feat(dataset/tag): upgrade knowledge base tag management

.agents/design/core/dataset/index.md

@@ -811,7 +811,7 @@ POST   /api/core/dataset/collection/create/link         # 链接导入
 POST   /api/core/dataset/collection/create/text         # 文本导入
 POST   /api/core/dataset/collection/create/images       # 图片导入
 PUT    /api/core/dataset/collection/update              # 更新集合
-GET    /api/core/dataset/collection/list                # 集合列表
+POST   /api/core/dataset/collection/listV2              # 集合列表
 GET    /api/core/dataset/collection/detail              # 集合详情
 POST   /api/core/dataset/collection/sync                # 同步集合
 GET    /api/core/dataset/collection/export              # 导出集合
@@ -893,9 +893,9 @@ MongoDatasetData.find(query, fields, {
 
 ### 3. 分页优化
 ```typescript
-// 使用 scrollList 而非传统分页
-// 避免深度分页性能问题
-GET /api/core/dataset/collection/scrollList?lastId=xxx&limit=20
+// 使用 offset 分页，避免深度 pageNum 分页性能问题
+POST /api/core/dataset/collection/listV2
+{ datasetId, offset, pageSize }
 ```
 
 ### 4. 缓存策略

.gitignore

@@ -50,3 +50,4 @@ pro/admin/worker/
 /pro/llm_benchmark/content_benchmark/eval-runs/
 /pro/llm_benchmark/content_benchmark/.cache/
 .gstack/
+.worktrees/

packages/global/common/error/code/dataset.ts

@@ -15,7 +15,15 @@ export enum DatasetErrEnum {
   notSupportSync = 'notSupportSync',
   sameApiCollection = 'sameApiCollection',
   noApiServer = 'noApiServer',
-  canNotEditAdminPermission = 'canNotEditAdminPermission'
+  canNotEditAdminPermission = 'canNotEditAdminPermission',
+
+  // Tag errors (501013+)
+  tagNameDuplicate = 'tagNameDuplicate',
+  tagNameEmpty = 'tagNameEmpty',
+  tagNotExist = 'tagNotExist',
+  tagValueInvalid = 'tagValueInvalid',
+  tagValueDatetimeInvalid = 'tagValueDatetimeInvalid',
+  arrayTagValueInvalid = 'arrayTagValueInvalid'
 }
 const datasetErr = [
   {
@@ -69,6 +77,32 @@ const datasetErr = [
   {
     statusText: DatasetErrEnum.noApiServer,
     message: i18nT('common:core.dataset.error.noApiServer')
+  },
+
+  // Tag errors
+  {
+    statusText: DatasetErrEnum.tagNameDuplicate,
+    message: i18nT('common:core.dataset.error.tagNameDuplicate')
+  },
+  {
+    statusText: DatasetErrEnum.tagNameEmpty,
+    message: i18nT('common:core.dataset.error.tagNameEmpty')
+  },
+  {
+    statusText: DatasetErrEnum.tagNotExist,
+    message: i18nT('common:core.dataset.error.tagNotExist')
+  },
+  {
+    statusText: DatasetErrEnum.tagValueInvalid,
+    message: i18nT('common:core.dataset.error.tagValueInvalid')
+  },
+  {
+    statusText: DatasetErrEnum.tagValueDatetimeInvalid,
+    message: i18nT('common:core.dataset.error.tagValueDatetimeInvalid')
+  },
+  {
+    statusText: DatasetErrEnum.arrayTagValueInvalid,
+    message: i18nT('common:core.dataset.error.arrayTagValueInvalid')
   }
 ];
 export default datasetErr.reduce((acc, cur, index) => {

packages/global/common/string/time.ts

@@ -83,7 +83,7 @@ export const cronParser2Fields = (cronString: string) => {
   try {
     const cronField = cronParser.parseExpression(cronString).fields;
     return cronField;
-  } catch (error) {
+  } catch {
     return null;
   }
 };

packages/global/core/app/formEdit/utils.ts

@@ -36,7 +36,8 @@ const agentGeneratedDenyRenderTypes = new Set<FlowNodeInputTypeEnum>([
   FlowNodeInputTypeEnum.selectTool,
   FlowNodeInputTypeEnum.selectDataset,
   FlowNodeInputTypeEnum.selectDatasetParamsModal,
-  FlowNodeInputTypeEnum.settingDatasetQuotePrompt
+  FlowNodeInputTypeEnum.settingDatasetQuotePrompt,
+  FlowNodeInputTypeEnum.datasetTagFilter
 ]);
 
 // 工具配置不能处理依赖文件、知识库、模型或外部动态上下文的输入。

packages/global/core/app/type.ts

@@ -2,6 +2,10 @@ import { StoreNodeItemTypeSchema } from '../workflow/type/node';
 import { AppTypeEnum } from './constants';
 import { NodeInputKeyEnum } from '../workflow/constants';
 import { DatasetSearchModeEnum } from '../dataset/constants';
+import {
+  DatasetTagFilterValueSchema,
+  DatasetTagFilterVersionSchema
+} from '../dataset/workflowTagFilter';
 import type { ReasoningEffort } from '../ai/llm/type';
 import { StoreEdgeItemTypeSchema } from '../workflow/type/edge';
 import type { AppPermission } from '../../support/permission/app/controller';
@@ -271,7 +275,8 @@ export const AppDatasetSearchParamsTypeSchema = z.object({
   datasetSearchExtensionBg: z.string().optional(),
   [NodeInputKeyEnum.authTmbId]: BoolSchema.optional(),
 
-  collectionFilterMatch: z.string().optional()
+  collectionFilterMatch: z.union([z.string(), DatasetTagFilterValueSchema]).optional(),
+  [NodeInputKeyEnum.collectionFilterVersion]: DatasetTagFilterVersionSchema.optional()
 });
 export type AppDatasetSearchParamsType = z.infer<typeof AppDatasetSearchParamsTypeSchema>;
 

packages/global/core/app/utils.ts

@@ -6,6 +6,7 @@ import { AppTypeEnum } from './constants';
 import appErrList from '../../common/error/code/app';
 import pluginErrList from '../../common/error/code/plugin';
 import { i18nT } from '../../common/i18n/utils';
+import { DatasetTagFilterVersionEnum } from '../dataset/workflowTagFilter';
 
 const deletedPluginErrorList = new Set([
   'plugin.team_not_installed',
@@ -34,6 +35,7 @@ export const getDefaultAppForm = (): AppFormEditFormType => {
       rerankWeight: 0.5,
       datasetSearchUsingExtensionQuery: true,
       datasetSearchExtensionBg: '',
+      [NodeInputKeyEnum.collectionFilterVersion]: DatasetTagFilterVersionEnum.structured,
       [NodeInputKeyEnum.authTmbId]: false
     },
     selectedTools: [],

packages/global/core/dataset/collection/utils.ts

@@ -1,7 +1,12 @@
 import { DatasetCollectionTypeEnum } from '../constants';
 import { type DatasetCollectionSchemaType } from '../type';
 
-export const getCollectionSourceData = (collection?: DatasetCollectionSchemaType) => {
+export const getCollectionSourceData = (
+  collection?: Pick<
+    DatasetCollectionSchemaType,
+    'fileId' | 'rawLink' | 'externalFileId' | 'externalFileUrl' | 'apiFileId' | 'name'
+  >
+) => {
   return {
     sourceId:
       collection?.fileId ||

packages/global/core/dataset/constants.ts

@@ -164,6 +164,29 @@ export const DatasetCollectionSyncResultMap = {
   }
 };
 
+/* ------------ collection tags -------------- */
+export enum DatasetCollectionTagTypeEnum {
+  string = 'string',
+  number = 'number',
+  datetime = 'datetime',
+  array = 'array'
+}
+
+export const DatasetCollectionTagTypeMap = {
+  [DatasetCollectionTagTypeEnum.string]: {
+    label: i18nT('dataset:core.dataset.tags.string')
+  },
+  [DatasetCollectionTagTypeEnum.number]: {
+    label: i18nT('dataset:core.dataset.tags.number')
+  },
+  [DatasetCollectionTagTypeEnum.datetime]: {
+    label: i18nT('dataset:core.dataset.tags.date')
+  },
+  [DatasetCollectionTagTypeEnum.array]: {
+    label: i18nT('dataset:core.dataset.tags.array')
+  }
+};
+
 export enum DatasetCollectionDataProcessModeEnum {
   chunk = 'chunk',
   qa = 'qa',

packages/global/core/dataset/tagUtils.ts

@@ -0,0 +1,19 @@
+import type { CollectionTagValueType } from './type';
+
+/** 同一标签类型固定：数字按大小，其余按字典序。前后端筛选项展示共用。 */
+export const sortCollectionTagValues = <T extends string | number>(values: T[]): T[] =>
+  [...values].sort((a, b) => {
+    if (typeof a === 'number' && typeof b === 'number') return a - b;
+    return String(a).localeCompare(String(b));
+  });
+
+/** 筛选项只保留非空字符串和有限数字。 */
+export const isUsableCollectionTagFilterValue = (value: unknown): value is string | number => {
+  if (typeof value === 'string') return value.length > 0;
+  if (typeof value === 'number') return Number.isFinite(value);
+  return false;
+};
+
+/** 新格式 { tagId, value }；旧字符串标签排除。 */
+export const isCollectionTagValue = (item: unknown): item is CollectionTagValueType =>
+  !!item && typeof item === 'object' && !Array.isArray(item) && 'tagId' in item && 'value' in item;

packages/global/core/dataset/type.ts

@@ -10,7 +10,9 @@ import {
   CollectionTrainingStatusEnum,
   ChunkSettingModeEnum,
   ChunkTriggerConfigTypeEnum,
-  ParagraphChunkAIModeEnum
+  ParagraphChunkAIModeEnum,
+  DatasetCollectionTagTypeEnum,
+  DatasetCollectionTagTypeMap
 } from './constants';
 import {
   ApiDatasetServerSchema,
@@ -27,6 +29,59 @@ import { ObjectIdSchema } from '../../common/type/mongo';
 import { PermissionSchema } from '../../support/permission/controller';
 import { NumSchema } from '../../common/zod';
 
+/* ===== Tag Type ===== */
+/** 标签类型枚举 */
+export { DatasetCollectionTagTypeEnum, DatasetCollectionTagTypeMap };
+export type DatasetCollectionTagType = `${DatasetCollectionTagTypeEnum}`;
+
+/** 选项类标签的预设选项，空选项由前端草稿过滤后再提交。 */
+export const DatasetCollectionTagOptionsSchema = z
+  .array(z.string().trim().min(1))
+  .meta({ description: '选项类标签的预设选项' });
+
+/** 旧字符串标签承载记录首次创建时使用的默认名称；身份只由 fromMigration 标识。 */
+export const DEFAULT_TAG = 'default_tag';
+
+/** Collection 标签值字段：string/number 存对应值，datetime 存 UTC 毫秒时间戳，array 存 string 数组 */
+export const CollectionTagValueFieldSchema = z.union([z.string(), z.number(), z.array(z.string())]);
+
+/** Collection API 的兼容标签输入/展示格式：旧标签名或带名称和值的新格式。 */
+export const CollectionTagLabelSchema = z.union([
+  z.string(),
+  z.object({
+    tag: z.string(),
+    value: CollectionTagValueFieldSchema
+  })
+]);
+export type CollectionTagLabelType = z.infer<typeof CollectionTagLabelSchema>;
+
+/** Collection 标签值类型（新格式） */
+export const CollectionTagValueSchema = z.object({
+  tagId: z.string().meta({ description: '引用 dataset_collection_tags_v2._id' }),
+  value: CollectionTagValueFieldSchema.meta({
+    description:
+      '标签值。string/number 类型存对应值，datetime 类型存 UTC 毫秒时间戳，array 类型存 string 数组'
+  })
+});
+export type CollectionTagValueType = z.infer<typeof CollectionTagValueSchema>;
+
+/** 详情页按标签值筛选的单条条件。同一标签多值为 OR，不同标签由调用方做 AND。 */
+export const CollectionTagFilterItemSchema = z.object({
+  tagId: ObjectIdSchema.meta({
+    example: '68ad85a7463006c963799a05',
+    description: '标签 ID'
+  }),
+  values: z
+    .array(z.union([z.string().min(1), z.number().finite()]))
+    .min(1)
+    .meta({
+      example: ['PRD'],
+      description:
+        '选中的标签值。文本/选项为字符串，数字为数值，日期为 UTC 毫秒时间戳。同一标签多值为 OR'
+    })
+});
+export type CollectionTagFilterItem = z.infer<typeof CollectionTagFilterItemSchema>;
+
 /* ===== Chunk ===== */
 export const ChunkSettingsSchema = z.object({
   trainingType: z
@@ -134,7 +189,10 @@ export const DatasetCollectionSchema = ChunkSettingsSchema.omit({
   parentId: ParentIdSchema.meta({ description: '父级 ID' }),
   name: z.string().meta({ description: '名称' }),
   type: z.enum(DatasetCollectionTypeEnum).meta({ description: '集合类型' }),
-  tags: z.array(z.string()).optional().meta({ description: '标签' }),
+  tags: z
+    .array(z.union([z.string(), CollectionTagValueSchema]))
+    .optional()
+    .meta({ description: '标签。支持混合格式：String(ObjectId) 旧格式 | { tagId, value } 新格式' }),
 
   createTime: z.coerce.date().meta({ description: '创建时间' }),
   updateTime: z.coerce.date().meta({ description: '更新时间' }),
@@ -165,7 +223,14 @@ export const DatasetCollectionTagsSchema = z.object({
   _id: ObjectIdSchema.meta({ description: '标签 ID' }),
   teamId: ObjectIdSchema.meta({ description: '团队 ID' }),
   datasetId: ObjectIdSchema.meta({ description: '数据集 ID' }),
-  tag: z.string().meta({ description: '标签' })
+  tag: z.string().meta({ description: '标签' }),
+  tagType: z.enum(DatasetCollectionTagTypeEnum).default(DatasetCollectionTagTypeEnum.string).meta({
+    description: '标签类型：string(默认)/number/datetime/array'
+  }),
+  options: DatasetCollectionTagOptionsSchema.optional(),
+  fromMigration: z.boolean().optional().meta({
+    description: '标识该记录是旧标签迁移/旧格式输入创建的 default_tag 承载记录'
+  })
 });
 export type DatasetCollectionTagsSchemaType = z.infer<typeof DatasetCollectionTagsSchema>;
 
@@ -353,7 +418,14 @@ export type DatasetItemType = z.infer<typeof DatasetItemSchema>;
 /* ================= tag ===================== */
 export const DatasetTagSchema = z.object({
   _id: ObjectIdSchema.meta({ description: '标签 ID' }),
-  tag: z.string().meta({ description: '标签' })
+  tag: z.string().meta({ description: '标签' }),
+  tagType: z.enum(DatasetCollectionTagTypeEnum).default(DatasetCollectionTagTypeEnum.string).meta({
+    description: '标签类型：string(默认)/number/datetime/array'
+  }),
+  options: DatasetCollectionTagOptionsSchema.optional(),
+  fromMigration: z.boolean().optional().meta({
+    description: '是否为旧标签数据的系统承载定义'
+  })
 });
 export type DatasetTagType = z.infer<typeof DatasetTagSchema>;
 
@@ -365,6 +437,11 @@ export type TagUsageType = z.infer<typeof TagUsageSchema>;
 
 /* ================= collection ===================== */
 export const DatasetCollectionItemSchema = CollectionWithDatasetSchema.extend({
+  // 详情接口的 tags 由 collectionTagsToTagLabel 解析为标签名格式（string | { tag, value }），区别于存储格式（string | { tagId, value }）
+  tags: z
+    .array(CollectionTagLabelSchema)
+    .optional()
+    .meta({ description: '标签。string 为标签名；新格式为 { tag, value }' }),
   sourceName: z.string().meta({ description: '来源名称' }),
   sourceId: z.string().optional().meta({ description: '来源 ID' }),
   file: z