Get Started

Lantern Cloud

The easiest way to get started with all of our tools is with Lantern Cloud.

In Lantern Cloud, you can create create a database with just a few clicks. Once you load your data into Lantern, you can generate embeddings with a single click from dozens of provided open source or proprietary embedding models. You can then create a vector index from your dashboard or run an index-tuning experiment to choose the best parameters for index creation.

Ubicloud

Lantern is also available as a managed offering on Ubicloud. You'll still receive direct support from the Lantern team for instances on Ubicloud.

Ubicloud is an open-source alternative to AWS. It offers cloud services like virtual machines, block storage, and managed Postgres at rates 2-10x lower than hyperscalers like AWS, Azure, and GCP.

Self-Host

Alternatively, you can also use our tools locally or self-host them. There are three tools that are provided out-of-the-box with Lantern Cloud.

• Lantern HNSW, our core Postgres extension, provides vector search in Postgres.
• Lantern Extras, which further extends Postgres to support embedding generation.
• Lantern CLI provides routines for generating embeddings and indexes.

You can install the tools individually by following the instructions linked.

Overview

Here is a non-comprehensive overview of what you can do with Lantern. The examples below use SQL, but we also provide examples for Python, JavaScript, Ruby, and Rust.

Create a table with an embedding column

CREATE TABLE books (id SERIAL PRIMARY KEY, book_embedding REAL[3]);

Generate embeddings

SELECT text_embedding('BAAI/bge-base-en', 'My text input');
SELECT openai_embedding('openai/text-embedding-ada-002', 'My text input');

Insert embeddings

INSERT INTO books (book_embedding) VALUES ('{0,1,0}'), ('{3,2,4}');

Calculate distance and select nearest rows

SELECT book_embedding <-> '{0,0,0}' FROM books
    ORDER BY book_embedding <-> '{0,0,0}' LIMIT 1;

Create an index to speed up nearest neighbor search

CREATE INDEX book_index ON books USING lantern_hnsw(book_embedding dist_l2sq_ops)
    WITH (M=2, ef_construction=10, ef=4, dim=3);