perf: dataset import params code (#4875)

* perf: dataset import params code * perf: api dataset code * model
2025-07-22 20:37:48 +00:00 · 2025-05-23 10:40:25 +08:00
parent 9af92d1eae
commit fae76e887a
23 changed files with 366 additions and 295 deletions
--- a/packages/global/core/dataset/api.d.ts
+++ b/packages/global/core/dataset/api.d.ts
@@ -1,9 +1,11 @@
-import type { DatasetDataIndexItemType, DatasetSchemaType } from './type';
+import type { ChunkSettingsType, DatasetDataIndexItemType, DatasetSchemaType } from './type';
 import type {
  DatasetCollectionTypeEnum,
  DatasetCollectionDataProcessModeEnum,
  ChunkSettingModeEnum,
-  DataChunkSplitModeEnum
+  DataChunkSplitModeEnum,
+  ChunkTriggerConfigTypeEnum,
+  ParagraphChunkAIModeEnum
 } from './constants';
 import type { LLMModelItemType } from '../ai/model.d';
 import type { ParentIdType } from 'common/parentFolder/type';
@@ -32,26 +34,16 @@ export type DatasetUpdateBody = {
 };

 /* ================= collection ===================== */
-export type DatasetCollectionChunkMetadataType = {
+// Input + store params
+type DatasetCollectionStoreDataType = ChunkSettingsType & {
  parentId?: string;
-  customPdfParse?: boolean;
-  trainingType?: DatasetCollectionDataProcessModeEnum;
-  imageIndex?: boolean;
-  autoIndexes?: boolean;
-
-  chunkSettingMode?: ChunkSettingModeEnum;
-  chunkSplitMode?: DataChunkSplitModeEnum;
-
-  chunkSize?: number;
-  indexSize?: number;
-
-  chunkSplitter?: string;
-  qaPrompt?: string;
  metadata?: Record<string, any>;
+
+  customPdfParse?: boolean;
 };

 // create collection params
-export type CreateDatasetCollectionParams = DatasetCollectionChunkMetadataType & {
+export type CreateDatasetCollectionParams = DatasetCollectionStoreDataType & {
  datasetId: string;
  name: string;
  type: DatasetCollectionTypeEnum;
@@ -72,7 +64,7 @@ export type CreateDatasetCollectionParams = DatasetCollectionChunkMetadataType &
  nextSyncTime?: Date;
 };

-export type ApiCreateDatasetCollectionParams = DatasetCollectionChunkMetadataType & {
+export type ApiCreateDatasetCollectionParams = DatasetCollectionStoreDataType & {
  datasetId: string;
  tags?: string[];
 };
@@ -90,7 +82,7 @@ export type ApiDatasetCreateDatasetCollectionParams = ApiCreateDatasetCollection
 export type FileIdCreateDatasetCollectionParams = ApiCreateDatasetCollectionParams & {
  fileId: string;
 };
-export type reTrainingDatasetFileCollectionParams = DatasetCollectionChunkMetadataType & {
+export type reTrainingDatasetFileCollectionParams = DatasetCollectionStoreDataType & {
  datasetId: string;
  collectionId: string;
 };
--- a/packages/global/core/dataset/constants.ts
+++ b/packages/global/core/dataset/constants.ts
@@ -143,15 +143,25 @@ export const DatasetCollectionDataProcessModeMap = {
  }
 };

+export enum ChunkTriggerConfigTypeEnum {
+  minSize = 'minSize',
+  forceChunk = 'forceChunk',
+  maxSize = 'maxSize'
+}
 export enum ChunkSettingModeEnum {
  auto = 'auto',
  custom = 'custom'
 }

 export enum DataChunkSplitModeEnum {
+  paragraph = 'paragraph',
  size = 'size',
  char = 'char'
 }
+export enum ParagraphChunkAIModeEnum {
+  auto = 'auto',
+  force = 'force'
+}

 /* ------------ data -------------- */

--- a/packages/global/core/dataset/data/constants.ts
+++ b/packages/global/core/dataset/data/constants.ts
@@ -32,7 +32,7 @@ export const DatasetDataIndexMap: Record<
    color: 'red'
  },
  [DatasetDataIndexTypeEnum.image]: {
-    label: i18nT('common:data_index_image'),
+    label: i18nT('dataset:data_index_image'),
    color: 'purple'
  }
 };
--- a/packages/global/core/dataset/type.d.ts
+++ b/packages/global/core/dataset/type.d.ts
@@ -8,26 +8,42 @@ import type {
  DatasetStatusEnum,
  DatasetTypeEnum,
  SearchScoreTypeEnum,
-  TrainingModeEnum
+  TrainingModeEnum,
+  ChunkSettingModeEnum
 } from './constants';
 import type { DatasetPermission } from '../../support/permission/dataset/controller';
-import { Permission } from '../../support/permission/controller';
 import type { APIFileServer, FeishuServer, YuqueServer } from './apiDataset';
 import type { SourceMemberType } from 'support/user/type';
 import type { DatasetDataIndexTypeEnum } from './data/constants';
-import type { ChunkSettingModeEnum } from './constants';

 export type ChunkSettingsType = {
-  trainingType: DatasetCollectionDataProcessModeEnum;
-  autoIndexes?: boolean;
+  trainingType?: DatasetCollectionDataProcessModeEnum;
+
+  // Chunk trigger
+  chunkTriggerType?: ChunkTriggerConfigTypeEnum;
+  chunkTriggerMinSize?: number; // maxSize from agent model, not store
+
+  // Data enhance
+  dataEnhanceCollectionName?: boolean; // Auto add collection name to data
+
+  // Index enhance
  imageIndex?: boolean;
+  autoIndexes?: boolean;

-  chunkSettingMode?: ChunkSettingModeEnum;
+  // Chunk setting
+  chunkSettingMode?: ChunkSettingModeEnum; // 系统参数/自定义参数
  chunkSplitMode?: DataChunkSplitModeEnum;
-
+  // Paragraph split
+  paragraphChunkAIMode?: ParagraphChunkAIModeEnum;
+  paragraphChunkDeep?: number; // Paragraph deep
+  paragraphChunkMinSize?: number; // Paragraph min size, if too small, it will merge
+  paragraphChunkMaxSize?: number; // Paragraph max size, if too large, it will split
+  // Size split
  chunkSize?: number;
-  indexSize?: number;
+  // Char split
  chunkSplitter?: string;
+  indexSize?: number;
+
  qaPrompt?: string;
 };

@@ -66,7 +82,7 @@ export type DatasetSchemaType = {
  defaultPermission?: number;
 };

-export type DatasetCollectionSchemaType = {
+export type DatasetCollectionSchemaType = ChunkSettingsType & {
  _id: string;
  teamId: string;
  tmbId: string;
@@ -101,18 +117,7 @@ export type DatasetCollectionSchemaType = {

  // Parse settings
  customPdfParse?: boolean;
-  // Chunk settings
-  autoIndexes?: boolean;
-  imageIndex?: boolean;
  trainingType: DatasetCollectionDataProcessModeEnum;
-
-  chunkSettingMode?: ChunkSettingModeEnum;
-  chunkSplitMode?: DataChunkSplitModeEnum;
-
-  chunkSize?: number;
-  indexSize?: number;
-  chunkSplitter?: string;
-  qaPrompt?: string;
 };

 export type DatasetCollectionTagsSchemaType = {
--- a/packages/service/core/ai/config/provider/Claude.json
+++ b/packages/service/core/ai/config/provider/Claude.json
@@ -1,6 +1,54 @@
 {
  "provider": "Claude",
  "list": [
+    {
+      "model": "claude-sonnet-4-20250514",
+      "name": "claude-sonnet-4-20250514",
+      "maxContext": 200000,
+      "maxResponse": 8000,
+      "quoteMaxToken": 100000,
+      "maxTemperature": 1,
+      "showTopP": true,
+      "showStopSign": true,
+      "vision": true,
+      "toolChoice": true,
+      "functionCall": false,
+      "defaultSystemChatPrompt": "",
+      "datasetProcess": true,
+      "usedInClassify": true,
+      "customCQPrompt": "",
+      "usedInExtractFields": true,
+      "usedInQueryExtension": true,
+      "customExtractPrompt": "",
+      "usedInToolCall": true,
+      "defaultConfig": {},
+      "fieldMap": {},
+      "type": "llm"
+    },
+    {
+      "model": "claude-opus-4-20250514",
+      "name": "claude-opus-4-20250514",
+      "maxContext": 200000,
+      "maxResponse": 4096,
+      "quoteMaxToken": 100000,
+      "maxTemperature": 1,
+      "showTopP": true,
+      "showStopSign": true,
+      "vision": true,
+      "toolChoice": true,
+      "functionCall": false,
+      "defaultSystemChatPrompt": "",
+      "datasetProcess": true,
+      "usedInClassify": true,
+      "customCQPrompt": "",
+      "usedInExtractFields": true,
+      "usedInQueryExtension": true,
+      "customExtractPrompt": "",
+      "usedInToolCall": true,
+      "defaultConfig": {},
+      "fieldMap": {},
+      "type": "llm"
+    },
    {
      "model": "claude-3-7-sonnet-20250219",
      "name": "claude-3-7-sonnet-20250219",
--- a/packages/service/core/ai/config/provider/Gemini.json
+++ b/packages/service/core/ai/config/provider/Gemini.json
@@ -25,6 +25,30 @@
      "showTopP": true,
      "showStopSign": true
    },
+    {
+      "model": "gemini-2.5-flash-preview-04-17",
+      "name": "gemini-2.5-flash-preview-04-17",
+      "maxContext": 1000000,
+      "maxResponse": 8000,
+      "quoteMaxToken": 60000,
+      "maxTemperature": 1,
+      "vision": true,
+      "toolChoice": true,
+      "functionCall": false,
+      "defaultSystemChatPrompt": "",
+      "datasetProcess": true,
+      "usedInClassify": true,
+      "customCQPrompt": "",
+      "usedInExtractFields": true,
+      "usedInQueryExtension": true,
+      "customExtractPrompt": "",
+      "usedInToolCall": true,
+      "defaultConfig": {},
+      "fieldMap": {},
+      "type": "llm",
+      "showTopP": true,
+      "showStopSign": true
+    },
    {
      "model": "gemini-2.0-flash",
      "name": "gemini-2.0-flash",
--- a/packages/service/core/dataset/collection/controller.ts
+++ b/packages/service/core/dataset/collection/controller.ts
@@ -74,6 +74,15 @@ export const createCollectionAndInsertData = async ({
    llmModel: getLLMModel(dataset.agentModel)
  });
  const chunkSplitter = computeChunkSplitter(createCollectionParams);
+  if (trainingType === DatasetCollectionDataProcessModeEnum.qa) {
+    delete createCollectionParams.chunkTriggerType;
+    delete createCollectionParams.chunkTriggerMinSize;
+    delete createCollectionParams.dataEnhanceCollectionName;
+    delete createCollectionParams.imageIndex;
+    delete createCollectionParams.autoIndexes;
+    delete createCollectionParams.indexSize;
+    delete createCollectionParams.qaPrompt;
+  }

  // 1. split chunks
  const chunks = rawText2Chunks({
--- a/packages/service/core/dataset/read.ts
+++ b/packages/service/core/dataset/read.ts
@@ -163,7 +163,7 @@ export const readApiServerFileContent = async ({
  title?: string;
  rawText: string;
 }> => {
-  const data = (
+  return (
    await getApiDatasetRequest({
      apiServer,
      yuqueServer,
@@ -175,10 +175,6 @@ export const readApiServerFileContent = async ({
    apiFileId,
    customPdfParse
  });
-  if (data) {
-    return data;
-  }
-  return Promise.reject(Error);
 };

 export const rawText2Chunks = ({
--- a/packages/service/core/dataset/schema.ts
+++ b/packages/service/core/dataset/schema.ts
@@ -1,10 +1,12 @@
 import { getMongoModel, Schema } from '../../common/mongo';
 import {
  ChunkSettingModeEnum,
+  ChunkTriggerConfigTypeEnum,
  DataChunkSplitModeEnum,
  DatasetCollectionDataProcessModeEnum,
  DatasetTypeEnum,
-  DatasetTypeMap
+  DatasetTypeMap,
+  ParagraphChunkAIModeEnum
 } from '@fastgpt/global/core/dataset/constants';
 import {
  TeamCollectionName,
@@ -15,12 +17,22 @@ import type { DatasetSchemaType } from '@fastgpt/global/core/dataset/type.d';
 export const DatasetCollectionName = 'datasets';

 export const ChunkSettings = {
-  imageIndex: Boolean,
-  autoIndexes: Boolean,
  trainingType: {
    type: String,
    enum: Object.values(DatasetCollectionDataProcessModeEnum)
  },
+
+  chunkTriggerType: {
+    type: String,
+    enum: Object.values(ChunkTriggerConfigTypeEnum)
+  },
+  chunkTriggerMinSize: Number,
+
+  dataEnhanceCollectionName: Boolean,
+
+  imageIndex: Boolean,
+  autoIndexes: Boolean,
+
  chunkSettingMode: {
    type: String,
    enum: Object.values(ChunkSettingModeEnum)
@@ -29,6 +41,13 @@ export const ChunkSettings = {
    type: String,
    enum: Object.values(DataChunkSplitModeEnum)
  },
+  paragraphChunkAIMode: {
+    type: String,
+    enum: Object.values(ParagraphChunkAIModeEnum)
+  },
+  paragraphChunkDeep: Number,
+  paragraphChunkMinSize: Number,
+  paragraphChunkMaxSize: Number,
  chunkSize: Number,
  chunkSplitter: String,

@@ -115,9 +134,7 @@ const DatasetSchema = new Schema({

  // abandoned
  autoSync: Boolean,
-  externalReadUrl: {
-    type: String
-  },
+  externalReadUrl: String,
  defaultPermission: Number
 });

--- a/packages/web/i18n/en/common.json
+++ b/packages/web/i18n/en/common.json
@@ -749,7 +749,6 @@
  "custom_title": "Custom Title",
  "data_index_custom": "Custom index",
  "data_index_default": "Default index",
-  "data_index_image": "Image Index",
  "data_index_question": "Inferred question index",
  "data_index_summary": "Summary Index",
  "data_not_found": "Data can't be found",
--- a/packages/web/i18n/en/dataset.json
+++ b/packages/web/i18n/en/dataset.json
@@ -22,7 +22,6 @@
  "collection.training_type": "Chunk type",
  "collection_data_count": "Data amount",
  "collection_metadata_custom_pdf_parse": "PDF enhancement analysis",
-  "collection_metadata_image_parse": "Image tagging",
  "collection_not_support_retraining": "This collection type does not support retuning parameters",
  "collection_not_support_sync": "This collection does not support synchronization",
  "collection_sync": "Sync data",
@@ -38,6 +37,7 @@
  "custom_split_sign_tip": "Allows you to chunk according to custom delimiters. \nUsually used for processed data, using specific separators for precise chunking. \nYou can use the | symbol to represent multiple splitters, such as: \".|.\" to represent a period in Chinese and English.\n\nTry to avoid using special symbols related to regular, such as: * () [] {}, etc.",
  "data_amount": "{{dataAmount}} Datas, {{indexAmount}} Indexes",
  "data_error_amount": "{{errorAmount}} Group training exception",
+  "data_index_image": "Image index",
  "data_index_num": "Index {{index}}",
  "data_process_params": "Params",
  "data_process_setting": "Processing config",
--- a/packages/web/i18n/zh-CN/common.json
+++ b/packages/web/i18n/zh-CN/common.json
@@ -749,7 +749,6 @@
  "custom_title": "自定义标题",
  "data_index_custom": "自定义索引",
  "data_index_default": "默认索引",
-  "data_index_image": "图片索引",
  "data_index_question": "推测问题索引",
  "data_index_summary": "摘要索引",
  "data_not_found": "数据找不到了",
--- a/packages/web/i18n/zh-CN/dataset.json
+++ b/packages/web/i18n/zh-CN/dataset.json
@@ -22,7 +22,6 @@
  "collection.training_type": "处理模式",
  "collection_data_count": "数据量",
  "collection_metadata_custom_pdf_parse": "PDF增强解析",
-  "collection_metadata_image_parse": "图片标注",
  "collection_not_support_retraining": "该集合类型不支持重新调整参数",
  "collection_not_support_sync": "该集合不支持同步",
  "collection_sync": "立即同步",
@@ -38,6 +37,7 @@
  "custom_split_sign_tip": "允许你根据自定义的分隔符进行分块。通常用于已处理好的数据，使用特定的分隔符来精确分块。可以使用 | 符号表示多个分割符，例如：“。|.” 表示中英文句号。\n尽量避免使用正则相关特殊符号，例如: * () [] {} 等。",
  "data_amount": "{{dataAmount}} 组数据, {{indexAmount}} 组索引",
  "data_error_amount": "{{errorAmount}} 组训练异常",
+  "data_index_image": "图片索引",
  "data_index_num": "索引 {{index}}",
  "data_process_params": "处理参数",
  "data_process_setting": "数据处理配置",
--- a/packages/web/i18n/zh-Hant/common.json
+++ b/packages/web/i18n/zh-Hant/common.json
@@ -749,7 +749,6 @@
  "custom_title": "自訂標題",
  "data_index_custom": "自定義索引",
  "data_index_default": "預設索引",
-  "data_index_image": "圖片索引",
  "data_index_question": "推測問題索引",
  "data_index_summary": "摘要索引",
  "data_not_found": "數據找不到了",
--- a/packages/web/i18n/zh-Hant/dataset.json
+++ b/packages/web/i18n/zh-Hant/dataset.json
@@ -21,7 +21,6 @@
  "collection.training_type": "處理模式",
  "collection_data_count": "資料量",
  "collection_metadata_custom_pdf_parse": "PDF 增強解析",
-  "collection_metadata_image_parse": "圖片標註",
  "collection_not_support_retraining": "此集合類型不支援重新調整參數",
  "collection_not_support_sync": "該集合不支援同步",
  "collection_sync": "立即同步",
@@ -37,6 +36,7 @@
  "custom_split_sign_tip": "允許你根據自定義的分隔符進行分塊。\n通常用於已處理好的資料，使用特定的分隔符來精確分塊。\n可以使用 | 符號表示多個分割符，例如：“。|.”表示中英文句號。\n\n盡量避免使用正則相關特殊符號，例如：* () [] {} 等。",
  "data_amount": "{{dataAmount}} 組資料，{{indexAmount}} 組索引",
  "data_error_amount": "{{errorAmount}} 組訓練異常",
+  "data_index_image": "圖片索引",
  "data_index_num": "索引 {{index}}",
  "data_process_params": "處理參數",
  "data_process_setting": "資料處理設定",