feat(server): lru cache for query embeddings (#19181)

This commit is contained in:
Mert 2025-06-16 11:03:49 -04:00 committed by GitHub
parent 176d53c1b3
commit 7347f64958
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 85 additions and 19 deletions

View File

@ -48,6 +48,7 @@
"kysely-postgres-js": "^2.0.0",
"lodash": "^4.17.21",
"luxon": "^3.4.2",
"mnemonist": "^0.40.3",
"nest-commander": "^3.16.0",
"nestjs-cls": "^5.0.0",
"nestjs-kysely": "^1.1.0",
@ -473,6 +474,13 @@
"lru-cache": "^10.4.3"
}
},
"node_modules/@asamuzakjp/css-color/node_modules/lru-cache": {
"version": "10.4.3",
"resolved": "https://registry.npmjs.org/lru-cache/-/lru-cache-10.4.3.tgz",
"integrity": "sha512-JNAzZcXrCt42VGLuYz0zfAzDfAvJWW6AfYlDBQyDV5DClI2m5sAmK+OIO7s59XfsRsWHp02jAJrRadPRGTt6SQ==",
"dev": true,
"license": "ISC"
},
"node_modules/@babel/code-frame": {
"version": "7.26.2",
"resolved": "https://registry.npmjs.org/@babel/code-frame/-/code-frame-7.26.2.tgz",
@ -3174,6 +3182,13 @@
"node": ">= 14"
}
},
"node_modules/@npmcli/agent/node_modules/lru-cache": {
"version": "10.4.3",
"resolved": "https://registry.npmjs.org/lru-cache/-/lru-cache-10.4.3.tgz",
"integrity": "sha512-JNAzZcXrCt42VGLuYz0zfAzDfAvJWW6AfYlDBQyDV5DClI2m5sAmK+OIO7s59XfsRsWHp02jAJrRadPRGTt6SQ==",
"dev": true,
"license": "ISC"
},
"node_modules/@npmcli/fs": {
"version": "4.0.0",
"resolved": "https://registry.npmjs.org/@npmcli/fs/-/fs-4.0.0.tgz",
@ -6859,6 +6874,12 @@
"@pkgjs/parseargs": "^0.11.0"
}
},
"node_modules/archiver-utils/node_modules/lru-cache": {
"version": "10.4.3",
"resolved": "https://registry.npmjs.org/lru-cache/-/lru-cache-10.4.3.tgz",
"integrity": "sha512-JNAzZcXrCt42VGLuYz0zfAzDfAvJWW6AfYlDBQyDV5DClI2m5sAmK+OIO7s59XfsRsWHp02jAJrRadPRGTt6SQ==",
"license": "ISC"
},
"node_modules/archiver-utils/node_modules/minimatch": {
"version": "9.0.5",
"resolved": "https://registry.npmjs.org/minimatch/-/minimatch-9.0.5.tgz",
@ -7484,6 +7505,13 @@
"@pkgjs/parseargs": "^0.11.0"
}
},
"node_modules/cacache/node_modules/lru-cache": {
"version": "10.4.3",
"resolved": "https://registry.npmjs.org/lru-cache/-/lru-cache-10.4.3.tgz",
"integrity": "sha512-JNAzZcXrCt42VGLuYz0zfAzDfAvJWW6AfYlDBQyDV5DClI2m5sAmK+OIO7s59XfsRsWHp02jAJrRadPRGTt6SQ==",
"dev": true,
"license": "ISC"
},
"node_modules/cacache/node_modules/minimatch": {
"version": "9.0.5",
"resolved": "https://registry.npmjs.org/minimatch/-/minimatch-9.0.5.tgz",
@ -11837,10 +11865,13 @@
"license": "MIT"
},
"node_modules/lru-cache": {
"version": "10.4.3",
"resolved": "https://registry.npmjs.org/lru-cache/-/lru-cache-10.4.3.tgz",
"integrity": "sha512-JNAzZcXrCt42VGLuYz0zfAzDfAvJWW6AfYlDBQyDV5DClI2m5sAmK+OIO7s59XfsRsWHp02jAJrRadPRGTt6SQ==",
"license": "ISC"
"version": "11.1.0",
"resolved": "https://registry.npmjs.org/lru-cache/-/lru-cache-11.1.0.tgz",
"integrity": "sha512-QIXZUBJUx+2zHUdQujWejBkcD9+cs94tLn0+YL8UrCh+D5sCXZ4c7LaEH48pNwRY3MLDgqUFyhlCyjJPf1WP0A==",
"license": "ISC",
"engines": {
"node": "20 || >=22"
}
},
"node_modules/luxon": {
"version": "3.6.1",
@ -12304,6 +12335,15 @@
"dev": true,
"license": "MIT"
},
"node_modules/mnemonist": {
"version": "0.40.3",
"resolved": "https://registry.npmjs.org/mnemonist/-/mnemonist-0.40.3.tgz",
"integrity": "sha512-Vjyr90sJ23CKKH/qPAgUKicw/v6pRoamxIEDFOF8uSgFME7DqPRpHgRTejWVjkdGg5dXj0/NyxZHZ9bcjH+2uQ==",
"license": "MIT",
"dependencies": {
"obliterator": "^2.0.4"
}
},
"node_modules/mock-fs": {
"version": "5.5.0",
"resolved": "https://registry.npmjs.org/mock-fs/-/mock-fs-5.5.0.tgz",
@ -13061,6 +13101,12 @@
"node": ">= 0.4"
}
},
"node_modules/obliterator": {
"version": "2.0.5",
"resolved": "https://registry.npmjs.org/obliterator/-/obliterator-2.0.5.tgz",
"integrity": "sha512-42CPE9AhahZRsMNslczq0ctAEtqk8Eka26QofnqC346BZdHDySk3LWka23LI7ULIw11NmltpiLagIq8gBozxTw==",
"license": "MIT"
},
"node_modules/on-finished": {
"version": "2.4.1",
"resolved": "https://registry.npmjs.org/on-finished/-/on-finished-2.4.1.tgz",
@ -13492,15 +13538,6 @@
"url": "https://github.com/sponsors/isaacs"
}
},
"node_modules/path-scurry/node_modules/lru-cache": {
"version": "11.1.0",
"resolved": "https://registry.npmjs.org/lru-cache/-/lru-cache-11.1.0.tgz",
"integrity": "sha512-QIXZUBJUx+2zHUdQujWejBkcD9+cs94tLn0+YL8UrCh+D5sCXZ4c7LaEH48pNwRY3MLDgqUFyhlCyjJPf1WP0A==",
"license": "ISC",
"engines": {
"node": "20 || >=22"
}
},
"node_modules/path-source": {
"version": "0.1.3",
"resolved": "https://registry.npmjs.org/path-source/-/path-source-0.1.3.tgz",
@ -16043,6 +16080,13 @@
"@pkgjs/parseargs": "^0.11.0"
}
},
"node_modules/sucrase/node_modules/lru-cache": {
"version": "10.4.3",
"resolved": "https://registry.npmjs.org/lru-cache/-/lru-cache-10.4.3.tgz",
"integrity": "sha512-JNAzZcXrCt42VGLuYz0zfAzDfAvJWW6AfYlDBQyDV5DClI2m5sAmK+OIO7s59XfsRsWHp02jAJrRadPRGTt6SQ==",
"license": "ISC",
"peer": true
},
"node_modules/sucrase/node_modules/minimatch": {
"version": "9.0.5",
"resolved": "https://registry.npmjs.org/minimatch/-/minimatch-9.0.5.tgz",
@ -16596,6 +16640,13 @@
"@pkgjs/parseargs": "^0.11.0"
}
},
"node_modules/test-exclude/node_modules/lru-cache": {
"version": "10.4.3",
"resolved": "https://registry.npmjs.org/lru-cache/-/lru-cache-10.4.3.tgz",
"integrity": "sha512-JNAzZcXrCt42VGLuYz0zfAzDfAvJWW6AfYlDBQyDV5DClI2m5sAmK+OIO7s59XfsRsWHp02jAJrRadPRGTt6SQ==",
"dev": true,
"license": "ISC"
},
"node_modules/test-exclude/node_modules/minimatch": {
"version": "9.0.5",
"resolved": "https://registry.npmjs.org/minimatch/-/minimatch-9.0.5.tgz",
@ -17196,6 +17247,12 @@
"@pkgjs/parseargs": "^0.11.0"
}
},
"node_modules/typeorm/node_modules/lru-cache": {
"version": "10.4.3",
"resolved": "https://registry.npmjs.org/lru-cache/-/lru-cache-10.4.3.tgz",
"integrity": "sha512-JNAzZcXrCt42VGLuYz0zfAzDfAvJWW6AfYlDBQyDV5DClI2m5sAmK+OIO7s59XfsRsWHp02jAJrRadPRGTt6SQ==",
"license": "ISC"
},
"node_modules/typeorm/node_modules/minimatch": {
"version": "9.0.5",
"resolved": "https://registry.npmjs.org/minimatch/-/minimatch-9.0.5.tgz",

View File

@ -74,6 +74,7 @@
"kysely-postgres-js": "^2.0.0",
"lodash": "^4.17.21",
"luxon": "^3.4.2",
"mnemonist": "^0.40.3",
"nest-commander": "^3.16.0",
"nestjs-cls": "^5.0.0",
"nestjs-kysely": "^1.1.0",

View File

@ -1,4 +1,5 @@
import { BadRequestException, Injectable } from '@nestjs/common';
import { LRUMap } from 'mnemonist';
import { AssetMapOptions, AssetResponseDto, MapAsset, mapAsset } from 'src/dtos/asset-response.dto';
import { AuthDto } from 'src/dtos/auth.dto';
import { mapPerson, PersonResponseDto } from 'src/dtos/person.dto';
@ -24,6 +25,8 @@ import { isSmartSearchEnabled } from 'src/utils/misc';
@Injectable()
export class SearchService extends BaseService {
private embeddingCache = new LRUMap<string, string>(100);
async searchPerson(auth: AuthDto, dto: SearchPeopleDto): Promise<PersonResponseDto[]> {
const people = await this.personRepository.getByName(auth.user.id, dto.name, { withHidden: dto.withHidden });
return people.map((person) => mapPerson(person));
@ -98,16 +101,21 @@ export class SearchService extends BaseService {
throw new BadRequestException('Smart search is not enabled');
}
const userIds = await this.getUserIdsToSearch(auth);
const embedding = await this.machineLearningRepository.encodeText(machineLearning.urls, dto.query, {
modelName: machineLearning.clip.modelName,
language: dto.language,
});
const userIds = this.getUserIdsToSearch(auth);
const key = machineLearning.clip.modelName + dto.query + dto.language;
let embedding = this.embeddingCache.get(key);
if (!embedding) {
embedding = await this.machineLearningRepository.encodeText(machineLearning.urls, dto.query, {
modelName: machineLearning.clip.modelName,
language: dto.language,
});
this.embeddingCache.set(key, embedding);
}
const page = dto.page ?? 1;
const size = dto.size || 100;
const { hasNextPage, items } = await this.searchRepository.searchSmart(
{ page, size },
{ ...dto, userIds, embedding },
{ ...dto, userIds: await userIds, embedding },
);
return this.mapResponse(items, hasNextPage ? (page + 1).toString() : null, { auth });