Eric Search

A local vector search engine built for speed and scalability.

• Fast: we use two-level IVF to effectively scale to millions of documents
• EricRanker(): powered by a cross-encoder model that extracts relevant information from the top documents.
• Accelerated compute: compatible with both MPS and CUDA.
• Easy to use: only a few lines of code are needed to train new datasets
• Lightweight: simple to install and run with a single Python script.
• Integrated with Hugging Face's Hub.
• Transferable: zip a single folder to move an entire dataset.

Install

pip install ericsearch

Maintainers

• Eric Fillion Lead Maintainer
• Ted Brownlow Maintainer