Shadowgraph

872commits

15contributors

10languages

Tech stack & purpose

Lantern is a PostgreSQL vector database extension designed for building AI applications that require approximate nearest-neighbor search and vector operations. The project provides multiple command-line tools and PostgreSQL extensions written in Rust to enable vector search, embedding generation, and index management within PostgreSQL. The Lantern CLI offers functionality for creating HNSW indexes outside of Postgres for faster parallel index building, generating vector embeddings from text and images using various models (including OpenAI and Cohere APIs), autotuning index parameters, and compressing vectors through product quantization. The Lantern Extras extension allows users to experiment with embeddings directly within PostgreSQL by providing streaming downloads of vector datasets, parsing vector benchmarks like SIFT and GIST, and generating embeddings using CLIP models and other embedding runtimes without leaving the database.