mirror of
https://github.com/labring/FastGPT.git
synced 2025-07-26 15:54:11 +00:00
V4.8.15 feature (#3331)
* feat: add customize toolkit (#3205) * chaoyang * fix-auth * add toolkit * add order * plugin usage * fix * delete console: * Fix: Fix fullscreen preview top positioning and improve Markdown rendering logic (#3247) * 完成任务：修复全屏预览顶部固定问题，优化 Markdown 渲染逻辑 * 有问题修改 * 问题再修改 * 修正问题 * fix: plugin standalone display issue (#3254) * 4.8.15 test (#3246) * o1 config * perf: system plugin code * 调整系统插件代码。增加html 渲染安全配置。 (#3258) * perf: base64 picker * perf: list app or dataset * perf: plugin config code * 小窗适配等问题 (#3257) * 小窗适配等问题 * git问题 * 小窗剩余问题 * feat: system plugin auth and lock version (#3265) * feat: system plugin auth and lock version * update comment * 4.8.15 test (#3267) * tmp log * perf: login direct * perf: iframe html code * remove log * fix: plugin standalone display (#3277) * refactor: 页面拆分&i18n拆分 (#3281) * refactor: account组件拆成独立页面 * script: 新增i18n json文件创建脚本 * refactor: 页面i18n拆分 * i18n: add en&hant * 4.8.15 test (#3285) * tmp log * remove log * fix: watch avatar refresh * perf: i18n code * fix(plugin): use intro instead of userguide (#3290) * Universal SSO (#3292) * tmp log * remove log * feat: common oauth * readme * perf: sso provider * remove sso code * perf: refresh plugins * feat: add api dataset (#3272) * add api-dataset * fix api-dataset * fix api dataset * fix ts * perf: create collection code (#3301) * tmp log * remove log * perf: i18n change * update version doc * feat: question guide from chatId * perf: create collection code * fix: request api * fix: request api * fix: tts auth and response type (#3303) * perf: md splitter * fix: tts auth and response type * fix: api file dataset (#3307) * perf: api dataset init (#3310) * perf: collection schema * perf: api dataset init * refactor: 团队管理独立页面 (#3302) * ui: 团队管理独立页面 * 代码优化 * fix * perf: sync collection and ui check (#3314) * perf: sync collection * remove script * perf: update api server * perf: api dataset parent * perf: team ui * perf: team 18n * update team ui * perf: ui check * perf: i18n * fix: debug variables & cronjob & system plugin callback load (#3315) * fix: debug variables & cronjob & system plugin callback load * fix type * fix * fix * fix: plugin dataset quote;perf: system variables init (#3316) * fix: plugin dataset quote * perf: system variables init * perf: node templates ui;fix: dataset import ui (#3318) * fix: dataset import ui * perf: node templates ui * perf: ui refresh * feat:套餐改名和套餐跳转配置 (#3309) * fixing:except Sidebar * 去除了多余的代码 * 修正了套餐说明的代码 * 修正了误删除的show_git代码 * 修正了名字部分等代码 * 修正了问题,遗留了其他和ui讨论不一致的部分 * 4.8.15 test (#3319) * remove log * pref: bill ui * pref: bill ui * perf: log * html渲染文档 (#3270) * html渲染文档 * 文档有点小问题 * feat: doc (#3322) * 集合重训练 (#3282) * rebaser * 一点补充 * 小问题 * 其他问题修正，删除集合保留文件的参数还没找到... * reTraining * delete uesless * 删除了一行错误代码 * 集合重训练部分 * fixing * 删除console代码 * feat: navbar item config (#3326) * perf: custom navbar code;perf: retraining code;feat: api dataset and dataset api doc (#3329) * feat: api dataset and dataset api doc * perf: retraining code * perf: custom navbar code * fix: ts (#3330) * fix: ts * fix: ts * retraining ui * perf: api collection filter * perf: retrining button --------- Co-authored-by: heheer <heheer@sealos.io> Co-authored-by: Jiangween <145003935+Jiangween@users.noreply.github.com> Co-authored-by: papapatrick <109422393+Patrickill@users.noreply.github.com>
This commit is contained in:
Archer
@@ -3,7 +3,8 @@ import type { CreateDatasetCollectionParams } from '@fastgpt/global/core/dataset
|
||||
import { MongoDatasetCollection } from './schema';
|
||||
import {
|
||||
CollectionWithDatasetType,
|
||||
DatasetCollectionSchemaType
|
||||
DatasetCollectionSchemaType,
|
||||
DatasetSchemaType
|
||||
} from '@fastgpt/global/core/dataset/type';
|
||||
import { MongoDatasetTraining } from '../training/schema';
|
||||
import { MongoDatasetData } from '../data/schema';
|
||||
@@ -13,7 +14,132 @@ import { delFileByFileIdList } from '../../../common/file/gridfs/controller';
|
||||
import { BucketNameEnum } from '@fastgpt/global/common/file/constants';
|
||||
import { ClientSession } from '../../../common/mongo';
|
||||
import { createOrGetCollectionTags } from './utils';
|
||||
import { rawText2Chunks } from '../read';
|
||||
import { checkDatasetLimit } from '../../../support/permission/teamLimit';
|
||||
import { predictDataLimitLength } from '../../../../global/core/dataset/utils';
|
||||
import { mongoSessionRun } from '../../../common/mongo/sessionRun';
|
||||
import { createTrainingUsage } from '../../../support/wallet/usage/controller';
|
||||
import { UsageSourceEnum } from '@fastgpt/global/support/wallet/usage/constants';
|
||||
import { getLLMModel, getVectorModel } from '../../ai/model';
|
||||
import { pushDataListToTrainingQueue } from '../training/controller';
|
||||
import { MongoImage } from '../../../common/file/image/schema';
|
||||
import { hashStr } from '@fastgpt/global/common/string/tools';
|
||||
|
||||
export const createCollectionAndInsertData = async ({
|
||||
dataset,
|
||||
rawText,
|
||||
relatedId,
|
||||
createCollectionParams,
|
||||
isQAImport = false,
|
||||
session
|
||||
}: {
|
||||
dataset: DatasetSchemaType;
|
||||
rawText: string;
|
||||
relatedId?: string;
|
||||
createCollectionParams: CreateOneCollectionParams;
|
||||
|
||||
isQAImport?: boolean;
|
||||
session?: ClientSession;
|
||||
}) => {
|
||||
const teamId = createCollectionParams.teamId;
|
||||
const tmbId = createCollectionParams.tmbId;
|
||||
// Chunk split params
|
||||
const trainingType = createCollectionParams.trainingType || TrainingModeEnum.chunk;
|
||||
const chunkSize = createCollectionParams.chunkSize;
|
||||
const chunkSplitter = createCollectionParams.chunkSplitter;
|
||||
const qaPrompt = createCollectionParams.qaPrompt;
|
||||
const usageName = createCollectionParams.name;
|
||||
|
||||
// 1. split chunks
|
||||
const chunks = rawText2Chunks({
|
||||
rawText,
|
||||
chunkLen: chunkSize,
|
||||
overlapRatio: trainingType === TrainingModeEnum.chunk ? 0.2 : 0,
|
||||
customReg: chunkSplitter ? [chunkSplitter] : [],
|
||||
isQAImport
|
||||
});
|
||||
|
||||
// 2. auth limit
|
||||
await checkDatasetLimit({
|
||||
teamId,
|
||||
insertLen: predictDataLimitLength(trainingType, chunks)
|
||||
});
|
||||
|
||||
const fn = async (session: ClientSession) => {
|
||||
// 3. create collection
|
||||
const { _id: collectionId } = await createOneCollection({
|
||||
...createCollectionParams,
|
||||
|
||||
hashRawText: hashStr(rawText),
|
||||
rawTextLength: rawText.length,
|
||||
session
|
||||
});
|
||||
|
||||
// 4. create training bill
|
||||
const { billId } = await createTrainingUsage({
|
||||
teamId,
|
||||
tmbId,
|
||||
appName: usageName,
|
||||
billSource: UsageSourceEnum.training,
|
||||
vectorModel: getVectorModel(dataset.vectorModel)?.name,
|
||||
agentModel: getLLMModel(dataset.agentModel)?.name,
|
||||
session
|
||||
});
|
||||
|
||||
// 5. insert to training queue
|
||||
const insertResults = await pushDataListToTrainingQueue({
|
||||
teamId,
|
||||
tmbId,
|
||||
datasetId: dataset._id,
|
||||
collectionId,
|
||||
agentModel: dataset.agentModel,
|
||||
vectorModel: dataset.vectorModel,
|
||||
trainingMode: trainingType,
|
||||
prompt: qaPrompt,
|
||||
billId,
|
||||
data: chunks.map((item, index) => ({
|
||||
...item,
|
||||
chunkIndex: index
|
||||
})),
|
||||
session
|
||||
});
|
||||
|
||||
// 6. remove related image ttl
|
||||
if (relatedId) {
|
||||
await MongoImage.updateMany(
|
||||
{
|
||||
teamId,
|
||||
'metadata.relatedId': relatedId
|
||||
},
|
||||
{
|
||||
// Remove expiredTime to avoid ttl expiration
|
||||
$unset: {
|
||||
expiredTime: 1
|
||||
}
|
||||
},
|
||||
{
|
||||
session
|
||||
}
|
||||
);
|
||||
}
|
||||
|
||||
return {
|
||||
collectionId,
|
||||
insertResults
|
||||
};
|
||||
};
|
||||
|
||||
if (session) {
|
||||
return fn(session);
|
||||
}
|
||||
return mongoSessionRun(fn);
|
||||
};
|
||||
|
||||
export type CreateOneCollectionParams = CreateDatasetCollectionParams & {
|
||||
teamId: string;
|
||||
tmbId: string;
|
||||
session?: ClientSession;
|
||||
};
|
||||
export async function createOneCollection({
|
||||
teamId,
|
||||
tmbId,
|
||||
@@ -33,18 +159,15 @@ export async function createOneCollection({
|
||||
externalFileId,
|
||||
externalFileUrl,
|
||||
|
||||
apiFileId,
|
||||
|
||||
hashRawText,
|
||||
rawTextLength,
|
||||
metadata = {},
|
||||
session,
|
||||
tags,
|
||||
...props
|
||||
}: CreateDatasetCollectionParams & {
|
||||
teamId: string;
|
||||
tmbId: string;
|
||||
[key: string]: any;
|
||||
session?: ClientSession;
|
||||
}) {
|
||||
createTime
|
||||
}: CreateOneCollectionParams) {
|
||||
// Create collection tags
|
||||
const collectionTags = await createOrGetCollectionTags({ tags, teamId, datasetId, session });
|
||||
|
||||
@@ -52,7 +175,6 @@ export async function createOneCollection({
|
||||
const [collection] = await MongoDatasetCollection.create(
|
||||
[
|
||||
{
|
||||
...props,
|
||||
teamId,
|
||||
tmbId,
|
||||
parentId: parentId || null,
|
||||
@@ -64,16 +186,18 @@ export async function createOneCollection({
|
||||
chunkSize,
|
||||
chunkSplitter,
|
||||
qaPrompt,
|
||||
metadata,
|
||||
|
||||
fileId,
|
||||
rawLink,
|
||||
...(fileId ? { fileId } : {}),
|
||||
...(rawLink ? { rawLink } : {}),
|
||||
...(externalFileId ? { externalFileId } : {}),
|
||||
externalFileUrl,
|
||||
...(externalFileUrl ? { externalFileUrl } : {}),
|
||||
...(apiFileId ? { apiFileId } : {}),
|
||||
|
||||
rawTextLength,
|
||||
hashRawText,
|
||||
metadata,
|
||||
tags: collectionTags
|
||||
tags: collectionTags,
|
||||
createTime
|
||||
}
|
||||
],
|
||||
{ session }
|
||||
@@ -116,7 +240,68 @@ export const delCollectionRelatedSource = async ({
|
||||
/**
|
||||
* delete collection and it related data
|
||||
*/
|
||||
export async function delCollectionAndRelatedSources({
|
||||
export async function delCollection({
|
||||
collections,
|
||||
session,
|
||||
delRelatedSource
|
||||
}: {
|
||||
collections: (CollectionWithDatasetType | DatasetCollectionSchemaType)[];
|
||||
session: ClientSession;
|
||||
delRelatedSource: boolean;
|
||||
}) {
|
||||
if (collections.length === 0) return;
|
||||
|
||||
const teamId = collections[0].teamId;
|
||||
|
||||
if (!teamId) return Promise.reject('teamId is not exist');
|
||||
|
||||
const datasetIds = Array.from(
|
||||
new Set(
|
||||
collections.map((item) => {
|
||||
if (typeof item.datasetId === 'string') {
|
||||
return String(item.datasetId);
|
||||
}
|
||||
return String(item.datasetId._id);
|
||||
})
|
||||
)
|
||||
);
|
||||
const collectionIds = collections.map((item) => String(item._id));
|
||||
|
||||
// delete training data
|
||||
await MongoDatasetTraining.deleteMany({
|
||||
teamId,
|
||||
datasetIds: { $in: datasetIds },
|
||||
collectionId: { $in: collectionIds }
|
||||
});
|
||||
|
||||
/* file and imgs */
|
||||
if (delRelatedSource) {
|
||||
await delCollectionRelatedSource({ collections, session });
|
||||
}
|
||||
|
||||
// delete dataset.datas
|
||||
await MongoDatasetData.deleteMany(
|
||||
{ teamId, datasetIds: { $in: datasetIds }, collectionId: { $in: collectionIds } },
|
||||
{ session }
|
||||
);
|
||||
|
||||
// delete collections
|
||||
await MongoDatasetCollection.deleteMany(
|
||||
{
|
||||
teamId,
|
||||
_id: { $in: collectionIds }
|
||||
},
|
||||
{ session }
|
||||
);
|
||||
|
||||
// no session delete: delete files, vector data
|
||||
await deleteDatasetDataVector({ teamId, datasetIds, collectionIds });
|
||||
}
|
||||
|
||||
/**
|
||||
* delete delOnlyCollection
|
||||
*/
|
||||
export async function delOnlyCollection({
|
||||
collections,
|
||||
session
|
||||
}: {
|
||||
@@ -148,9 +333,6 @@ export async function delCollectionAndRelatedSources({
|
||||
collectionId: { $in: collectionIds }
|
||||
});
|
||||
|
||||
/* file and imgs */
|
||||
await delCollectionRelatedSource({ collections, session });
|
||||
|
||||
// delete dataset.datas
|
||||
await MongoDatasetData.deleteMany(
|
||||
{ teamId, datasetIds: { $in: datasetIds }, collectionId: { $in: collectionIds } },
|
||||
|
||||
@@ -10,90 +10,100 @@ import {
|
||||
|
||||
export const DatasetColCollectionName = 'dataset_collections';
|
||||
|
||||
const DatasetCollectionSchema = new Schema({
|
||||
parentId: {
|
||||
type: Schema.Types.ObjectId,
|
||||
ref: DatasetColCollectionName,
|
||||
default: null
|
||||
},
|
||||
teamId: {
|
||||
type: Schema.Types.ObjectId,
|
||||
ref: TeamCollectionName,
|
||||
required: true
|
||||
},
|
||||
tmbId: {
|
||||
type: Schema.Types.ObjectId,
|
||||
ref: TeamMemberCollectionName,
|
||||
required: true
|
||||
},
|
||||
datasetId: {
|
||||
type: Schema.Types.ObjectId,
|
||||
ref: DatasetCollectionName,
|
||||
required: true
|
||||
},
|
||||
type: {
|
||||
type: String,
|
||||
enum: Object.keys(DatasetCollectionTypeMap),
|
||||
required: true
|
||||
},
|
||||
name: {
|
||||
type: String,
|
||||
required: true
|
||||
},
|
||||
createTime: {
|
||||
type: Date,
|
||||
default: () => new Date()
|
||||
},
|
||||
updateTime: {
|
||||
type: Date,
|
||||
default: () => new Date()
|
||||
},
|
||||
forbid: {
|
||||
type: Boolean,
|
||||
default: false
|
||||
},
|
||||
const DatasetCollectionSchema = new Schema(
|
||||
{
|
||||
parentId: {
|
||||
type: Schema.Types.ObjectId,
|
||||
ref: DatasetColCollectionName,
|
||||
default: null
|
||||
},
|
||||
teamId: {
|
||||
type: Schema.Types.ObjectId,
|
||||
ref: TeamCollectionName,
|
||||
required: true
|
||||
},
|
||||
tmbId: {
|
||||
type: Schema.Types.ObjectId,
|
||||
ref: TeamMemberCollectionName,
|
||||
required: true
|
||||
},
|
||||
datasetId: {
|
||||
type: Schema.Types.ObjectId,
|
||||
ref: DatasetCollectionName,
|
||||
required: true
|
||||
},
|
||||
type: {
|
||||
type: String,
|
||||
enum: Object.keys(DatasetCollectionTypeMap),
|
||||
required: true
|
||||
},
|
||||
name: {
|
||||
type: String,
|
||||
required: true
|
||||
},
|
||||
createTime: {
|
||||
type: Date,
|
||||
default: () => new Date()
|
||||
},
|
||||
updateTime: {
|
||||
type: Date,
|
||||
default: () => new Date()
|
||||
},
|
||||
forbid: {
|
||||
type: Boolean,
|
||||
default: false
|
||||
},
|
||||
|
||||
// chunk filed
|
||||
trainingType: {
|
||||
type: String,
|
||||
enum: Object.keys(TrainingTypeMap)
|
||||
},
|
||||
chunkSize: {
|
||||
type: Number,
|
||||
required: true
|
||||
},
|
||||
chunkSplitter: {
|
||||
type: String
|
||||
},
|
||||
qaPrompt: {
|
||||
type: String
|
||||
},
|
||||
ocrParse: Boolean,
|
||||
// chunk filed
|
||||
trainingType: {
|
||||
type: String,
|
||||
enum: Object.keys(TrainingTypeMap)
|
||||
},
|
||||
chunkSize: {
|
||||
type: Number,
|
||||
required: true
|
||||
},
|
||||
chunkSplitter: {
|
||||
type: String
|
||||
},
|
||||
qaPrompt: {
|
||||
type: String
|
||||
},
|
||||
ocrParse: Boolean,
|
||||
|
||||
tags: {
|
||||
type: [String],
|
||||
default: []
|
||||
},
|
||||
tags: {
|
||||
type: [String],
|
||||
default: []
|
||||
},
|
||||
|
||||
// local file collection
|
||||
fileId: {
|
||||
type: Schema.Types.ObjectId,
|
||||
ref: 'dataset.files'
|
||||
},
|
||||
// web link collection
|
||||
rawLink: String,
|
||||
// external collection
|
||||
externalFileId: String,
|
||||
// local file collection
|
||||
fileId: {
|
||||
type: Schema.Types.ObjectId,
|
||||
ref: 'dataset.files'
|
||||
},
|
||||
// web link collection
|
||||
rawLink: String,
|
||||
// api collection
|
||||
apiFileId: String,
|
||||
// external collection
|
||||
externalFileId: String,
|
||||
externalFileUrl: String, // external import url
|
||||
|
||||
// metadata
|
||||
rawTextLength: Number,
|
||||
hashRawText: String,
|
||||
externalFileUrl: String, // external import url
|
||||
metadata: {
|
||||
type: Object,
|
||||
default: {}
|
||||
// metadata
|
||||
rawTextLength: Number,
|
||||
hashRawText: String,
|
||||
metadata: {
|
||||
type: Object,
|
||||
default: {}
|
||||
}
|
||||
},
|
||||
{
|
||||
// Auto update updateTime
|
||||
timestamps: {
|
||||
updatedAt: 'updateTime'
|
||||
}
|
||||
}
|
||||
});
|
||||
);
|
||||
|
||||
try {
|
||||
// auth file
|
||||
|
||||
@@ -1,17 +1,19 @@
|
||||
import type { CollectionWithDatasetType } from '@fastgpt/global/core/dataset/type.d';
|
||||
import { MongoDatasetCollection } from './schema';
|
||||
import { splitText2Chunks } from '@fastgpt/global/common/string/textSplitter';
|
||||
import { MongoDatasetTraining } from '../training/schema';
|
||||
import { urlsFetch } from '../../../common/string/cheerio';
|
||||
import {
|
||||
DatasetCollectionTypeEnum,
|
||||
TrainingModeEnum
|
||||
} from '@fastgpt/global/core/dataset/constants';
|
||||
import { hashStr } from '@fastgpt/global/common/string/tools';
|
||||
import { ClientSession } from '../../../common/mongo';
|
||||
import { PushDatasetDataResponse } from '@fastgpt/global/core/dataset/api';
|
||||
import { MongoDatasetCollectionTags } from '../tag/schema';
|
||||
import { readFromSecondary } from '../../../common/mongo/utils';
|
||||
import { CollectionWithDatasetType } from '@fastgpt/global/core/dataset/type';
|
||||
import {
|
||||
DatasetCollectionSyncResultEnum,
|
||||
DatasetCollectionTypeEnum,
|
||||
DatasetSourceReadTypeEnum,
|
||||
DatasetTypeEnum
|
||||
} from '@fastgpt/global/core/dataset/constants';
|
||||
import { DatasetErrEnum } from '@fastgpt/global/common/error/code/dataset';
|
||||
import { readDatasetSourceRawText } from '../read';
|
||||
import { hashStr } from '@fastgpt/global/common/string/tools';
|
||||
import { mongoSessionRun } from '../../../common/mongo/sessionRun';
|
||||
import { createCollectionAndInsertData, delCollection } from './controller';
|
||||
|
||||
/**
|
||||
* get all collection by top collectionId
|
||||
@@ -61,148 +63,6 @@ export function getCollectionUpdateTime({ name, time }: { time?: Date; name: str
|
||||
return new Date();
|
||||
}
|
||||
|
||||
/**
|
||||
* Get collection raw text by Collection or collectionId
|
||||
*/
|
||||
export const getCollectionAndRawText = async ({
|
||||
collectionId,
|
||||
collection,
|
||||
newRawText
|
||||
}: {
|
||||
collectionId?: string;
|
||||
collection?: CollectionWithDatasetType;
|
||||
newRawText?: string;
|
||||
}) => {
|
||||
const col = await (async () => {
|
||||
if (collection) return collection;
|
||||
if (collectionId) {
|
||||
return (await MongoDatasetCollection.findById(collectionId).populate(
|
||||
'datasetId'
|
||||
)) as CollectionWithDatasetType;
|
||||
}
|
||||
|
||||
return null;
|
||||
})();
|
||||
|
||||
if (!col) {
|
||||
return Promise.reject('Collection not found');
|
||||
}
|
||||
|
||||
const { title, rawText } = await (async () => {
|
||||
if (newRawText)
|
||||
return {
|
||||
title: '',
|
||||
rawText: newRawText
|
||||
};
|
||||
// link
|
||||
if (col.type === DatasetCollectionTypeEnum.link && col.rawLink) {
|
||||
// crawl new data
|
||||
const result = await urlsFetch({
|
||||
urlList: [col.rawLink],
|
||||
selector: col.datasetId?.websiteConfig?.selector || col?.metadata?.webPageSelector
|
||||
});
|
||||
|
||||
return {
|
||||
title: result[0]?.title,
|
||||
rawText: result[0]?.content
|
||||
};
|
||||
}
|
||||
|
||||
// file
|
||||
|
||||
return {
|
||||
title: '',
|
||||
rawText: ''
|
||||
};
|
||||
})();
|
||||
|
||||
const hashRawText = hashStr(rawText);
|
||||
const isSameRawText = rawText && col.hashRawText === hashRawText;
|
||||
|
||||
return {
|
||||
collection: col,
|
||||
title,
|
||||
rawText,
|
||||
isSameRawText
|
||||
};
|
||||
};
|
||||
|
||||
/* link collection start load data */
|
||||
export const reloadCollectionChunks = async ({
|
||||
collection,
|
||||
tmbId,
|
||||
billId,
|
||||
rawText,
|
||||
session
|
||||
}: {
|
||||
collection: CollectionWithDatasetType;
|
||||
tmbId: string;
|
||||
billId?: string;
|
||||
rawText?: string;
|
||||
session: ClientSession;
|
||||
}): Promise<PushDatasetDataResponse> => {
|
||||
const {
|
||||
title,
|
||||
rawText: newRawText,
|
||||
collection: col,
|
||||
isSameRawText
|
||||
} = await getCollectionAndRawText({
|
||||
collection,
|
||||
newRawText: rawText
|
||||
});
|
||||
|
||||
if (isSameRawText)
|
||||
return {
|
||||
insertLen: 0
|
||||
};
|
||||
|
||||
// split data
|
||||
const { chunks } = splitText2Chunks({
|
||||
text: newRawText,
|
||||
chunkLen: col.chunkSize || 512,
|
||||
customReg: col.chunkSplitter ? [col.chunkSplitter] : []
|
||||
});
|
||||
|
||||
// insert to training queue
|
||||
const model = await (() => {
|
||||
if (col.trainingType === TrainingModeEnum.chunk) return col.datasetId.vectorModel;
|
||||
if (col.trainingType === TrainingModeEnum.qa) return col.datasetId.agentModel;
|
||||
return Promise.reject('Training model error');
|
||||
})();
|
||||
|
||||
const result = await MongoDatasetTraining.insertMany(
|
||||
chunks.map((item, i) => ({
|
||||
teamId: col.teamId,
|
||||
tmbId,
|
||||
datasetId: col.datasetId._id,
|
||||
collectionId: col._id,
|
||||
billId,
|
||||
mode: col.trainingType,
|
||||
prompt: '',
|
||||
model,
|
||||
q: item,
|
||||
a: '',
|
||||
chunkIndex: i
|
||||
})),
|
||||
{ session }
|
||||
);
|
||||
|
||||
// update raw text
|
||||
await MongoDatasetCollection.findByIdAndUpdate(
|
||||
col._id,
|
||||
{
|
||||
...(title && { name: title }),
|
||||
rawTextLength: newRawText.length,
|
||||
hashRawText: hashStr(newRawText)
|
||||
},
|
||||
{ session }
|
||||
);
|
||||
|
||||
return {
|
||||
insertLen: result.length
|
||||
};
|
||||
};
|
||||
|
||||
export const createOrGetCollectionTags = async ({
|
||||
tags,
|
||||
datasetId,
|
||||
@@ -268,3 +128,88 @@ export const collectionTagsToTagLabel = async ({
|
||||
})
|
||||
.filter(Boolean);
|
||||
};
|
||||
|
||||
export const syncCollection = async (collection: CollectionWithDatasetType) => {
|
||||
const dataset = collection.datasetId;
|
||||
|
||||
if (
|
||||
collection.type !== DatasetCollectionTypeEnum.link &&
|
||||
dataset.type !== DatasetTypeEnum.apiDataset
|
||||
) {
|
||||
return Promise.reject(DatasetErrEnum.notSupportSync);
|
||||
}
|
||||
|
||||
// Get new text
|
||||
const sourceReadType = await (async () => {
|
||||
if (collection.type === DatasetCollectionTypeEnum.link) {
|
||||
if (!collection.rawLink) return Promise.reject('rawLink is missing');
|
||||
return {
|
||||
type: DatasetSourceReadTypeEnum.link,
|
||||
sourceId: collection.rawLink,
|
||||
selector: collection.metadata?.webPageSelector
|
||||
};
|
||||
}
|
||||
|
||||
if (!collection.apiFileId) return Promise.reject('apiFileId is missing');
|
||||
if (!dataset.apiServer) return Promise.reject('apiServer not found');
|
||||
return {
|
||||
type: DatasetSourceReadTypeEnum.apiFile,
|
||||
sourceId: collection.apiFileId,
|
||||
apiServer: dataset.apiServer
|
||||
};
|
||||
})();
|
||||
const rawText = await readDatasetSourceRawText({
|
||||
teamId: collection.teamId,
|
||||
...sourceReadType
|
||||
});
|
||||
|
||||
// Check if the original text is the same: skip if same
|
||||
const hashRawText = hashStr(rawText);
|
||||
if (collection.hashRawText && hashRawText === collection.hashRawText) {
|
||||
return DatasetCollectionSyncResultEnum.sameRaw;
|
||||
}
|
||||
|
||||
await mongoSessionRun(async (session) => {
|
||||
// Create new collection
|
||||
await createCollectionAndInsertData({
|
||||
session,
|
||||
dataset,
|
||||
rawText: rawText,
|
||||
createCollectionParams: {
|
||||
teamId: collection.teamId,
|
||||
tmbId: collection.tmbId,
|
||||
datasetId: collection.datasetId._id,
|
||||
name: collection.name,
|
||||
type: collection.type,
|
||||
|
||||
fileId: collection.fileId,
|
||||
rawLink: collection.rawLink,
|
||||
externalFileId: collection.externalFileId,
|
||||
externalFileUrl: collection.externalFileUrl,
|
||||
apiFileId: collection.apiFileId,
|
||||
|
||||
rawTextLength: rawText.length,
|
||||
hashRawText,
|
||||
|
||||
tags: collection.tags,
|
||||
createTime: collection.createTime,
|
||||
|
||||
parentId: collection.parentId,
|
||||
trainingType: collection.trainingType,
|
||||
chunkSize: collection.chunkSize,
|
||||
chunkSplitter: collection.chunkSplitter,
|
||||
qaPrompt: collection.qaPrompt,
|
||||
metadata: collection.metadata
|
||||
}
|
||||
});
|
||||
|
||||
// Delete old collection
|
||||
await delCollection({
|
||||
collections: [collection],
|
||||
delRelatedSource: false,
|
||||
session
|
||||
});
|
||||
});
|
||||
|
||||
return DatasetCollectionSyncResultEnum.success;
|
||||
};
|
||||
|
||||
Reference in New Issue
Block a user