FastEmbed: Lightweight Python Library for Embeddings

If you need embeddings but don't want to deal with the overhead of PyTorch or a massive cloud API, FastEmbed is the answer. Developed by the Qdrant team, it is a focused library designed for extreme efficiency and ease of use.

Optimized for Local Inference

FastEmbed uses ONNX Runtime under the hood, allowing it to generate embeddings on your CPU with incredible speed. It is significantly faster than standard Python-based embedding libraries and has a minimal memory footprint, making it perfect for edge devices and local development.

Ready-to-Use Models

The library comes with pre-packaged, quantized versions of the best open-source embedding models (like BGE and FlagEmbedding). You can get a working embedding pipeline up and running in seconds, with zero external dependencies and zero API costs.