README.md

lanterndata/mikro-orm

---

Setting up LanternDB Extension

If you've already executed this through a raw query, then skip this step.

import { MikroORM } from '@mikro-orm/postgresql';
import { extend } from 'lanterndata/mikro-orm';

const orm = await MikroORM.init({});
const em = orm.em.fork();

extend(em);

await em.createLanternExtension();

// For the LanternExtras, you need to run this instead
await em.createLanternExtrasExtension();

Create the Table and add an Index

import { toSql } from 'lanterndata/mikro-orm';

const Book = new EntitySchema({
  name: 'Book',
  tableName: 'books',
  properties: {
    id: { type: 'number', primary: true },
    name: { type: 'string', nullable: true },
    url: { type: 'string', nullable: true },
    embedding: { type: 'Array<number>', nullable: true },
  },
});

await em.execute('CREATE INDEX book_index ON books USING lantern_hnsw(embedding dist_l2sq_ops');

const books = booksToInsert.map((book) =>
  em.create(Book, {
    ...book,
    // use toSql method to conver [1,2,3] inro '{1,2,3}'
    embedding: toSql(book.embedding),
  }),
);

await em.persistAndFlush(books);

Vector search methods

You can performe vectore search using those distance methods.

await em
  .qb(Book)
  .orderBy({ [em.l2Distance('embedding', [1, 1, 1])]: 'ASC' })
  .limit(5)
  .getResult();

await em
  .qb(Book)
  .orderBy({ [em.cosineDistance('embedding', [1, 1, 1])]: 'ASC' })
  .limit(5)
  .getResult();

await em
  .qb(Book)
  .orderBy({ [em.hammingDistance('embedding', [1, 1, 1])]: 'ASC' })
  .limit(5)
  .getResult();

Embedding generation

import { MikroORM } from '@mikro-orm/postgresql';
import { extend } from 'lanterndata/mikro-orm';

import { TextEmbeddingModels, ImageEmbeddingModels } from 'lanterndata/embeddings';

const orm = await MikroORM.init({});
const em = orm.em.fork();

extend(em);

const { BAAI_BGE_BASE_EN } = TextEmbeddingModels;
const { CLIP_VIT_B_32_VISUAL } = ImageEmbeddingModels;

// text embedding
const text = 'hello world';
const result = await em.generateTextEmbedding(BAAI_BGE_BASE_EN, text);
console.log(result[0].text_embedding);

// image embedding
const imageUrl = 'https://lantern.dev/images/home/footer.png';
const result = await em.generateImageEmbedding(CLIP_VIT_B_32_VISUAL, imageUrl);
console.log(result[0].image_embedding);

Vector searche with embedding generation

import { MikroORM } from '@mikro-orm/postgresql';
import { extend } from 'lanterndata/mikro-orm';

import { TextEmbeddingModels, ImageEmbeddingModels } from 'lanterndata/embeddings';

const orm = await MikroORM.init({});
const em = orm.em.fork();

extend(em);

const { BAAI_BGE_BASE_EN } = TextEmbeddingModels;
const { CLIP_VIT_B_32_VISUAL } = ImageEmbeddingModels;

const text = 'hello world';
const imageUrl = 'https://lantern.dev/images/home/footer.png';

// distance search with text embedding generation
await em
  .qb(Book)
  .select('*')
  .orderBy({ [em.cosineDistance('embedding', em.textEmbedding(BAAI_BGE_BASE_EN, text))]: 'DESC' })
  .limit(2)
  .execute('all');

// distance search with image embedding generation
await em
  .qb(Book)
  .select('*')
  .orderBy({ [em.l2Distance('embedding', em.imageEmbedding(CLIP_VIT_B_32_VISUAL, imageUrl))]: 'DESC' })
  .limit(2)
  .execute('all');

Corresponding SQL code (example):

SELECT * FROM "books"
ORDER BY "embedding" <-> image_embedding('clip/ViT-B-32-visual', "...") DESC
LIMIT 2;

Other methods

• openaiEmbedding(OpenAITextEmbeddingModelType, text, [dimension])
• cohereEmbedding(CohereTextEmbeddingModelType, text)

import { OpenAITextEmbeddingModelType, CohereTextEmbeddingModelType } from 'lanterndata/embeddings';

em.openaiEmbedding(OpenAITextEmbeddingModelType.ADA_002, 'hello world', 256);
em.cohereEmbedding(CohereTextEmbeddingModelType.ENGLISH_V3_0, 'hello world');