mirror of
https://github.com/labring/FastGPT.git
synced 2025-07-24 22:03:54 +00:00

* 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>
354 lines
8.9 KiB
TypeScript
354 lines
8.9 KiB
TypeScript
import { TrainingModeEnum } from '@fastgpt/global/core/dataset/constants';
|
|
import type { CreateDatasetCollectionParams } from '@fastgpt/global/core/dataset/api.d';
|
|
import { MongoDatasetCollection } from './schema';
|
|
import {
|
|
CollectionWithDatasetType,
|
|
DatasetCollectionSchemaType,
|
|
DatasetSchemaType
|
|
} from '@fastgpt/global/core/dataset/type';
|
|
import { MongoDatasetTraining } from '../training/schema';
|
|
import { MongoDatasetData } from '../data/schema';
|
|
import { delImgByRelatedId } from '../../../common/file/image/controller';
|
|
import { deleteDatasetDataVector } from '../../../common/vectorStore/controller';
|
|
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,
|
|
name,
|
|
parentId,
|
|
datasetId,
|
|
type,
|
|
|
|
trainingType = TrainingModeEnum.chunk,
|
|
chunkSize = 512,
|
|
chunkSplitter,
|
|
qaPrompt,
|
|
|
|
fileId,
|
|
rawLink,
|
|
|
|
externalFileId,
|
|
externalFileUrl,
|
|
|
|
apiFileId,
|
|
|
|
hashRawText,
|
|
rawTextLength,
|
|
metadata = {},
|
|
session,
|
|
tags,
|
|
createTime
|
|
}: CreateOneCollectionParams) {
|
|
// Create collection tags
|
|
const collectionTags = await createOrGetCollectionTags({ tags, teamId, datasetId, session });
|
|
|
|
// Create collection
|
|
const [collection] = await MongoDatasetCollection.create(
|
|
[
|
|
{
|
|
teamId,
|
|
tmbId,
|
|
parentId: parentId || null,
|
|
datasetId,
|
|
name,
|
|
type,
|
|
|
|
trainingType,
|
|
chunkSize,
|
|
chunkSplitter,
|
|
qaPrompt,
|
|
metadata,
|
|
|
|
...(fileId ? { fileId } : {}),
|
|
...(rawLink ? { rawLink } : {}),
|
|
...(externalFileId ? { externalFileId } : {}),
|
|
...(externalFileUrl ? { externalFileUrl } : {}),
|
|
...(apiFileId ? { apiFileId } : {}),
|
|
|
|
rawTextLength,
|
|
hashRawText,
|
|
tags: collectionTags,
|
|
createTime
|
|
}
|
|
],
|
|
{ session }
|
|
);
|
|
|
|
return collection;
|
|
}
|
|
|
|
/* delete collection related images/files */
|
|
export const delCollectionRelatedSource = async ({
|
|
collections,
|
|
session
|
|
}: {
|
|
collections: (CollectionWithDatasetType | DatasetCollectionSchemaType)[];
|
|
session: ClientSession;
|
|
}) => {
|
|
if (collections.length === 0) return;
|
|
|
|
const teamId = collections[0].teamId;
|
|
|
|
if (!teamId) return Promise.reject('teamId is not exist');
|
|
|
|
const fileIdList = collections.map((item) => item?.fileId || '').filter(Boolean);
|
|
const relatedImageIds = collections
|
|
.map((item) => item?.metadata?.relatedImgId || '')
|
|
.filter(Boolean);
|
|
|
|
// delete files
|
|
await delFileByFileIdList({
|
|
bucketName: BucketNameEnum.dataset,
|
|
fileIdList
|
|
});
|
|
// delete images
|
|
await delImgByRelatedId({
|
|
teamId,
|
|
relateIds: relatedImageIds,
|
|
session
|
|
});
|
|
};
|
|
/**
|
|
* delete collection and it related data
|
|
*/
|
|
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
|
|
}: {
|
|
collections: (CollectionWithDatasetType | DatasetCollectionSchemaType)[];
|
|
session: ClientSession;
|
|
}) {
|
|
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 }
|
|
});
|
|
|
|
// 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 });
|
|
}
|