,IAAI;gCAC9C,cAAcA,MAAM,YAAY,IAAI;gCACpC,WAAWC,YAAY;gCACvB,YAAYxD;gCACZ,mBAAmBc;gCACnB,QAAQjB,YAAY,MAAM;4BAC5B;wBACF;wBACA2C,QAAQ,OAAO,CAAE6B;wBACjB;oBACF;gBACF;gBACAhB,UAAUC;gBACVN,kBACE,CAAC,iBAAiB,EAAEhD,UAAU,QAAQ,EAAEgB,UAAU,UAAU,WAAW,EAAEwC,UAAU;YAEvF,OAAO;oBAUqGc,eAAyDC,gBAAwDC;gBAT3N,MAAMC,SAAS,MAAMhC,WAAW,MAAM,CAAC;oBACrC,OAAOzC;oBACPuC;oBACA,iBAAiBI;oBACjB,GAAGc,YAAY;gBACjB;gBACAD,WAAWL,KAAK,GAAG,KAAKD;gBAExBF,kBACE,CAAC,OAAO,EAAEhD,UAAU,QAAQ,EAAEgB,UAAU,UAAU,mBAAmB,EAAED,cAAc,iBAAiB,EAAEuD,AAAAA,SAAAA,CAAAA,gBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,cAAc,aAAa,AAAD,KAAK,GAAG,qBAAqB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,iBAAiB,AAAD,KAAK,GAAG,gBAAgB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,YAAY,AAAD,KAAK,GAAG,WAAW,EAAEhB,SAAS,aAAa,EAAEiB,OAAO,WAAW,IAAI,IAAI;gBAG3TxB,mBACE,CAAC,oBAAoB,EAAEyB,KAAK,SAAS,CAACD,OAAO,KAAK,GAAG;gBAGvDhD,OACEgD,OAAO,OAAO,EACd,CAAC,mCAAmC,EAAEC,KAAK,SAAS,CAACD,SAAS;gBAEhEpB,UAAUoB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;gBAC3ClB,QAAQkB,OAAO,KAAK;YACtB;YAEA1B,UAAU,CAAC,UAAU,EAAEM,SAAS;YAChC5B,OAAO4B,SAAS;QAClB,OAAO,IAAIX,AAAU,gBAAVA,OAAuB;YAChC,MAAMiC,sBAAsB,CAACtB;gBAC3B,IAAIA,AAAiB,gBAAjBA,QAAQ,IAAI,EAAkB;oBAChC,MAAMuB,YAAYvB,QAAQ,SAAS,CAAC,GAAG;oBACvC5B,OAAOmD,WAAW;oBAClB,MAAM,EAAEC,QAAQ,EAAEC,IAAI,EAAE,GAAGC,YAAY1B,QAAQ,SAAS,CAAC,GAAG;oBAC5D,OAAO;wBACL,QAAQ;4BACN,MAAM;4BACN,YAAYwB;4BACZ,MAAMC;wBACR;wBACA,MAAM;oBACR;gBACF;gBACA,OAAOzB;YACT;YAEA,IAAID,aAAa;gBACf,MAAMO,SAAU,MAAMlB,WAAW,MAAM,CAAC;oBACtC,OAAOzC;oBACP,QAAQ;oBACR,UAAUuC,SAAS,GAAG,CAAC,CAACyC,IAAO;4BAC7B,MAAM;4BACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;wBACf;oBACA,iBAAiBrC;oBACjB,GAAGc,YAAY;gBACjB;gBAEA,WAAW,MAAMG,SAASD,OAAQ;wBAChBuB;oBAAhB,MAAM7B,UAAU6B,AAAAA,SAAAA,CAAAA,eAAAA,MAAM,KAAK,AAAD,IAAVA,KAAAA,IAAAA,aAAa,IAAI,AAAD,KAAK;oBACrC,IAAI7B,SAAS;wBACXC,eAAeD;wBACf,MAAMY,YAAiC;4BACrCZ;4BACAC;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAOY;wBACT;wBACA1B,QAAQ,OAAO,CAAEyB;oBACnB;oBAGA,IAAIL,AAAe,mBAAfA,MAAM,IAAI,EAAqB;wBACjCJ,WAAWL,KAAK,GAAG,KAAKD;wBACxB,MAAMiC,iBAAiBvB,MAAM,KAAK;wBAGlC,MAAMS,aAAkC;4BACtC,SAAS;4BACTf;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAO6B,iBACH;gCACE,eAAeA,eAAe,YAAY,IAAI;gCAC9C,mBAAmBA,eAAe,aAAa,IAAI;gCACnD,cACGA,AAAAA,CAAAA,eAAe,YAAY,IAAI,KAC/BA,CAAAA,eAAe,aAAa,IAAI;gCACnC,WAAW3B,YAAY;gCACvB,YAAYxD;gCACZ,mBAAmBc;gCACnB,QAAQjB,YAAY,MAAM;4BAC5B,IACAqE;wBACN;wBACA1B,QAAQ,OAAO,CAAE6B;wBACjB;oBACF;gBACF;gBACAhB,UAAUC;YACZ,OAAO;gBACL,MAAMmB,SAAS,MAAMhC,WAAW,MAAM,CAAC;oBACrC,OAAOzC;oBACP,QAAQ;oBACR,UAAUuC,SAAS,GAAG,CAAC,CAACyC,IAAO;4BAC7B,MAAM;4BACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;wBACf;oBACA,iBAAiBrC;oBACjB,GAAGc,YAAY;gBACjB;gBACAD,WAAWL,KAAK,GAAG,KAAKD;gBACxBG,UAAWoB,OAAe,OAAO,CAAC,EAAE,CAAC,IAAI;gBACzClB,QAAQkB,OAAO,KAAK;YACtB;YAEAhD,OAAO4B,SAAS;QAClB;QAEA,IAAID,eAAe,CAACG,OAAO;YAEzB,MAAMY,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEf,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;YAEtCE,QAAQ;gBACN,eAAeY;gBACf,mBAAmBA;gBACnB,cAAcA,AAAkB,IAAlBA;YAChB;QACF;QAEA,OAAO;YACL,SAASd,WAAW;YACpB,OAAOE,QACH;gBACE,eAAeA,MAAM,aAAa,IAAI;gBACtC,mBAAmBA,MAAM,iBAAiB,IAAI;gBAC9C,cAAcA,MAAM,YAAY,IAAI;gBACpC,WAAWC,YAAY;gBACvB,YAAYxD;gBACZ,mBAAmBc;gBACnB,QAAQjB,YAAY,MAAM;YAC5B,IACAqE;YACJ,YAAY,CAAC,CAACd;QAChB;IACF,EAAE,OAAOgC,GAAQ;QACfjD,QAAQ,KAAK,CAAC,kBAAkBiD;QAChC,MAAMC,WAAW,IAAInD,MACnB,CAAC,eAAe,EAAEkB,cAAc,eAAe,GAAG,kBAAkB,EAAEgC,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC/I;YACE,OAAOA;QACT;QAEF,MAAMC;IACR;AACF;AAEO,MAAMzC,oBAAoB,CAC/B5C,WACAJ;IAIA,IAAI+C;IAKJ,IAAI3C,UAAU,QAAQ,CAAC,UACrB,OAAQJ;QACN,KAAK0F,aAAa,MAAM;YACtB3C,iBAAiB4C;YACjB;QACF,KAAKD,aAAa,eAAe;YAC/B3C,iBAAiB6C;YACjB;QACF,KAAKF,aAAa,IAAI;YACpB3C,iBAAiB8C;YACjB;QACF,KAAKH,aAAa,YAAY;QAC9B,KAAKA,aAAa,gBAAgB;YAChC3C,iBAAiB;gBAAE,MAAM+C,iBAAiB,IAAI;YAAC;YAC/C;QACF,KAAKJ,aAAa,IAAI;YAEpB3C,iBAAiBuB;YACjB;IACJ;IAKF,IACElE,AAAc,wBAAdA,aACAJ,sBAAsB0F,aAAa,IAAI,EAEvC3C,iBAAiB;QAAE,MAAM+C,iBAAiB,IAAI;IAAC;IAGjD,OAAO/C;AACT;AAEO,eAAegD,yBACpBpD,QAAsC,EACtC3C,iBAA+B,EAC/BC,WAAyB;IAEzB,MAAM+F,WAAW,MAAMtD,OAAOC,UAAU3C,mBAAmBC;IAC3D4B,OAAOmE,UAAU;IACjB,MAAM5E,SAASnB,YAAY,MAAM;IACjC,MAAMgG,cAAcC,cAAcF,SAAS,OAAO,EAAE5E;IACpD,OAAO;QAAE,SAAS6E;QAAa,OAAOD,SAAS,KAAK;IAAC;AACvD;AAEO,eAAeG,yBACpBC,IAAY,EACZpG,iBAA+B,EAC/BC,WAAyB;IAEzB,MAAM,EAAEwD,OAAO,EAAEE,KAAK,EAAE,GAAG,MAAMjB,OAAO0D,MAAMpG,mBAAmBC;IACjE,OAAO;QAAEwD;QAASE;IAAM;AAC1B;AAEO,SAAS0C,yBAAyBL,QAAgB;IACvD,IAAI;QAEF,MAAMM,YAAYN,SAAS,KAAK,CAAC;QACjC,IAAIM,WACF,OAAOA,SAAS,CAAC,EAAE;QAIrB,MAAMC,iBAAiBP,SAAS,KAAK,CACnC;QAEF,IAAIO,gBACF,OAAOA,cAAc,CAAC,EAAE;QAI1B,MAAMC,gBAAgBR,SAAS,KAAK,CAAC;QACrC,IAAIQ,eACF,OAAOA,aAAa,CAAC,EAAE;IAE3B,EAAE,OAAM,CAAC;IAET,OAAOR;AACT;AAEO,SAASS,yBAAyBC,KAAa;IACpD,IAAIA,MAAM,QAAQ,CAAC,SAEjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;IAG5C,OAAOA;AACT;AAEO,SAASR,cAAcQ,KAAa,EAAEtF,MAAgC;IAC3E,MAAMuF,kBAAkBN,yBAAyBK;IAEjD,IAAIC,QAAAA,kBAAAA,KAAAA,IAAAA,gBAAiB,KAAK,CAAC,oBAAoB;YACtCC;QAAP,OAAO,QAAAA,CAAAA,yBAAAA,gBACJ,KAAK,CAAC,kBAAiB,IADnBA,KAAAA,IAAAA,uBAEH,KAAK,CAAC,GACP,GAAG,CAAC9C;IACT;IACA,IAAI;QACF,OAAOgB,KAAK,KAAK,CAAC6B;IACpB,EAAE,OAAM,CAAC;IACT,IAAI;QACF,OAAO7B,KAAK,KAAK,CAAC+B,WAAWF;IAC/B,EAAE,OAAOnB,GAAG,CAAC;IAEb,IAAIpE,AAAW,oBAAXA,UAA8BA,AAAW,kBAAXA,QAA0B;QAC1D,MAAM0F,aAAaL,yBAAyBE;QAC5C,OAAO7B,KAAK,KAAK,CAAC+B,WAAWC;IAC/B;IACA,MAAMxE,MAAM,CAAC,+BAA+B,EAAEoE,OAAO;AACvD"}
1
+ {"version":3,"file":"ai-model/service-caller/index.mjs","sources":["webpack://@midscene/core/./src/ai-model/service-caller/index.ts"],"sourcesContent":["import { AIResponseFormat, type AIUsageInfo } from '@/types';\nimport type { CodeGenerationChunk, StreamingCallback } from '@/types';\nimport { Anthropic } from '@anthropic-ai/sdk';\nimport {\n DefaultAzureCredential,\n getBearerTokenProvider,\n} from '@azure/identity';\nimport {\n type IModelConfig,\n MIDSCENE_API_TYPE,\n MIDSCENE_LANGSMITH_DEBUG,\n OPENAI_MAX_TOKENS,\n type TVlModeTypes,\n type UITarsModelVersion,\n globalConfigManager,\n} from '@midscene/shared/env';\n\nimport { parseBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { ifInBrowser } from '@midscene/shared/utils';\nimport { HttpsProxyAgent } from 'https-proxy-agent';\nimport { jsonrepair } from 'jsonrepair';\nimport OpenAI, { AzureOpenAI } from 'openai';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { Stream } from 'openai/streaming';\nimport { SocksProxyAgent } from 'socks-proxy-agent';\nimport { AIActionType, type AIArgs } from '../common';\nimport { assertSchema } from '../prompt/assertion';\nimport { locatorSchema } from '../prompt/llm-locator';\nimport { planSchema } from '../prompt/llm-planning';\n\nasync function createChatClient({\n AIActionTypeValue,\n modelConfig,\n}: {\n AIActionTypeValue: AIActionType;\n modelConfig: IModelConfig;\n}): Promise<{\n completion: OpenAI.Chat.Completions;\n style: 'openai' | 'anthropic';\n modelName: string;\n modelDescription: string;\n uiTarsVersion?: UITarsModelVersion;\n vlMode: TVlModeTypes | undefined;\n}> {\n const {\n socksProxy,\n httpProxy,\n modelName,\n openaiBaseURL,\n openaiApiKey,\n openaiExtraConfig,\n openaiUseAzureDeprecated,\n useAzureOpenai,\n azureOpenaiScope,\n azureOpenaiKey,\n azureOpenaiEndpoint,\n azureOpenaiApiVersion,\n azureOpenaiDeployment,\n azureExtraConfig,\n useAnthropicSdk,\n anthropicApiKey,\n modelDescription,\n uiTarsModelVersion: uiTarsVersion,\n vlMode,\n } = modelConfig;\n\n let openai: OpenAI | AzureOpenAI | undefined;\n\n let proxyAgent = undefined;\n const debugProxy = getDebug('ai:call:proxy');\n if (httpProxy) {\n debugProxy('using http proxy', httpProxy);\n proxyAgent = new HttpsProxyAgent(httpProxy);\n } else if (socksProxy) {\n debugProxy('using socks proxy', socksProxy);\n proxyAgent = new SocksProxyAgent(socksProxy);\n }\n\n if (openaiUseAzureDeprecated) {\n // this is deprecated\n openai = new AzureOpenAI({\n baseURL: openaiBaseURL,\n apiKey: openaiApiKey,\n httpAgent: proxyAgent,\n ...openaiExtraConfig,\n dangerouslyAllowBrowser: true,\n }) as OpenAI;\n } else if (useAzureOpenai) {\n // https://learn.microsoft.com/en-us/azure/ai-services/openai/chatgpt-quickstart?tabs=bash%2Cjavascript-key%2Ctypescript-keyless%2Cpython&pivots=programming-language-javascript#rest-api\n // keyless authentication\n let tokenProvider: any = undefined;\n if (azureOpenaiScope) {\n assert(\n !ifInBrowser,\n 'Azure OpenAI is not supported in browser with Midscene.',\n );\n const credential = new DefaultAzureCredential();\n\n tokenProvider = getBearerTokenProvider(credential, azureOpenaiScope);\n\n openai = new AzureOpenAI({\n azureADTokenProvider: tokenProvider,\n endpoint: azureOpenaiEndpoint,\n apiVersion: azureOpenaiApiVersion,\n deployment: azureOpenaiDeployment,\n ...openaiExtraConfig,\n ...azureExtraConfig,\n });\n } else {\n // endpoint, apiKey, apiVersion, deployment\n openai = new AzureOpenAI({\n apiKey: azureOpenaiKey,\n endpoint: azureOpenaiEndpoint,\n apiVersion: azureOpenaiApiVersion,\n deployment: azureOpenaiDeployment,\n dangerouslyAllowBrowser: true,\n ...openaiExtraConfig,\n ...azureExtraConfig,\n });\n }\n } else if (!useAnthropicSdk) {\n openai = new OpenAI({\n baseURL: openaiBaseURL,\n apiKey: openaiApiKey,\n httpAgent: proxyAgent,\n ...openaiExtraConfig,\n defaultHeaders: {\n ...(openaiExtraConfig?.defaultHeaders || {}),\n [MIDSCENE_API_TYPE]: AIActionTypeValue.toString(),\n },\n dangerouslyAllowBrowser: true,\n /**\n * By default, OpenAI uses the built-in fetch function of Node.js as the implementation of fetch.\n * However, the built-in error handling logic of OpenAI does not throw out all the text in the HTTP response.\n * This will prevent users from seeing the most crucial error messages, especially when using third-party models.\n */\n fetch: async (...args) => {\n const result = await fetch(...args);\n if (!result.ok) {\n const clone = result.clone();\n try {\n const text = await clone.text();\n console.log(\n `call AI model service error with status code ${result.status} and response text: ${text}`,\n );\n } catch (e) {\n console.log(\n `call AI model service error with status code ${result.status} but get response text failed.`,\n e,\n );\n }\n }\n return result;\n },\n });\n }\n\n if (\n openai &&\n globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGSMITH_DEBUG)\n ) {\n if (ifInBrowser) {\n throw new Error('langsmith is not supported in browser');\n }\n console.log('DEBUGGING MODE: langsmith wrapper enabled');\n const { wrapOpenAI } = await import('langsmith/wrappers');\n openai = wrapOpenAI(openai);\n }\n\n if (typeof openai !== 'undefined') {\n return {\n completion: openai.chat.completions,\n style: 'openai',\n modelName,\n modelDescription,\n uiTarsVersion,\n vlMode,\n };\n }\n\n // Anthropic\n if (useAnthropicSdk) {\n openai = new Anthropic({\n apiKey: anthropicApiKey,\n httpAgent: proxyAgent,\n dangerouslyAllowBrowser: true,\n }) as any;\n }\n\n if (typeof openai !== 'undefined' && (openai as any).messages) {\n return {\n completion: (openai as any).messages,\n style: 'anthropic',\n modelName,\n modelDescription,\n uiTarsVersion,\n vlMode,\n };\n }\n\n throw new Error('Openai SDK or Anthropic SDK is not initialized');\n}\n\nexport async function callAI(\n messages: ChatCompletionMessageParam[],\n AIActionTypeValue: AIActionType,\n modelConfig: IModelConfig,\n options?: {\n stream?: boolean;\n onChunk?: StreamingCallback;\n },\n): Promise<{ content: string; usage?: AIUsageInfo; isStreamed: boolean }> {\n const {\n completion,\n style,\n modelName,\n modelDescription,\n uiTarsVersion,\n vlMode,\n } = await createChatClient({\n AIActionTypeValue,\n modelConfig,\n });\n\n const responseFormat = getResponseFormat(modelName, AIActionTypeValue);\n\n const maxTokens = globalConfigManager.getEnvConfigValue(OPENAI_MAX_TOKENS);\n const debugCall = getDebug('ai:call');\n const debugProfileStats = getDebug('ai:profile:stats');\n const debugProfileDetail = getDebug('ai:profile:detail');\n\n const startTime = Date.now();\n\n const isStreaming = options?.stream && options?.onChunk;\n let content: string | undefined;\n let accumulated = '';\n let usage: OpenAI.CompletionUsage | undefined;\n let timeCost: number | undefined;\n\n const commonConfig = {\n temperature: vlMode === 'vlm-ui-tars' ? 0.0 : 0.1,\n stream: !!isStreaming,\n max_tokens:\n typeof maxTokens === 'number'\n ? maxTokens\n : Number.parseInt(maxTokens || '2048', 10),\n ...(vlMode === 'qwen-vl' || vlMode === 'qwen3-vl' // qwen specific config\n ? {\n vl_high_resolution_images: true,\n }\n : {}),\n };\n\n try {\n if (style === 'openai') {\n debugCall(\n `sending ${isStreaming ? 'streaming ' : ''}request to ${modelName}`,\n );\n\n if (isStreaming) {\n const stream = (await completion.create(\n {\n model: modelName,\n messages,\n response_format: responseFormat,\n ...commonConfig,\n },\n {\n stream: true,\n },\n )) as Stream<OpenAI.Chat.Completions.ChatCompletionChunk> & {\n _request_id?: string | null;\n };\n\n for await (const chunk of stream) {\n const content = chunk.choices?.[0]?.delta?.content || '';\n const reasoning_content =\n (chunk.choices?.[0]?.delta as any)?.reasoning_content || '';\n\n // Check for usage info in any chunk (OpenAI provides usage in separate chunks)\n if (chunk.usage) {\n usage = chunk.usage;\n }\n\n if (content || reasoning_content) {\n accumulated += content;\n const chunkData: CodeGenerationChunk = {\n content,\n reasoning_content,\n accumulated,\n isComplete: false,\n usage: undefined,\n };\n options.onChunk!(chunkData);\n }\n\n // Check if stream is complete\n if (chunk.choices?.[0]?.finish_reason) {\n timeCost = Date.now() - startTime;\n\n // If usage is not available from the stream, provide a basic usage info\n if (!usage) {\n // Estimate token counts based on content length (rough approximation)\n const estimatedTokens = Math.max(\n 1,\n Math.floor(accumulated.length / 4),\n );\n usage = {\n prompt_tokens: estimatedTokens,\n completion_tokens: estimatedTokens,\n total_tokens: estimatedTokens * 2,\n };\n }\n\n // Send final chunk\n const finalChunk: CodeGenerationChunk = {\n content: '',\n accumulated,\n reasoning_content: '',\n isComplete: true,\n usage: {\n prompt_tokens: usage.prompt_tokens ?? 0,\n completion_tokens: usage.completion_tokens ?? 0,\n total_tokens: usage.total_tokens ?? 0,\n time_cost: timeCost ?? 0,\n model_name: modelName,\n model_description: modelDescription,\n intent: modelConfig.intent,\n },\n };\n options.onChunk!(finalChunk);\n break;\n }\n }\n content = accumulated;\n debugProfileStats(\n `streaming model, ${modelName}, mode, ${vlMode || 'default'}, cost-ms, ${timeCost}`,\n );\n } else {\n const result = await completion.create({\n model: modelName,\n messages,\n response_format: responseFormat,\n ...commonConfig,\n } as any);\n timeCost = Date.now() - startTime;\n\n debugProfileStats(\n `model, ${modelName}, mode, ${vlMode || 'default'}, ui-tars-version, ${uiTarsVersion}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}`,\n );\n\n debugProfileDetail(\n `model usage detail: ${JSON.stringify(result.usage)}`,\n );\n\n assert(\n result.choices,\n `invalid response from LLM service: ${JSON.stringify(result)}`,\n );\n content = result.choices[0].message.content!;\n usage = result.usage;\n }\n\n debugCall(`response: ${content}`);\n assert(content, 'empty content');\n } else if (style === 'anthropic') {\n const convertImageContent = (content: any) => {\n if (content.type === 'image_url') {\n const imgBase64 = content.image_url.url;\n assert(imgBase64, 'image_url is required');\n const { mimeType, body } = parseBase64(content.image_url.url);\n return {\n source: {\n type: 'base64',\n media_type: mimeType,\n data: body,\n },\n type: 'image',\n };\n }\n return content;\n };\n\n if (isStreaming) {\n const stream = (await completion.create({\n model: modelName,\n system: 'You are a versatile professional in software UI automation',\n messages: messages.map((m) => ({\n role: 'user',\n content: Array.isArray(m.content)\n ? (m.content as any).map(convertImageContent)\n : m.content,\n })),\n response_format: responseFormat,\n ...commonConfig,\n } as any)) as any;\n\n for await (const chunk of stream) {\n const content = chunk.delta?.text || '';\n if (content) {\n accumulated += content;\n const chunkData: CodeGenerationChunk = {\n content,\n accumulated,\n reasoning_content: '',\n isComplete: false,\n usage: undefined,\n };\n options.onChunk!(chunkData);\n }\n\n // Check if stream is complete\n if (chunk.type === 'message_stop') {\n timeCost = Date.now() - startTime;\n const anthropicUsage = chunk.usage;\n\n // Send final chunk\n const finalChunk: CodeGenerationChunk = {\n content: '',\n accumulated,\n reasoning_content: '',\n isComplete: true,\n usage: anthropicUsage\n ? {\n prompt_tokens: anthropicUsage.input_tokens ?? 0,\n completion_tokens: anthropicUsage.output_tokens ?? 0,\n total_tokens:\n (anthropicUsage.input_tokens ?? 0) +\n (anthropicUsage.output_tokens ?? 0),\n time_cost: timeCost ?? 0,\n model_name: modelName,\n model_description: modelDescription,\n intent: modelConfig.intent,\n }\n : undefined,\n };\n options.onChunk!(finalChunk);\n break;\n }\n }\n content = accumulated;\n } else {\n const result = await completion.create({\n model: modelName,\n system: 'You are a versatile professional in software UI automation',\n messages: messages.map((m) => ({\n role: 'user',\n content: Array.isArray(m.content)\n ? (m.content as any).map(convertImageContent)\n : m.content,\n })),\n response_format: responseFormat,\n ...commonConfig,\n } as any);\n timeCost = Date.now() - startTime;\n content = (result as any).content[0].text as string;\n usage = result.usage;\n }\n\n assert(content, 'empty content');\n }\n // Ensure we always have usage info for streaming responses\n if (isStreaming && !usage) {\n // Estimate token counts based on content length (rough approximation)\n const estimatedTokens = Math.max(\n 1,\n Math.floor((content || '').length / 4),\n );\n usage = {\n prompt_tokens: estimatedTokens,\n completion_tokens: estimatedTokens,\n total_tokens: estimatedTokens * 2,\n };\n }\n\n return {\n content: content || '',\n usage: usage\n ? {\n prompt_tokens: usage.prompt_tokens ?? 0,\n completion_tokens: usage.completion_tokens ?? 0,\n total_tokens: usage.total_tokens ?? 0,\n time_cost: timeCost ?? 0,\n model_name: modelName,\n model_description: modelDescription,\n intent: modelConfig.intent,\n }\n : undefined,\n isStreamed: !!isStreaming,\n };\n } catch (e: any) {\n console.error('call AI model service error', e);\n const newError = new Error(\n `failed to call ${isStreaming ? 'streaming ' : ''}AI model service: ${e.message}. Trouble shooting: https://midscenejs.com/model-provider.html`,\n {\n cause: e,\n },\n );\n throw newError;\n }\n}\n\nexport const getResponseFormat = (\n modelName: string,\n AIActionTypeValue: AIActionType,\n):\n | OpenAI.ChatCompletionCreateParams['response_format']\n | OpenAI.ResponseFormatJSONObject => {\n let responseFormat:\n | OpenAI.ChatCompletionCreateParams['response_format']\n | OpenAI.ResponseFormatJSONObject\n | undefined;\n\n if (modelName.includes('gpt-4')) {\n switch (AIActionTypeValue) {\n case AIActionType.ASSERT:\n responseFormat = assertSchema;\n break;\n case AIActionType.INSPECT_ELEMENT:\n responseFormat = locatorSchema;\n break;\n case AIActionType.PLAN:\n responseFormat = planSchema;\n break;\n case AIActionType.EXTRACT_DATA:\n case AIActionType.DESCRIBE_ELEMENT:\n responseFormat = { type: AIResponseFormat.JSON };\n break;\n case AIActionType.TEXT:\n // No response format for plain text - return as-is\n responseFormat = undefined;\n break;\n }\n }\n\n // gpt-4o-2024-05-13 only supports json_object response format\n // Skip for plain text to allow string output\n if (\n modelName === 'gpt-4o-2024-05-13' &&\n AIActionTypeValue !== AIActionType.TEXT\n ) {\n responseFormat = { type: AIResponseFormat.JSON };\n }\n\n return responseFormat;\n};\n\nexport async function callAIWithObjectResponse<T>(\n messages: ChatCompletionMessageParam[],\n AIActionTypeValue: AIActionType,\n modelConfig: IModelConfig,\n): Promise<{ content: T; usage?: AIUsageInfo }> {\n const response = await callAI(messages, AIActionTypeValue, modelConfig);\n assert(response, 'empty response');\n const vlMode = modelConfig.vlMode;\n const jsonContent = safeParseJson(response.content, vlMode);\n return { content: jsonContent, usage: response.usage };\n}\n\nexport async function callAIWithStringResponse(\n msgs: AIArgs,\n AIActionTypeValue: AIActionType,\n modelConfig: IModelConfig,\n): Promise<{ content: string; usage?: AIUsageInfo }> {\n const { content, usage } = await callAI(msgs, AIActionTypeValue, modelConfig);\n return { content, usage };\n}\n\nexport function extractJSONFromCodeBlock(response: string) {\n try {\n // First, try to match a JSON object directly in the response\n const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\n if (jsonMatch) {\n return jsonMatch[1];\n }\n\n // If no direct JSON object is found, try to extract JSON from a code block\n const codeBlockMatch = response.match(\n /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\n );\n if (codeBlockMatch) {\n return codeBlockMatch[1];\n }\n\n // If no code block is found, try to find a JSON-like structure in the text\n const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\n if (jsonLikeMatch) {\n return jsonLikeMatch[0];\n }\n } catch {}\n // If no JSON-like structure is found, return the original response\n return response;\n}\n\nexport function preprocessDoubaoBboxJson(input: string) {\n if (input.includes('bbox')) {\n // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\n while (/\\d+\\s+\\d+/.test(input)) {\n input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n }\n }\n return input;\n}\n\nexport function safeParseJson(input: string, vlMode: TVlModeTypes | undefined) {\n const cleanJsonString = extractJSONFromCodeBlock(input);\n // match the point\n if (cleanJsonString?.match(/\\((\\d+),(\\d+)\\)/)) {\n return cleanJsonString\n .match(/\\((\\d+),(\\d+)\\)/)\n ?.slice(1)\n .map(Number);\n }\n try {\n return JSON.parse(cleanJsonString);\n } catch {}\n try {\n return JSON.parse(jsonrepair(cleanJsonString));\n } catch (e) {}\n\n if (vlMode === 'doubao-vision' || vlMode === 'vlm-ui-tars') {\n const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\n return JSON.parse(jsonrepair(jsonString));\n }\n throw Error(`failed to parse json response: ${input}`);\n}\n"],"names":["createChatClient","AIActionTypeValue","modelConfig","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","openaiUseAzureDeprecated","useAzureOpenai","azureOpenaiScope","azureOpenaiKey","azureOpenaiEndpoint","azureOpenaiApiVersion","azureOpenaiDeployment","azureExtraConfig","useAnthropicSdk","anthropicApiKey","modelDescription","uiTarsVersion","vlMode","openai","proxyAgent","debugProxy","getDebug","HttpsProxyAgent","SocksProxyAgent","AzureOpenAI","tokenProvider","assert","ifInBrowser","credential","DefaultAzureCredential","getBearerTokenProvider","OpenAI","MIDSCENE_API_TYPE","args","result","fetch","clone","text","console","e","globalConfigManager","MIDSCENE_LANGSMITH_DEBUG","Error","wrapOpenAI","Anthropic","callAI","messages","options","completion","style","responseFormat","getResponseFormat","maxTokens","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","isStreaming","content","accumulated","usage","timeCost","commonConfig","Number","stream","chunk","_chunk_choices__delta","_chunk_choices__delta1","_chunk_choices_2","reasoning_content","chunkData","undefined","estimatedTokens","Math","finalChunk","_result_usage","_result_usage1","_result_usage2","JSON","convertImageContent","imgBase64","mimeType","body","parseBase64","m","Array","_chunk_delta","anthropicUsage","newError","AIActionType","assertSchema","locatorSchema","planSchema","AIResponseFormat","callAIWithObjectResponse","response","jsonContent","safeParseJson","callAIWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","cleanJsonString","_cleanJsonString_match","jsonrepair","jsonString"],"mappings":";;;;;;;;;;;;;;;AAgCA,eAAeA,iBAAiB,EAC9BC,iBAAiB,EACjBC,WAAW,EAIZ;IAQC,MAAM,EACJC,UAAU,EACVC,SAAS,EACTC,SAAS,EACTC,aAAa,EACbC,YAAY,EACZC,iBAAiB,EACjBC,wBAAwB,EACxBC,cAAc,EACdC,gBAAgB,EAChBC,cAAc,EACdC,mBAAmB,EACnBC,qBAAqB,EACrBC,qBAAqB,EACrBC,gBAAgB,EAChBC,eAAe,EACfC,eAAe,EACfC,gBAAgB,EAChB,oBAAoBC,aAAa,EACjCC,MAAM,EACP,GAAGnB;IAEJ,IAAIoB;IAEJ,IAAIC;IACJ,MAAMC,aAAaC,SAAS;IAC5B,IAAIrB,WAAW;QACboB,WAAW,oBAAoBpB;QAC/BmB,aAAa,IAAIG,gBAAgBtB;IACnC,OAAO,IAAID,YAAY;QACrBqB,WAAW,qBAAqBrB;QAChCoB,aAAa,IAAII,gBAAgBxB;IACnC;IAEA,IAAIM,0BAEFa,SAAS,IAAIM,YAAY;QACvB,SAAStB;QACT,QAAQC;QACR,WAAWgB;QACX,GAAGf,iBAAiB;QACpB,yBAAyB;IAC3B;SACK,IAAIE,gBAAgB;QAGzB,IAAImB;QACJ,IAAIlB,kBAAkB;YACpBmB,OACE,CAACC,aACD;YAEF,MAAMC,aAAa,IAAIC;YAEvBJ,gBAAgBK,uBAAuBF,YAAYrB;YAEnDW,SAAS,IAAIM,YAAY;gBACvB,sBAAsBC;gBACtB,UAAUhB;gBACV,YAAYC;gBACZ,YAAYC;gBACZ,GAAGP,iBAAiB;gBACpB,GAAGQ,gBAAgB;YACrB;QACF,OAEEM,SAAS,IAAIM,YAAY;YACvB,QAAQhB;YACR,UAAUC;YACV,YAAYC;YACZ,YAAYC;YACZ,yBAAyB;YACzB,GAAGP,iBAAiB;YACpB,GAAGQ,gBAAgB;QACrB;IAEJ,OAAO,IAAI,CAACC,iBACVK,SAAS,IAAIa,SAAO;QAClB,SAAS7B;QACT,QAAQC;QACR,WAAWgB;QACX,GAAGf,iBAAiB;QACpB,gBAAgB;YACd,GAAIA,AAAAA,CAAAA,QAAAA,oBAAAA,KAAAA,IAAAA,kBAAmB,cAAc,AAAD,KAAK,CAAC,CAAC;YAC3C,CAAC4B,kBAAkB,EAAEnC,kBAAkB,QAAQ;QACjD;QACA,yBAAyB;QAMzB,OAAO,OAAO,GAAGoC;YACf,MAAMC,SAAS,MAAMC,SAASF;YAC9B,IAAI,CAACC,OAAO,EAAE,EAAE;gBACd,MAAME,QAAQF,OAAO,KAAK;gBAC1B,IAAI;oBACF,MAAMG,OAAO,MAAMD,MAAM,IAAI;oBAC7BE,QAAQ,GAAG,CACT,CAAC,6CAA6C,EAAEJ,OAAO,MAAM,CAAC,oBAAoB,EAAEG,MAAM;gBAE9F,EAAE,OAAOE,GAAG;oBACVD,QAAQ,GAAG,CACT,CAAC,6CAA6C,EAAEJ,OAAO,MAAM,CAAC,8BAA8B,CAAC,EAC7FK;gBAEJ;YACF;YACA,OAAOL;QACT;IACF;IAGF,IACEhB,UACAsB,oBAAoB,qBAAqB,CAACC,2BAC1C;QACA,IAAId,aACF,MAAM,IAAIe,MAAM;QAElBJ,QAAQ,GAAG,CAAC;QACZ,MAAM,EAAEK,UAAU,EAAE,GAAG,MAAM,MAAM,CAAC;QACpCzB,SAASyB,WAAWzB;IACtB;IAEA,IAAI,AAAkB,WAAXA,QACT,OAAO;QACL,YAAYA,OAAO,IAAI,CAAC,WAAW;QACnC,OAAO;QACPjB;QACAc;QACAC;QACAC;IACF;IAIF,IAAIJ,iBACFK,SAAS,IAAI0B,UAAU;QACrB,QAAQ9B;QACR,WAAWK;QACX,yBAAyB;IAC3B;IAGF,IAAI,AAAkB,WAAXD,UAA2BA,OAAe,QAAQ,EAC3D,OAAO;QACL,YAAaA,OAAe,QAAQ;QACpC,OAAO;QACPjB;QACAc;QACAC;QACAC;IACF;IAGF,MAAM,IAAIyB,MAAM;AAClB;AAEO,eAAeG,OACpBC,QAAsC,EACtCjD,iBAA+B,EAC/BC,WAAyB,EACzBiD,OAGC;IAED,MAAM,EACJC,UAAU,EACVC,KAAK,EACLhD,SAAS,EACTc,gBAAgB,EAChBC,aAAa,EACbC,MAAM,EACP,GAAG,MAAMrB,iBAAiB;QACzBC;QACAC;IACF;IAEA,MAAMoD,iBAAiBC,kBAAkBlD,WAAWJ;IAEpD,MAAMuD,YAAYZ,oBAAoB,iBAAiB,CAACa;IACxD,MAAMC,YAAYjC,SAAS;IAC3B,MAAMkC,oBAAoBlC,SAAS;IACnC,MAAMmC,qBAAqBnC,SAAS;IAEpC,MAAMoC,YAAYC,KAAK,GAAG;IAE1B,MAAMC,cAAcZ,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM,AAAD,KAAKA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO,AAAD;IACtD,IAAIa;IACJ,IAAIC,cAAc;IAClB,IAAIC;IACJ,IAAIC;IAEJ,MAAMC,eAAe;QACnB,aAAa/C,AAAW,kBAAXA,SAA2B,MAAM;QAC9C,QAAQ,CAAC,CAAC0C;QACV,YACE,AAAqB,YAArB,OAAOP,YACHA,YACAa,OAAO,QAAQ,CAACb,aAAa,QAAQ;QAC3C,GAAInC,AAAW,cAAXA,UAAwBA,AAAW,eAAXA,SACxB;YACE,2BAA2B;QAC7B,IACA,CAAC,CAAC;IACR;IAEA,IAAI;QACF,IAAIgC,AAAU,aAAVA,OAAoB;YACtBK,UACE,CAAC,QAAQ,EAAEK,cAAc,eAAe,GAAG,WAAW,EAAE1D,WAAW;YAGrE,IAAI0D,aAAa;gBACf,MAAMO,SAAU,MAAMlB,WAAW,MAAM,CACrC;oBACE,OAAO/C;oBACP6C;oBACA,iBAAiBI;oBACjB,GAAGc,YAAY;gBACjB,GACA;oBACE,QAAQ;gBACV;gBAKF,WAAW,MAAMG,SAASD,OAAQ;wBAChBE,uBAAAA,iBAAAA,gBAEbC,wBAAAA,kBAAAA,iBAoBCC,kBAAAA;oBAtBJ,MAAMV,UAAUQ,AAAAA,SAAAA,CAAAA,iBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,kBAAAA,cAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,QAAAA,CAAAA,wBAAAA,gBAAoB,KAAK,AAAD,IAAxBA,KAAAA,IAAAA,sBAA2B,OAAO,AAAD,KAAK;oBACtD,MAAMG,oBACJ,AAAC,SAAAF,CAAAA,kBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,mBAAAA,eAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,QAAAA,CAAAA,yBAAAA,iBAAoB,KAAK,AAAD,IAAxBA,KAAAA,IAAAA,uBAAmC,iBAAiB,AAAD,KAAK;oBAG3D,IAAIF,MAAM,KAAK,EACbL,QAAQK,MAAM,KAAK;oBAGrB,IAAIP,WAAWW,mBAAmB;wBAChCV,eAAeD;wBACf,MAAMY,YAAiC;4BACrCZ;4BACAW;4BACAV;4BACA,YAAY;4BACZ,OAAOY;wBACT;wBACA1B,QAAQ,OAAO,CAAEyB;oBACnB;oBAGA,IAAI,QAAAF,CAAAA,kBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,mBAAAA,eAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,iBAAoB,aAAa,EAAE;wBACrCP,WAAWL,KAAK,GAAG,KAAKD;wBAGxB,IAAI,CAACK,OAAO;4BAEV,MAAMY,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAACd,YAAY,MAAM,GAAG;4BAElCC,QAAQ;gCACN,eAAeY;gCACf,mBAAmBA;gCACnB,cAAcA,AAAkB,IAAlBA;4BAChB;wBACF;wBAGA,MAAME,aAAkC;4BACtC,SAAS;4BACTf;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAO;gCACL,eAAeC,MAAM,aAAa,IAAI;gCACtC,mBAAmBA,MAAM,iBAAiB,IAAI;gCAC9C,cAAcA,MAAM,YAAY,IAAI;gCACpC,WAAWC,YAAY;gCACvB,YAAY9D;gCACZ,mBAAmBc;gCACnB,QAAQjB,YAAY,MAAM;4BAC5B;wBACF;wBACAiD,QAAQ,OAAO,CAAE6B;wBACjB;oBACF;gBACF;gBACAhB,UAAUC;gBACVN,kBACE,CAAC,iBAAiB,EAAEtD,UAAU,QAAQ,EAAEgB,UAAU,UAAU,WAAW,EAAE8C,UAAU;YAEvF,OAAO;oBAUqGc,eAAyDC,gBAAwDC;gBAT3N,MAAM7C,SAAS,MAAMc,WAAW,MAAM,CAAC;oBACrC,OAAO/C;oBACP6C;oBACA,iBAAiBI;oBACjB,GAAGc,YAAY;gBACjB;gBACAD,WAAWL,KAAK,GAAG,KAAKD;gBAExBF,kBACE,CAAC,OAAO,EAAEtD,UAAU,QAAQ,EAAEgB,UAAU,UAAU,mBAAmB,EAAED,cAAc,iBAAiB,EAAE6D,AAAAA,SAAAA,CAAAA,gBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,cAAc,aAAa,AAAD,KAAK,GAAG,qBAAqB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,iBAAiB,AAAD,KAAK,GAAG,gBAAgB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,YAAY,AAAD,KAAK,GAAG,WAAW,EAAEhB,SAAS,aAAa,EAAE7B,OAAO,WAAW,IAAI,IAAI;gBAG3TsB,mBACE,CAAC,oBAAoB,EAAEwB,KAAK,SAAS,CAAC9C,OAAO,KAAK,GAAG;gBAGvDR,OACEQ,OAAO,OAAO,EACd,CAAC,mCAAmC,EAAE8C,KAAK,SAAS,CAAC9C,SAAS;gBAEhE0B,UAAU1B,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;gBAC3C4B,QAAQ5B,OAAO,KAAK;YACtB;YAEAoB,UAAU,CAAC,UAAU,EAAEM,SAAS;YAChClC,OAAOkC,SAAS;QAClB,OAAO,IAAIX,AAAU,gBAAVA,OAAuB;YAChC,MAAMgC,sBAAsB,CAACrB;gBAC3B,IAAIA,AAAiB,gBAAjBA,QAAQ,IAAI,EAAkB;oBAChC,MAAMsB,YAAYtB,QAAQ,SAAS,CAAC,GAAG;oBACvClC,OAAOwD,WAAW;oBAClB,MAAM,EAAEC,QAAQ,EAAEC,IAAI,EAAE,GAAGC,YAAYzB,QAAQ,SAAS,CAAC,GAAG;oBAC5D,OAAO;wBACL,QAAQ;4BACN,MAAM;4BACN,YAAYuB;4BACZ,MAAMC;wBACR;wBACA,MAAM;oBACR;gBACF;gBACA,OAAOxB;YACT;YAEA,IAAID,aAAa;gBACf,MAAMO,SAAU,MAAMlB,WAAW,MAAM,CAAC;oBACtC,OAAO/C;oBACP,QAAQ;oBACR,UAAU6C,SAAS,GAAG,CAAC,CAACwC,IAAO;4BAC7B,MAAM;4BACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;wBACf;oBACA,iBAAiBpC;oBACjB,GAAGc,YAAY;gBACjB;gBAEA,WAAW,MAAMG,SAASD,OAAQ;wBAChBsB;oBAAhB,MAAM5B,UAAU4B,AAAAA,SAAAA,CAAAA,eAAAA,MAAM,KAAK,AAAD,IAAVA,KAAAA,IAAAA,aAAa,IAAI,AAAD,KAAK;oBACrC,IAAI5B,SAAS;wBACXC,eAAeD;wBACf,MAAMY,YAAiC;4BACrCZ;4BACAC;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAOY;wBACT;wBACA1B,QAAQ,OAAO,CAAEyB;oBACnB;oBAGA,IAAIL,AAAe,mBAAfA,MAAM,IAAI,EAAqB;wBACjCJ,WAAWL,KAAK,GAAG,KAAKD;wBACxB,MAAMgC,iBAAiBtB,MAAM,KAAK;wBAGlC,MAAMS,aAAkC;4BACtC,SAAS;4BACTf;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAO4B,iBACH;gCACE,eAAeA,eAAe,YAAY,IAAI;gCAC9C,mBAAmBA,eAAe,aAAa,IAAI;gCACnD,cACGA,AAAAA,CAAAA,eAAe,YAAY,IAAI,KAC/BA,CAAAA,eAAe,aAAa,IAAI;gCACnC,WAAW1B,YAAY;gCACvB,YAAY9D;gCACZ,mBAAmBc;gCACnB,QAAQjB,YAAY,MAAM;4BAC5B,IACA2E;wBACN;wBACA1B,QAAQ,OAAO,CAAE6B;wBACjB;oBACF;gBACF;gBACAhB,UAAUC;YACZ,OAAO;gBACL,MAAM3B,SAAS,MAAMc,WAAW,MAAM,CAAC;oBACrC,OAAO/C;oBACP,QAAQ;oBACR,UAAU6C,SAAS,GAAG,CAAC,CAACwC,IAAO;4BAC7B,MAAM;4BACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;wBACf;oBACA,iBAAiBpC;oBACjB,GAAGc,YAAY;gBACjB;gBACAD,WAAWL,KAAK,GAAG,KAAKD;gBACxBG,UAAW1B,OAAe,OAAO,CAAC,EAAE,CAAC,IAAI;gBACzC4B,QAAQ5B,OAAO,KAAK;YACtB;YAEAR,OAAOkC,SAAS;QAClB;QAEA,IAAID,eAAe,CAACG,OAAO;YAEzB,MAAMY,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEf,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;YAEtCE,QAAQ;gBACN,eAAeY;gBACf,mBAAmBA;gBACnB,cAAcA,AAAkB,IAAlBA;YAChB;QACF;QAEA,OAAO;YACL,SAASd,WAAW;YACpB,OAAOE,QACH;gBACE,eAAeA,MAAM,aAAa,IAAI;gBACtC,mBAAmBA,MAAM,iBAAiB,IAAI;gBAC9C,cAAcA,MAAM,YAAY,IAAI;gBACpC,WAAWC,YAAY;gBACvB,YAAY9D;gBACZ,mBAAmBc;gBACnB,QAAQjB,YAAY,MAAM;YAC5B,IACA2E;YACJ,YAAY,CAAC,CAACd;QAChB;IACF,EAAE,OAAOpB,GAAQ;QACfD,QAAQ,KAAK,CAAC,+BAA+BC;QAC7C,MAAMmD,WAAW,IAAIhD,MACnB,CAAC,eAAe,EAAEiB,cAAc,eAAe,GAAG,kBAAkB,EAAEpB,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC/I;YACE,OAAOA;QACT;QAEF,MAAMmD;IACR;AACF;AAEO,MAAMvC,oBAAoB,CAC/BlD,WACAJ;IAIA,IAAIqD;IAKJ,IAAIjD,UAAU,QAAQ,CAAC,UACrB,OAAQJ;QACN,KAAK8F,aAAa,MAAM;YACtBzC,iBAAiB0C;YACjB;QACF,KAAKD,aAAa,eAAe;YAC/BzC,iBAAiB2C;YACjB;QACF,KAAKF,aAAa,IAAI;YACpBzC,iBAAiB4C;YACjB;QACF,KAAKH,aAAa,YAAY;QAC9B,KAAKA,aAAa,gBAAgB;YAChCzC,iBAAiB;gBAAE,MAAM6C,iBAAiB,IAAI;YAAC;YAC/C;QACF,KAAKJ,aAAa,IAAI;YAEpBzC,iBAAiBuB;YACjB;IACJ;IAKF,IACExE,AAAc,wBAAdA,aACAJ,sBAAsB8F,aAAa,IAAI,EAEvCzC,iBAAiB;QAAE,MAAM6C,iBAAiB,IAAI;IAAC;IAGjD,OAAO7C;AACT;AAEO,eAAe8C,yBACpBlD,QAAsC,EACtCjD,iBAA+B,EAC/BC,WAAyB;IAEzB,MAAMmG,WAAW,MAAMpD,OAAOC,UAAUjD,mBAAmBC;IAC3D4B,OAAOuE,UAAU;IACjB,MAAMhF,SAASnB,YAAY,MAAM;IACjC,MAAMoG,cAAcC,cAAcF,SAAS,OAAO,EAAEhF;IACpD,OAAO;QAAE,SAASiF;QAAa,OAAOD,SAAS,KAAK;IAAC;AACvD;AAEO,eAAeG,yBACpBC,IAAY,EACZxG,iBAA+B,EAC/BC,WAAyB;IAEzB,MAAM,EAAE8D,OAAO,EAAEE,KAAK,EAAE,GAAG,MAAMjB,OAAOwD,MAAMxG,mBAAmBC;IACjE,OAAO;QAAE8D;QAASE;IAAM;AAC1B;AAEO,SAASwC,yBAAyBL,QAAgB;IACvD,IAAI;QAEF,MAAMM,YAAYN,SAAS,KAAK,CAAC;QACjC,IAAIM,WACF,OAAOA,SAAS,CAAC,EAAE;QAIrB,MAAMC,iBAAiBP,SAAS,KAAK,CACnC;QAEF,IAAIO,gBACF,OAAOA,cAAc,CAAC,EAAE;QAI1B,MAAMC,gBAAgBR,SAAS,KAAK,CAAC;QACrC,IAAIQ,eACF,OAAOA,aAAa,CAAC,EAAE;IAE3B,EAAE,OAAM,CAAC;IAET,OAAOR;AACT;AAEO,SAASS,yBAAyBC,KAAa;IACpD,IAAIA,MAAM,QAAQ,CAAC,SAEjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;IAG5C,OAAOA;AACT;AAEO,SAASR,cAAcQ,KAAa,EAAE1F,MAAgC;IAC3E,MAAM2F,kBAAkBN,yBAAyBK;IAEjD,IAAIC,QAAAA,kBAAAA,KAAAA,IAAAA,gBAAiB,KAAK,CAAC,oBAAoB;YACtCC;QAAP,OAAO,QAAAA,CAAAA,yBAAAA,gBACJ,KAAK,CAAC,kBAAiB,IADnBA,KAAAA,IAAAA,uBAEH,KAAK,CAAC,GACP,GAAG,CAAC5C;IACT;IACA,IAAI;QACF,OAAOe,KAAK,KAAK,CAAC4B;IACpB,EAAE,OAAM,CAAC;IACT,IAAI;QACF,OAAO5B,KAAK,KAAK,CAAC8B,WAAWF;IAC/B,EAAE,OAAOrE,GAAG,CAAC;IAEb,IAAItB,AAAW,oBAAXA,UAA8BA,AAAW,kBAAXA,QAA0B;QAC1D,MAAM8F,aAAaL,yBAAyBE;QAC5C,OAAO5B,KAAK,KAAK,CAAC8B,WAAWC;IAC/B;IACA,MAAMrE,MAAM,CAAC,+BAA+B,EAAEiE,OAAO;AACvD"}
@@ -1 +1 @@
1
- {"version":3,"file":"types.mjs","sources":["webpack://@midscene/core/./src/types.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n\nimport type { NodeType } from '@midscene/shared/constants';\nimport type { TModelConfigFn } from '@midscene/shared/env';\nimport type {\n BaseElement,\n ElementTreeNode,\n Rect,\n Size,\n} from '@midscene/shared/types';\nimport type { z } from 'zod';\nimport type { TUserPrompt } from './ai-model/common';\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 model_description: string | undefined;\n intent: 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 abstract _isFrozen?: boolean;\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 ElementCacheFeature = Record<string, unknown>;\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 PlanningActionParamSleep {\n timeMs: number;\n}\n\nexport interface PlanningActionParamError {\n thought: string;\n}\n\nexport type PlanningActionParamWaitFor = AgentWaitForOpt & {};\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 uiContext?: 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 searchAreaUsage?: 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 },\n PlanningAIResponse\n>;\n\nexport type ExecutionTaskPlanning = ExecutionTask<ExecutionTaskPlanningApply>;\n\n/*\nGrouped dump\n*/\nexport interface GroupedActionDump {\n groupName: string;\n groupDescription?: string;\n modelBriefs: string[];\n executions: ExecutionDump[];\n}\n\nexport type InterfaceType =\n | 'puppeteer'\n | 'playwright'\n | 'static'\n | 'chrome-extension-proxy'\n | 'android'\n | string;\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 interface DeviceAction<T = any> {\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\n/**\n * Web-specific types\n */\nexport interface WebElementInfo extends BaseElement {\n id: string;\n attributes: {\n nodeType: NodeType;\n [key: string]: string;\n };\n}\n\nexport type WebUIContext = UIContext<WebElementInfo>;\n\n/**\n * Agent\n */\n\nexport type CacheConfig =\n | { strategy: 'read-only'; id: string }\n | { strategy?: undefined; id: string };\n\nexport type Cache =\n | false // No read, no write\n | true // Will throw error at runtime - deprecated\n | CacheConfig; // Object configuration (requires explicit id)\n\nexport interface AgentOpt {\n testId?: string;\n // @deprecated\n cacheId?: string; // Keep backward compatibility, but marked as deprecated\n groupName?: string;\n groupDescription?: string;\n /* if auto generate report, default true */\n generateReport?: boolean;\n /* if auto print report msg, default true */\n autoPrintReportMsg?: boolean;\n onTaskStartTip?: OnTaskStartTip;\n aiActionContext?: string;\n /* custom report file name */\n reportFileName?: string;\n modelConfig?: TModelConfigFn;\n cache?: Cache;\n replanningCycleLimit?: number;\n}\n"],"names":["AIResponseFormat","UIContext"],"mappings":";AAqCO,IAAKA,yBAAgBA,WAAAA,GAAAA,SAAhBA,gBAAgB;;;WAAhBA;;AAwFL,MAAeC;AAQtB"}
1
+ {"version":3,"file":"types.mjs","sources":["webpack://@midscene/core/./src/types.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n\nimport type { NodeType } from '@midscene/shared/constants';\nimport type { TModelConfigFn } from '@midscene/shared/env';\nimport type {\n BaseElement,\n ElementTreeNode,\n Rect,\n Size,\n} from '@midscene/shared/types';\nimport type { z } from 'zod';\nimport type { TUserPrompt } from './ai-model/common';\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 model_description: string | undefined;\n intent: 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 abstract _isFrozen?: boolean;\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 ElementCacheFeature = Record<string, unknown>;\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 PlanningActionParamSleep {\n timeMs: number;\n}\n\nexport interface PlanningActionParamError {\n thought: string;\n}\n\nexport type PlanningActionParamWaitFor = AgentWaitForOpt & {};\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 uiContext?: 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 searchAreaUsage?: 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 },\n PlanningAIResponse\n>;\n\nexport type ExecutionTaskPlanning = ExecutionTask<ExecutionTaskPlanningApply>;\n\n/*\nGrouped dump\n*/\nexport interface GroupedActionDump {\n groupName: string;\n groupDescription?: string;\n modelBriefs: string[];\n executions: ExecutionDump[];\n}\n\nexport type InterfaceType =\n | 'puppeteer'\n | 'playwright'\n | 'static'\n | 'chrome-extension-proxy'\n | 'android'\n | string;\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 interface DeviceAction<T = any> {\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\n/**\n * Web-specific types\n */\nexport interface WebElementInfo extends BaseElement {\n id: string;\n attributes: {\n nodeType: NodeType;\n [key: string]: string;\n };\n}\n\nexport type WebUIContext = UIContext<WebElementInfo>;\n\n/**\n * Agent\n */\n\nexport type CacheConfig = { strategy?: 'read-only' | 'read-write'; id: string };\n\nexport type Cache =\n | false // No read, no write\n | true // Will throw error at runtime - deprecated\n | CacheConfig; // Object configuration (requires explicit id)\n\nexport interface AgentOpt {\n testId?: string;\n // @deprecated\n cacheId?: string; // Keep backward compatibility, but marked as deprecated\n groupName?: string;\n groupDescription?: string;\n /* if auto generate report, default true */\n generateReport?: boolean;\n /* if auto print report msg, default true */\n autoPrintReportMsg?: boolean;\n onTaskStartTip?: OnTaskStartTip;\n aiActionContext?: string;\n /* custom report file name */\n reportFileName?: string;\n modelConfig?: TModelConfigFn;\n cache?: Cache;\n replanningCycleLimit?: number;\n}\n"],"names":["AIResponseFormat","UIContext"],"mappings":";AAqCO,IAAKA,yBAAgBA,WAAAA,GAAAA,SAAhBA,gBAAgB;;;WAAhBA;;AAwFL,MAAeC;AAQtB"}