Fastembed – Rust library for generating vector embeddings, reranking locally (github.com)

Fastembed, a new Rust library, has been launched to streamline the generation of vector embeddings and local reranking of results, without requiring asynchronous dependencies like Tokio. Its efficient architecture leverages the ONNX runtime for high-performance inference, utilizing models from Hugging Face for rapid text and image encoding. With support for various embedding models—ranging from BAAI's BGE series to sentence-transformers and even multilingual options—Fastembed opens new possibilities for developers looking to incorporate advanced embedding capabilities into their applications. This development is significant for the AI/ML community as it provides a robust, easy-to-use toolset for generating and manipulating embeddings, crucial for various AI tasks such as search and recommendation systems. Fastembed's ability to produce embeddings in bulk, handle images and text simultaneously, and the option for local model use enhances its appeal for researchers and businesses needing performant and scalable solutions. With features like direct integration with vector search engines and GPU compatibility, Fastembed positions itself as a strong contender in the growing landscape of embedding libraries.