@midscene/core 1.0.5-beta-20260104042628.0 → 1.0.5-beta-20260105023603.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (42) hide show
  1. package/dist/es/agent/agent.mjs +3 -4
  2. package/dist/es/agent/agent.mjs.map +1 -1
  3. package/dist/es/agent/task-builder.mjs +1 -0
  4. package/dist/es/agent/task-builder.mjs.map +1 -1
  5. package/dist/es/agent/tasks.mjs +7 -4
  6. package/dist/es/agent/tasks.mjs.map +1 -1
  7. package/dist/es/agent/utils.mjs +1 -1
  8. package/dist/es/ai-model/inspect.mjs +4 -2
  9. package/dist/es/ai-model/inspect.mjs.map +1 -1
  10. package/dist/es/ai-model/llm-planning.mjs +4 -1
  11. package/dist/es/ai-model/llm-planning.mjs.map +1 -1
  12. package/dist/es/ai-model/service-caller/index.mjs +52 -7
  13. package/dist/es/ai-model/service-caller/index.mjs.map +1 -1
  14. package/dist/es/service/index.mjs +7 -4
  15. package/dist/es/service/index.mjs.map +1 -1
  16. package/dist/es/types.mjs.map +1 -1
  17. package/dist/es/utils.mjs +2 -2
  18. package/dist/lib/agent/agent.js +3 -4
  19. package/dist/lib/agent/agent.js.map +1 -1
  20. package/dist/lib/agent/task-builder.js +1 -0
  21. package/dist/lib/agent/task-builder.js.map +1 -1
  22. package/dist/lib/agent/tasks.js +7 -4
  23. package/dist/lib/agent/tasks.js.map +1 -1
  24. package/dist/lib/agent/utils.js +1 -1
  25. package/dist/lib/ai-model/inspect.js +4 -2
  26. package/dist/lib/ai-model/inspect.js.map +1 -1
  27. package/dist/lib/ai-model/llm-planning.js +4 -1
  28. package/dist/lib/ai-model/llm-planning.js.map +1 -1
  29. package/dist/lib/ai-model/service-caller/index.js +54 -6
  30. package/dist/lib/ai-model/service-caller/index.js.map +1 -1
  31. package/dist/lib/service/index.js +7 -4
  32. package/dist/lib/service/index.js.map +1 -1
  33. package/dist/lib/types.js.map +1 -1
  34. package/dist/lib/utils.js +2 -2
  35. package/dist/types/agent/agent.d.ts +2 -1
  36. package/dist/types/agent/tasks.d.ts +2 -2
  37. package/dist/types/ai-model/inspect.d.ts +2 -0
  38. package/dist/types/ai-model/llm-planning.d.ts +2 -1
  39. package/dist/types/ai-model/service-caller/index.d.ts +15 -2
  40. package/dist/types/types.d.ts +5 -0
  41. package/dist/types/yaml.d.ts +0 -1
  42. package/package.json +2 -2
@@ -1,4 +1,4 @@
1
- import { type ActionParam, type ActionReturn, type AgentAssertOpt, type AgentDescribeElementAtPointResult, type AgentOpt, type AgentWaitForOpt, type DeviceAction, type ExecutionDump, type GroupedActionDump, type LocateOption, type LocateResultElement, type LocateValidatorResult, type LocatorValidatorOption, type OnTaskStartTip, type ScrollParam, Service, type ServiceAction, type ServiceExtractOption, type ServiceExtractParam, type TUserPrompt, type UIContext } from '../index';
1
+ import { type ActionParam, type ActionReturn, type AgentAssertOpt, type AgentDescribeElementAtPointResult, type AgentOpt, type AgentWaitForOpt, type DeepThinkOption, type DeviceAction, type ExecutionDump, type GroupedActionDump, type LocateOption, type LocateResultElement, type LocateValidatorResult, type LocatorValidatorOption, type OnTaskStartTip, type ScrollParam, Service, type ServiceAction, type ServiceExtractOption, type ServiceExtractParam, type TUserPrompt, type UIContext } from '../index';
2
2
  export type TestStatus = 'passed' | 'failed' | 'timedOut' | 'skipped' | 'interrupted';
3
3
  import type { AbstractInterface } from '../device';
4
4
  import type { TaskRunner } from '../task-runner';
@@ -7,6 +7,7 @@ import { TaskCache } from './task-cache';
7
7
  import { TaskExecutor } from './tasks';
8
8
  export type AiActOptions = {
9
9
  cacheable?: boolean;
10
+ deepThink?: DeepThinkOption;
10
11
  };
11
12
  export declare class Agent<InterfaceType extends AbstractInterface = AbstractInterface> {
12
13
  interface: InterfaceType;
@@ -3,7 +3,7 @@ import type { AbstractInterface } from '../device';
3
3
  import type Service from '../service';
4
4
  import type { TaskRunner } from '../task-runner';
5
5
  import { TaskExecutionError } from '../task-runner';
6
- import type { DeviceAction, ExecutionTaskApply, ExecutionTaskProgressOptions, MidsceneYamlFlowItem, PlanningAction, PlanningActionParamSleep, PlanningActionParamWaitFor, ServiceExtractOption, ServiceExtractParam } from '../types';
6
+ import type { DeepThinkOption, DeviceAction, ExecutionTaskApply, ExecutionTaskProgressOptions, MidsceneYamlFlowItem, PlanningAction, PlanningActionParamSleep, PlanningActionParamWaitFor, ServiceExtractOption, ServiceExtractParam } from '../types';
7
7
  import type { IModelConfig } from '@midscene/shared/env';
8
8
  import type { TaskCache } from './task-cache';
9
9
  export { locatePlanForLocate } from './task-builder';
@@ -46,7 +46,7 @@ export declare class TaskExecutor {
46
46
  runner: TaskRunner;
47
47
  }>;
48
48
  runPlans(title: string, plans: PlanningAction[], modelConfigForPlanning: IModelConfig, modelConfigForDefaultIntent: IModelConfig): Promise<ExecutionResult>;
49
- action(userPrompt: string, modelConfigForPlanning: IModelConfig, modelConfigForDefaultIntent: IModelConfig, includeBboxInPlanning: boolean, aiActContext?: string, cacheable?: boolean, replanningCycleLimitOverride?: number, imagesIncludeCount?: number): Promise<ExecutionResult<{
49
+ action(userPrompt: string, modelConfigForPlanning: IModelConfig, modelConfigForDefaultIntent: IModelConfig, includeBboxInPlanning: boolean, aiActContext?: string, cacheable?: boolean, replanningCycleLimitOverride?: number, imagesIncludeCount?: number, deepThink?: DeepThinkOption): Promise<ExecutionResult<{
50
50
  yamlFlow?: MidsceneYamlFlowItem[];
51
51
  } | undefined>>;
52
52
  private createTypeQueryTask;
@@ -23,6 +23,7 @@ export declare function AiLocateElement(options: {
23
23
  rect?: Rect;
24
24
  rawResponse: string;
25
25
  usage?: AIUsageInfo;
26
+ reasoning_content?: string;
26
27
  }>;
27
28
  export declare function AiLocateSection(options: {
28
29
  context: UIContext;
@@ -45,6 +46,7 @@ export declare function AiExtractElementInfo<T>(options: {
45
46
  }): Promise<{
46
47
  parseResult: AIDataExtractionResponse<T>;
47
48
  usage: AIUsageInfo | undefined;
49
+ reasoning_content: string | undefined;
48
50
  }>;
49
51
  export declare function AiJudgeOrderSensitive(description: string, callAIFn: typeof callAIWithObjectResponse<{
50
52
  isOrderSensitive: boolean;
@@ -1,4 +1,4 @@
1
- import type { DeviceAction, InterfaceType, PlanningAIResponse, UIContext } from '../types';
1
+ import type { DeepThinkOption, DeviceAction, InterfaceType, PlanningAIResponse, UIContext } from '../types';
2
2
  import type { IModelConfig } from '@midscene/shared/env';
3
3
  import type { ConversationHistory } from './conversation-history';
4
4
  export declare function plan(userInstruction: string, opts: {
@@ -10,4 +10,5 @@ export declare function plan(userInstruction: string, opts: {
10
10
  conversationHistory: ConversationHistory;
11
11
  includeBbox: boolean;
12
12
  imagesIncludeCount?: number;
13
+ deepThink?: DeepThinkOption;
13
14
  }): Promise<PlanningAIResponse>;
@@ -1,4 +1,4 @@
1
- import { type AIUsageInfo } from '../../types';
1
+ import { type AIUsageInfo, type DeepThinkOption } from '../../types';
2
2
  import type { StreamingCallback } from '../../types';
3
3
  import { type IModelConfig, type TVlModeTypes } from '@midscene/shared/env';
4
4
  import type { ChatCompletionMessageParam } from 'openai/resources/index';
@@ -6,15 +6,20 @@ import type { AIActionType, AIArgs } from '../../common';
6
6
  export declare function callAI(messages: ChatCompletionMessageParam[], AIActionTypeValue: AIActionType, modelConfig: IModelConfig, options?: {
7
7
  stream?: boolean;
8
8
  onChunk?: StreamingCallback;
9
+ deepThink?: DeepThinkOption;
9
10
  }): Promise<{
10
11
  content: string;
12
+ reasoning_content?: string;
11
13
  usage?: AIUsageInfo;
12
14
  isStreamed: boolean;
13
15
  }>;
14
- export declare function callAIWithObjectResponse<T>(messages: ChatCompletionMessageParam[], AIActionTypeValue: AIActionType, modelConfig: IModelConfig): Promise<{
16
+ export declare function callAIWithObjectResponse<T>(messages: ChatCompletionMessageParam[], AIActionTypeValue: AIActionType, modelConfig: IModelConfig, options?: {
17
+ deepThink?: DeepThinkOption;
18
+ }): Promise<{
15
19
  content: T;
16
20
  contentString: string;
17
21
  usage?: AIUsageInfo;
22
+ reasoning_content?: string;
18
23
  }>;
19
24
  export declare function callAIWithStringResponse(msgs: AIArgs, AIActionTypeValue: AIActionType, modelConfig: IModelConfig): Promise<{
20
25
  content: string;
@@ -22,4 +27,12 @@ export declare function callAIWithStringResponse(msgs: AIArgs, AIActionTypeValue
22
27
  }>;
23
28
  export declare function extractJSONFromCodeBlock(response: string): string;
24
29
  export declare function preprocessDoubaoBboxJson(input: string): string;
30
+ export declare function resolveDeepThinkConfig({ deepThink, vlMode, }: {
31
+ deepThink?: DeepThinkOption;
32
+ vlMode?: TVlModeTypes;
33
+ }): {
34
+ config: Record<string, unknown>;
35
+ debugMessage?: string;
36
+ warningMessage?: string;
37
+ };
25
38
  export declare function safeParseJson(input: string, vlMode: TVlModeTypes | undefined): any;
@@ -97,6 +97,7 @@ export interface LocateResult {
97
97
  rect?: Rect;
98
98
  }
99
99
  export type ThinkingLevel = 'off' | 'medium' | 'high';
100
+ export type DeepThinkOption = 'unset' | true | false;
100
101
  export interface ServiceTaskInfo {
101
102
  durationMs: number;
102
103
  formatResponse?: string;
@@ -105,6 +106,7 @@ export interface ServiceTaskInfo {
105
106
  searchArea?: Rect;
106
107
  searchAreaRawResponse?: string;
107
108
  searchAreaUsage?: AIUsageInfo;
109
+ reasoning_content?: string;
108
110
  }
109
111
  export interface DumpMeta {
110
112
  logTime: number;
@@ -140,6 +142,7 @@ export interface ServiceExtractResult<T> extends ServiceResultBase {
140
142
  data: T;
141
143
  thought?: string;
142
144
  usage?: AIUsageInfo;
145
+ reasoning_content?: string;
143
146
  }
144
147
  export declare class ServiceError extends Error {
145
148
  dump: ServiceDump;
@@ -193,6 +196,7 @@ export interface PlanningAIResponse extends Omit<RawResponsePlanningAIResponse,
193
196
  yamlFlow?: MidsceneYamlFlowItem[];
194
197
  yamlString?: string;
195
198
  error?: string;
199
+ reasoning_content?: string;
196
200
  }
197
201
  export interface PlanningActionParamSleep {
198
202
  timeMs: number;
@@ -269,6 +273,7 @@ export type ExecutionTask<E extends ExecutionTaskApply<any, any, any> = Executio
269
273
  };
270
274
  usage?: AIUsageInfo;
271
275
  searchAreaUsage?: AIUsageInfo;
276
+ reasoning_content?: string;
272
277
  };
273
278
  export interface ExecutionDump extends DumpMeta {
274
279
  name: string;
@@ -129,7 +129,6 @@ export interface MidsceneYamlFlowItemAIAction {
129
129
  aiAct?: string;
130
130
  aiActionProgressTips?: string[];
131
131
  cacheable?: boolean;
132
- _deepThink?: boolean;
133
132
  [key: string]: unknown;
134
133
  }
135
134
  export interface MidsceneYamlFlowItemAIAssert {
package/package.json CHANGED
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "name": "@midscene/core",
3
3
  "description": "Automate browser actions, extract data, and perform assertions using AI. It offers JavaScript SDK, Chrome extension, and support for scripting in YAML. See https://midscenejs.com/ for details.",
4
- "version": "1.0.5-beta-20260104042628.0",
4
+ "version": "1.0.5-beta-20260105023603.0",
5
5
  "repository": "https://github.com/web-infra-dev/midscene",
6
6
  "homepage": "https://midscenejs.com/",
7
7
  "main": "./dist/lib/index.js",
@@ -89,7 +89,7 @@
89
89
  "semver": "7.5.2",
90
90
  "js-yaml": "4.1.0",
91
91
  "zod": "3.24.3",
92
- "@midscene/shared": "1.0.5-beta-20260104042628.0"
92
+ "@midscene/shared": "1.0.5-beta-20260105023603.0"
93
93
  },
94
94
  "devDependencies": {
95
95
  "@rslib/core": "^0.18.3",