Lantern

Lantern is an open-source PostgreSQL extension and toolkit that provides efficient vector search capabilities and enables developers to build production-ready AI applications directly within their PostgreSQL databases. It allows users to perform similarity searches, advanced text searches, and integrate AI workflows without requiring separate vector databases or search engines.

Features

• Vector Search with pgvector: Supports searching over sparse and dense vectors, including binary, scalar, and product compression. It scales seamlessly to millions of vectors with serverless indexing.
• Text Search with BM25: Utilizes the BM25 ranking algorithm for more relevant text search results, surpassing the default full-text search capabilities in PostgreSQL.
• Hybrid Search: Combines vector search and BM25 text search using RRF or other reranking algorithms for enhanced results.
• Serverless Indexing: Offloads vector index creation to a separate machine from the main database, allowing effortless scaling without compromising database performance.
  • Provides infinite scalability, adapting resource usage to needs.
  • Enables building or updating indexes without slowing down the database.
  • Offers seamless integration by simply adding external=true to vector index creation.
• Embedding Generation and LLM Integrations: Simplifies AI workflows by allowing direct generation of embeddings and running LLMs within the database using simple SQL commands.
  • Supports over 20 embedding models and LLMs, including OpenAI, Cohere, and Jina AI.
  • Automatically generates new vector or LLM columns based on existing data.
• Easy to Use: Integrates directly with SQL and popular ORMs, eliminating the need to learn new APIs or frameworks.
• Deployment Options: Available as open-source for self-hosting or as a managed service (Lantern Cloud).

Pricing

• Free tier: $0.00/month. This is a multi-tenant setup that never pauses and includes vector search and indexing capabilities.