@midscene/core 1.2.2-beta-20260115092052.0 → 1.2.2-beta-20260116060040.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/es/agent/agent.mjs +27 -24
- package/dist/es/agent/agent.mjs.map +1 -1
- package/dist/es/agent/tasks.mjs +15 -29
- package/dist/es/agent/tasks.mjs.map +1 -1
- package/dist/es/agent/utils.mjs +4 -2
- package/dist/es/agent/utils.mjs.map +1 -1
- package/dist/es/ai-model/inspect.mjs +3 -3
- package/dist/es/ai-model/inspect.mjs.map +1 -1
- package/dist/es/ai-model/llm-planning.mjs +14 -4
- package/dist/es/ai-model/llm-planning.mjs.map +1 -1
- package/dist/es/ai-model/prompt/llm-planning.mjs +44 -9
- package/dist/es/ai-model/prompt/llm-planning.mjs.map +1 -1
- package/dist/es/ai-model/service-caller/index.mjs +11 -10
- package/dist/es/ai-model/service-caller/index.mjs.map +1 -1
- package/dist/es/ai-model/ui-tars-planning.mjs +3 -2
- package/dist/es/ai-model/ui-tars-planning.mjs.map +1 -1
- package/dist/es/common.mjs +14 -5
- package/dist/es/common.mjs.map +1 -1
- package/dist/es/device/index.mjs +29 -4
- package/dist/es/device/index.mjs.map +1 -1
- package/dist/es/dump/html-utils.mjs +42 -0
- package/dist/es/dump/html-utils.mjs.map +1 -0
- package/dist/es/dump/image-restoration.mjs +31 -0
- package/dist/es/dump/image-restoration.mjs.map +1 -0
- package/dist/es/dump/index.mjs +3 -0
- package/dist/es/index.mjs +4 -2
- package/dist/es/index.mjs.map +1 -1
- package/dist/es/screenshot-item.mjs +34 -0
- package/dist/es/screenshot-item.mjs.map +1 -0
- package/dist/es/service/index.mjs +2 -1
- package/dist/es/service/index.mjs.map +1 -1
- package/dist/es/task-runner.mjs +7 -9
- package/dist/es/task-runner.mjs.map +1 -1
- package/dist/es/types.mjs +78 -1
- package/dist/es/types.mjs.map +1 -1
- package/dist/es/utils.mjs +3 -2
- package/dist/es/utils.mjs.map +1 -1
- package/dist/lib/agent/agent.js +24 -21
- package/dist/lib/agent/agent.js.map +1 -1
- package/dist/lib/agent/tasks.js +15 -29
- package/dist/lib/agent/tasks.js.map +1 -1
- package/dist/lib/agent/utils.js +4 -2
- package/dist/lib/agent/utils.js.map +1 -1
- package/dist/lib/ai-model/inspect.js +3 -3
- package/dist/lib/ai-model/inspect.js.map +1 -1
- package/dist/lib/ai-model/llm-planning.js +13 -3
- package/dist/lib/ai-model/llm-planning.js.map +1 -1
- package/dist/lib/ai-model/prompt/llm-planning.js +44 -9
- package/dist/lib/ai-model/prompt/llm-planning.js.map +1 -1
- package/dist/lib/ai-model/service-caller/index.js +11 -10
- package/dist/lib/ai-model/service-caller/index.js.map +1 -1
- package/dist/lib/ai-model/ui-tars-planning.js +3 -2
- package/dist/lib/ai-model/ui-tars-planning.js.map +1 -1
- package/dist/lib/common.js +20 -5
- package/dist/lib/common.js.map +1 -1
- package/dist/lib/device/index.js +53 -16
- package/dist/lib/device/index.js.map +1 -1
- package/dist/lib/dump/html-utils.js +94 -0
- package/dist/lib/dump/html-utils.js.map +1 -0
- package/dist/lib/dump/image-restoration.js +65 -0
- package/dist/lib/dump/image-restoration.js.map +1 -0
- package/dist/lib/dump/index.js +60 -0
- package/dist/lib/dump/index.js.map +1 -0
- package/dist/lib/index.js +42 -7
- package/dist/lib/index.js.map +1 -1
- package/dist/lib/screenshot-item.js +68 -0
- package/dist/lib/screenshot-item.js.map +1 -0
- package/dist/lib/service/index.js +2 -1
- package/dist/lib/service/index.js.map +1 -1
- package/dist/lib/task-runner.js +7 -9
- package/dist/lib/task-runner.js.map +1 -1
- package/dist/lib/types.js +91 -3
- package/dist/lib/types.js.map +1 -1
- package/dist/lib/utils.js +3 -2
- package/dist/lib/utils.js.map +1 -1
- package/dist/types/agent/agent.d.ts +5 -16
- package/dist/types/agent/tasks.d.ts +4 -4
- package/dist/types/ai-model/service-caller/index.d.ts +3 -3
- package/dist/types/common.d.ts +8 -1
- package/dist/types/device/index.d.ts +85 -23
- package/dist/types/dump/html-utils.d.ts +7 -0
- package/dist/types/dump/image-restoration.d.ts +6 -0
- package/dist/types/dump/index.d.ts +5 -0
- package/dist/types/index.d.ts +3 -1
- package/dist/types/screenshot-item.d.ts +27 -0
- package/dist/types/task-runner.d.ts +1 -1
- package/dist/types/types.d.ts +61 -6
- package/package.json +2 -2
package/dist/lib/agent/tasks.js
CHANGED
|
@@ -30,6 +30,7 @@ __webpack_require__.d(__webpack_exports__, {
|
|
|
30
30
|
withFileChooser: ()=>withFileChooser
|
|
31
31
|
});
|
|
32
32
|
const index_js_namespaceObject = require("../ai-model/index.js");
|
|
33
|
+
const external_common_js_namespaceObject = require("../common.js");
|
|
33
34
|
const external_task_runner_js_namespaceObject = require("../task-runner.js");
|
|
34
35
|
const external_types_js_namespaceObject = require("../types.js");
|
|
35
36
|
const logger_namespaceObject = require("@midscene/shared/logger");
|
|
@@ -82,7 +83,7 @@ class TaskExecutor {
|
|
|
82
83
|
return {
|
|
83
84
|
output: {
|
|
84
85
|
actions: [],
|
|
85
|
-
|
|
86
|
+
shouldContinuePlanning: false,
|
|
86
87
|
log: '',
|
|
87
88
|
yamlString
|
|
88
89
|
},
|
|
@@ -127,6 +128,7 @@ class TaskExecutor {
|
|
|
127
128
|
const replanningCycleLimit = replanningCycleLimitOverride ?? this.replanningCycleLimit;
|
|
128
129
|
(0, utils_namespaceObject.assert)(void 0 !== replanningCycleLimit, 'replanningCycleLimit is required for TaskExecutor.action');
|
|
129
130
|
let errorCountInOnePlanningLoop = 0;
|
|
131
|
+
let outputString;
|
|
130
132
|
while(true){
|
|
131
133
|
const result = await session.appendAndRun({
|
|
132
134
|
type: 'Planning',
|
|
@@ -138,7 +140,6 @@ class TaskExecutor {
|
|
|
138
140
|
deepThink
|
|
139
141
|
},
|
|
140
142
|
executor: async (param, executorContext)=>{
|
|
141
|
-
const startTime = Date.now();
|
|
142
143
|
const { uiContext } = executorContext;
|
|
143
144
|
(0, utils_namespaceObject.assert)(uiContext, 'uiContext is required for Planning task');
|
|
144
145
|
const { vlMode } = modelConfigForPlanning;
|
|
@@ -159,7 +160,7 @@ class TaskExecutor {
|
|
|
159
160
|
deepThink
|
|
160
161
|
});
|
|
161
162
|
debug('planResult', JSON.stringify(planResult, null, 2));
|
|
162
|
-
const { actions, log,
|
|
163
|
+
const { actions, log, note, error, usage, rawResponse, reasoning_content } = planResult;
|
|
163
164
|
executorContext.task.log = {
|
|
164
165
|
...executorContext.task.log || {},
|
|
165
166
|
rawResponse
|
|
@@ -168,19 +169,13 @@ class TaskExecutor {
|
|
|
168
169
|
executorContext.task.reasoning_content = reasoning_content;
|
|
169
170
|
executorContext.task.output = {
|
|
170
171
|
actions: actions || [],
|
|
171
|
-
more_actions_needed_by_instruction,
|
|
172
172
|
log,
|
|
173
|
-
|
|
173
|
+
note,
|
|
174
|
+
yamlFlow: planResult.yamlFlow,
|
|
175
|
+
output: outputString,
|
|
176
|
+
shouldContinuePlanning: planResult.shouldContinuePlanning
|
|
174
177
|
};
|
|
175
178
|
executorContext.uiContext = uiContext;
|
|
176
|
-
const finalActions = [
|
|
177
|
-
...actions || []
|
|
178
|
-
];
|
|
179
|
-
if (sleep) {
|
|
180
|
-
const timeNow = Date.now();
|
|
181
|
-
const timeRemaining = sleep - (timeNow - startTime);
|
|
182
|
-
if (timeRemaining > 0) finalActions.push(this.sleepPlan(timeRemaining));
|
|
183
|
-
}
|
|
184
179
|
(0, utils_namespaceObject.assert)(!error, `Failed to continue: ${error}\n${log || ''}`);
|
|
185
180
|
return {
|
|
186
181
|
cache: {
|
|
@@ -204,28 +199,27 @@ class TaskExecutor {
|
|
|
204
199
|
return session.appendErrorPlan(`Error converting plans to executable tasks: ${error}, plans: ${JSON.stringify(plans)}`);
|
|
205
200
|
}
|
|
206
201
|
if (this.conversationHistory.pendingFeedbackMessage) console.warn('unconsumed pending feedback message detected, this may lead to unexpected planning result:', this.conversationHistory.pendingFeedbackMessage);
|
|
207
|
-
let errorFlag = false;
|
|
208
202
|
try {
|
|
209
|
-
await session.appendAndRun(executables.tasks);
|
|
203
|
+
const result = await session.appendAndRun(executables.tasks);
|
|
204
|
+
outputString = result?.output;
|
|
210
205
|
} catch (error) {
|
|
211
|
-
errorFlag = true;
|
|
212
206
|
errorCountInOnePlanningLoop++;
|
|
213
|
-
this.conversationHistory.pendingFeedbackMessage = `Error executing running tasks: ${error?.message || String(error)}`;
|
|
207
|
+
this.conversationHistory.pendingFeedbackMessage = `Time: ${(0, external_common_js_namespaceObject.getReadableTimeString)()}, Error executing running tasks: ${error?.message || String(error)}`;
|
|
214
208
|
debug('error when executing running tasks, but continue to run if it is not too many errors:', error instanceof Error ? error.message : String(error), 'current error count in one planning loop:', errorCountInOnePlanningLoop);
|
|
215
209
|
}
|
|
216
210
|
if (errorCountInOnePlanningLoop > maxErrorCountAllowedInOnePlanningLoop) return session.appendErrorPlan('Too many errors in one planning loop');
|
|
217
|
-
if (!planResult?.
|
|
218
|
-
else break;
|
|
211
|
+
if (!planResult?.shouldContinuePlanning) break;
|
|
219
212
|
++replanCount;
|
|
220
213
|
if (replanCount > replanningCycleLimit) {
|
|
221
214
|
const errorMsg = `Replanned ${replanningCycleLimit} times, exceeding the limit. Please configure a larger value for replanningCycleLimit (or use MIDSCENE_REPLANNING_CYCLE_LIMIT) to handle more complex tasks.`;
|
|
222
215
|
return session.appendErrorPlan(errorMsg);
|
|
223
216
|
}
|
|
224
|
-
if (!this.conversationHistory.pendingFeedbackMessage) this.conversationHistory.pendingFeedbackMessage =
|
|
217
|
+
if (!this.conversationHistory.pendingFeedbackMessage) this.conversationHistory.pendingFeedbackMessage = `Time: ${(0, external_common_js_namespaceObject.getReadableTimeString)()}, I have finished the action previously planned.`;
|
|
225
218
|
}
|
|
226
219
|
return {
|
|
227
220
|
output: {
|
|
228
|
-
yamlFlow
|
|
221
|
+
yamlFlow,
|
|
222
|
+
output: outputString
|
|
229
223
|
},
|
|
230
224
|
runner
|
|
231
225
|
};
|
|
@@ -315,14 +309,6 @@ class TaskExecutor {
|
|
|
315
309
|
runner
|
|
316
310
|
};
|
|
317
311
|
}
|
|
318
|
-
sleepPlan(timeMs) {
|
|
319
|
-
return {
|
|
320
|
-
type: 'Sleep',
|
|
321
|
-
param: {
|
|
322
|
-
timeMs
|
|
323
|
-
}
|
|
324
|
-
};
|
|
325
|
-
}
|
|
326
312
|
async taskForSleep(timeMs, _modelConfig) {
|
|
327
313
|
return this.taskBuilder.createSleepTask({
|
|
328
314
|
timeMs
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"agent/tasks.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../src/agent/tasks.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import { ConversationHistory, plan, uiTarsPlanning } from '@/ai-model';\nimport type { TMultimodalPrompt, TUserPrompt } from '@/common';\nimport type { AbstractInterface, FileChooserHandler } from '@/device';\nimport type Service from '@/service';\nimport type { TaskRunner } from '@/task-runner';\nimport { TaskExecutionError } from '@/task-runner';\nimport type {\n DeepThinkOption,\n DeviceAction,\n ExecutionTaskApply,\n ExecutionTaskInsightQueryApply,\n ExecutionTaskPlanningApply,\n ExecutionTaskProgressOptions,\n InterfaceType,\n MidsceneYamlFlowItem,\n PlanningAIResponse,\n PlanningAction,\n PlanningActionParamSleep,\n PlanningActionParamWaitFor,\n ServiceDump,\n ServiceExtractOption,\n ServiceExtractParam,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { ExecutionSession } from './execution-session';\nimport { TaskBuilder } from './task-builder';\nimport type { TaskCache } from './task-cache';\nexport { locatePlanForLocate } from './task-builder';\nimport { descriptionOfTree } from '@midscene/shared/extractor';\nimport { taskTitleStr } from './ui-utils';\nimport { parsePrompt } from './utils';\n\ninterface ExecutionResult<OutputType = any> {\n output: OutputType;\n thought?: string;\n runner: TaskRunner;\n}\n\ninterface TaskExecutorHooks {\n onTaskUpdate?: (\n runner: TaskRunner,\n error?: TaskExecutionError,\n ) => Promise<void> | void;\n}\n\nconst debug = getDebug('device-task-executor');\nconst maxErrorCountAllowedInOnePlanningLoop = 5;\n\nexport { TaskExecutionError };\n\nexport class TaskExecutor {\n interface: AbstractInterface;\n\n service: Service;\n\n taskCache?: TaskCache;\n\n private readonly providedActionSpace: DeviceAction[];\n\n private readonly taskBuilder: TaskBuilder;\n\n private conversationHistory: ConversationHistory;\n\n onTaskStartCallback?: ExecutionTaskProgressOptions['onTaskStart'];\n\n private readonly hooks?: TaskExecutorHooks;\n\n replanningCycleLimit?: number;\n\n // @deprecated use .interface instead\n get page() {\n return this.interface;\n }\n\n constructor(\n interfaceInstance: AbstractInterface,\n service: Service,\n opts: {\n taskCache?: TaskCache;\n onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n replanningCycleLimit?: number;\n hooks?: TaskExecutorHooks;\n actionSpace: DeviceAction[];\n },\n ) {\n this.interface = interfaceInstance;\n this.service = service;\n this.taskCache = opts.taskCache;\n this.onTaskStartCallback = opts?.onTaskStart;\n this.replanningCycleLimit = opts.replanningCycleLimit;\n this.hooks = opts.hooks;\n this.conversationHistory = new ConversationHistory();\n this.providedActionSpace = opts.actionSpace;\n this.taskBuilder = new TaskBuilder({\n interfaceInstance,\n service,\n taskCache: opts.taskCache,\n actionSpace: this.getActionSpace(),\n });\n }\n\n private createExecutionSession(\n title: string,\n options?: { tasks?: ExecutionTaskApply[] },\n ) {\n return new ExecutionSession(\n title,\n () => Promise.resolve(this.service.contextRetrieverFn()),\n {\n onTaskStart: this.onTaskStartCallback,\n tasks: options?.tasks,\n onTaskUpdate: this.hooks?.onTaskUpdate,\n },\n );\n }\n\n private getActionSpace(): DeviceAction[] {\n return this.providedActionSpace;\n }\n\n public async convertPlanToExecutable(\n plans: PlanningAction[],\n modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n options?: {\n cacheable?: boolean;\n subTask?: boolean;\n },\n ) {\n return this.taskBuilder.build(\n plans,\n modelConfigForPlanning,\n modelConfigForDefaultIntent,\n options,\n );\n }\n\n async loadYamlFlowAsPlanning(userInstruction: string, yamlString: string) {\n const session = this.createExecutionSession(\n taskTitleStr('Action', userInstruction),\n );\n\n const task: ExecutionTaskPlanningApply = {\n type: 'Planning',\n subType: 'LoadYaml',\n param: {\n userInstruction,\n },\n executor: async (param, executorContext) => {\n const { uiContext } = executorContext;\n assert(uiContext, 'uiContext is required for Planning task');\n return {\n output: {\n actions: [],\n more_actions_needed_by_instruction: false,\n log: '',\n yamlString,\n },\n cache: {\n hit: true,\n },\n hitBy: {\n from: 'Cache',\n context: {\n yamlString,\n },\n },\n };\n },\n };\n const runner = session.getRunner();\n await session.appendAndRun(task);\n\n return {\n runner,\n };\n }\n\n async runPlans(\n title: string,\n plans: PlanningAction[],\n modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n ): Promise<ExecutionResult> {\n const session = this.createExecutionSession(title);\n const { tasks } = await this.convertPlanToExecutable(\n plans,\n modelConfigForPlanning,\n modelConfigForDefaultIntent,\n );\n const runner = session.getRunner();\n const result = await session.appendAndRun(tasks);\n const { output } = result ?? {};\n return {\n output,\n runner,\n };\n }\n\n async action(\n userPrompt: string,\n modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n includeBboxInPlanning: boolean,\n aiActContext?: string,\n cacheable?: boolean,\n replanningCycleLimitOverride?: number,\n imagesIncludeCount?: number,\n deepThink?: DeepThinkOption,\n fileChooserAccept?: string[],\n ): Promise<\n ExecutionResult<\n | {\n yamlFlow?: MidsceneYamlFlowItem[]; // for cache use\n }\n | undefined\n >\n > {\n return withFileChooser(this.interface, fileChooserAccept, async () => {\n return this.runAction(\n userPrompt,\n modelConfigForPlanning,\n modelConfigForDefaultIntent,\n includeBboxInPlanning,\n aiActContext,\n cacheable,\n replanningCycleLimitOverride,\n imagesIncludeCount,\n deepThink,\n );\n });\n }\n\n private async runAction(\n userPrompt: string,\n modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n includeBboxInPlanning: boolean,\n aiActContext?: string,\n cacheable?: boolean,\n replanningCycleLimitOverride?: number,\n imagesIncludeCount?: number,\n deepThink?: DeepThinkOption,\n ): Promise<\n ExecutionResult<\n | {\n yamlFlow?: MidsceneYamlFlowItem[]; // for cache use\n }\n | undefined\n >\n > {\n this.conversationHistory.reset();\n\n const session = this.createExecutionSession(\n taskTitleStr('Action', userPrompt),\n );\n const runner = session.getRunner();\n\n let replanCount = 0;\n const yamlFlow: MidsceneYamlFlowItem[] = [];\n const replanningCycleLimit =\n replanningCycleLimitOverride ?? this.replanningCycleLimit;\n assert(\n replanningCycleLimit !== undefined,\n 'replanningCycleLimit is required for TaskExecutor.action',\n );\n\n let errorCountInOnePlanningLoop = 0; // count the number of errors in one planning loop\n\n // Main planning loop - unified plan/replan logic\n while (true) {\n const result = await session.appendAndRun(\n {\n type: 'Planning',\n subType: 'Plan',\n param: {\n userInstruction: userPrompt,\n aiActContext,\n imagesIncludeCount,\n deepThink,\n },\n executor: async (param, executorContext) => {\n const startTime = Date.now();\n const { uiContext } = executorContext;\n assert(uiContext, 'uiContext is required for Planning task');\n const { vlMode } = modelConfigForPlanning;\n const uiTarsModelVersion =\n vlMode === 'vlm-ui-tars'\n ? modelConfigForPlanning.uiTarsModelVersion\n : undefined;\n\n const actionSpace = this.getActionSpace();\n debug(\n 'actionSpace for this interface is:',\n actionSpace.map((action) => action.name).join(', '),\n );\n assert(Array.isArray(actionSpace), 'actionSpace must be an array');\n if (actionSpace.length === 0) {\n console.warn(\n `ActionSpace for ${this.interface.interfaceType} is empty. This may lead to unexpected behavior.`,\n );\n }\n\n const planResult = await (uiTarsModelVersion\n ? uiTarsPlanning\n : plan)(param.userInstruction, {\n context: uiContext,\n actionContext: param.aiActContext,\n interfaceType: this.interface.interfaceType as InterfaceType,\n actionSpace,\n modelConfig: modelConfigForPlanning,\n conversationHistory: this.conversationHistory,\n includeBbox: includeBboxInPlanning,\n imagesIncludeCount,\n deepThink,\n });\n debug('planResult', JSON.stringify(planResult, null, 2));\n\n const {\n actions,\n log,\n more_actions_needed_by_instruction,\n error,\n usage,\n rawResponse,\n sleep,\n reasoning_content,\n } = planResult;\n\n executorContext.task.log = {\n ...(executorContext.task.log || {}),\n rawResponse,\n };\n executorContext.task.usage = usage;\n executorContext.task.reasoning_content = reasoning_content;\n executorContext.task.output = {\n actions: actions || [],\n more_actions_needed_by_instruction,\n log,\n yamlFlow: planResult.yamlFlow,\n };\n executorContext.uiContext = uiContext;\n\n const finalActions = [...(actions || [])];\n\n if (sleep) {\n const timeNow = Date.now();\n const timeRemaining = sleep - (timeNow - startTime);\n if (timeRemaining > 0) {\n finalActions.push(this.sleepPlan(timeRemaining));\n }\n }\n\n assert(!error, `Failed to continue: ${error}\\n${log || ''}`);\n\n return {\n cache: {\n hit: false,\n },\n } as any;\n },\n },\n {\n allowWhenError: true,\n },\n );\n\n const planResult = result?.output as PlanningAIResponse | undefined;\n\n // Execute planned actions\n const plans = planResult?.actions || [];\n yamlFlow.push(...(planResult?.yamlFlow || []));\n\n let executables: Awaited<ReturnType<typeof this.convertPlanToExecutable>>;\n try {\n executables = await this.convertPlanToExecutable(\n plans,\n modelConfigForPlanning,\n modelConfigForDefaultIntent,\n {\n cacheable,\n subTask: true,\n },\n );\n } catch (error) {\n return session.appendErrorPlan(\n `Error converting plans to executable tasks: ${error}, plans: ${JSON.stringify(\n plans,\n )}`,\n );\n }\n if (this.conversationHistory.pendingFeedbackMessage) {\n console.warn(\n 'unconsumed pending feedback message detected, this may lead to unexpected planning result:',\n this.conversationHistory.pendingFeedbackMessage,\n );\n }\n let errorFlag = false;\n try {\n await session.appendAndRun(executables.tasks);\n } catch (error: any) {\n errorFlag = true;\n errorCountInOnePlanningLoop++;\n this.conversationHistory.pendingFeedbackMessage = `Error executing running tasks: ${error?.message || String(error)}`;\n debug(\n 'error when executing running tasks, but continue to run if it is not too many errors:',\n error instanceof Error ? error.message : String(error),\n 'current error count in one planning loop:',\n errorCountInOnePlanningLoop,\n );\n }\n\n if (errorCountInOnePlanningLoop > maxErrorCountAllowedInOnePlanningLoop) {\n return session.appendErrorPlan('Too many errors in one planning loop');\n }\n\n // Check if task is complete\n if (!planResult?.more_actions_needed_by_instruction) {\n if (errorFlag) {\n debug(\n 'more_actions_needed_by_instruction is false, but there are errors in one planning loop, continue to run',\n );\n } else {\n break;\n }\n }\n\n // Increment replan count for next iteration\n ++replanCount;\n\n if (replanCount > replanningCycleLimit) {\n const errorMsg = `Replanned ${replanningCycleLimit} times, exceeding the limit. Please configure a larger value for replanningCycleLimit (or use MIDSCENE_REPLANNING_CYCLE_LIMIT) to handle more complex tasks.`;\n return session.appendErrorPlan(errorMsg);\n }\n\n if (!this.conversationHistory.pendingFeedbackMessage) {\n this.conversationHistory.pendingFeedbackMessage =\n 'I have finished the action previously planned.';\n }\n }\n\n return {\n output: {\n yamlFlow,\n },\n runner,\n };\n }\n\n private createTypeQueryTask(\n type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert' | 'WaitFor',\n demand: ServiceExtractParam,\n modelConfig: IModelConfig,\n opt?: ServiceExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ) {\n const queryTask: ExecutionTaskInsightQueryApply = {\n type: 'Insight',\n subType: type,\n param: {\n dataDemand: multimodalPrompt\n ? ({\n demand,\n multimodalPrompt,\n } as never)\n : demand, // for user param presentation in report right sidebar\n },\n executor: async (param, taskContext) => {\n const { task } = taskContext;\n let queryDump: ServiceDump | undefined;\n const applyDump = (dump: ServiceDump) => {\n queryDump = dump;\n task.log = {\n dump,\n };\n };\n\n // Get context for query operations\n const uiContext = taskContext.uiContext;\n assert(uiContext, 'uiContext is required for Query task');\n\n const ifTypeRestricted = type !== 'Query';\n let demandInput = demand;\n let keyOfResult = 'result';\n if (ifTypeRestricted && (type === 'Assert' || type === 'WaitFor')) {\n keyOfResult = 'StatementIsTruthy';\n const booleanPrompt =\n type === 'Assert'\n ? `Boolean, whether the following statement is true: ${demand}`\n : `Boolean, the user wants to do some 'wait for' operation, please check whether the following statement is true: ${demand}`;\n demandInput = {\n [keyOfResult]: booleanPrompt,\n };\n } else if (ifTypeRestricted) {\n demandInput = {\n [keyOfResult]: `${type}, ${demand}`,\n };\n }\n\n let extractResult;\n\n let extraPageDescription = '';\n if (opt?.domIncluded && this.interface.getElementsNodeTree) {\n debug('appending tree info for page');\n const tree = await this.interface.getElementsNodeTree();\n extraPageDescription = await descriptionOfTree(\n tree,\n 200,\n false,\n opt?.domIncluded === 'visible-only',\n );\n }\n\n try {\n extractResult = await this.service.extract<any>(\n demandInput,\n modelConfig,\n opt,\n extraPageDescription,\n multimodalPrompt,\n );\n } catch (error) {\n if (error instanceof ServiceError) {\n applyDump(error.dump);\n }\n throw error;\n }\n\n const { data, usage, thought, dump, reasoning_content } = extractResult;\n applyDump(dump);\n task.reasoning_content = reasoning_content;\n\n let outputResult = data;\n if (ifTypeRestricted) {\n // If AI returned a plain string instead of structured format, use it directly\n if (typeof data === 'string') {\n outputResult = data;\n } else if (type === 'WaitFor') {\n if (data === null || data === undefined) {\n outputResult = false;\n } else {\n outputResult = (data as any)[keyOfResult];\n }\n } else if (data === null || data === undefined) {\n outputResult = null;\n } else {\n assert(\n data?.[keyOfResult] !== undefined,\n 'No result in query data',\n );\n outputResult = (data as any)[keyOfResult];\n }\n }\n\n if (type === 'Assert' && !outputResult) {\n task.usage = usage;\n task.thought = thought;\n throw new Error(`Assertion failed: ${thought}`);\n }\n\n return {\n output: outputResult,\n log: queryDump,\n usage,\n thought,\n };\n },\n };\n\n return queryTask;\n }\n async createTypeQueryExecution<T>(\n type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert',\n demand: ServiceExtractParam,\n modelConfig: IModelConfig,\n opt?: ServiceExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ): Promise<ExecutionResult<T>> {\n const session = this.createExecutionSession(\n taskTitleStr(\n type,\n typeof demand === 'string' ? demand : JSON.stringify(demand),\n ),\n );\n\n const queryTask = await this.createTypeQueryTask(\n type,\n demand,\n modelConfig,\n opt,\n multimodalPrompt,\n );\n\n const runner = session.getRunner();\n const result = await session.appendAndRun(queryTask);\n\n if (!result) {\n throw new Error(\n 'result of taskExecutor.flush() is undefined in function createTypeQueryTask',\n );\n }\n\n const { output, thought } = result;\n\n return {\n output,\n thought,\n runner,\n };\n }\n\n private sleepPlan(timeMs: number): PlanningAction<PlanningActionParamSleep> {\n return {\n type: 'Sleep',\n param: {\n timeMs,\n },\n };\n }\n\n async taskForSleep(timeMs: number, _modelConfig: IModelConfig) {\n return this.taskBuilder.createSleepTask({\n timeMs,\n });\n }\n\n async waitFor(\n assertion: TUserPrompt,\n opt: PlanningActionParamWaitFor,\n modelConfig: IModelConfig,\n ): Promise<ExecutionResult<void>> {\n const { textPrompt, multimodalPrompt } = parsePrompt(assertion);\n\n const description = `waitFor: ${textPrompt}`;\n const session = this.createExecutionSession(\n taskTitleStr('WaitFor', description),\n );\n const runner = session.getRunner();\n const {\n timeoutMs,\n checkIntervalMs,\n domIncluded,\n screenshotIncluded,\n ...restOpt\n } = opt;\n const serviceExtractOpt: ServiceExtractOption = {\n domIncluded,\n screenshotIncluded,\n ...restOpt,\n };\n\n assert(assertion, 'No assertion for waitFor');\n assert(timeoutMs, 'No timeoutMs for waitFor');\n assert(checkIntervalMs, 'No checkIntervalMs for waitFor');\n\n assert(\n checkIntervalMs <= timeoutMs,\n `wrong config for waitFor: checkIntervalMs must be less than timeoutMs, config: {checkIntervalMs: ${checkIntervalMs}, timeoutMs: ${timeoutMs}}`,\n );\n\n const overallStartTime = Date.now();\n let lastCheckStart = overallStartTime;\n let errorThought = '';\n // Continue checking as long as the previous iteration began within the timeout window.\n while (lastCheckStart - overallStartTime <= timeoutMs) {\n const currentCheckStart = Date.now();\n lastCheckStart = currentCheckStart;\n const queryTask = await this.createTypeQueryTask(\n 'WaitFor',\n textPrompt,\n modelConfig,\n serviceExtractOpt,\n multimodalPrompt,\n );\n\n const result = (await session.appendAndRun(queryTask)) as\n | {\n output: boolean;\n thought?: string;\n }\n | undefined;\n\n if (result?.output) {\n return {\n output: undefined,\n runner,\n };\n }\n\n errorThought =\n result?.thought ||\n (!result && `No result from assertion: ${textPrompt}`) ||\n `unknown error when waiting for assertion: ${textPrompt}`;\n const now = Date.now();\n if (now - currentCheckStart < checkIntervalMs) {\n const timeRemaining = checkIntervalMs - (now - currentCheckStart);\n const sleepTask = this.taskBuilder.createSleepTask({\n timeMs: timeRemaining,\n });\n await session.append(sleepTask);\n }\n }\n\n return session.appendErrorPlan(`waitFor timeout: ${errorThought}`);\n }\n}\n\nexport async function withFileChooser<T>(\n interfaceInstance: AbstractInterface,\n fileChooserAccept: string[] | undefined,\n action: () => Promise<T>,\n): Promise<T> {\n if (!fileChooserAccept?.length) {\n return action();\n }\n\n if (!interfaceInstance.registerFileChooserListener) {\n throw new Error(\n `File upload is not supported on ${interfaceInstance.interfaceType}`,\n );\n }\n\n const handler = async (chooser: FileChooserHandler) => {\n await chooser.accept(fileChooserAccept);\n };\n\n const { dispose, getError } =\n await interfaceInstance.registerFileChooserListener(handler);\n try {\n const result = await action();\n // Check for errors that occurred during file chooser handling\n const error = getError();\n if (error) {\n throw error;\n }\n return result;\n } finally {\n dispose();\n }\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","maxErrorCountAllowedInOnePlanningLoop","TaskExecutor","title","options","ExecutionSession","Promise","plans","modelConfigForPlanning","modelConfigForDefaultIntent","userInstruction","yamlString","session","taskTitleStr","task","param","executorContext","uiContext","assert","runner","tasks","result","output","userPrompt","includeBboxInPlanning","aiActContext","cacheable","replanningCycleLimitOverride","imagesIncludeCount","deepThink","fileChooserAccept","withFileChooser","replanCount","yamlFlow","replanningCycleLimit","undefined","errorCountInOnePlanningLoop","startTime","Date","vlMode","uiTarsModelVersion","actionSpace","action","Array","console","planResult","uiTarsPlanning","plan","JSON","actions","log","more_actions_needed_by_instruction","error","usage","rawResponse","sleep","reasoning_content","finalActions","timeNow","timeRemaining","executables","errorFlag","String","Error","errorMsg","type","demand","modelConfig","opt","multimodalPrompt","queryTask","taskContext","queryDump","applyDump","dump","ifTypeRestricted","demandInput","keyOfResult","booleanPrompt","extractResult","extraPageDescription","tree","descriptionOfTree","ServiceError","data","thought","outputResult","timeMs","_modelConfig","assertion","textPrompt","parsePrompt","description","timeoutMs","checkIntervalMs","domIncluded","screenshotIncluded","restOpt","serviceExtractOpt","overallStartTime","lastCheckStart","errorThought","currentCheckStart","now","sleepTask","interfaceInstance","service","opts","ConversationHistory","TaskBuilder","handler","chooser","dispose","getError"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AC0CA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AACvB,MAAMC,wCAAwC;AAIvC,MAAMC;IAoBX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IA6BQ,uBACNC,KAAa,EACbC,OAA0C,EAC1C;QACA,OAAO,IAAIC,8CAAAA,gBAAgBA,CACzBF,OACA,IAAMG,QAAQ,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,kBAAkB,KACrD;YACE,aAAa,IAAI,CAAC,mBAAmB;YACrC,OAAOF,SAAS;YAChB,cAAc,IAAI,CAAC,KAAK,EAAE;QAC5B;IAEJ;IAEQ,iBAAiC;QACvC,OAAO,IAAI,CAAC,mBAAmB;IACjC;IAEA,MAAa,wBACXG,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACzCL,OAGC,EACD;QACA,OAAO,IAAI,CAAC,WAAW,CAAC,KAAK,CAC3BG,OACAC,wBACAC,6BACAL;IAEJ;IAEA,MAAM,uBAAuBM,eAAuB,EAAEC,UAAkB,EAAE;QACxE,MAAMC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,AAAAA,IAAAA,qCAAAA,YAAAA,AAAAA,EAAa,UAAUH;QAGzB,MAAMI,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,OAAO;gBACLJ;YACF;YACA,UAAU,OAAOK,OAAOC;gBACtB,MAAM,EAAEC,SAAS,EAAE,GAAGD;gBACtBE,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,WAAW;gBAClB,OAAO;oBACL,QAAQ;wBACN,SAAS,EAAE;wBACX,oCAAoC;wBACpC,KAAK;wBACLN;oBACF;oBACA,OAAO;wBACL,KAAK;oBACP;oBACA,OAAO;wBACL,MAAM;wBACN,SAAS;4BACPA;wBACF;oBACF;gBACF;YACF;QACF;QACA,MAAMQ,SAASP,QAAQ,SAAS;QAChC,MAAMA,QAAQ,YAAY,CAACE;QAE3B,OAAO;YACLK;QACF;IACF;IAEA,MAAM,SACJhB,KAAa,EACbI,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACf;QAC1B,MAAMG,UAAU,IAAI,CAAC,sBAAsB,CAACT;QAC5C,MAAM,EAAEiB,KAAK,EAAE,GAAG,MAAM,IAAI,CAAC,uBAAuB,CAClDb,OACAC,wBACAC;QAEF,MAAMU,SAASP,QAAQ,SAAS;QAChC,MAAMS,SAAS,MAAMT,QAAQ,YAAY,CAACQ;QAC1C,MAAM,EAAEE,MAAM,EAAE,GAAGD,UAAU,CAAC;QAC9B,OAAO;YACLC;YACAH;QACF;IACF;IAEA,MAAM,OACJI,UAAkB,EAClBf,sBAAoC,EACpCC,2BAAyC,EACzCe,qBAA8B,EAC9BC,YAAqB,EACrBC,SAAmB,EACnBC,4BAAqC,EACrCC,kBAA2B,EAC3BC,SAA2B,EAC3BC,iBAA4B,EAQ5B;QACA,OAAOC,gBAAgB,IAAI,CAAC,SAAS,EAAED,mBAAmB,UACjD,IAAI,CAAC,SAAS,CACnBP,YACAf,wBACAC,6BACAe,uBACAC,cACAC,WACAC,8BACAC,oBACAC;IAGN;IAEA,MAAc,UACZN,UAAkB,EAClBf,sBAAoC,EACpCC,2BAAyC,EACzCe,qBAA8B,EAC9BC,YAAqB,EACrBC,SAAmB,EACnBC,4BAAqC,EACrCC,kBAA2B,EAC3BC,SAA2B,EAQ3B;QACA,IAAI,CAAC,mBAAmB,CAAC,KAAK;QAE9B,MAAMjB,UAAU,IAAI,CAAC,sBAAsB,CACzCC,AAAAA,IAAAA,qCAAAA,YAAAA,AAAAA,EAAa,UAAUU;QAEzB,MAAMJ,SAASP,QAAQ,SAAS;QAEhC,IAAIoB,cAAc;QAClB,MAAMC,WAAmC,EAAE;QAC3C,MAAMC,uBACJP,gCAAgC,IAAI,CAAC,oBAAoB;QAC3DT,IAAAA,sBAAAA,MAAAA,AAAAA,EACEgB,AAAyBC,WAAzBD,sBACA;QAGF,IAAIE,8BAA8B;QAGlC,MAAO,KAAM;YACX,MAAMf,SAAS,MAAMT,QAAQ,YAAY,CACvC;gBACE,MAAM;gBACN,SAAS;gBACT,OAAO;oBACL,iBAAiBW;oBACjBE;oBACAG;oBACAC;gBACF;gBACA,UAAU,OAAOd,OAAOC;oBACtB,MAAMqB,YAAYC,KAAK,GAAG;oBAC1B,MAAM,EAAErB,SAAS,EAAE,GAAGD;oBACtBE,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,WAAW;oBAClB,MAAM,EAAEsB,MAAM,EAAE,GAAG/B;oBACnB,MAAMgC,qBACJD,AAAW,kBAAXA,SACI/B,uBAAuB,kBAAkB,GACzC2B;oBAEN,MAAMM,cAAc,IAAI,CAAC,cAAc;oBACvC1C,MACE,sCACA0C,YAAY,GAAG,CAAC,CAACC,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;oBAEhDxB,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOyB,MAAM,OAAO,CAACF,cAAc;oBACnC,IAAIA,AAAuB,MAAvBA,YAAY,MAAM,EACpBG,QAAQ,IAAI,CACV,CAAC,gBAAgB,EAAE,IAAI,CAAC,SAAS,CAAC,aAAa,CAAC,gDAAgD,CAAC;oBAIrG,MAAMC,aAAa,MAAOL,AAAAA,CAAAA,qBACtBM,yBAAAA,cAAcA,GACdC,yBAAAA,IAAG,AAAHA,EAAMhC,MAAM,eAAe,EAAE;wBAC/B,SAASE;wBACT,eAAeF,MAAM,YAAY;wBACjC,eAAe,IAAI,CAAC,SAAS,CAAC,aAAa;wBAC3C0B;wBACA,aAAajC;wBACb,qBAAqB,IAAI,CAAC,mBAAmB;wBAC7C,aAAagB;wBACbI;wBACAC;oBACF;oBACA9B,MAAM,cAAciD,KAAK,SAAS,CAACH,YAAY,MAAM;oBAErD,MAAM,EACJI,OAAO,EACPC,GAAG,EACHC,kCAAkC,EAClCC,KAAK,EACLC,KAAK,EACLC,WAAW,EACXC,KAAK,EACLC,iBAAiB,EAClB,GAAGX;oBAEJ7B,gBAAgB,IAAI,CAAC,GAAG,GAAG;wBACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;wBAClCsC;oBACF;oBACAtC,gBAAgB,IAAI,CAAC,KAAK,GAAGqC;oBAC7BrC,gBAAgB,IAAI,CAAC,iBAAiB,GAAGwC;oBACzCxC,gBAAgB,IAAI,CAAC,MAAM,GAAG;wBAC5B,SAASiC,WAAW,EAAE;wBACtBE;wBACAD;wBACA,UAAUL,WAAW,QAAQ;oBAC/B;oBACA7B,gBAAgB,SAAS,GAAGC;oBAE5B,MAAMwC,eAAe;2BAAKR,WAAW,EAAE;qBAAE;oBAEzC,IAAIM,OAAO;wBACT,MAAMG,UAAUpB,KAAK,GAAG;wBACxB,MAAMqB,gBAAgBJ,QAASG,CAAAA,UAAUrB,SAAQ;wBACjD,IAAIsB,gBAAgB,GAClBF,aAAa,IAAI,CAAC,IAAI,CAAC,SAAS,CAACE;oBAErC;oBAEAzC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,CAACkC,OAAO,CAAC,oBAAoB,EAAEA,MAAM,EAAE,EAAEF,OAAO,IAAI;oBAE3D,OAAO;wBACL,OAAO;4BACL,KAAK;wBACP;oBACF;gBACF;YACF,GACA;gBACE,gBAAgB;YAClB;YAGF,MAAML,aAAaxB,QAAQ;YAG3B,MAAMd,QAAQsC,YAAY,WAAW,EAAE;YACvCZ,SAAS,IAAI,IAAKY,YAAY,YAAY,EAAE;YAE5C,IAAIe;YACJ,IAAI;gBACFA,cAAc,MAAM,IAAI,CAAC,uBAAuB,CAC9CrD,OACAC,wBACAC,6BACA;oBACEiB;oBACA,SAAS;gBACX;YAEJ,EAAE,OAAO0B,OAAO;gBACd,OAAOxC,QAAQ,eAAe,CAC5B,CAAC,4CAA4C,EAAEwC,MAAM,SAAS,EAAEJ,KAAK,SAAS,CAC5EzC,QACC;YAEP;YACA,IAAI,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,EACjDqC,QAAQ,IAAI,CACV,8FACA,IAAI,CAAC,mBAAmB,CAAC,sBAAsB;YAGnD,IAAIiB,YAAY;YAChB,IAAI;gBACF,MAAMjD,QAAQ,YAAY,CAACgD,YAAY,KAAK;YAC9C,EAAE,OAAOR,OAAY;gBACnBS,YAAY;gBACZzB;gBACA,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,GAAG,CAAC,+BAA+B,EAAEgB,OAAO,WAAWU,OAAOV,QAAQ;gBACrHrD,MACE,yFACAqD,iBAAiBW,QAAQX,MAAM,OAAO,GAAGU,OAAOV,QAChD,6CACAhB;YAEJ;YAEA,IAAIA,8BAA8BnC,uCAChC,OAAOW,QAAQ,eAAe,CAAC;YAIjC,IAAI,CAACiC,YAAY,oCACf,IAAIgB,WACF9D,MACE;iBAGF;YAKJ,EAAEiC;YAEF,IAAIA,cAAcE,sBAAsB;gBACtC,MAAM8B,WAAW,CAAC,UAAU,EAAE9B,qBAAqB,4JAA4J,CAAC;gBAChN,OAAOtB,QAAQ,eAAe,CAACoD;YACjC;YAEA,IAAI,CAAC,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,EAClD,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,GAC7C;QAEN;QAEA,OAAO;YACL,QAAQ;gBACN/B;YACF;YACAd;QACF;IACF;IAEQ,oBACN8C,IAAsE,EACtEC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACpC;QACA,MAAMC,YAA4C;YAChD,MAAM;YACN,SAASL;YACT,OAAO;gBACL,YAAYI,mBACP;oBACCH;oBACAG;gBACF,IACAH;YACN;YACA,UAAU,OAAOnD,OAAOwD;gBACtB,MAAM,EAAEzD,IAAI,EAAE,GAAGyD;gBACjB,IAAIC;gBACJ,MAAMC,YAAY,CAACC;oBACjBF,YAAYE;oBACZ5D,KAAK,GAAG,GAAG;wBACT4D;oBACF;gBACF;gBAGA,MAAMzD,YAAYsD,YAAY,SAAS;gBACvCrD,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,WAAW;gBAElB,MAAM0D,mBAAmBV,AAAS,YAATA;gBACzB,IAAIW,cAAcV;gBAClB,IAAIW,cAAc;gBAClB,IAAIF,oBAAqBV,CAAAA,AAAS,aAATA,QAAqBA,AAAS,cAATA,IAAiB,GAAI;oBACjEY,cAAc;oBACd,MAAMC,gBACJb,AAAS,aAATA,OACI,CAAC,kDAAkD,EAAEC,QAAQ,GAC7D,CAAC,+GAA+G,EAAEA,QAAQ;oBAChIU,cAAc;wBACZ,CAACC,YAAY,EAAEC;oBACjB;gBACF,OAAO,IAAIH,kBACTC,cAAc;oBACZ,CAACC,YAAY,EAAE,GAAGZ,KAAK,EAAE,EAAEC,QAAQ;gBACrC;gBAGF,IAAIa;gBAEJ,IAAIC,uBAAuB;gBAC3B,IAAIZ,KAAK,eAAe,IAAI,CAAC,SAAS,CAAC,mBAAmB,EAAE;oBAC1DrE,MAAM;oBACN,MAAMkF,OAAO,MAAM,IAAI,CAAC,SAAS,CAAC,mBAAmB;oBACrDD,uBAAuB,MAAME,AAAAA,IAAAA,0BAAAA,iBAAAA,AAAAA,EAC3BD,MACA,KACA,OACAb,KAAK,gBAAgB;gBAEzB;gBAEA,IAAI;oBACFW,gBAAgB,MAAM,IAAI,CAAC,OAAO,CAAC,OAAO,CACxCH,aACAT,aACAC,KACAY,sBACAX;gBAEJ,EAAE,OAAOjB,OAAO;oBACd,IAAIA,iBAAiB+B,kCAAAA,YAAYA,EAC/BV,UAAUrB,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAEA,MAAM,EAAEgC,IAAI,EAAE/B,KAAK,EAAEgC,OAAO,EAAEX,IAAI,EAAElB,iBAAiB,EAAE,GAAGuB;gBAC1DN,UAAUC;gBACV5D,KAAK,iBAAiB,GAAG0C;gBAEzB,IAAI8B,eAAeF;gBACnB,IAAIT,kBAEF,IAAI,AAAgB,YAAhB,OAAOS,MACTE,eAAeF;qBACV,IAAInB,AAAS,cAATA,MAEPqB,eADEF,QAAAA,OACa,QAECA,IAAY,CAACP,YAAY;qBAEtC,IAAIO,QAAAA,MACTE,eAAe;qBACV;oBACLpE,IAAAA,sBAAAA,MAAAA,AAAAA,EACEkE,MAAM,CAACP,YAAY,KAAK1C,QACxB;oBAEFmD,eAAgBF,IAAY,CAACP,YAAY;gBAC3C;gBAGF,IAAIZ,AAAS,aAATA,QAAqB,CAACqB,cAAc;oBACtCxE,KAAK,KAAK,GAAGuC;oBACbvC,KAAK,OAAO,GAAGuE;oBACf,MAAM,IAAItB,MAAM,CAAC,kBAAkB,EAAEsB,SAAS;gBAChD;gBAEA,OAAO;oBACL,QAAQC;oBACR,KAAKd;oBACLnB;oBACAgC;gBACF;YACF;QACF;QAEA,OAAOf;IACT;IACA,MAAM,yBACJL,IAA0D,EAC1DC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACP;QAC7B,MAAMzD,UAAU,IAAI,CAAC,sBAAsB,CACzCC,AAAAA,IAAAA,qCAAAA,YAAAA,AAAAA,EACEoD,MACA,AAAkB,YAAlB,OAAOC,SAAsBA,SAASlB,KAAK,SAAS,CAACkB;QAIzD,MAAMI,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9CL,MACAC,QACAC,aACAC,KACAC;QAGF,MAAMlD,SAASP,QAAQ,SAAS;QAChC,MAAMS,SAAS,MAAMT,QAAQ,YAAY,CAAC0D;QAE1C,IAAI,CAACjD,QACH,MAAM,IAAI0C,MACR;QAIJ,MAAM,EAAEzC,MAAM,EAAE+D,OAAO,EAAE,GAAGhE;QAE5B,OAAO;YACLC;YACA+D;YACAlE;QACF;IACF;IAEQ,UAAUoE,MAAc,EAA4C;QAC1E,OAAO;YACL,MAAM;YACN,OAAO;gBACLA;YACF;QACF;IACF;IAEA,MAAM,aAAaA,MAAc,EAAEC,YAA0B,EAAE;QAC7D,OAAO,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;YACtCD;QACF;IACF;IAEA,MAAM,QACJE,SAAsB,EACtBrB,GAA+B,EAC/BD,WAAyB,EACO;QAChC,MAAM,EAAEuB,UAAU,EAAErB,gBAAgB,EAAE,GAAGsB,AAAAA,IAAAA,kCAAAA,WAAAA,AAAAA,EAAYF;QAErD,MAAMG,cAAc,CAAC,SAAS,EAAEF,YAAY;QAC5C,MAAM9E,UAAU,IAAI,CAAC,sBAAsB,CACzCC,AAAAA,IAAAA,qCAAAA,YAAAA,AAAAA,EAAa,WAAW+E;QAE1B,MAAMzE,SAASP,QAAQ,SAAS;QAChC,MAAM,EACJiF,SAAS,EACTC,eAAe,EACfC,WAAW,EACXC,kBAAkB,EAClB,GAAGC,SACJ,GAAG7B;QACJ,MAAM8B,oBAA0C;YAC9CH;YACAC;YACA,GAAGC,OAAO;QACZ;QAEA/E,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOuE,WAAW;QAClBvE,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO2E,WAAW;QAClB3E,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO4E,iBAAiB;QAExB5E,IAAAA,sBAAAA,MAAAA,AAAAA,EACE4E,mBAAmBD,WACnB,CAAC,iGAAiG,EAAEC,gBAAgB,aAAa,EAAED,UAAU,CAAC,CAAC;QAGjJ,MAAMM,mBAAmB7D,KAAK,GAAG;QACjC,IAAI8D,iBAAiBD;QACrB,IAAIE,eAAe;QAEnB,MAAOD,iBAAiBD,oBAAoBN,UAAW;YACrD,MAAMS,oBAAoBhE,KAAK,GAAG;YAClC8D,iBAAiBE;YACjB,MAAMhC,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9C,WACAoB,YACAvB,aACA+B,mBACA7B;YAGF,MAAMhD,SAAU,MAAMT,QAAQ,YAAY,CAAC0D;YAO3C,IAAIjD,QAAQ,QACV,OAAO;gBACL,QAAQc;gBACRhB;YACF;YAGFkF,eACEhF,QAAQ,WACP,CAACA,UAAU,CAAC,0BAA0B,EAAEqE,YAAY,IACrD,CAAC,0CAA0C,EAAEA,YAAY;YAC3D,MAAMa,MAAMjE,KAAK,GAAG;YACpB,IAAIiE,MAAMD,oBAAoBR,iBAAiB;gBAC7C,MAAMnC,gBAAgBmC,kBAAmBS,CAAAA,MAAMD,iBAAgB;gBAC/D,MAAME,YAAY,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;oBACjD,QAAQ7C;gBACV;gBACA,MAAM/C,QAAQ,MAAM,CAAC4F;YACvB;QACF;QAEA,OAAO5F,QAAQ,eAAe,CAAC,CAAC,iBAAiB,EAAEyF,cAAc;IACnE;IAtnBA,YACEI,iBAAoC,EACpCC,OAAgB,EAChBC,IAMC,CACD;QAjCF;QAEA;QAEA;QAEA,uBAAiB,uBAAjB;QAEA,uBAAiB,eAAjB;QAEA,uBAAQ,uBAAR;QAEA;QAEA,uBAAiB,SAAjB;QAEA;QAkBE,IAAI,CAAC,SAAS,GAAGF;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC,KAAK,SAAS;QAC/B,IAAI,CAAC,mBAAmB,GAAGA,MAAM;QACjC,IAAI,CAAC,oBAAoB,GAAGA,KAAK,oBAAoB;QACrD,IAAI,CAAC,KAAK,GAAGA,KAAK,KAAK;QACvB,IAAI,CAAC,mBAAmB,GAAG,IAAIC,yBAAAA,mBAAmBA;QAClD,IAAI,CAAC,mBAAmB,GAAGD,KAAK,WAAW;QAC3C,IAAI,CAAC,WAAW,GAAG,IAAIE,yCAAAA,WAAWA,CAAC;YACjCJ;YACAC;YACA,WAAWC,KAAK,SAAS;YACzB,aAAa,IAAI,CAAC,cAAc;QAClC;IACF;AA8lBF;AAEO,eAAe5E,gBACpB0E,iBAAoC,EACpC3E,iBAAuC,EACvCY,MAAwB;IAExB,IAAI,CAACZ,mBAAmB,QACtB,OAAOY;IAGT,IAAI,CAAC+D,kBAAkB,2BAA2B,EAChD,MAAM,IAAI1C,MACR,CAAC,gCAAgC,EAAE0C,kBAAkB,aAAa,EAAE;IAIxE,MAAMK,UAAU,OAAOC;QACrB,MAAMA,QAAQ,MAAM,CAACjF;IACvB;IAEA,MAAM,EAAEkF,OAAO,EAAEC,QAAQ,EAAE,GACzB,MAAMR,kBAAkB,2BAA2B,CAACK;IACtD,IAAI;QACF,MAAMzF,SAAS,MAAMqB;QAErB,MAAMU,QAAQ6D;QACd,IAAI7D,OACF,MAAMA;QAER,OAAO/B;IACT,SAAU;QACR2F;IACF;AACF"}
|
|
1
|
+
{"version":3,"file":"agent/tasks.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../src/agent/tasks.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import { ConversationHistory, plan, uiTarsPlanning } from '@/ai-model';\nimport {\n type TMultimodalPrompt,\n type TUserPrompt,\n getReadableTimeString,\n} from '@/common';\nimport type { AbstractInterface, FileChooserHandler } from '@/device';\nimport type Service from '@/service';\nimport type { TaskRunner } from '@/task-runner';\nimport { TaskExecutionError } from '@/task-runner';\nimport type {\n DeepThinkOption,\n DeviceAction,\n ExecutionTaskApply,\n ExecutionTaskInsightQueryApply,\n ExecutionTaskPlanningApply,\n ExecutionTaskProgressOptions,\n InterfaceType,\n MidsceneYamlFlowItem,\n PlanningAIResponse,\n PlanningAction,\n PlanningActionParamWaitFor,\n ServiceDump,\n ServiceExtractOption,\n ServiceExtractParam,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { ExecutionSession } from './execution-session';\nimport { TaskBuilder } from './task-builder';\nimport type { TaskCache } from './task-cache';\nexport { locatePlanForLocate } from './task-builder';\nimport { descriptionOfTree } from '@midscene/shared/extractor';\nimport { taskTitleStr } from './ui-utils';\nimport { parsePrompt } from './utils';\n\ninterface ExecutionResult<OutputType = any> {\n output: OutputType;\n thought?: string;\n runner: TaskRunner;\n}\n\ninterface TaskExecutorHooks {\n onTaskUpdate?: (\n runner: TaskRunner,\n error?: TaskExecutionError,\n ) => Promise<void> | void;\n}\n\nconst debug = getDebug('device-task-executor');\nconst maxErrorCountAllowedInOnePlanningLoop = 5;\n\nexport { TaskExecutionError };\n\nexport class TaskExecutor {\n interface: AbstractInterface;\n\n service: Service;\n\n taskCache?: TaskCache;\n\n private readonly providedActionSpace: DeviceAction[];\n\n private readonly taskBuilder: TaskBuilder;\n\n private conversationHistory: ConversationHistory;\n\n onTaskStartCallback?: ExecutionTaskProgressOptions['onTaskStart'];\n\n private readonly hooks?: TaskExecutorHooks;\n\n replanningCycleLimit?: number;\n\n // @deprecated use .interface instead\n get page() {\n return this.interface;\n }\n\n constructor(\n interfaceInstance: AbstractInterface,\n service: Service,\n opts: {\n taskCache?: TaskCache;\n onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n replanningCycleLimit?: number;\n hooks?: TaskExecutorHooks;\n actionSpace: DeviceAction[];\n },\n ) {\n this.interface = interfaceInstance;\n this.service = service;\n this.taskCache = opts.taskCache;\n this.onTaskStartCallback = opts?.onTaskStart;\n this.replanningCycleLimit = opts.replanningCycleLimit;\n this.hooks = opts.hooks;\n this.conversationHistory = new ConversationHistory();\n this.providedActionSpace = opts.actionSpace;\n this.taskBuilder = new TaskBuilder({\n interfaceInstance,\n service,\n taskCache: opts.taskCache,\n actionSpace: this.getActionSpace(),\n });\n }\n\n private createExecutionSession(\n title: string,\n options?: { tasks?: ExecutionTaskApply[] },\n ) {\n return new ExecutionSession(\n title,\n () => Promise.resolve(this.service.contextRetrieverFn()),\n {\n onTaskStart: this.onTaskStartCallback,\n tasks: options?.tasks,\n onTaskUpdate: this.hooks?.onTaskUpdate,\n },\n );\n }\n\n private getActionSpace(): DeviceAction[] {\n return this.providedActionSpace;\n }\n\n public async convertPlanToExecutable(\n plans: PlanningAction[],\n modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n options?: {\n cacheable?: boolean;\n subTask?: boolean;\n },\n ) {\n return this.taskBuilder.build(\n plans,\n modelConfigForPlanning,\n modelConfigForDefaultIntent,\n options,\n );\n }\n\n async loadYamlFlowAsPlanning(userInstruction: string, yamlString: string) {\n const session = this.createExecutionSession(\n taskTitleStr('Action', userInstruction),\n );\n\n const task: ExecutionTaskPlanningApply = {\n type: 'Planning',\n subType: 'LoadYaml',\n param: {\n userInstruction,\n },\n executor: async (param, executorContext) => {\n const { uiContext } = executorContext;\n assert(uiContext, 'uiContext is required for Planning task');\n return {\n output: {\n actions: [],\n shouldContinuePlanning: false,\n log: '',\n yamlString,\n },\n cache: {\n hit: true,\n },\n hitBy: {\n from: 'Cache',\n context: {\n yamlString,\n },\n },\n };\n },\n };\n const runner = session.getRunner();\n await session.appendAndRun(task);\n\n return {\n runner,\n };\n }\n\n async runPlans(\n title: string,\n plans: PlanningAction[],\n modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n ): Promise<ExecutionResult> {\n const session = this.createExecutionSession(title);\n const { tasks } = await this.convertPlanToExecutable(\n plans,\n modelConfigForPlanning,\n modelConfigForDefaultIntent,\n );\n const runner = session.getRunner();\n const result = await session.appendAndRun(tasks);\n const { output } = result ?? {};\n return {\n output,\n runner,\n };\n }\n\n async action(\n userPrompt: string,\n modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n includeBboxInPlanning: boolean,\n aiActContext?: string,\n cacheable?: boolean,\n replanningCycleLimitOverride?: number,\n imagesIncludeCount?: number,\n deepThink?: DeepThinkOption,\n fileChooserAccept?: string[],\n ): Promise<\n ExecutionResult<\n | {\n yamlFlow?: MidsceneYamlFlowItem[]; // for cache use\n output?: string;\n }\n | undefined\n >\n > {\n return withFileChooser(this.interface, fileChooserAccept, async () => {\n return this.runAction(\n userPrompt,\n modelConfigForPlanning,\n modelConfigForDefaultIntent,\n includeBboxInPlanning,\n aiActContext,\n cacheable,\n replanningCycleLimitOverride,\n imagesIncludeCount,\n deepThink,\n );\n });\n }\n\n private async runAction(\n userPrompt: string,\n modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n includeBboxInPlanning: boolean,\n aiActContext?: string,\n cacheable?: boolean,\n replanningCycleLimitOverride?: number,\n imagesIncludeCount?: number,\n deepThink?: DeepThinkOption,\n ): Promise<\n ExecutionResult<\n | {\n yamlFlow?: MidsceneYamlFlowItem[]; // for cache use\n output?: string;\n }\n | undefined\n >\n > {\n this.conversationHistory.reset();\n\n const session = this.createExecutionSession(\n taskTitleStr('Action', userPrompt),\n );\n const runner = session.getRunner();\n\n let replanCount = 0;\n const yamlFlow: MidsceneYamlFlowItem[] = [];\n const replanningCycleLimit =\n replanningCycleLimitOverride ?? this.replanningCycleLimit;\n assert(\n replanningCycleLimit !== undefined,\n 'replanningCycleLimit is required for TaskExecutor.action',\n );\n\n let errorCountInOnePlanningLoop = 0; // count the number of errors in one planning loop\n let outputString: string | undefined;\n\n // Main planning loop - unified plan/replan logic\n while (true) {\n const result = await session.appendAndRun(\n {\n type: 'Planning',\n subType: 'Plan',\n param: {\n userInstruction: userPrompt,\n aiActContext,\n imagesIncludeCount,\n deepThink,\n },\n executor: async (param, executorContext) => {\n const { uiContext } = executorContext;\n assert(uiContext, 'uiContext is required for Planning task');\n const { vlMode } = modelConfigForPlanning;\n const uiTarsModelVersion =\n vlMode === 'vlm-ui-tars'\n ? modelConfigForPlanning.uiTarsModelVersion\n : undefined;\n\n const actionSpace = this.getActionSpace();\n debug(\n 'actionSpace for this interface is:',\n actionSpace.map((action) => action.name).join(', '),\n );\n assert(Array.isArray(actionSpace), 'actionSpace must be an array');\n if (actionSpace.length === 0) {\n console.warn(\n `ActionSpace for ${this.interface.interfaceType} is empty. This may lead to unexpected behavior.`,\n );\n }\n\n const planResult = await (uiTarsModelVersion\n ? uiTarsPlanning\n : plan)(param.userInstruction, {\n context: uiContext,\n actionContext: param.aiActContext,\n interfaceType: this.interface.interfaceType as InterfaceType,\n actionSpace,\n modelConfig: modelConfigForPlanning,\n conversationHistory: this.conversationHistory,\n includeBbox: includeBboxInPlanning,\n imagesIncludeCount,\n deepThink,\n });\n debug('planResult', JSON.stringify(planResult, null, 2));\n\n const {\n actions,\n log,\n note,\n error,\n usage,\n rawResponse,\n reasoning_content,\n } = planResult;\n\n executorContext.task.log = {\n ...(executorContext.task.log || {}),\n rawResponse,\n };\n executorContext.task.usage = usage;\n executorContext.task.reasoning_content = reasoning_content;\n executorContext.task.output = {\n actions: actions || [],\n log,\n note,\n yamlFlow: planResult.yamlFlow,\n output: outputString,\n shouldContinuePlanning: planResult.shouldContinuePlanning,\n };\n executorContext.uiContext = uiContext;\n\n assert(!error, `Failed to continue: ${error}\\n${log || ''}`);\n\n return {\n cache: {\n hit: false,\n },\n } as any;\n },\n },\n {\n allowWhenError: true,\n },\n );\n\n const planResult = result?.output as PlanningAIResponse | undefined;\n\n // Execute planned actions\n const plans = planResult?.actions || [];\n yamlFlow.push(...(planResult?.yamlFlow || []));\n\n let executables: Awaited<ReturnType<typeof this.convertPlanToExecutable>>;\n try {\n executables = await this.convertPlanToExecutable(\n plans,\n modelConfigForPlanning,\n modelConfigForDefaultIntent,\n {\n cacheable,\n subTask: true,\n },\n );\n } catch (error) {\n return session.appendErrorPlan(\n `Error converting plans to executable tasks: ${error}, plans: ${JSON.stringify(\n plans,\n )}`,\n );\n }\n if (this.conversationHistory.pendingFeedbackMessage) {\n console.warn(\n 'unconsumed pending feedback message detected, this may lead to unexpected planning result:',\n this.conversationHistory.pendingFeedbackMessage,\n );\n }\n // todo: set time string\n\n try {\n const result = await session.appendAndRun(executables.tasks);\n outputString = result?.output;\n } catch (error: any) {\n // errorFlag = true;\n errorCountInOnePlanningLoop++;\n this.conversationHistory.pendingFeedbackMessage = `Time: ${getReadableTimeString()}, Error executing running tasks: ${error?.message || String(error)}`;\n debug(\n 'error when executing running tasks, but continue to run if it is not too many errors:',\n error instanceof Error ? error.message : String(error),\n 'current error count in one planning loop:',\n errorCountInOnePlanningLoop,\n );\n }\n\n if (errorCountInOnePlanningLoop > maxErrorCountAllowedInOnePlanningLoop) {\n return session.appendErrorPlan('Too many errors in one planning loop');\n }\n\n // // Check if task is complete\n if (!planResult?.shouldContinuePlanning) {\n break;\n }\n\n // Increment replan count for next iteration\n ++replanCount;\n\n if (replanCount > replanningCycleLimit) {\n const errorMsg = `Replanned ${replanningCycleLimit} times, exceeding the limit. Please configure a larger value for replanningCycleLimit (or use MIDSCENE_REPLANNING_CYCLE_LIMIT) to handle more complex tasks.`;\n return session.appendErrorPlan(errorMsg);\n }\n\n if (!this.conversationHistory.pendingFeedbackMessage) {\n this.conversationHistory.pendingFeedbackMessage = `Time: ${getReadableTimeString()}, I have finished the action previously planned.`;\n }\n }\n\n return {\n output: {\n yamlFlow,\n output: outputString,\n },\n runner,\n };\n }\n\n private createTypeQueryTask(\n type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert' | 'WaitFor',\n demand: ServiceExtractParam,\n modelConfig: IModelConfig,\n opt?: ServiceExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ) {\n const queryTask: ExecutionTaskInsightQueryApply = {\n type: 'Insight',\n subType: type,\n param: {\n dataDemand: multimodalPrompt\n ? ({\n demand,\n multimodalPrompt,\n } as never)\n : demand, // for user param presentation in report right sidebar\n },\n executor: async (param, taskContext) => {\n const { task } = taskContext;\n let queryDump: ServiceDump | undefined;\n const applyDump = (dump: ServiceDump) => {\n queryDump = dump;\n task.log = {\n dump,\n };\n };\n\n // Get context for query operations\n const uiContext = taskContext.uiContext;\n assert(uiContext, 'uiContext is required for Query task');\n\n const ifTypeRestricted = type !== 'Query';\n let demandInput = demand;\n let keyOfResult = 'result';\n if (ifTypeRestricted && (type === 'Assert' || type === 'WaitFor')) {\n keyOfResult = 'StatementIsTruthy';\n const booleanPrompt =\n type === 'Assert'\n ? `Boolean, whether the following statement is true: ${demand}`\n : `Boolean, the user wants to do some 'wait for' operation, please check whether the following statement is true: ${demand}`;\n demandInput = {\n [keyOfResult]: booleanPrompt,\n };\n } else if (ifTypeRestricted) {\n demandInput = {\n [keyOfResult]: `${type}, ${demand}`,\n };\n }\n\n let extractResult;\n\n let extraPageDescription = '';\n if (opt?.domIncluded && this.interface.getElementsNodeTree) {\n debug('appending tree info for page');\n const tree = await this.interface.getElementsNodeTree();\n extraPageDescription = await descriptionOfTree(\n tree,\n 200,\n false,\n opt?.domIncluded === 'visible-only',\n );\n }\n\n try {\n extractResult = await this.service.extract<any>(\n demandInput,\n modelConfig,\n opt,\n extraPageDescription,\n multimodalPrompt,\n );\n } catch (error) {\n if (error instanceof ServiceError) {\n applyDump(error.dump);\n }\n throw error;\n }\n\n const { data, usage, thought, dump, reasoning_content } = extractResult;\n applyDump(dump);\n task.reasoning_content = reasoning_content;\n\n let outputResult = data;\n if (ifTypeRestricted) {\n // If AI returned a plain string instead of structured format, use it directly\n if (typeof data === 'string') {\n outputResult = data;\n } else if (type === 'WaitFor') {\n if (data === null || data === undefined) {\n outputResult = false;\n } else {\n outputResult = (data as any)[keyOfResult];\n }\n } else if (data === null || data === undefined) {\n outputResult = null;\n } else {\n assert(\n data?.[keyOfResult] !== undefined,\n 'No result in query data',\n );\n outputResult = (data as any)[keyOfResult];\n }\n }\n\n if (type === 'Assert' && !outputResult) {\n task.usage = usage;\n task.thought = thought;\n throw new Error(`Assertion failed: ${thought}`);\n }\n\n return {\n output: outputResult,\n log: queryDump,\n usage,\n thought,\n };\n },\n };\n\n return queryTask;\n }\n async createTypeQueryExecution<T>(\n type: 'Query' | 'Boolean' | 'Number' | 'String' | 'Assert',\n demand: ServiceExtractParam,\n modelConfig: IModelConfig,\n opt?: ServiceExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ): Promise<ExecutionResult<T>> {\n const session = this.createExecutionSession(\n taskTitleStr(\n type,\n typeof demand === 'string' ? demand : JSON.stringify(demand),\n ),\n );\n\n const queryTask = await this.createTypeQueryTask(\n type,\n demand,\n modelConfig,\n opt,\n multimodalPrompt,\n );\n\n const runner = session.getRunner();\n const result = await session.appendAndRun(queryTask);\n\n if (!result) {\n throw new Error(\n 'result of taskExecutor.flush() is undefined in function createTypeQueryTask',\n );\n }\n\n const { output, thought } = result;\n\n return {\n output,\n thought,\n runner,\n };\n }\n\n async taskForSleep(timeMs: number, _modelConfig: IModelConfig) {\n return this.taskBuilder.createSleepTask({\n timeMs,\n });\n }\n\n async waitFor(\n assertion: TUserPrompt,\n opt: PlanningActionParamWaitFor,\n modelConfig: IModelConfig,\n ): Promise<ExecutionResult<void>> {\n const { textPrompt, multimodalPrompt } = parsePrompt(assertion);\n\n const description = `waitFor: ${textPrompt}`;\n const session = this.createExecutionSession(\n taskTitleStr('WaitFor', description),\n );\n const runner = session.getRunner();\n const {\n timeoutMs,\n checkIntervalMs,\n domIncluded,\n screenshotIncluded,\n ...restOpt\n } = opt;\n const serviceExtractOpt: ServiceExtractOption = {\n domIncluded,\n screenshotIncluded,\n ...restOpt,\n };\n\n assert(assertion, 'No assertion for waitFor');\n assert(timeoutMs, 'No timeoutMs for waitFor');\n assert(checkIntervalMs, 'No checkIntervalMs for waitFor');\n\n assert(\n checkIntervalMs <= timeoutMs,\n `wrong config for waitFor: checkIntervalMs must be less than timeoutMs, config: {checkIntervalMs: ${checkIntervalMs}, timeoutMs: ${timeoutMs}}`,\n );\n\n const overallStartTime = Date.now();\n let lastCheckStart = overallStartTime;\n let errorThought = '';\n // Continue checking as long as the previous iteration began within the timeout window.\n while (lastCheckStart - overallStartTime <= timeoutMs) {\n const currentCheckStart = Date.now();\n lastCheckStart = currentCheckStart;\n const queryTask = await this.createTypeQueryTask(\n 'WaitFor',\n textPrompt,\n modelConfig,\n serviceExtractOpt,\n multimodalPrompt,\n );\n\n const result = (await session.appendAndRun(queryTask)) as\n | {\n output: boolean;\n thought?: string;\n }\n | undefined;\n\n if (result?.output) {\n return {\n output: undefined,\n runner,\n };\n }\n\n errorThought =\n result?.thought ||\n (!result && `No result from assertion: ${textPrompt}`) ||\n `unknown error when waiting for assertion: ${textPrompt}`;\n const now = Date.now();\n if (now - currentCheckStart < checkIntervalMs) {\n const timeRemaining = checkIntervalMs - (now - currentCheckStart);\n const sleepTask = this.taskBuilder.createSleepTask({\n timeMs: timeRemaining,\n });\n await session.append(sleepTask);\n }\n }\n\n return session.appendErrorPlan(`waitFor timeout: ${errorThought}`);\n }\n}\n\nexport async function withFileChooser<T>(\n interfaceInstance: AbstractInterface,\n fileChooserAccept: string[] | undefined,\n action: () => Promise<T>,\n): Promise<T> {\n if (!fileChooserAccept?.length) {\n return action();\n }\n\n if (!interfaceInstance.registerFileChooserListener) {\n throw new Error(\n `File upload is not supported on ${interfaceInstance.interfaceType}`,\n );\n }\n\n const handler = async (chooser: FileChooserHandler) => {\n await chooser.accept(fileChooserAccept);\n };\n\n const { dispose, getError } =\n await interfaceInstance.registerFileChooserListener(handler);\n try {\n const result = await action();\n // Check for errors that occurred during file chooser handling\n const error = getError();\n if (error) {\n throw error;\n }\n return result;\n } finally {\n dispose();\n }\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","maxErrorCountAllowedInOnePlanningLoop","TaskExecutor","title","options","ExecutionSession","Promise","plans","modelConfigForPlanning","modelConfigForDefaultIntent","userInstruction","yamlString","session","taskTitleStr","task","param","executorContext","uiContext","assert","runner","tasks","result","output","userPrompt","includeBboxInPlanning","aiActContext","cacheable","replanningCycleLimitOverride","imagesIncludeCount","deepThink","fileChooserAccept","withFileChooser","replanCount","yamlFlow","replanningCycleLimit","undefined","errorCountInOnePlanningLoop","outputString","vlMode","uiTarsModelVersion","actionSpace","action","Array","console","planResult","uiTarsPlanning","plan","JSON","actions","log","note","error","usage","rawResponse","reasoning_content","executables","getReadableTimeString","String","Error","errorMsg","type","demand","modelConfig","opt","multimodalPrompt","queryTask","taskContext","queryDump","applyDump","dump","ifTypeRestricted","demandInput","keyOfResult","booleanPrompt","extractResult","extraPageDescription","tree","descriptionOfTree","ServiceError","data","thought","outputResult","timeMs","_modelConfig","assertion","textPrompt","parsePrompt","description","timeoutMs","checkIntervalMs","domIncluded","screenshotIncluded","restOpt","serviceExtractOpt","overallStartTime","Date","lastCheckStart","errorThought","currentCheckStart","now","timeRemaining","sleepTask","interfaceInstance","service","opts","ConversationHistory","TaskBuilder","handler","chooser","dispose","getError"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AC6CA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AACvB,MAAMC,wCAAwC;AAIvC,MAAMC;IAoBX,IAAI,OAAO;QACT,OAAO,IAAI,CAAC,SAAS;IACvB;IA6BQ,uBACNC,KAAa,EACbC,OAA0C,EAC1C;QACA,OAAO,IAAIC,8CAAAA,gBAAgBA,CACzBF,OACA,IAAMG,QAAQ,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,kBAAkB,KACrD;YACE,aAAa,IAAI,CAAC,mBAAmB;YACrC,OAAOF,SAAS;YAChB,cAAc,IAAI,CAAC,KAAK,EAAE;QAC5B;IAEJ;IAEQ,iBAAiC;QACvC,OAAO,IAAI,CAAC,mBAAmB;IACjC;IAEA,MAAa,wBACXG,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACzCL,OAGC,EACD;QACA,OAAO,IAAI,CAAC,WAAW,CAAC,KAAK,CAC3BG,OACAC,wBACAC,6BACAL;IAEJ;IAEA,MAAM,uBAAuBM,eAAuB,EAAEC,UAAkB,EAAE;QACxE,MAAMC,UAAU,IAAI,CAAC,sBAAsB,CACzCC,AAAAA,IAAAA,qCAAAA,YAAAA,AAAAA,EAAa,UAAUH;QAGzB,MAAMI,OAAmC;YACvC,MAAM;YACN,SAAS;YACT,OAAO;gBACLJ;YACF;YACA,UAAU,OAAOK,OAAOC;gBACtB,MAAM,EAAEC,SAAS,EAAE,GAAGD;gBACtBE,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,WAAW;gBAClB,OAAO;oBACL,QAAQ;wBACN,SAAS,EAAE;wBACX,wBAAwB;wBACxB,KAAK;wBACLN;oBACF;oBACA,OAAO;wBACL,KAAK;oBACP;oBACA,OAAO;wBACL,MAAM;wBACN,SAAS;4BACPA;wBACF;oBACF;gBACF;YACF;QACF;QACA,MAAMQ,SAASP,QAAQ,SAAS;QAChC,MAAMA,QAAQ,YAAY,CAACE;QAE3B,OAAO;YACLK;QACF;IACF;IAEA,MAAM,SACJhB,KAAa,EACbI,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACf;QAC1B,MAAMG,UAAU,IAAI,CAAC,sBAAsB,CAACT;QAC5C,MAAM,EAAEiB,KAAK,EAAE,GAAG,MAAM,IAAI,CAAC,uBAAuB,CAClDb,OACAC,wBACAC;QAEF,MAAMU,SAASP,QAAQ,SAAS;QAChC,MAAMS,SAAS,MAAMT,QAAQ,YAAY,CAACQ;QAC1C,MAAM,EAAEE,MAAM,EAAE,GAAGD,UAAU,CAAC;QAC9B,OAAO;YACLC;YACAH;QACF;IACF;IAEA,MAAM,OACJI,UAAkB,EAClBf,sBAAoC,EACpCC,2BAAyC,EACzCe,qBAA8B,EAC9BC,YAAqB,EACrBC,SAAmB,EACnBC,4BAAqC,EACrCC,kBAA2B,EAC3BC,SAA2B,EAC3BC,iBAA4B,EAS5B;QACA,OAAOC,gBAAgB,IAAI,CAAC,SAAS,EAAED,mBAAmB,UACjD,IAAI,CAAC,SAAS,CACnBP,YACAf,wBACAC,6BACAe,uBACAC,cACAC,WACAC,8BACAC,oBACAC;IAGN;IAEA,MAAc,UACZN,UAAkB,EAClBf,sBAAoC,EACpCC,2BAAyC,EACzCe,qBAA8B,EAC9BC,YAAqB,EACrBC,SAAmB,EACnBC,4BAAqC,EACrCC,kBAA2B,EAC3BC,SAA2B,EAS3B;QACA,IAAI,CAAC,mBAAmB,CAAC,KAAK;QAE9B,MAAMjB,UAAU,IAAI,CAAC,sBAAsB,CACzCC,AAAAA,IAAAA,qCAAAA,YAAAA,AAAAA,EAAa,UAAUU;QAEzB,MAAMJ,SAASP,QAAQ,SAAS;QAEhC,IAAIoB,cAAc;QAClB,MAAMC,WAAmC,EAAE;QAC3C,MAAMC,uBACJP,gCAAgC,IAAI,CAAC,oBAAoB;QAC3DT,IAAAA,sBAAAA,MAAAA,AAAAA,EACEgB,AAAyBC,WAAzBD,sBACA;QAGF,IAAIE,8BAA8B;QAClC,IAAIC;QAGJ,MAAO,KAAM;YACX,MAAMhB,SAAS,MAAMT,QAAQ,YAAY,CACvC;gBACE,MAAM;gBACN,SAAS;gBACT,OAAO;oBACL,iBAAiBW;oBACjBE;oBACAG;oBACAC;gBACF;gBACA,UAAU,OAAOd,OAAOC;oBACtB,MAAM,EAAEC,SAAS,EAAE,GAAGD;oBACtBE,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,WAAW;oBAClB,MAAM,EAAEqB,MAAM,EAAE,GAAG9B;oBACnB,MAAM+B,qBACJD,AAAW,kBAAXA,SACI9B,uBAAuB,kBAAkB,GACzC2B;oBAEN,MAAMK,cAAc,IAAI,CAAC,cAAc;oBACvCzC,MACE,sCACAyC,YAAY,GAAG,CAAC,CAACC,SAAWA,OAAO,IAAI,EAAE,IAAI,CAAC;oBAEhDvB,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOwB,MAAM,OAAO,CAACF,cAAc;oBACnC,IAAIA,AAAuB,MAAvBA,YAAY,MAAM,EACpBG,QAAQ,IAAI,CACV,CAAC,gBAAgB,EAAE,IAAI,CAAC,SAAS,CAAC,aAAa,CAAC,gDAAgD,CAAC;oBAIrG,MAAMC,aAAa,MAAOL,AAAAA,CAAAA,qBACtBM,yBAAAA,cAAcA,GACdC,yBAAAA,IAAG,AAAHA,EAAM/B,MAAM,eAAe,EAAE;wBAC/B,SAASE;wBACT,eAAeF,MAAM,YAAY;wBACjC,eAAe,IAAI,CAAC,SAAS,CAAC,aAAa;wBAC3CyB;wBACA,aAAahC;wBACb,qBAAqB,IAAI,CAAC,mBAAmB;wBAC7C,aAAagB;wBACbI;wBACAC;oBACF;oBACA9B,MAAM,cAAcgD,KAAK,SAAS,CAACH,YAAY,MAAM;oBAErD,MAAM,EACJI,OAAO,EACPC,GAAG,EACHC,IAAI,EACJC,KAAK,EACLC,KAAK,EACLC,WAAW,EACXC,iBAAiB,EAClB,GAAGV;oBAEJ5B,gBAAgB,IAAI,CAAC,GAAG,GAAG;wBACzB,GAAIA,gBAAgB,IAAI,CAAC,GAAG,IAAI,CAAC,CAAC;wBAClCqC;oBACF;oBACArC,gBAAgB,IAAI,CAAC,KAAK,GAAGoC;oBAC7BpC,gBAAgB,IAAI,CAAC,iBAAiB,GAAGsC;oBACzCtC,gBAAgB,IAAI,CAAC,MAAM,GAAG;wBAC5B,SAASgC,WAAW,EAAE;wBACtBC;wBACAC;wBACA,UAAUN,WAAW,QAAQ;wBAC7B,QAAQP;wBACR,wBAAwBO,WAAW,sBAAsB;oBAC3D;oBACA5B,gBAAgB,SAAS,GAAGC;oBAE5BC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,CAACiC,OAAO,CAAC,oBAAoB,EAAEA,MAAM,EAAE,EAAEF,OAAO,IAAI;oBAE3D,OAAO;wBACL,OAAO;4BACL,KAAK;wBACP;oBACF;gBACF;YACF,GACA;gBACE,gBAAgB;YAClB;YAGF,MAAML,aAAavB,QAAQ;YAG3B,MAAMd,QAAQqC,YAAY,WAAW,EAAE;YACvCX,SAAS,IAAI,IAAKW,YAAY,YAAY,EAAE;YAE5C,IAAIW;YACJ,IAAI;gBACFA,cAAc,MAAM,IAAI,CAAC,uBAAuB,CAC9ChD,OACAC,wBACAC,6BACA;oBACEiB;oBACA,SAAS;gBACX;YAEJ,EAAE,OAAOyB,OAAO;gBACd,OAAOvC,QAAQ,eAAe,CAC5B,CAAC,4CAA4C,EAAEuC,MAAM,SAAS,EAAEJ,KAAK,SAAS,CAC5ExC,QACC;YAEP;YACA,IAAI,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,EACjDoC,QAAQ,IAAI,CACV,8FACA,IAAI,CAAC,mBAAmB,CAAC,sBAAsB;YAKnD,IAAI;gBACF,MAAMtB,SAAS,MAAMT,QAAQ,YAAY,CAAC2C,YAAY,KAAK;gBAC3DlB,eAAehB,QAAQ;YACzB,EAAE,OAAO8B,OAAY;gBAEnBf;gBACA,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,GAAG,CAAC,MAAM,EAAEoB,AAAAA,IAAAA,mCAAAA,qBAAAA,AAAAA,IAAwB,iCAAiC,EAAEL,OAAO,WAAWM,OAAON,QAAQ;gBACvJpD,MACE,yFACAoD,iBAAiBO,QAAQP,MAAM,OAAO,GAAGM,OAAON,QAChD,6CACAf;YAEJ;YAEA,IAAIA,8BAA8BnC,uCAChC,OAAOW,QAAQ,eAAe,CAAC;YAIjC,IAAI,CAACgC,YAAY,wBACf;YAIF,EAAEZ;YAEF,IAAIA,cAAcE,sBAAsB;gBACtC,MAAMyB,WAAW,CAAC,UAAU,EAAEzB,qBAAqB,4JAA4J,CAAC;gBAChN,OAAOtB,QAAQ,eAAe,CAAC+C;YACjC;YAEA,IAAI,CAAC,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,EAClD,IAAI,CAAC,mBAAmB,CAAC,sBAAsB,GAAG,CAAC,MAAM,EAAEH,AAAAA,IAAAA,mCAAAA,qBAAAA,AAAAA,IAAwB,gDAAgD,CAAC;QAExI;QAEA,OAAO;YACL,QAAQ;gBACNvB;gBACA,QAAQI;YACV;YACAlB;QACF;IACF;IAEQ,oBACNyC,IAAsE,EACtEC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACpC;QACA,MAAMC,YAA4C;YAChD,MAAM;YACN,SAASL;YACT,OAAO;gBACL,YAAYI,mBACP;oBACCH;oBACAG;gBACF,IACAH;YACN;YACA,UAAU,OAAO9C,OAAOmD;gBACtB,MAAM,EAAEpD,IAAI,EAAE,GAAGoD;gBACjB,IAAIC;gBACJ,MAAMC,YAAY,CAACC;oBACjBF,YAAYE;oBACZvD,KAAK,GAAG,GAAG;wBACTuD;oBACF;gBACF;gBAGA,MAAMpD,YAAYiD,YAAY,SAAS;gBACvChD,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,WAAW;gBAElB,MAAMqD,mBAAmBV,AAAS,YAATA;gBACzB,IAAIW,cAAcV;gBAClB,IAAIW,cAAc;gBAClB,IAAIF,oBAAqBV,CAAAA,AAAS,aAATA,QAAqBA,AAAS,cAATA,IAAiB,GAAI;oBACjEY,cAAc;oBACd,MAAMC,gBACJb,AAAS,aAATA,OACI,CAAC,kDAAkD,EAAEC,QAAQ,GAC7D,CAAC,+GAA+G,EAAEA,QAAQ;oBAChIU,cAAc;wBACZ,CAACC,YAAY,EAAEC;oBACjB;gBACF,OAAO,IAAIH,kBACTC,cAAc;oBACZ,CAACC,YAAY,EAAE,GAAGZ,KAAK,EAAE,EAAEC,QAAQ;gBACrC;gBAGF,IAAIa;gBAEJ,IAAIC,uBAAuB;gBAC3B,IAAIZ,KAAK,eAAe,IAAI,CAAC,SAAS,CAAC,mBAAmB,EAAE;oBAC1DhE,MAAM;oBACN,MAAM6E,OAAO,MAAM,IAAI,CAAC,SAAS,CAAC,mBAAmB;oBACrDD,uBAAuB,MAAME,AAAAA,IAAAA,0BAAAA,iBAAAA,AAAAA,EAC3BD,MACA,KACA,OACAb,KAAK,gBAAgB;gBAEzB;gBAEA,IAAI;oBACFW,gBAAgB,MAAM,IAAI,CAAC,OAAO,CAAC,OAAO,CACxCH,aACAT,aACAC,KACAY,sBACAX;gBAEJ,EAAE,OAAOb,OAAO;oBACd,IAAIA,iBAAiB2B,kCAAAA,YAAYA,EAC/BV,UAAUjB,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAEA,MAAM,EAAE4B,IAAI,EAAE3B,KAAK,EAAE4B,OAAO,EAAEX,IAAI,EAAEf,iBAAiB,EAAE,GAAGoB;gBAC1DN,UAAUC;gBACVvD,KAAK,iBAAiB,GAAGwC;gBAEzB,IAAI2B,eAAeF;gBACnB,IAAIT,kBAEF,IAAI,AAAgB,YAAhB,OAAOS,MACTE,eAAeF;qBACV,IAAInB,AAAS,cAATA,MAEPqB,eADEF,QAAAA,OACa,QAECA,IAAY,CAACP,YAAY;qBAEtC,IAAIO,QAAAA,MACTE,eAAe;qBACV;oBACL/D,IAAAA,sBAAAA,MAAAA,AAAAA,EACE6D,MAAM,CAACP,YAAY,KAAKrC,QACxB;oBAEF8C,eAAgBF,IAAY,CAACP,YAAY;gBAC3C;gBAGF,IAAIZ,AAAS,aAATA,QAAqB,CAACqB,cAAc;oBACtCnE,KAAK,KAAK,GAAGsC;oBACbtC,KAAK,OAAO,GAAGkE;oBACf,MAAM,IAAItB,MAAM,CAAC,kBAAkB,EAAEsB,SAAS;gBAChD;gBAEA,OAAO;oBACL,QAAQC;oBACR,KAAKd;oBACLf;oBACA4B;gBACF;YACF;QACF;QAEA,OAAOf;IACT;IACA,MAAM,yBACJL,IAA0D,EAC1DC,MAA2B,EAC3BC,WAAyB,EACzBC,GAA0B,EAC1BC,gBAAoC,EACP;QAC7B,MAAMpD,UAAU,IAAI,CAAC,sBAAsB,CACzCC,AAAAA,IAAAA,qCAAAA,YAAAA,AAAAA,EACE+C,MACA,AAAkB,YAAlB,OAAOC,SAAsBA,SAASd,KAAK,SAAS,CAACc;QAIzD,MAAMI,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9CL,MACAC,QACAC,aACAC,KACAC;QAGF,MAAM7C,SAASP,QAAQ,SAAS;QAChC,MAAMS,SAAS,MAAMT,QAAQ,YAAY,CAACqD;QAE1C,IAAI,CAAC5C,QACH,MAAM,IAAIqC,MACR;QAIJ,MAAM,EAAEpC,MAAM,EAAE0D,OAAO,EAAE,GAAG3D;QAE5B,OAAO;YACLC;YACA0D;YACA7D;QACF;IACF;IAEA,MAAM,aAAa+D,MAAc,EAAEC,YAA0B,EAAE;QAC7D,OAAO,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;YACtCD;QACF;IACF;IAEA,MAAM,QACJE,SAAsB,EACtBrB,GAA+B,EAC/BD,WAAyB,EACO;QAChC,MAAM,EAAEuB,UAAU,EAAErB,gBAAgB,EAAE,GAAGsB,AAAAA,IAAAA,kCAAAA,WAAAA,AAAAA,EAAYF;QAErD,MAAMG,cAAc,CAAC,SAAS,EAAEF,YAAY;QAC5C,MAAMzE,UAAU,IAAI,CAAC,sBAAsB,CACzCC,AAAAA,IAAAA,qCAAAA,YAAAA,AAAAA,EAAa,WAAW0E;QAE1B,MAAMpE,SAASP,QAAQ,SAAS;QAChC,MAAM,EACJ4E,SAAS,EACTC,eAAe,EACfC,WAAW,EACXC,kBAAkB,EAClB,GAAGC,SACJ,GAAG7B;QACJ,MAAM8B,oBAA0C;YAC9CH;YACAC;YACA,GAAGC,OAAO;QACZ;QAEA1E,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOkE,WAAW;QAClBlE,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOsE,WAAW;QAClBtE,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOuE,iBAAiB;QAExBvE,IAAAA,sBAAAA,MAAAA,AAAAA,EACEuE,mBAAmBD,WACnB,CAAC,iGAAiG,EAAEC,gBAAgB,aAAa,EAAED,UAAU,CAAC,CAAC;QAGjJ,MAAMM,mBAAmBC,KAAK,GAAG;QACjC,IAAIC,iBAAiBF;QACrB,IAAIG,eAAe;QAEnB,MAAOD,iBAAiBF,oBAAoBN,UAAW;YACrD,MAAMU,oBAAoBH,KAAK,GAAG;YAClCC,iBAAiBE;YACjB,MAAMjC,YAAY,MAAM,IAAI,CAAC,mBAAmB,CAC9C,WACAoB,YACAvB,aACA+B,mBACA7B;YAGF,MAAM3C,SAAU,MAAMT,QAAQ,YAAY,CAACqD;YAO3C,IAAI5C,QAAQ,QACV,OAAO;gBACL,QAAQc;gBACRhB;YACF;YAGF8E,eACE5E,QAAQ,WACP,CAACA,UAAU,CAAC,0BAA0B,EAAEgE,YAAY,IACrD,CAAC,0CAA0C,EAAEA,YAAY;YAC3D,MAAMc,MAAMJ,KAAK,GAAG;YACpB,IAAII,MAAMD,oBAAoBT,iBAAiB;gBAC7C,MAAMW,gBAAgBX,kBAAmBU,CAAAA,MAAMD,iBAAgB;gBAC/D,MAAMG,YAAY,IAAI,CAAC,WAAW,CAAC,eAAe,CAAC;oBACjD,QAAQD;gBACV;gBACA,MAAMxF,QAAQ,MAAM,CAACyF;YACvB;QACF;QAEA,OAAOzF,QAAQ,eAAe,CAAC,CAAC,iBAAiB,EAAEqF,cAAc;IACnE;IAlmBA,YACEK,iBAAoC,EACpCC,OAAgB,EAChBC,IAMC,CACD;QAjCF;QAEA;QAEA;QAEA,uBAAiB,uBAAjB;QAEA,uBAAiB,eAAjB;QAEA,uBAAQ,uBAAR;QAEA;QAEA,uBAAiB,SAAjB;QAEA;QAkBE,IAAI,CAAC,SAAS,GAAGF;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC,KAAK,SAAS;QAC/B,IAAI,CAAC,mBAAmB,GAAGA,MAAM;QACjC,IAAI,CAAC,oBAAoB,GAAGA,KAAK,oBAAoB;QACrD,IAAI,CAAC,KAAK,GAAGA,KAAK,KAAK;QACvB,IAAI,CAAC,mBAAmB,GAAG,IAAIC,yBAAAA,mBAAmBA;QAClD,IAAI,CAAC,mBAAmB,GAAGD,KAAK,WAAW;QAC3C,IAAI,CAAC,WAAW,GAAG,IAAIE,yCAAAA,WAAWA,CAAC;YACjCJ;YACAC;YACA,WAAWC,KAAK,SAAS;YACzB,aAAa,IAAI,CAAC,cAAc;QAClC;IACF;AA0kBF;AAEO,eAAezE,gBACpBuE,iBAAoC,EACpCxE,iBAAuC,EACvCW,MAAwB;IAExB,IAAI,CAACX,mBAAmB,QACtB,OAAOW;IAGT,IAAI,CAAC6D,kBAAkB,2BAA2B,EAChD,MAAM,IAAI5C,MACR,CAAC,gCAAgC,EAAE4C,kBAAkB,aAAa,EAAE;IAIxE,MAAMK,UAAU,OAAOC;QACrB,MAAMA,QAAQ,MAAM,CAAC9E;IACvB;IAEA,MAAM,EAAE+E,OAAO,EAAEC,QAAQ,EAAE,GACzB,MAAMR,kBAAkB,2BAA2B,CAACK;IACtD,IAAI;QACF,MAAMtF,SAAS,MAAMoB;QAErB,MAAMU,QAAQ2D;QACd,IAAI3D,OACF,MAAMA;QAER,OAAO9B;IACT,SAAU;QACRwF;IACF;AACF"}
|
package/dist/lib/agent/utils.js
CHANGED
|
@@ -44,6 +44,7 @@ __webpack_require__.d(__webpack_exports__, {
|
|
|
44
44
|
ifPlanLocateParamIsBbox: ()=>ifPlanLocateParamIsBbox,
|
|
45
45
|
printReportMsg: ()=>printReportMsg
|
|
46
46
|
});
|
|
47
|
+
const external_screenshot_item_js_namespaceObject = require("../screenshot-item.js");
|
|
47
48
|
const external_utils_js_namespaceObject = require("../utils.js");
|
|
48
49
|
const env_namespaceObject = require("@midscene/shared/env");
|
|
49
50
|
const extractor_namespaceObject = require("@midscene/shared/extractor");
|
|
@@ -69,9 +70,10 @@ async function commonContextParser(interfaceInstance, _opt) {
|
|
|
69
70
|
debugProfile('will get size');
|
|
70
71
|
const size = await interfaceInstance.size();
|
|
71
72
|
debugProfile(`size: ${size.width}x${size.height} dpr: ${size.dpr}`);
|
|
73
|
+
const screenshot = external_screenshot_item_js_namespaceObject.ScreenshotItem.create(screenshotBase64);
|
|
72
74
|
return {
|
|
73
75
|
size,
|
|
74
|
-
|
|
76
|
+
screenshot
|
|
75
77
|
};
|
|
76
78
|
}
|
|
77
79
|
function getReportFileName(tag = 'web') {
|
|
@@ -148,7 +150,7 @@ async function matchElementFromCache(context, cacheEntry, cachePrompt, cacheable
|
|
|
148
150
|
return;
|
|
149
151
|
}
|
|
150
152
|
}
|
|
151
|
-
const getMidsceneVersion = ()=>"1.2.2-beta-
|
|
153
|
+
const getMidsceneVersion = ()=>"1.2.2-beta-20260116060040.0";
|
|
152
154
|
const parsePrompt = (prompt)=>{
|
|
153
155
|
if ('string' == typeof prompt) return {
|
|
154
156
|
textPrompt: prompt,
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"agent/utils.js","sources":["webpack/runtime/compat_get_default_export","webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../src/agent/utils.ts"],"sourcesContent":["// getDefaultExport function for compatibility with non-ESM modules\n__webpack_require__.n = (module) => {\n\tvar getter = module && module.__esModule ?\n\t\t() => (module['default']) :\n\t\t() => (module);\n\t__webpack_require__.d(getter, { a: getter });\n\treturn getter;\n};\n","__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type { TMultimodalPrompt, TUserPrompt } from '@/common';\nimport type { AbstractInterface } from '@/device';\nimport type {\n ElementCacheFeature,\n LocateResultElement,\n PlanningLocateParam,\n UIContext,\n} from '@/types';\nimport { uploadTestInfoToServer } from '@/utils';\nimport {\n MIDSCENE_REPORT_TAG_NAME,\n globalConfigManager,\n} from '@midscene/shared/env';\nimport { generateElementByPosition } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { assert, logMsg, uuid } from '@midscene/shared/utils';\nimport dayjs from 'dayjs';\nimport type { TaskCache } from './task-cache';\nimport { debug as cacheDebug } from './task-cache';\n\nconst debugProfile = getDebug('web:tool:profile');\n\nexport async function commonContextParser(\n interfaceInstance: AbstractInterface,\n _opt: { uploadServerUrl?: string },\n): Promise<UIContext> {\n assert(interfaceInstance, 'interfaceInstance is required');\n\n debugProfile('Getting interface description');\n const description = interfaceInstance.describe?.() || '';\n debugProfile('Interface description end');\n\n debugProfile('Uploading test info to server');\n uploadTestInfoToServer({\n testUrl: description,\n serverUrl: _opt.uploadServerUrl,\n });\n debugProfile('UploadTestInfoToServer end');\n\n const screenshotBase64 = await interfaceInstance.screenshotBase64();\n assert(screenshotBase64!, 'screenshotBase64 is required');\n\n debugProfile('will get size');\n const size = await interfaceInstance.size();\n debugProfile(`size: ${size.width}x${size.height} dpr: ${size.dpr}`);\n\n return {\n size,\n screenshotBase64: screenshotBase64!,\n };\n}\n\nexport function getReportFileName(tag = 'web') {\n const reportTagName = globalConfigManager.getEnvConfigValue(\n MIDSCENE_REPORT_TAG_NAME,\n );\n const dateTimeInFileName = dayjs().format('YYYY-MM-DD_HH-mm-ss');\n // ensure uniqueness at the same time\n const uniqueId = uuid().substring(0, 8);\n return `${reportTagName || tag}-${dateTimeInFileName}-${uniqueId}`;\n}\n\nexport function printReportMsg(filepath: string) {\n logMsg(`Midscene - report file updated: ${filepath}`);\n}\n\n/**\n * Get the current execution file name\n * @returns The name of the current execution file\n */\nexport function getCurrentExecutionFile(trace?: string): string | false {\n const error = new Error();\n const stackTrace = trace || error.stack;\n const pkgDir = process.cwd() || '';\n if (stackTrace) {\n const stackLines = stackTrace.split('\\n');\n for (const line of stackLines) {\n if (\n line.includes('.spec.') ||\n line.includes('.test.') ||\n line.includes('.ts') ||\n line.includes('.js')\n ) {\n const match = line.match(/(?:at\\s+)?(.*?\\.(?:spec|test)\\.[jt]s)/);\n if (match?.[1]) {\n const targetFileName = match[1]\n .replace(pkgDir, '')\n .trim()\n .replace('at ', '');\n return targetFileName;\n }\n }\n }\n }\n return false;\n}\n\nconst testFileIndex = new Map<string, number>();\n\nexport function generateCacheId(fileName?: string): string {\n let taskFile = fileName || getCurrentExecutionFile();\n if (!taskFile) {\n taskFile = uuid();\n console.warn(\n 'Midscene - using random UUID for cache id. Cache may be invalid.',\n );\n }\n\n if (testFileIndex.has(taskFile)) {\n const currentIndex = testFileIndex.get(taskFile);\n if (currentIndex !== undefined) {\n testFileIndex.set(taskFile, currentIndex + 1);\n }\n } else {\n testFileIndex.set(taskFile, 1);\n }\n return `${taskFile}-${testFileIndex.get(taskFile)}`;\n}\n\nexport function ifPlanLocateParamIsBbox(\n planLocateParam: PlanningLocateParam,\n): boolean {\n return !!(\n planLocateParam.bbox &&\n Array.isArray(planLocateParam.bbox) &&\n planLocateParam.bbox.length === 4\n );\n}\n\nexport function matchElementFromPlan(\n planLocateParam: PlanningLocateParam,\n): LocateResultElement | undefined {\n if (!planLocateParam) {\n return undefined;\n }\n\n if (planLocateParam.bbox) {\n const centerPosition = {\n x: Math.floor((planLocateParam.bbox[0] + planLocateParam.bbox[2]) / 2),\n y: Math.floor((planLocateParam.bbox[1] + planLocateParam.bbox[3]) / 2),\n };\n\n const element = generateElementByPosition(\n centerPosition,\n typeof planLocateParam.prompt === 'string'\n ? planLocateParam.prompt\n : planLocateParam.prompt?.prompt || '',\n );\n return element;\n }\n\n return undefined;\n}\n\nexport async function matchElementFromCache(\n context: {\n taskCache?: TaskCache;\n interfaceInstance: AbstractInterface;\n },\n cacheEntry: ElementCacheFeature | undefined,\n cachePrompt: TUserPrompt,\n cacheable: boolean | undefined,\n): Promise<LocateResultElement | undefined> {\n if (!cacheEntry) {\n return undefined;\n }\n\n if (cacheable === false) {\n cacheDebug('cache disabled for prompt: %s', cachePrompt);\n return undefined;\n }\n\n if (!context.taskCache?.isCacheResultUsed) {\n return undefined;\n }\n\n if (!context.interfaceInstance.rectMatchesCacheFeature) {\n cacheDebug(\n 'interface does not implement rectMatchesCacheFeature, skip cache',\n );\n return undefined;\n }\n\n try {\n const rect =\n await context.interfaceInstance.rectMatchesCacheFeature(cacheEntry);\n const element: LocateResultElement = {\n center: [\n Math.round(rect.left + rect.width / 2),\n Math.round(rect.top + rect.height / 2),\n ],\n rect,\n description:\n typeof cachePrompt === 'string'\n ? cachePrompt\n : cachePrompt.prompt || '',\n };\n\n cacheDebug('cache hit, prompt: %s', cachePrompt);\n return element;\n } catch (error) {\n cacheDebug('rectMatchesCacheFeature error: %s', error);\n return undefined;\n }\n}\n\ndeclare const __VERSION__: string | undefined;\n\nexport const getMidsceneVersion = (): string => {\n if (typeof __VERSION__ !== 'undefined') {\n return __VERSION__;\n } else if (\n process.env.__VERSION__ &&\n process.env.__VERSION__ !== 'undefined'\n ) {\n return process.env.__VERSION__;\n }\n throw new Error('__VERSION__ inject failed during build');\n};\n\nexport const parsePrompt = (\n prompt: TUserPrompt,\n): {\n textPrompt: string;\n multimodalPrompt?: TMultimodalPrompt;\n} => {\n if (typeof prompt === 'string') {\n return {\n textPrompt: prompt,\n multimodalPrompt: undefined,\n };\n }\n return {\n textPrompt: prompt.prompt,\n multimodalPrompt: prompt.images\n ? {\n images: prompt.images,\n convertHttpImage2Base64: !!prompt.convertHttpImage2Base64,\n }\n : undefined,\n };\n};\n"],"names":["__webpack_require__","module","getter","definition","key","Object","obj","prop","Symbol","debugProfile","getDebug","commonContextParser","interfaceInstance","_opt","assert","description","uploadTestInfoToServer","screenshotBase64","size","getReportFileName","tag","reportTagName","globalConfigManager","MIDSCENE_REPORT_TAG_NAME","dateTimeInFileName","dayjs","uniqueId","uuid","printReportMsg","filepath","logMsg","getCurrentExecutionFile","trace","error","Error","stackTrace","pkgDir","process","stackLines","line","match","targetFileName","testFileIndex","Map","generateCacheId","fileName","taskFile","console","currentIndex","undefined","ifPlanLocateParamIsBbox","planLocateParam","Array","matchElementFromPlan","centerPosition","Math","element","generateElementByPosition","matchElementFromCache","context","cacheEntry","cachePrompt","cacheable","cacheDebug","rect","getMidsceneVersion","__VERSION__","parsePrompt","prompt"],"mappings":";;;IACAA,oBAAoB,CAAC,GAAG,CAACC;QACxB,IAAIC,SAASD,UAAUA,OAAO,UAAU,GACvC,IAAOA,MAAM,CAAC,UAAU,GACxB,IAAOA;QACRD,oBAAoB,CAAC,CAACE,QAAQ;YAAE,GAAGA;QAAO;QAC1C,OAAOA;IACR;;;ICPAF,oBAAoB,CAAC,GAAG,CAAC,UAASG;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGH,oBAAoB,CAAC,CAACG,YAAYC,QAAQ,CAACJ,oBAAoB,CAAC,CAAC,UAASI,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAJ,oBAAoB,CAAC,GAAG,CAACM,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFP,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOQ,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;;;;;ACeA,MAAMI,eAAeC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAEvB,eAAeC,oBACpBC,iBAAoC,EACpCC,IAAkC;IAElCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,mBAAmB;IAE1BH,aAAa;IACb,MAAMM,cAAcH,kBAAkB,QAAQ,QAAQ;IACtDH,aAAa;IAEbA,aAAa;IACbO,IAAAA,kCAAAA,sBAAAA,AAAAA,EAAuB;QACrB,SAASD;QACT,WAAWF,KAAK,eAAe;IACjC;IACAJ,aAAa;IAEb,MAAMQ,mBAAmB,MAAML,kBAAkB,gBAAgB;IACjEE,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOG,kBAAmB;IAE1BR,aAAa;IACb,MAAMS,OAAO,MAAMN,kBAAkB,IAAI;IACzCH,aAAa,CAAC,MAAM,EAAES,KAAK,KAAK,CAAC,CAAC,EAAEA,KAAK,MAAM,CAAC,MAAM,EAAEA,KAAK,GAAG,EAAE;IAElE,OAAO;QACLA;QACA,kBAAkBD;IACpB;AACF;AAEO,SAASE,kBAAkBC,MAAM,KAAK;IAC3C,MAAMC,gBAAgBC,oBAAAA,mBAAAA,CAAAA,iBAAqC,CACzDC,oBAAAA,wBAAwBA;IAE1B,MAAMC,qBAAqBC,2BAAQ,MAAM,CAAC;IAE1C,MAAMC,WAAWC,AAAAA,IAAAA,sBAAAA,IAAAA,AAAAA,IAAO,SAAS,CAAC,GAAG;IACrC,OAAO,GAAGN,iBAAiBD,IAAI,CAAC,EAAEI,mBAAmB,CAAC,EAAEE,UAAU;AACpE;AAEO,SAASE,eAAeC,QAAgB;IAC7CC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,CAAC,gCAAgC,EAAED,UAAU;AACtD;AAMO,SAASE,wBAAwBC,KAAc;IACpD,MAAMC,QAAQ,IAAIC;IAClB,MAAMC,aAAaH,SAASC,MAAM,KAAK;IACvC,MAAMG,SAASC,QAAQ,GAAG,MAAM;IAChC,IAAIF,YAAY;QACd,MAAMG,aAAaH,WAAW,KAAK,CAAC;QACpC,KAAK,MAAMI,QAAQD,WACjB,IACEC,KAAK,QAAQ,CAAC,aACdA,KAAK,QAAQ,CAAC,aACdA,KAAK,QAAQ,CAAC,UACdA,KAAK,QAAQ,CAAC,QACd;YACA,MAAMC,QAAQD,KAAK,KAAK,CAAC;YACzB,IAAIC,OAAO,CAAC,EAAE,EAAE;gBACd,MAAMC,iBAAiBD,KAAK,CAAC,EAAE,CAC5B,OAAO,CAACJ,QAAQ,IAChB,IAAI,GACJ,OAAO,CAAC,OAAO;gBAClB,OAAOK;YACT;QACF;IAEJ;IACA,OAAO;AACT;AAEA,MAAMC,gBAAgB,IAAIC;AAEnB,SAASC,gBAAgBC,QAAiB;IAC/C,IAAIC,WAAWD,YAAYd;IAC3B,IAAI,CAACe,UAAU;QACbA,WAAWnB,AAAAA,IAAAA,sBAAAA,IAAAA,AAAAA;QACXoB,QAAQ,IAAI,CACV;IAEJ;IAEA,IAAIL,cAAc,GAAG,CAACI,WAAW;QAC/B,MAAME,eAAeN,cAAc,GAAG,CAACI;QACvC,IAAIE,AAAiBC,WAAjBD,cACFN,cAAc,GAAG,CAACI,UAAUE,eAAe;IAE/C,OACEN,cAAc,GAAG,CAACI,UAAU;IAE9B,OAAO,GAAGA,SAAS,CAAC,EAAEJ,cAAc,GAAG,CAACI,WAAW;AACrD;AAEO,SAASI,wBACdC,eAAoC;IAEpC,OAAO,CAAC,CACNA,CAAAA,gBAAgB,IAAI,IACpBC,MAAM,OAAO,CAACD,gBAAgB,IAAI,KAClCA,AAAgC,MAAhCA,gBAAgB,IAAI,CAAC,MAAM,AAAK;AAEpC;AAEO,SAASE,qBACdF,eAAoC;IAEpC,IAAI,CAACA,iBACH;IAGF,IAAIA,gBAAgB,IAAI,EAAE;QACxB,MAAMG,iBAAiB;YACrB,GAAGC,KAAK,KAAK,CAAEJ,AAAAA,CAAAA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAC,IAAK;YACpE,GAAGI,KAAK,KAAK,CAAEJ,AAAAA,CAAAA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAC,IAAK;QACtE;QAEA,MAAMK,UAAUC,AAAAA,IAAAA,0BAAAA,yBAAAA,AAAAA,EACdH,gBACA,AAAkC,YAAlC,OAAOH,gBAAgB,MAAM,GACzBA,gBAAgB,MAAM,GACtBA,gBAAgB,MAAM,EAAE,UAAU;QAExC,OAAOK;IACT;AAGF;AAEO,eAAeE,sBACpBC,OAGC,EACDC,UAA2C,EAC3CC,WAAwB,EACxBC,SAA8B;IAE9B,IAAI,CAACF,YACH;IAGF,IAAIE,AAAc,UAAdA,WAAqB,YACvBC,AAAAA,IAAAA,uCAAAA,KAAAA,AAAAA,EAAW,iCAAiCF;IAI9C,IAAI,CAACF,QAAQ,SAAS,EAAE,mBACtB;IAGF,IAAI,CAACA,QAAQ,iBAAiB,CAAC,uBAAuB,EAAE,YACtDI,AAAAA,IAAAA,uCAAAA,KAAAA,AAAAA,EACE;IAKJ,IAAI;QACF,MAAMC,OACJ,MAAML,QAAQ,iBAAiB,CAAC,uBAAuB,CAACC;QAC1D,MAAMJ,UAA+B;YACnC,QAAQ;gBACND,KAAK,KAAK,CAACS,KAAK,IAAI,GAAGA,KAAK,KAAK,GAAG;gBACpCT,KAAK,KAAK,CAACS,KAAK,GAAG,GAAGA,KAAK,MAAM,GAAG;aACrC;YACDA;YACA,aACE,AAAuB,YAAvB,OAAOH,cACHA,cACAA,YAAY,MAAM,IAAI;QAC9B;QAEAE,IAAAA,uCAAAA,KAAAA,AAAAA,EAAW,yBAAyBF;QACpC,OAAOL;IACT,EAAE,OAAOvB,OAAO;QACd8B,IAAAA,uCAAAA,KAAAA,AAAAA,EAAW,qCAAqC9B;QAChD;IACF;AACF;AAIO,MAAMgC,qBAAqB,IAEvBC;AAUJ,MAAMC,cAAc,CACzBC;IAKA,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAO;QACL,YAAYA;QACZ,kBAAkBnB;IACpB;IAEF,OAAO;QACL,YAAYmB,OAAO,MAAM;QACzB,kBAAkBA,OAAO,MAAM,GAC3B;YACE,QAAQA,OAAO,MAAM;YACrB,yBAAyB,CAAC,CAACA,OAAO,uBAAuB;QAC3D,IACAnB;IACN;AACF"}
|
|
1
|
+
{"version":3,"file":"agent/utils.js","sources":["webpack/runtime/compat_get_default_export","webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../src/agent/utils.ts"],"sourcesContent":["// getDefaultExport function for compatibility with non-ESM modules\n__webpack_require__.n = (module) => {\n\tvar getter = module && module.__esModule ?\n\t\t() => (module['default']) :\n\t\t() => (module);\n\t__webpack_require__.d(getter, { a: getter });\n\treturn getter;\n};\n","__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type { TMultimodalPrompt, TUserPrompt } from '@/common';\nimport type { AbstractInterface } from '@/device';\nimport { ScreenshotItem } from '@/screenshot-item';\nimport type {\n ElementCacheFeature,\n LocateResultElement,\n PlanningLocateParam,\n UIContext,\n} from '@/types';\nimport { uploadTestInfoToServer } from '@/utils';\nimport {\n MIDSCENE_REPORT_TAG_NAME,\n globalConfigManager,\n} from '@midscene/shared/env';\nimport { generateElementByPosition } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { assert, logMsg, uuid } from '@midscene/shared/utils';\nimport dayjs from 'dayjs';\nimport type { TaskCache } from './task-cache';\nimport { debug as cacheDebug } from './task-cache';\n\nconst debugProfile = getDebug('web:tool:profile');\n\nexport async function commonContextParser(\n interfaceInstance: AbstractInterface,\n _opt: { uploadServerUrl?: string },\n): Promise<UIContext> {\n assert(interfaceInstance, 'interfaceInstance is required');\n\n debugProfile('Getting interface description');\n const description = interfaceInstance.describe?.() || '';\n debugProfile('Interface description end');\n\n debugProfile('Uploading test info to server');\n uploadTestInfoToServer({\n testUrl: description,\n serverUrl: _opt.uploadServerUrl,\n });\n debugProfile('UploadTestInfoToServer end');\n\n const screenshotBase64 = await interfaceInstance.screenshotBase64();\n assert(screenshotBase64!, 'screenshotBase64 is required');\n\n debugProfile('will get size');\n const size = await interfaceInstance.size();\n debugProfile(`size: ${size.width}x${size.height} dpr: ${size.dpr}`);\n\n const screenshot = ScreenshotItem.create(screenshotBase64!);\n\n return {\n size,\n screenshot,\n };\n}\n\nexport function getReportFileName(tag = 'web') {\n const reportTagName = globalConfigManager.getEnvConfigValue(\n MIDSCENE_REPORT_TAG_NAME,\n );\n const dateTimeInFileName = dayjs().format('YYYY-MM-DD_HH-mm-ss');\n // ensure uniqueness at the same time\n const uniqueId = uuid().substring(0, 8);\n return `${reportTagName || tag}-${dateTimeInFileName}-${uniqueId}`;\n}\n\nexport function printReportMsg(filepath: string) {\n logMsg(`Midscene - report file updated: ${filepath}`);\n}\n\n/**\n * Get the current execution file name\n * @returns The name of the current execution file\n */\nexport function getCurrentExecutionFile(trace?: string): string | false {\n const error = new Error();\n const stackTrace = trace || error.stack;\n const pkgDir = process.cwd() || '';\n if (stackTrace) {\n const stackLines = stackTrace.split('\\n');\n for (const line of stackLines) {\n if (\n line.includes('.spec.') ||\n line.includes('.test.') ||\n line.includes('.ts') ||\n line.includes('.js')\n ) {\n const match = line.match(/(?:at\\s+)?(.*?\\.(?:spec|test)\\.[jt]s)/);\n if (match?.[1]) {\n const targetFileName = match[1]\n .replace(pkgDir, '')\n .trim()\n .replace('at ', '');\n return targetFileName;\n }\n }\n }\n }\n return false;\n}\n\nconst testFileIndex = new Map<string, number>();\n\nexport function generateCacheId(fileName?: string): string {\n let taskFile = fileName || getCurrentExecutionFile();\n if (!taskFile) {\n taskFile = uuid();\n console.warn(\n 'Midscene - using random UUID for cache id. Cache may be invalid.',\n );\n }\n\n if (testFileIndex.has(taskFile)) {\n const currentIndex = testFileIndex.get(taskFile);\n if (currentIndex !== undefined) {\n testFileIndex.set(taskFile, currentIndex + 1);\n }\n } else {\n testFileIndex.set(taskFile, 1);\n }\n return `${taskFile}-${testFileIndex.get(taskFile)}`;\n}\n\nexport function ifPlanLocateParamIsBbox(\n planLocateParam: PlanningLocateParam,\n): boolean {\n return !!(\n planLocateParam.bbox &&\n Array.isArray(planLocateParam.bbox) &&\n planLocateParam.bbox.length === 4\n );\n}\n\nexport function matchElementFromPlan(\n planLocateParam: PlanningLocateParam,\n): LocateResultElement | undefined {\n if (!planLocateParam) {\n return undefined;\n }\n\n if (planLocateParam.bbox) {\n const centerPosition = {\n x: Math.floor((planLocateParam.bbox[0] + planLocateParam.bbox[2]) / 2),\n y: Math.floor((planLocateParam.bbox[1] + planLocateParam.bbox[3]) / 2),\n };\n\n const element = generateElementByPosition(\n centerPosition,\n typeof planLocateParam.prompt === 'string'\n ? planLocateParam.prompt\n : planLocateParam.prompt?.prompt || '',\n );\n return element;\n }\n\n return undefined;\n}\n\nexport async function matchElementFromCache(\n context: {\n taskCache?: TaskCache;\n interfaceInstance: AbstractInterface;\n },\n cacheEntry: ElementCacheFeature | undefined,\n cachePrompt: TUserPrompt,\n cacheable: boolean | undefined,\n): Promise<LocateResultElement | undefined> {\n if (!cacheEntry) {\n return undefined;\n }\n\n if (cacheable === false) {\n cacheDebug('cache disabled for prompt: %s', cachePrompt);\n return undefined;\n }\n\n if (!context.taskCache?.isCacheResultUsed) {\n return undefined;\n }\n\n if (!context.interfaceInstance.rectMatchesCacheFeature) {\n cacheDebug(\n 'interface does not implement rectMatchesCacheFeature, skip cache',\n );\n return undefined;\n }\n\n try {\n const rect =\n await context.interfaceInstance.rectMatchesCacheFeature(cacheEntry);\n const element: LocateResultElement = {\n center: [\n Math.round(rect.left + rect.width / 2),\n Math.round(rect.top + rect.height / 2),\n ],\n rect,\n description:\n typeof cachePrompt === 'string'\n ? cachePrompt\n : cachePrompt.prompt || '',\n };\n\n cacheDebug('cache hit, prompt: %s', cachePrompt);\n return element;\n } catch (error) {\n cacheDebug('rectMatchesCacheFeature error: %s', error);\n return undefined;\n }\n}\n\ndeclare const __VERSION__: string | undefined;\n\nexport const getMidsceneVersion = (): string => {\n if (typeof __VERSION__ !== 'undefined') {\n return __VERSION__;\n } else if (\n process.env.__VERSION__ &&\n process.env.__VERSION__ !== 'undefined'\n ) {\n return process.env.__VERSION__;\n }\n throw new Error('__VERSION__ inject failed during build');\n};\n\nexport const parsePrompt = (\n prompt: TUserPrompt,\n): {\n textPrompt: string;\n multimodalPrompt?: TMultimodalPrompt;\n} => {\n if (typeof prompt === 'string') {\n return {\n textPrompt: prompt,\n multimodalPrompt: undefined,\n };\n }\n return {\n textPrompt: prompt.prompt,\n multimodalPrompt: prompt.images\n ? {\n images: prompt.images,\n convertHttpImage2Base64: !!prompt.convertHttpImage2Base64,\n }\n : undefined,\n };\n};\n"],"names":["__webpack_require__","module","getter","definition","key","Object","obj","prop","Symbol","debugProfile","getDebug","commonContextParser","interfaceInstance","_opt","assert","description","uploadTestInfoToServer","screenshotBase64","size","screenshot","ScreenshotItem","getReportFileName","tag","reportTagName","globalConfigManager","MIDSCENE_REPORT_TAG_NAME","dateTimeInFileName","dayjs","uniqueId","uuid","printReportMsg","filepath","logMsg","getCurrentExecutionFile","trace","error","Error","stackTrace","pkgDir","process","stackLines","line","match","targetFileName","testFileIndex","Map","generateCacheId","fileName","taskFile","console","currentIndex","undefined","ifPlanLocateParamIsBbox","planLocateParam","Array","matchElementFromPlan","centerPosition","Math","element","generateElementByPosition","matchElementFromCache","context","cacheEntry","cachePrompt","cacheable","cacheDebug","rect","getMidsceneVersion","__VERSION__","parsePrompt","prompt"],"mappings":";;;IACAA,oBAAoB,CAAC,GAAG,CAACC;QACxB,IAAIC,SAASD,UAAUA,OAAO,UAAU,GACvC,IAAOA,MAAM,CAAC,UAAU,GACxB,IAAOA;QACRD,oBAAoB,CAAC,CAACE,QAAQ;YAAE,GAAGA;QAAO;QAC1C,OAAOA;IACR;;;ICPAF,oBAAoB,CAAC,GAAG,CAAC,UAASG;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGH,oBAAoB,CAAC,CAACG,YAAYC,QAAQ,CAACJ,oBAAoB,CAAC,CAAC,UAASI,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAJ,oBAAoB,CAAC,GAAG,CAACM,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFP,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOQ,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;;;;;;ACgBA,MAAMI,eAAeC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAEvB,eAAeC,oBACpBC,iBAAoC,EACpCC,IAAkC;IAElCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,mBAAmB;IAE1BH,aAAa;IACb,MAAMM,cAAcH,kBAAkB,QAAQ,QAAQ;IACtDH,aAAa;IAEbA,aAAa;IACbO,IAAAA,kCAAAA,sBAAAA,AAAAA,EAAuB;QACrB,SAASD;QACT,WAAWF,KAAK,eAAe;IACjC;IACAJ,aAAa;IAEb,MAAMQ,mBAAmB,MAAML,kBAAkB,gBAAgB;IACjEE,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOG,kBAAmB;IAE1BR,aAAa;IACb,MAAMS,OAAO,MAAMN,kBAAkB,IAAI;IACzCH,aAAa,CAAC,MAAM,EAAES,KAAK,KAAK,CAAC,CAAC,EAAEA,KAAK,MAAM,CAAC,MAAM,EAAEA,KAAK,GAAG,EAAE;IAElE,MAAMC,aAAaC,4CAAAA,cAAAA,CAAAA,MAAqB,CAACH;IAEzC,OAAO;QACLC;QACAC;IACF;AACF;AAEO,SAASE,kBAAkBC,MAAM,KAAK;IAC3C,MAAMC,gBAAgBC,oBAAAA,mBAAAA,CAAAA,iBAAqC,CACzDC,oBAAAA,wBAAwBA;IAE1B,MAAMC,qBAAqBC,2BAAQ,MAAM,CAAC;IAE1C,MAAMC,WAAWC,AAAAA,IAAAA,sBAAAA,IAAAA,AAAAA,IAAO,SAAS,CAAC,GAAG;IACrC,OAAO,GAAGN,iBAAiBD,IAAI,CAAC,EAAEI,mBAAmB,CAAC,EAAEE,UAAU;AACpE;AAEO,SAASE,eAAeC,QAAgB;IAC7CC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,CAAC,gCAAgC,EAAED,UAAU;AACtD;AAMO,SAASE,wBAAwBC,KAAc;IACpD,MAAMC,QAAQ,IAAIC;IAClB,MAAMC,aAAaH,SAASC,MAAM,KAAK;IACvC,MAAMG,SAASC,QAAQ,GAAG,MAAM;IAChC,IAAIF,YAAY;QACd,MAAMG,aAAaH,WAAW,KAAK,CAAC;QACpC,KAAK,MAAMI,QAAQD,WACjB,IACEC,KAAK,QAAQ,CAAC,aACdA,KAAK,QAAQ,CAAC,aACdA,KAAK,QAAQ,CAAC,UACdA,KAAK,QAAQ,CAAC,QACd;YACA,MAAMC,QAAQD,KAAK,KAAK,CAAC;YACzB,IAAIC,OAAO,CAAC,EAAE,EAAE;gBACd,MAAMC,iBAAiBD,KAAK,CAAC,EAAE,CAC5B,OAAO,CAACJ,QAAQ,IAChB,IAAI,GACJ,OAAO,CAAC,OAAO;gBAClB,OAAOK;YACT;QACF;IAEJ;IACA,OAAO;AACT;AAEA,MAAMC,gBAAgB,IAAIC;AAEnB,SAASC,gBAAgBC,QAAiB;IAC/C,IAAIC,WAAWD,YAAYd;IAC3B,IAAI,CAACe,UAAU;QACbA,WAAWnB,AAAAA,IAAAA,sBAAAA,IAAAA,AAAAA;QACXoB,QAAQ,IAAI,CACV;IAEJ;IAEA,IAAIL,cAAc,GAAG,CAACI,WAAW;QAC/B,MAAME,eAAeN,cAAc,GAAG,CAACI;QACvC,IAAIE,AAAiBC,WAAjBD,cACFN,cAAc,GAAG,CAACI,UAAUE,eAAe;IAE/C,OACEN,cAAc,GAAG,CAACI,UAAU;IAE9B,OAAO,GAAGA,SAAS,CAAC,EAAEJ,cAAc,GAAG,CAACI,WAAW;AACrD;AAEO,SAASI,wBACdC,eAAoC;IAEpC,OAAO,CAAC,CACNA,CAAAA,gBAAgB,IAAI,IACpBC,MAAM,OAAO,CAACD,gBAAgB,IAAI,KAClCA,AAAgC,MAAhCA,gBAAgB,IAAI,CAAC,MAAM,AAAK;AAEpC;AAEO,SAASE,qBACdF,eAAoC;IAEpC,IAAI,CAACA,iBACH;IAGF,IAAIA,gBAAgB,IAAI,EAAE;QACxB,MAAMG,iBAAiB;YACrB,GAAGC,KAAK,KAAK,CAAEJ,AAAAA,CAAAA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAC,IAAK;YACpE,GAAGI,KAAK,KAAK,CAAEJ,AAAAA,CAAAA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAC,IAAK;QACtE;QAEA,MAAMK,UAAUC,AAAAA,IAAAA,0BAAAA,yBAAAA,AAAAA,EACdH,gBACA,AAAkC,YAAlC,OAAOH,gBAAgB,MAAM,GACzBA,gBAAgB,MAAM,GACtBA,gBAAgB,MAAM,EAAE,UAAU;QAExC,OAAOK;IACT;AAGF;AAEO,eAAeE,sBACpBC,OAGC,EACDC,UAA2C,EAC3CC,WAAwB,EACxBC,SAA8B;IAE9B,IAAI,CAACF,YACH;IAGF,IAAIE,AAAc,UAAdA,WAAqB,YACvBC,AAAAA,IAAAA,uCAAAA,KAAAA,AAAAA,EAAW,iCAAiCF;IAI9C,IAAI,CAACF,QAAQ,SAAS,EAAE,mBACtB;IAGF,IAAI,CAACA,QAAQ,iBAAiB,CAAC,uBAAuB,EAAE,YACtDI,AAAAA,IAAAA,uCAAAA,KAAAA,AAAAA,EACE;IAKJ,IAAI;QACF,MAAMC,OACJ,MAAML,QAAQ,iBAAiB,CAAC,uBAAuB,CAACC;QAC1D,MAAMJ,UAA+B;YACnC,QAAQ;gBACND,KAAK,KAAK,CAACS,KAAK,IAAI,GAAGA,KAAK,KAAK,GAAG;gBACpCT,KAAK,KAAK,CAACS,KAAK,GAAG,GAAGA,KAAK,MAAM,GAAG;aACrC;YACDA;YACA,aACE,AAAuB,YAAvB,OAAOH,cACHA,cACAA,YAAY,MAAM,IAAI;QAC9B;QAEAE,IAAAA,uCAAAA,KAAAA,AAAAA,EAAW,yBAAyBF;QACpC,OAAOL;IACT,EAAE,OAAOvB,OAAO;QACd8B,IAAAA,uCAAAA,KAAAA,AAAAA,EAAW,qCAAqC9B;QAChD;IACF;AACF;AAIO,MAAMgC,qBAAqB,IAEvBC;AAUJ,MAAMC,cAAc,CACzBC;IAKA,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAO;QACL,YAAYA;QACZ,kBAAkBnB;IACpB;IAEF,OAAO;QACL,YAAYmB,OAAO,MAAM;QACzB,kBAAkBA,OAAO,MAAM,GAC3B;YACE,QAAQA,OAAO,MAAM;YACrB,yBAAyB,CAAC,CAACA,OAAO,uBAAuB;QAC3D,IACAnB;IACN;AACF"}
|
|
@@ -87,7 +87,7 @@ const promptsToChatParam = async (multimodalPrompt)=>{
|
|
|
87
87
|
async function AiLocateElement(options) {
|
|
88
88
|
const { context, targetElementDescription, callAIFn, modelConfig } = options;
|
|
89
89
|
const { vlMode } = modelConfig;
|
|
90
|
-
const
|
|
90
|
+
const screenshotBase64 = context.screenshot.getData();
|
|
91
91
|
(0, utils_namespaceObject.assert)(targetElementDescription, "cannot find the target element description");
|
|
92
92
|
const targetElementDescriptionText = extraTextFromUserPrompt(targetElementDescription);
|
|
93
93
|
const userInstructionPrompt = (0, llm_locator_js_namespaceObject.findElementPrompt)(targetElementDescriptionText);
|
|
@@ -180,7 +180,7 @@ async function AiLocateElement(options) {
|
|
|
180
180
|
async function AiLocateSection(options) {
|
|
181
181
|
const { context, sectionDescription, modelConfig } = options;
|
|
182
182
|
const { vlMode } = modelConfig;
|
|
183
|
-
const
|
|
183
|
+
const screenshotBase64 = context.screenshot.getData();
|
|
184
184
|
const systemPrompt = (0, llm_section_locator_js_namespaceObject.systemPromptToLocateSection)(vlMode);
|
|
185
185
|
const sectionLocatorInstructionText = (0, llm_section_locator_js_namespaceObject.sectionLocatorInstruction)(extraTextFromUserPrompt(sectionDescription));
|
|
186
186
|
const msgs = [
|
|
@@ -248,7 +248,7 @@ async function AiLocateSection(options) {
|
|
|
248
248
|
async function AiExtractElementInfo(options) {
|
|
249
249
|
const { dataQuery, context, extractOption, multimodalPrompt, modelConfig } = options;
|
|
250
250
|
const systemPrompt = (0, extraction_js_namespaceObject.systemPromptToExtract)();
|
|
251
|
-
const
|
|
251
|
+
const screenshotBase64 = context.screenshot.getData();
|
|
252
252
|
const extractDataPromptText = (0, extraction_js_namespaceObject.extractDataQueryPrompt)(options.pageDescription || '', dataQuery);
|
|
253
253
|
const userContent = [];
|
|
254
254
|
if (extractOption?.screenshotIncluded !== false) userContent.push({
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"ai-model/inspect.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../src/ai-model/inspect.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n AIDataExtractionResponse,\n AIElementResponse,\n AISectionLocatorResponse,\n AIUsageInfo,\n Rect,\n ReferenceImage,\n ServiceExtractOption,\n UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { generateElementByPosition } from '@midscene/shared/extractor/dom-util';\nimport {\n cropByRect,\n paddingToMatchBlockByBase64,\n preProcessImageUrl,\n} from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport type { LocateResultElement } from '@midscene/shared/types';\nimport { assert } from '@midscene/shared/utils';\nimport type {\n ChatCompletionSystemMessageParam,\n ChatCompletionUserMessageParam,\n} from 'openai/resources/index';\nimport type { TMultimodalPrompt, TUserPrompt } from '../common';\nimport { adaptBboxToRect, expandSearchArea, mergeRects } from '../common';\nimport {\n extractDataQueryPrompt,\n systemPromptToExtract,\n} from './prompt/extraction';\nimport {\n findElementPrompt,\n systemPromptToLocateElement,\n} from './prompt/llm-locator';\nimport {\n sectionLocatorInstruction,\n systemPromptToLocateSection,\n} from './prompt/llm-section-locator';\nimport {\n orderSensitiveJudgePrompt,\n systemPromptToJudgeOrderSensitive,\n} from './prompt/order-sensitive-judge';\nimport { callAIWithObjectResponse } from './service-caller/index';\n\nexport type AIArgs = [\n ChatCompletionSystemMessageParam,\n ...ChatCompletionUserMessageParam[],\n];\n\nconst debugInspect = getDebug('ai:inspect');\nconst debugSection = getDebug('ai:section');\n\nconst extraTextFromUserPrompt = (prompt: TUserPrompt): string => {\n if (typeof prompt === 'string') {\n return prompt;\n } else {\n return prompt.prompt;\n }\n};\n\nconst promptsToChatParam = async (\n multimodalPrompt: TMultimodalPrompt,\n): Promise<ChatCompletionUserMessageParam[]> => {\n const msgs: ChatCompletionUserMessageParam[] = [];\n if (multimodalPrompt?.images?.length) {\n msgs.push({\n role: 'user',\n content: [\n {\n type: 'text',\n text: 'Next, I will provide all the reference images.',\n },\n ],\n });\n\n for (const item of multimodalPrompt.images) {\n const base64 = await preProcessImageUrl(\n item.url,\n !!multimodalPrompt.convertHttpImage2Base64,\n );\n\n msgs.push({\n role: 'user',\n content: [\n {\n type: 'text',\n text: `this is the reference image named '${item.name}':`,\n },\n ],\n });\n\n msgs.push({\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: base64,\n detail: 'high',\n },\n },\n ],\n });\n }\n }\n return msgs;\n};\n\nexport async function AiLocateElement(options: {\n context: UIContext;\n targetElementDescription: TUserPrompt;\n referenceImage?: ReferenceImage;\n callAIFn: typeof callAIWithObjectResponse<\n AIElementResponse | [number, number]\n >;\n searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;\n modelConfig: IModelConfig;\n}): Promise<{\n parseResult: {\n elements: LocateResultElement[];\n errors?: string[];\n };\n rect?: Rect;\n rawResponse: string;\n usage?: AIUsageInfo;\n reasoning_content?: string;\n}> {\n const { context, targetElementDescription, callAIFn, modelConfig } = options;\n const { vlMode } = modelConfig;\n const { screenshotBase64 } = context;\n\n assert(\n targetElementDescription,\n 'cannot find the target element description',\n );\n const targetElementDescriptionText = extraTextFromUserPrompt(\n targetElementDescription,\n );\n const userInstructionPrompt = findElementPrompt(targetElementDescriptionText);\n const systemPrompt = systemPromptToLocateElement(vlMode);\n\n let imagePayload = screenshotBase64;\n let imageWidth = context.size.width;\n let imageHeight = context.size.height;\n let originalImageWidth = imageWidth;\n let originalImageHeight = imageHeight;\n\n if (options.searchConfig) {\n assert(\n options.searchConfig.rect,\n 'searchArea is provided but its rect cannot be found. Failed to locate element',\n );\n assert(\n options.searchConfig.imageBase64,\n 'searchArea is provided but its imageBase64 cannot be found. Failed to locate element',\n );\n\n imagePayload = options.searchConfig.imageBase64;\n imageWidth = options.searchConfig.rect?.width;\n imageHeight = options.searchConfig.rect?.height;\n originalImageWidth = imageWidth;\n originalImageHeight = imageHeight;\n } else if (vlMode === 'qwen2.5-vl') {\n const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n imageWidth = paddedResult.width;\n imageHeight = paddedResult.height;\n imagePayload = paddedResult.imageBase64;\n }\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n detail: 'high',\n },\n },\n {\n type: 'text',\n text: userInstructionPrompt,\n },\n ],\n },\n ];\n\n if (typeof targetElementDescription !== 'string') {\n const addOns = await promptsToChatParam({\n images: targetElementDescription.images,\n convertHttpImage2Base64: targetElementDescription.convertHttpImage2Base64,\n });\n msgs.push(...addOns);\n }\n\n const res = await callAIFn(msgs, modelConfig);\n\n const rawResponse = JSON.stringify(res.content);\n\n let resRect: Rect | undefined;\n let matchedElements: LocateResultElement[] = [];\n let errors: string[] | undefined =\n 'errors' in res.content ? res.content.errors : [];\n try {\n if (\n 'bbox' in res.content &&\n Array.isArray(res.content.bbox) &&\n res.content.bbox.length >= 1\n ) {\n resRect = adaptBboxToRect(\n res.content.bbox,\n imageWidth,\n imageHeight,\n options.searchConfig?.rect?.left,\n options.searchConfig?.rect?.top,\n originalImageWidth,\n originalImageHeight,\n vlMode,\n );\n\n debugInspect('resRect', resRect);\n\n const rectCenter = {\n x: resRect.left + resRect.width / 2,\n y: resRect.top + resRect.height / 2,\n };\n\n const element: LocateResultElement = generateElementByPosition(\n rectCenter,\n targetElementDescriptionText as string,\n );\n errors = [];\n\n if (element) {\n matchedElements = [element];\n }\n }\n } catch (e) {\n const msg =\n e instanceof Error\n ? `Failed to parse bbox: ${e.message}`\n : 'unknown error in locate';\n if (!errors || errors?.length === 0) {\n errors = [msg];\n } else {\n errors.push(`(${msg})`);\n }\n }\n\n return {\n rect: resRect,\n parseResult: {\n elements: matchedElements as LocateResultElement[],\n errors: errors as string[],\n },\n rawResponse,\n usage: res.usage,\n reasoning_content: res.reasoning_content,\n };\n}\n\nexport async function AiLocateSection(options: {\n context: UIContext;\n sectionDescription: TUserPrompt;\n modelConfig: IModelConfig;\n}): Promise<{\n rect?: Rect;\n imageBase64?: string;\n error?: string;\n rawResponse: string;\n usage?: AIUsageInfo;\n}> {\n const { context, sectionDescription, modelConfig } = options;\n const { vlMode } = modelConfig;\n const { screenshotBase64 } = context;\n\n const systemPrompt = systemPromptToLocateSection(vlMode);\n const sectionLocatorInstructionText = sectionLocatorInstruction(\n extraTextFromUserPrompt(sectionDescription),\n );\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: screenshotBase64,\n detail: 'high',\n },\n },\n {\n type: 'text',\n text: sectionLocatorInstructionText,\n },\n ],\n },\n ];\n\n if (typeof sectionDescription !== 'string') {\n const addOns = await promptsToChatParam({\n images: sectionDescription.images,\n convertHttpImage2Base64: sectionDescription.convertHttpImage2Base64,\n });\n msgs.push(...addOns);\n }\n\n const result = await callAIWithObjectResponse<AISectionLocatorResponse>(\n msgs,\n modelConfig,\n );\n\n let sectionRect: Rect | undefined;\n const sectionBbox = result.content.bbox;\n if (sectionBbox) {\n const targetRect = adaptBboxToRect(\n sectionBbox,\n context.size.width,\n context.size.height,\n 0,\n 0,\n context.size.width,\n context.size.height,\n vlMode,\n );\n debugSection('original targetRect %j', targetRect);\n\n const referenceBboxList = result.content.references_bbox || [];\n debugSection('referenceBboxList %j', referenceBboxList);\n\n const referenceRects = referenceBboxList\n .filter((bbox) => Array.isArray(bbox))\n .map((bbox) => {\n return adaptBboxToRect(\n bbox,\n context.size.width,\n context.size.height,\n 0,\n 0,\n context.size.width,\n context.size.height,\n vlMode,\n );\n });\n debugSection('referenceRects %j', referenceRects);\n\n // merge the sectionRect and referenceRects\n const mergedRect = mergeRects([targetRect, ...referenceRects]);\n debugSection('mergedRect %j', mergedRect);\n\n // expand search area to at least 200 x 200\n sectionRect = expandSearchArea(mergedRect, context.size, vlMode);\n debugSection('expanded sectionRect %j', sectionRect);\n }\n\n let imageBase64 = screenshotBase64;\n if (sectionRect) {\n const croppedResult = await cropByRect(\n screenshotBase64,\n sectionRect,\n vlMode === 'qwen2.5-vl',\n );\n imageBase64 = croppedResult.imageBase64;\n sectionRect.width = croppedResult.width;\n sectionRect.height = croppedResult.height;\n }\n\n return {\n rect: sectionRect,\n imageBase64,\n error: result.content.error,\n rawResponse: JSON.stringify(result.content),\n usage: result.usage,\n };\n}\n\nexport async function AiExtractElementInfo<T>(options: {\n dataQuery: string | Record<string, string>;\n multimodalPrompt?: TMultimodalPrompt;\n context: UIContext;\n pageDescription?: string;\n extractOption?: ServiceExtractOption;\n modelConfig: IModelConfig;\n}) {\n const { dataQuery, context, extractOption, multimodalPrompt, modelConfig } =\n options;\n const systemPrompt = systemPromptToExtract();\n const { screenshotBase64 } = context;\n\n const extractDataPromptText = extractDataQueryPrompt(\n options.pageDescription || '',\n dataQuery,\n );\n\n const userContent: ChatCompletionUserMessageParam['content'] = [];\n\n if (extractOption?.screenshotIncluded !== false) {\n userContent.push({\n type: 'image_url',\n image_url: {\n url: screenshotBase64,\n detail: 'high',\n },\n });\n }\n\n userContent.push({\n type: 'text',\n text: extractDataPromptText,\n });\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: userContent,\n },\n ];\n\n if (multimodalPrompt) {\n const addOns = await promptsToChatParam({\n images: multimodalPrompt.images,\n convertHttpImage2Base64: multimodalPrompt.convertHttpImage2Base64,\n });\n msgs.push(...addOns);\n }\n\n const result = await callAIWithObjectResponse<AIDataExtractionResponse<T>>(\n msgs,\n modelConfig,\n );\n return {\n parseResult: result.content,\n usage: result.usage,\n reasoning_content: result.reasoning_content,\n };\n}\n\nexport async function AiJudgeOrderSensitive(\n description: string,\n callAIFn: typeof callAIWithObjectResponse<{ isOrderSensitive: boolean }>,\n modelConfig: IModelConfig,\n): Promise<{\n isOrderSensitive: boolean;\n usage?: AIUsageInfo;\n}> {\n const systemPrompt = systemPromptToJudgeOrderSensitive();\n const userPrompt = orderSensitiveJudgePrompt(description);\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: userPrompt,\n },\n ];\n\n const result = await callAIFn(msgs, modelConfig);\n\n return {\n isOrderSensitive: result.content.isOrderSensitive ?? false,\n usage: result.usage,\n };\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debugInspect","getDebug","debugSection","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","msgs","item","base64","preProcessImageUrl","AiLocateElement","options","context","targetElementDescription","callAIFn","modelConfig","vlMode","screenshotBase64","assert","targetElementDescriptionText","userInstructionPrompt","findElementPrompt","systemPrompt","systemPromptToLocateElement","imagePayload","imageWidth","imageHeight","originalImageWidth","originalImageHeight","paddedResult","paddingToMatchBlockByBase64","addOns","res","rawResponse","JSON","resRect","matchedElements","errors","Array","adaptBboxToRect","rectCenter","element","generateElementByPosition","e","msg","Error","AiLocateSection","sectionDescription","systemPromptToLocateSection","sectionLocatorInstructionText","sectionLocatorInstruction","result","callAIWithObjectResponse","sectionRect","sectionBbox","targetRect","referenceBboxList","referenceRects","bbox","mergedRect","mergeRects","expandSearchArea","imageBase64","croppedResult","cropByRect","AiExtractElementInfo","dataQuery","extractOption","systemPromptToExtract","extractDataPromptText","extractDataQueryPrompt","userContent","AiJudgeOrderSensitive","description","systemPromptToJudgeOrderSensitive","userPrompt","orderSensitiveJudgePrompt"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;AC2CA,MAAMI,eAAeC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAC9B,MAAMC,eAAeD,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAE9B,MAAME,0BAA0B,CAACC;IAC/B,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAEP,OAAOA,OAAO,MAAM;AAExB;AAEA,MAAMC,qBAAqB,OACzBC;IAEA,MAAMC,OAAyC,EAAE;IACjD,IAAID,kBAAkB,QAAQ,QAAQ;QACpCC,KAAK,IAAI,CAAC;YACR,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM;gBACR;aACD;QACH;QAEA,KAAK,MAAMC,QAAQF,iBAAiB,MAAM,CAAE;YAC1C,MAAMG,SAAS,MAAMC,AAAAA,IAAAA,oBAAAA,kBAAAA,AAAAA,EACnBF,KAAK,GAAG,EACR,CAAC,CAACF,iBAAiB,uBAAuB;YAG5CC,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,MAAM,CAAC,mCAAmC,EAAEC,KAAK,IAAI,CAAC,EAAE,CAAC;oBAC3D;iBACD;YACH;YAEAD,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKE;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;QACF;IACF;IACA,OAAOF;AACT;AAEO,eAAeI,gBAAgBC,OASrC;IAUC,MAAM,EAAEC,OAAO,EAAEC,wBAAwB,EAAEC,QAAQ,EAAEC,WAAW,EAAE,GAAGJ;IACrE,MAAM,EAAEK,MAAM,EAAE,GAAGD;IACnB,MAAM,EAAEE,gBAAgB,EAAE,GAAGL;IAE7BM,IAAAA,sBAAAA,MAAAA,AAAAA,EACEL,0BACA;IAEF,MAAMM,+BAA+BjB,wBACnCW;IAEF,MAAMO,wBAAwBC,AAAAA,IAAAA,+BAAAA,iBAAAA,AAAAA,EAAkBF;IAChD,MAAMG,eAAeC,AAAAA,IAAAA,+BAAAA,2BAAAA,AAAAA,EAA4BP;IAEjD,IAAIQ,eAAeP;IACnB,IAAIQ,aAAab,QAAQ,IAAI,CAAC,KAAK;IACnC,IAAIc,cAAcd,QAAQ,IAAI,CAAC,MAAM;IACrC,IAAIe,qBAAqBF;IACzB,IAAIG,sBAAsBF;IAE1B,IAAIf,QAAQ,YAAY,EAAE;QACxBO,IAAAA,sBAAAA,MAAAA,AAAAA,EACEP,QAAQ,YAAY,CAAC,IAAI,EACzB;QAEFO,IAAAA,sBAAAA,MAAAA,AAAAA,EACEP,QAAQ,YAAY,CAAC,WAAW,EAChC;QAGFa,eAAeb,QAAQ,YAAY,CAAC,WAAW;QAC/Cc,aAAad,QAAQ,YAAY,CAAC,IAAI,EAAE;QACxCe,cAAcf,QAAQ,YAAY,CAAC,IAAI,EAAE;QACzCgB,qBAAqBF;QACrBG,sBAAsBF;IACxB,OAAO,IAAIV,AAAW,iBAAXA,QAAyB;QAClC,MAAMa,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,2BAAAA,AAAAA,EAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAMvB,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKE;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMJ;gBACR;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAOP,0BAAuC;QAChD,MAAMkB,SAAS,MAAM3B,mBAAmB;YACtC,QAAQS,yBAAyB,MAAM;YACvC,yBAAyBA,yBAAyB,uBAAuB;QAC3E;QACAP,KAAK,IAAI,IAAIyB;IACf;IAEA,MAAMC,MAAM,MAAMlB,SAASR,MAAMS;IAEjC,MAAMkB,cAAcC,KAAK,SAAS,CAACF,IAAI,OAAO;IAE9C,IAAIG;IACJ,IAAIC,kBAAyC,EAAE;IAC/C,IAAIC,SACF,YAAYL,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,MAAM,GAAG,EAAE;IACnD,IAAI;QACF,IACE,UAAUA,IAAI,OAAO,IACrBM,MAAM,OAAO,CAACN,IAAI,OAAO,CAAC,IAAI,KAC9BA,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,IAAI,GAC3B;YACAG,UAAUI,AAAAA,IAAAA,mCAAAA,eAAAA,AAAAA,EACRP,IAAI,OAAO,CAAC,IAAI,EAChBP,YACAC,aACAf,QAAQ,YAAY,EAAE,MAAM,MAC5BA,QAAQ,YAAY,EAAE,MAAM,KAC5BgB,oBACAC,qBACAZ;YAGFjB,aAAa,WAAWoC;YAExB,MAAMK,aAAa;gBACjB,GAAGL,QAAQ,IAAI,GAAGA,QAAQ,KAAK,GAAG;gBAClC,GAAGA,QAAQ,GAAG,GAAGA,QAAQ,MAAM,GAAG;YACpC;YAEA,MAAMM,UAA+BC,AAAAA,IAAAA,yBAAAA,yBAAAA,AAAAA,EACnCF,YACArB;YAEFkB,SAAS,EAAE;YAEX,IAAII,SACFL,kBAAkB;gBAACK;aAAQ;QAE/B;IACF,EAAE,OAAOE,GAAG;QACV,MAAMC,MACJD,aAAaE,QACT,CAAC,sBAAsB,EAAEF,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACN,UAAUA,QAAQ,WAAW,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAEO,IAAI,CAAC,CAAC;aAFtBP,SAAS;YAACO;SAAI;IAIlB;IAEA,OAAO;QACL,MAAMT;QACN,aAAa;YACX,UAAUC;YACV,QAAQC;QACV;QACAJ;QACA,OAAOD,IAAI,KAAK;QAChB,mBAAmBA,IAAI,iBAAiB;IAC1C;AACF;AAEO,eAAec,gBAAgBnC,OAIrC;IAOC,MAAM,EAAEC,OAAO,EAAEmC,kBAAkB,EAAEhC,WAAW,EAAE,GAAGJ;IACrD,MAAM,EAAEK,MAAM,EAAE,GAAGD;IACnB,MAAM,EAAEE,gBAAgB,EAAE,GAAGL;IAE7B,MAAMU,eAAe0B,AAAAA,IAAAA,uCAAAA,2BAAAA,AAAAA,EAA4BhC;IACjD,MAAMiC,gCAAgCC,AAAAA,IAAAA,uCAAAA,yBAAAA,AAAAA,EACpChD,wBAAwB6C;IAE1B,MAAMzC,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKL;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMgC;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAMhB,SAAS,MAAM3B,mBAAmB;YACtC,QAAQ2C,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACAzC,KAAK,IAAI,IAAIyB;IACf;IAEA,MAAMoB,SAAS,MAAMC,AAAAA,IAAAA,yBAAAA,wBAAAA,AAAAA,EACnB9C,MACAS;IAGF,IAAIsC;IACJ,MAAMC,cAAcH,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIG,aAAa;QACf,MAAMC,aAAahB,AAAAA,IAAAA,mCAAAA,eAAAA,AAAAA,EACjBe,aACA1C,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAEFf,aAAa,0BAA0BsD;QAEvC,MAAMC,oBAAoBL,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9DlD,aAAa,wBAAwBuD;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAASpB,MAAM,OAAO,CAACoB,OAC/B,GAAG,CAAC,CAACA,OACGnB,AAAAA,IAAAA,mCAAAA,eAAAA,AAAAA,EACLmB,MACA9C,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAGNf,aAAa,qBAAqBwD;QAGlC,MAAME,aAAaC,AAAAA,IAAAA,mCAAAA,UAAAA,AAAAA,EAAW;YAACL;eAAeE;SAAe;QAC7DxD,aAAa,iBAAiB0D;QAG9BN,cAAcQ,AAAAA,IAAAA,mCAAAA,gBAAAA,AAAAA,EAAiBF,YAAY/C,QAAQ,IAAI,EAAEI;QACzDf,aAAa,2BAA2BoD;IAC1C;IAEA,IAAIS,cAAc7C;IAClB,IAAIoC,aAAa;QACf,MAAMU,gBAAgB,MAAMC,AAAAA,IAAAA,oBAAAA,UAAAA,AAAAA,EAC1B/C,kBACAoC,aACArC,AAAW,iBAAXA;QAEF8C,cAAcC,cAAc,WAAW;QACvCV,YAAY,KAAK,GAAGU,cAAc,KAAK;QACvCV,YAAY,MAAM,GAAGU,cAAc,MAAM;IAC3C;IAEA,OAAO;QACL,MAAMV;QACNS;QACA,OAAOX,OAAO,OAAO,CAAC,KAAK;QAC3B,aAAajB,KAAK,SAAS,CAACiB,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAec,qBAAwBtD,OAO7C;IACC,MAAM,EAAEuD,SAAS,EAAEtD,OAAO,EAAEuD,aAAa,EAAE9D,gBAAgB,EAAEU,WAAW,EAAE,GACxEJ;IACF,MAAMW,eAAe8C,AAAAA,IAAAA,8BAAAA,qBAAAA,AAAAA;IACrB,MAAM,EAAEnD,gBAAgB,EAAE,GAAGL;IAE7B,MAAMyD,wBAAwBC,AAAAA,IAAAA,8BAAAA,sBAAAA,AAAAA,EAC5B3D,QAAQ,eAAe,IAAI,IAC3BuD;IAGF,MAAMK,cAAyD,EAAE;IAEjE,IAAIJ,eAAe,uBAAuB,OACxCI,YAAY,IAAI,CAAC;QACf,MAAM;QACN,WAAW;YACT,KAAKtD;YACL,QAAQ;QACV;IACF;IAGFsD,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAM/D,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAASiD;QACX;KACD;IAED,IAAIlE,kBAAkB;QACpB,MAAM0B,SAAS,MAAM3B,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAC,KAAK,IAAI,IAAIyB;IACf;IAEA,MAAMoB,SAAS,MAAMC,AAAAA,IAAAA,yBAAAA,wBAAAA,AAAAA,EACnB9C,MACAS;IAEF,OAAO;QACL,aAAaoC,OAAO,OAAO;QAC3B,OAAOA,OAAO,KAAK;QACnB,mBAAmBA,OAAO,iBAAiB;IAC7C;AACF;AAEO,eAAeqB,sBACpBC,WAAmB,EACnB3D,QAAwE,EACxEC,WAAyB;IAKzB,MAAMO,eAAeoD,AAAAA,IAAAA,yCAAAA,iCAAAA,AAAAA;IACrB,MAAMC,aAAaC,AAAAA,IAAAA,yCAAAA,yBAAAA,AAAAA,EAA0BH;IAE7C,MAAMnE,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAASqD;QACX;KACD;IAED,MAAMxB,SAAS,MAAMrC,SAASR,MAAMS;IAEpC,OAAO;QACL,kBAAkBoC,OAAO,OAAO,CAAC,gBAAgB,IAAI;QACrD,OAAOA,OAAO,KAAK;IACrB;AACF"}
|
|
1
|
+
{"version":3,"file":"ai-model/inspect.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../src/ai-model/inspect.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n AIDataExtractionResponse,\n AIElementResponse,\n AISectionLocatorResponse,\n AIUsageInfo,\n Rect,\n ReferenceImage,\n ServiceExtractOption,\n UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { generateElementByPosition } from '@midscene/shared/extractor/dom-util';\nimport {\n cropByRect,\n paddingToMatchBlockByBase64,\n preProcessImageUrl,\n} from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport type { LocateResultElement } from '@midscene/shared/types';\nimport { assert } from '@midscene/shared/utils';\nimport type {\n ChatCompletionSystemMessageParam,\n ChatCompletionUserMessageParam,\n} from 'openai/resources/index';\nimport type { TMultimodalPrompt, TUserPrompt } from '../common';\nimport { adaptBboxToRect, expandSearchArea, mergeRects } from '../common';\nimport {\n extractDataQueryPrompt,\n systemPromptToExtract,\n} from './prompt/extraction';\nimport {\n findElementPrompt,\n systemPromptToLocateElement,\n} from './prompt/llm-locator';\nimport {\n sectionLocatorInstruction,\n systemPromptToLocateSection,\n} from './prompt/llm-section-locator';\nimport {\n orderSensitiveJudgePrompt,\n systemPromptToJudgeOrderSensitive,\n} from './prompt/order-sensitive-judge';\nimport { callAIWithObjectResponse } from './service-caller/index';\n\nexport type AIArgs = [\n ChatCompletionSystemMessageParam,\n ...ChatCompletionUserMessageParam[],\n];\n\nconst debugInspect = getDebug('ai:inspect');\nconst debugSection = getDebug('ai:section');\n\nconst extraTextFromUserPrompt = (prompt: TUserPrompt): string => {\n if (typeof prompt === 'string') {\n return prompt;\n } else {\n return prompt.prompt;\n }\n};\n\nconst promptsToChatParam = async (\n multimodalPrompt: TMultimodalPrompt,\n): Promise<ChatCompletionUserMessageParam[]> => {\n const msgs: ChatCompletionUserMessageParam[] = [];\n if (multimodalPrompt?.images?.length) {\n msgs.push({\n role: 'user',\n content: [\n {\n type: 'text',\n text: 'Next, I will provide all the reference images.',\n },\n ],\n });\n\n for (const item of multimodalPrompt.images) {\n const base64 = await preProcessImageUrl(\n item.url,\n !!multimodalPrompt.convertHttpImage2Base64,\n );\n\n msgs.push({\n role: 'user',\n content: [\n {\n type: 'text',\n text: `this is the reference image named '${item.name}':`,\n },\n ],\n });\n\n msgs.push({\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: base64,\n detail: 'high',\n },\n },\n ],\n });\n }\n }\n return msgs;\n};\n\nexport async function AiLocateElement(options: {\n context: UIContext;\n targetElementDescription: TUserPrompt;\n referenceImage?: ReferenceImage;\n callAIFn: typeof callAIWithObjectResponse<\n AIElementResponse | [number, number]\n >;\n searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;\n modelConfig: IModelConfig;\n}): Promise<{\n parseResult: {\n elements: LocateResultElement[];\n errors?: string[];\n };\n rect?: Rect;\n rawResponse: string;\n usage?: AIUsageInfo;\n reasoning_content?: string;\n}> {\n const { context, targetElementDescription, callAIFn, modelConfig } = options;\n const { vlMode } = modelConfig;\n const screenshotBase64 = context.screenshot.getData();\n\n assert(\n targetElementDescription,\n 'cannot find the target element description',\n );\n const targetElementDescriptionText = extraTextFromUserPrompt(\n targetElementDescription,\n );\n const userInstructionPrompt = findElementPrompt(targetElementDescriptionText);\n const systemPrompt = systemPromptToLocateElement(vlMode);\n\n let imagePayload = screenshotBase64;\n let imageWidth = context.size.width;\n let imageHeight = context.size.height;\n let originalImageWidth = imageWidth;\n let originalImageHeight = imageHeight;\n\n if (options.searchConfig) {\n assert(\n options.searchConfig.rect,\n 'searchArea is provided but its rect cannot be found. Failed to locate element',\n );\n assert(\n options.searchConfig.imageBase64,\n 'searchArea is provided but its imageBase64 cannot be found. Failed to locate element',\n );\n\n imagePayload = options.searchConfig.imageBase64;\n imageWidth = options.searchConfig.rect?.width;\n imageHeight = options.searchConfig.rect?.height;\n originalImageWidth = imageWidth;\n originalImageHeight = imageHeight;\n } else if (vlMode === 'qwen2.5-vl') {\n const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n imageWidth = paddedResult.width;\n imageHeight = paddedResult.height;\n imagePayload = paddedResult.imageBase64;\n }\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n detail: 'high',\n },\n },\n {\n type: 'text',\n text: userInstructionPrompt,\n },\n ],\n },\n ];\n\n if (typeof targetElementDescription !== 'string') {\n const addOns = await promptsToChatParam({\n images: targetElementDescription.images,\n convertHttpImage2Base64: targetElementDescription.convertHttpImage2Base64,\n });\n msgs.push(...addOns);\n }\n\n const res = await callAIFn(msgs, modelConfig);\n\n const rawResponse = JSON.stringify(res.content);\n\n let resRect: Rect | undefined;\n let matchedElements: LocateResultElement[] = [];\n let errors: string[] | undefined =\n 'errors' in res.content ? res.content.errors : [];\n try {\n if (\n 'bbox' in res.content &&\n Array.isArray(res.content.bbox) &&\n res.content.bbox.length >= 1\n ) {\n resRect = adaptBboxToRect(\n res.content.bbox,\n imageWidth,\n imageHeight,\n options.searchConfig?.rect?.left,\n options.searchConfig?.rect?.top,\n originalImageWidth,\n originalImageHeight,\n vlMode,\n );\n\n debugInspect('resRect', resRect);\n\n const rectCenter = {\n x: resRect.left + resRect.width / 2,\n y: resRect.top + resRect.height / 2,\n };\n\n const element: LocateResultElement = generateElementByPosition(\n rectCenter,\n targetElementDescriptionText as string,\n );\n errors = [];\n\n if (element) {\n matchedElements = [element];\n }\n }\n } catch (e) {\n const msg =\n e instanceof Error\n ? `Failed to parse bbox: ${e.message}`\n : 'unknown error in locate';\n if (!errors || errors?.length === 0) {\n errors = [msg];\n } else {\n errors.push(`(${msg})`);\n }\n }\n\n return {\n rect: resRect,\n parseResult: {\n elements: matchedElements as LocateResultElement[],\n errors: errors as string[],\n },\n rawResponse,\n usage: res.usage,\n reasoning_content: res.reasoning_content,\n };\n}\n\nexport async function AiLocateSection(options: {\n context: UIContext;\n sectionDescription: TUserPrompt;\n modelConfig: IModelConfig;\n}): Promise<{\n rect?: Rect;\n imageBase64?: string;\n error?: string;\n rawResponse: string;\n usage?: AIUsageInfo;\n}> {\n const { context, sectionDescription, modelConfig } = options;\n const { vlMode } = modelConfig;\n const screenshotBase64 = context.screenshot.getData();\n\n const systemPrompt = systemPromptToLocateSection(vlMode);\n const sectionLocatorInstructionText = sectionLocatorInstruction(\n extraTextFromUserPrompt(sectionDescription),\n );\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: screenshotBase64,\n detail: 'high',\n },\n },\n {\n type: 'text',\n text: sectionLocatorInstructionText,\n },\n ],\n },\n ];\n\n if (typeof sectionDescription !== 'string') {\n const addOns = await promptsToChatParam({\n images: sectionDescription.images,\n convertHttpImage2Base64: sectionDescription.convertHttpImage2Base64,\n });\n msgs.push(...addOns);\n }\n\n const result = await callAIWithObjectResponse<AISectionLocatorResponse>(\n msgs,\n modelConfig,\n );\n\n let sectionRect: Rect | undefined;\n const sectionBbox = result.content.bbox;\n if (sectionBbox) {\n const targetRect = adaptBboxToRect(\n sectionBbox,\n context.size.width,\n context.size.height,\n 0,\n 0,\n context.size.width,\n context.size.height,\n vlMode,\n );\n debugSection('original targetRect %j', targetRect);\n\n const referenceBboxList = result.content.references_bbox || [];\n debugSection('referenceBboxList %j', referenceBboxList);\n\n const referenceRects = referenceBboxList\n .filter((bbox) => Array.isArray(bbox))\n .map((bbox) => {\n return adaptBboxToRect(\n bbox,\n context.size.width,\n context.size.height,\n 0,\n 0,\n context.size.width,\n context.size.height,\n vlMode,\n );\n });\n debugSection('referenceRects %j', referenceRects);\n\n // merge the sectionRect and referenceRects\n const mergedRect = mergeRects([targetRect, ...referenceRects]);\n debugSection('mergedRect %j', mergedRect);\n\n // expand search area to at least 200 x 200\n sectionRect = expandSearchArea(mergedRect, context.size, vlMode);\n debugSection('expanded sectionRect %j', sectionRect);\n }\n\n let imageBase64 = screenshotBase64;\n if (sectionRect) {\n const croppedResult = await cropByRect(\n screenshotBase64,\n sectionRect,\n vlMode === 'qwen2.5-vl',\n );\n imageBase64 = croppedResult.imageBase64;\n sectionRect.width = croppedResult.width;\n sectionRect.height = croppedResult.height;\n }\n\n return {\n rect: sectionRect,\n imageBase64,\n error: result.content.error,\n rawResponse: JSON.stringify(result.content),\n usage: result.usage,\n };\n}\n\nexport async function AiExtractElementInfo<T>(options: {\n dataQuery: string | Record<string, string>;\n multimodalPrompt?: TMultimodalPrompt;\n context: UIContext;\n pageDescription?: string;\n extractOption?: ServiceExtractOption;\n modelConfig: IModelConfig;\n}) {\n const { dataQuery, context, extractOption, multimodalPrompt, modelConfig } =\n options;\n const systemPrompt = systemPromptToExtract();\n const screenshotBase64 = context.screenshot.getData();\n\n const extractDataPromptText = extractDataQueryPrompt(\n options.pageDescription || '',\n dataQuery,\n );\n\n const userContent: ChatCompletionUserMessageParam['content'] = [];\n\n if (extractOption?.screenshotIncluded !== false) {\n userContent.push({\n type: 'image_url',\n image_url: {\n url: screenshotBase64,\n detail: 'high',\n },\n });\n }\n\n userContent.push({\n type: 'text',\n text: extractDataPromptText,\n });\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: userContent,\n },\n ];\n\n if (multimodalPrompt) {\n const addOns = await promptsToChatParam({\n images: multimodalPrompt.images,\n convertHttpImage2Base64: multimodalPrompt.convertHttpImage2Base64,\n });\n msgs.push(...addOns);\n }\n\n const result = await callAIWithObjectResponse<AIDataExtractionResponse<T>>(\n msgs,\n modelConfig,\n );\n return {\n parseResult: result.content,\n usage: result.usage,\n reasoning_content: result.reasoning_content,\n };\n}\n\nexport async function AiJudgeOrderSensitive(\n description: string,\n callAIFn: typeof callAIWithObjectResponse<{ isOrderSensitive: boolean }>,\n modelConfig: IModelConfig,\n): Promise<{\n isOrderSensitive: boolean;\n usage?: AIUsageInfo;\n}> {\n const systemPrompt = systemPromptToJudgeOrderSensitive();\n const userPrompt = orderSensitiveJudgePrompt(description);\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: userPrompt,\n },\n ];\n\n const result = await callAIFn(msgs, modelConfig);\n\n return {\n isOrderSensitive: result.content.isOrderSensitive ?? false,\n usage: result.usage,\n };\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debugInspect","getDebug","debugSection","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","msgs","item","base64","preProcessImageUrl","AiLocateElement","options","context","targetElementDescription","callAIFn","modelConfig","vlMode","screenshotBase64","assert","targetElementDescriptionText","userInstructionPrompt","findElementPrompt","systemPrompt","systemPromptToLocateElement","imagePayload","imageWidth","imageHeight","originalImageWidth","originalImageHeight","paddedResult","paddingToMatchBlockByBase64","addOns","res","rawResponse","JSON","resRect","matchedElements","errors","Array","adaptBboxToRect","rectCenter","element","generateElementByPosition","e","msg","Error","AiLocateSection","sectionDescription","systemPromptToLocateSection","sectionLocatorInstructionText","sectionLocatorInstruction","result","callAIWithObjectResponse","sectionRect","sectionBbox","targetRect","referenceBboxList","referenceRects","bbox","mergedRect","mergeRects","expandSearchArea","imageBase64","croppedResult","cropByRect","AiExtractElementInfo","dataQuery","extractOption","systemPromptToExtract","extractDataPromptText","extractDataQueryPrompt","userContent","AiJudgeOrderSensitive","description","systemPromptToJudgeOrderSensitive","userPrompt","orderSensitiveJudgePrompt"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;AC2CA,MAAMI,eAAeC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAC9B,MAAMC,eAAeD,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAE9B,MAAME,0BAA0B,CAACC;IAC/B,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAEP,OAAOA,OAAO,MAAM;AAExB;AAEA,MAAMC,qBAAqB,OACzBC;IAEA,MAAMC,OAAyC,EAAE;IACjD,IAAID,kBAAkB,QAAQ,QAAQ;QACpCC,KAAK,IAAI,CAAC;YACR,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM;gBACR;aACD;QACH;QAEA,KAAK,MAAMC,QAAQF,iBAAiB,MAAM,CAAE;YAC1C,MAAMG,SAAS,MAAMC,AAAAA,IAAAA,oBAAAA,kBAAAA,AAAAA,EACnBF,KAAK,GAAG,EACR,CAAC,CAACF,iBAAiB,uBAAuB;YAG5CC,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,MAAM,CAAC,mCAAmC,EAAEC,KAAK,IAAI,CAAC,EAAE,CAAC;oBAC3D;iBACD;YACH;YAEAD,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKE;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;QACF;IACF;IACA,OAAOF;AACT;AAEO,eAAeI,gBAAgBC,OASrC;IAUC,MAAM,EAAEC,OAAO,EAAEC,wBAAwB,EAAEC,QAAQ,EAAEC,WAAW,EAAE,GAAGJ;IACrE,MAAM,EAAEK,MAAM,EAAE,GAAGD;IACnB,MAAME,mBAAmBL,QAAQ,UAAU,CAAC,OAAO;IAEnDM,IAAAA,sBAAAA,MAAAA,AAAAA,EACEL,0BACA;IAEF,MAAMM,+BAA+BjB,wBACnCW;IAEF,MAAMO,wBAAwBC,AAAAA,IAAAA,+BAAAA,iBAAAA,AAAAA,EAAkBF;IAChD,MAAMG,eAAeC,AAAAA,IAAAA,+BAAAA,2BAAAA,AAAAA,EAA4BP;IAEjD,IAAIQ,eAAeP;IACnB,IAAIQ,aAAab,QAAQ,IAAI,CAAC,KAAK;IACnC,IAAIc,cAAcd,QAAQ,IAAI,CAAC,MAAM;IACrC,IAAIe,qBAAqBF;IACzB,IAAIG,sBAAsBF;IAE1B,IAAIf,QAAQ,YAAY,EAAE;QACxBO,IAAAA,sBAAAA,MAAAA,AAAAA,EACEP,QAAQ,YAAY,CAAC,IAAI,EACzB;QAEFO,IAAAA,sBAAAA,MAAAA,AAAAA,EACEP,QAAQ,YAAY,CAAC,WAAW,EAChC;QAGFa,eAAeb,QAAQ,YAAY,CAAC,WAAW;QAC/Cc,aAAad,QAAQ,YAAY,CAAC,IAAI,EAAE;QACxCe,cAAcf,QAAQ,YAAY,CAAC,IAAI,EAAE;QACzCgB,qBAAqBF;QACrBG,sBAAsBF;IACxB,OAAO,IAAIV,AAAW,iBAAXA,QAAyB;QAClC,MAAMa,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,2BAAAA,AAAAA,EAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAMvB,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKE;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMJ;gBACR;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAOP,0BAAuC;QAChD,MAAMkB,SAAS,MAAM3B,mBAAmB;YACtC,QAAQS,yBAAyB,MAAM;YACvC,yBAAyBA,yBAAyB,uBAAuB;QAC3E;QACAP,KAAK,IAAI,IAAIyB;IACf;IAEA,MAAMC,MAAM,MAAMlB,SAASR,MAAMS;IAEjC,MAAMkB,cAAcC,KAAK,SAAS,CAACF,IAAI,OAAO;IAE9C,IAAIG;IACJ,IAAIC,kBAAyC,EAAE;IAC/C,IAAIC,SACF,YAAYL,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,MAAM,GAAG,EAAE;IACnD,IAAI;QACF,IACE,UAAUA,IAAI,OAAO,IACrBM,MAAM,OAAO,CAACN,IAAI,OAAO,CAAC,IAAI,KAC9BA,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,IAAI,GAC3B;YACAG,UAAUI,AAAAA,IAAAA,mCAAAA,eAAAA,AAAAA,EACRP,IAAI,OAAO,CAAC,IAAI,EAChBP,YACAC,aACAf,QAAQ,YAAY,EAAE,MAAM,MAC5BA,QAAQ,YAAY,EAAE,MAAM,KAC5BgB,oBACAC,qBACAZ;YAGFjB,aAAa,WAAWoC;YAExB,MAAMK,aAAa;gBACjB,GAAGL,QAAQ,IAAI,GAAGA,QAAQ,KAAK,GAAG;gBAClC,GAAGA,QAAQ,GAAG,GAAGA,QAAQ,MAAM,GAAG;YACpC;YAEA,MAAMM,UAA+BC,AAAAA,IAAAA,yBAAAA,yBAAAA,AAAAA,EACnCF,YACArB;YAEFkB,SAAS,EAAE;YAEX,IAAII,SACFL,kBAAkB;gBAACK;aAAQ;QAE/B;IACF,EAAE,OAAOE,GAAG;QACV,MAAMC,MACJD,aAAaE,QACT,CAAC,sBAAsB,EAAEF,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACN,UAAUA,QAAQ,WAAW,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAEO,IAAI,CAAC,CAAC;aAFtBP,SAAS;YAACO;SAAI;IAIlB;IAEA,OAAO;QACL,MAAMT;QACN,aAAa;YACX,UAAUC;YACV,QAAQC;QACV;QACAJ;QACA,OAAOD,IAAI,KAAK;QAChB,mBAAmBA,IAAI,iBAAiB;IAC1C;AACF;AAEO,eAAec,gBAAgBnC,OAIrC;IAOC,MAAM,EAAEC,OAAO,EAAEmC,kBAAkB,EAAEhC,WAAW,EAAE,GAAGJ;IACrD,MAAM,EAAEK,MAAM,EAAE,GAAGD;IACnB,MAAME,mBAAmBL,QAAQ,UAAU,CAAC,OAAO;IAEnD,MAAMU,eAAe0B,AAAAA,IAAAA,uCAAAA,2BAAAA,AAAAA,EAA4BhC;IACjD,MAAMiC,gCAAgCC,AAAAA,IAAAA,uCAAAA,yBAAAA,AAAAA,EACpChD,wBAAwB6C;IAE1B,MAAMzC,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKL;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMgC;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAMhB,SAAS,MAAM3B,mBAAmB;YACtC,QAAQ2C,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACAzC,KAAK,IAAI,IAAIyB;IACf;IAEA,MAAMoB,SAAS,MAAMC,AAAAA,IAAAA,yBAAAA,wBAAAA,AAAAA,EACnB9C,MACAS;IAGF,IAAIsC;IACJ,MAAMC,cAAcH,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIG,aAAa;QACf,MAAMC,aAAahB,AAAAA,IAAAA,mCAAAA,eAAAA,AAAAA,EACjBe,aACA1C,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAEFf,aAAa,0BAA0BsD;QAEvC,MAAMC,oBAAoBL,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9DlD,aAAa,wBAAwBuD;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAASpB,MAAM,OAAO,CAACoB,OAC/B,GAAG,CAAC,CAACA,OACGnB,AAAAA,IAAAA,mCAAAA,eAAAA,AAAAA,EACLmB,MACA9C,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAGNf,aAAa,qBAAqBwD;QAGlC,MAAME,aAAaC,AAAAA,IAAAA,mCAAAA,UAAAA,AAAAA,EAAW;YAACL;eAAeE;SAAe;QAC7DxD,aAAa,iBAAiB0D;QAG9BN,cAAcQ,AAAAA,IAAAA,mCAAAA,gBAAAA,AAAAA,EAAiBF,YAAY/C,QAAQ,IAAI,EAAEI;QACzDf,aAAa,2BAA2BoD;IAC1C;IAEA,IAAIS,cAAc7C;IAClB,IAAIoC,aAAa;QACf,MAAMU,gBAAgB,MAAMC,AAAAA,IAAAA,oBAAAA,UAAAA,AAAAA,EAC1B/C,kBACAoC,aACArC,AAAW,iBAAXA;QAEF8C,cAAcC,cAAc,WAAW;QACvCV,YAAY,KAAK,GAAGU,cAAc,KAAK;QACvCV,YAAY,MAAM,GAAGU,cAAc,MAAM;IAC3C;IAEA,OAAO;QACL,MAAMV;QACNS;QACA,OAAOX,OAAO,OAAO,CAAC,KAAK;QAC3B,aAAajB,KAAK,SAAS,CAACiB,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAec,qBAAwBtD,OAO7C;IACC,MAAM,EAAEuD,SAAS,EAAEtD,OAAO,EAAEuD,aAAa,EAAE9D,gBAAgB,EAAEU,WAAW,EAAE,GACxEJ;IACF,MAAMW,eAAe8C,AAAAA,IAAAA,8BAAAA,qBAAAA,AAAAA;IACrB,MAAMnD,mBAAmBL,QAAQ,UAAU,CAAC,OAAO;IAEnD,MAAMyD,wBAAwBC,AAAAA,IAAAA,8BAAAA,sBAAAA,AAAAA,EAC5B3D,QAAQ,eAAe,IAAI,IAC3BuD;IAGF,MAAMK,cAAyD,EAAE;IAEjE,IAAIJ,eAAe,uBAAuB,OACxCI,YAAY,IAAI,CAAC;QACf,MAAM;QACN,WAAW;YACT,KAAKtD;YACL,QAAQ;QACV;IACF;IAGFsD,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAM/D,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAASiD;QACX;KACD;IAED,IAAIlE,kBAAkB;QACpB,MAAM0B,SAAS,MAAM3B,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAC,KAAK,IAAI,IAAIyB;IACf;IAEA,MAAMoB,SAAS,MAAMC,AAAAA,IAAAA,yBAAAA,wBAAAA,AAAAA,EACnB9C,MACAS;IAEF,OAAO;QACL,aAAaoC,OAAO,OAAO;QAC3B,OAAOA,OAAO,KAAK;QACnB,mBAAmBA,OAAO,iBAAiB;IAC7C;AACF;AAEO,eAAeqB,sBACpBC,WAAmB,EACnB3D,QAAwE,EACxEC,WAAyB;IAKzB,MAAMO,eAAeoD,AAAAA,IAAAA,yCAAAA,iCAAAA,AAAAA;IACrB,MAAMC,aAAaC,AAAAA,IAAAA,yCAAAA,yBAAAA,AAAAA,EAA0BH;IAE7C,MAAMnE,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAASqD;QACX;KACD;IAED,MAAMxB,SAAS,MAAMrC,SAASR,MAAMS;IAEpC,OAAO;QACL,kBAAkBoC,OAAO,OAAO,CAAC,gBAAgB,IAAI;QACrD,OAAOA,OAAO,KAAK;IACrB;AACF"}
|
|
@@ -35,7 +35,8 @@ const index_js_namespaceObject = require("./service-caller/index.js");
|
|
|
35
35
|
const debug = (0, logger_namespaceObject.getDebug)('planning');
|
|
36
36
|
async function plan(userInstruction, opts) {
|
|
37
37
|
const { context, modelConfig, conversationHistory } = opts;
|
|
38
|
-
const {
|
|
38
|
+
const { size } = context;
|
|
39
|
+
const screenshotBase64 = context.screenshot.getData();
|
|
39
40
|
const { vlMode } = modelConfig;
|
|
40
41
|
const systemPrompt = await (0, llm_planning_js_namespaceObject.systemPromptToTaskPlanning)({
|
|
41
42
|
actionSpace: opts.actionSpace,
|
|
@@ -116,13 +117,23 @@ async function plan(userInstruction, opts) {
|
|
|
116
117
|
const actions = planFromAI.action ? [
|
|
117
118
|
planFromAI.action
|
|
118
119
|
] : [];
|
|
120
|
+
let shouldContinuePlanning = true;
|
|
121
|
+
if (!actions.length) {
|
|
122
|
+
debug('no actions planned and no sleep instruction, stop planning');
|
|
123
|
+
shouldContinuePlanning = false;
|
|
124
|
+
}
|
|
125
|
+
if (actions[0]?.type === external_common_js_namespaceObject.finalizeActionName) {
|
|
126
|
+
debug('finalize action planned, stop planning');
|
|
127
|
+
shouldContinuePlanning = false;
|
|
128
|
+
}
|
|
119
129
|
const returnValue = {
|
|
120
130
|
...planFromAI,
|
|
121
131
|
actions,
|
|
122
132
|
rawResponse,
|
|
123
133
|
usage,
|
|
124
134
|
reasoning_content,
|
|
125
|
-
yamlFlow: (0, external_common_js_namespaceObject.buildYamlFlowFromPlans)(actions, opts.actionSpace,
|
|
135
|
+
yamlFlow: (0, external_common_js_namespaceObject.buildYamlFlowFromPlans)(actions, opts.actionSpace),
|
|
136
|
+
shouldContinuePlanning
|
|
126
137
|
};
|
|
127
138
|
(0, utils_namespaceObject.assert)(planFromAI, "can't get plans from AI");
|
|
128
139
|
actions.forEach((action)=>{
|
|
@@ -136,7 +147,6 @@ async function plan(userInstruction, opts) {
|
|
|
136
147
|
if (locateResult && void 0 !== vlMode) action.param[field] = (0, external_common_js_namespaceObject.fillBboxParam)(locateResult, imageWidth, imageHeight, rightLimit, bottomLimit, vlMode);
|
|
137
148
|
});
|
|
138
149
|
});
|
|
139
|
-
if (0 === actions.length && returnValue.more_actions_needed_by_instruction && !returnValue.sleep) console.warn('No actions planned for the prompt, but model said more actions are needed:', userInstruction);
|
|
140
150
|
conversationHistory.append({
|
|
141
151
|
role: 'assistant',
|
|
142
152
|
content: [
|