At Ubicloud, we are building an open source alternative to AWS. You can use our managed service to reduce your cloud spend by 2x - 10x, and regain control of your cloud infrastructure with open source. 

Today, we're excited to announce our partnership with Lantern, making it easy to build high-performance AI applications with the simplicity of Ubicloud PostgreSQL.

The Problem

Vector databases have become essential for AI applications, while the two common ways to run them each have their limitations:

• Using pgvector on PostgreSQL, commonly available on AWS RDS, Azure and many managed PostgreSQL services: This makes for a good starting point for many applications. However, it does not scale well to larger datasets. 
  
  Importantly, when using services like RDS Postgres, you get only generic PostgreSQL operational support, and not AI workload-specific expertise about PostgreSQL. Questions like what the advantages and limitations of creating HNSW indexes are, how to create and maintain your embeddings, or how to tune your Postgres parameters for AI have no first party support.
• Using a vector database, like Pinecone: Introducing a newly-built database management system to your database stack comes with its own complexity and added costs. You need to understand and instrument ways to keep data in sync between your primary database (often PostgreSQL) and the new vector database. Importantly, you get locked into a proprietary ecosystem, with high prices; and more moving parts.

The Solution

With Lantern on Ubicloud, we take a different approach:

• Lantern turns PostgreSQL into an AI database, and is packaged as a native PostgreSQL extension.
• Ubicloud PostgreSQL runs the Lantern extension in the cloud with automated backups, point-in-time-recovery, high-availability, and all the ease of use that comes with managed PostgreSQL.
• Your primary PostgreSQL database remains unchanged, whether it was already running on Ubicloud, or managed on a different cloud. A new replica of it running Lantern is created, and gets managed by Ubicloud.
  
• High-speed NVMe disks on Ubicloud maximize your search performance -- sometimes by 10x, or more.
  
• When you have a question about Lantern or pgvector, you get immediate and direct access to the Lantern team.
  

And of course, all of this is implemented with open source; at both Ubicloud and Lantern. 

Key features and performance

Lantern on Ubicloud PostgreSQL is available starting today in our Germany and US-East (Virginia) regions. The cost-effectiveness of bare metal instances we use allows us to offer competitive pricing, without sacrificing performance. Lantern on Ubicloud offers:

• Automatic embedding generation from unstructured data
• Automatic LLM completion from unstructured data
• Support for 20+ embedding models (OpenAI, Cohere, etc.)
  
• Serverless index construction for pgvector HNSW indexes
  
• Coming soon: BM25 text search
  

Costs less by combining Ubicloud's cost-effective PostgreSQL service with Lantern's efficient vector operations, Lantern on Ubicloud gives you top performance for the cost. For example, operating at a rate of 20 queries per second, you can save ~10x compared to specialized vector databases like Pinecone.

• Ubicloud PostgreSQL starts at just $97/month ($0.14/hour) for a standard-2 instance.

What workloads are best suited for Lantern on Ubicloud PostgreSQL?

Lantern on Ubicloud is applicable to a broad range of users and AI workloads, including:

• Developers building AI-powered search, recommendation systems, or similarity matching features who want the power of vector search with the familiarity of PostgreSQL
• Teams looking to reduce cloud costs for their vector database workloads without sacrificing performance or ease of use

Getting started