freedmand/semantra

Converts documents into vector embeddings using local transformer models (or OpenAI's API) and enables interactive semantic queries through a web interface, supporting query arithmetic (adding/subtracting terms to refine meaning). Built as a Python CLI tool that processes PDFs and text files once, then launches a local server for exploration with relevance scoring and document highlighting. Configurable embedding windows, multiple pre-trained models, and private local processing make it suitable for journalists, researchers, and analysts working with sensitive document collections.

2,705 stars and 178 monthly downloads. No commits in the last 6 months. Available on PyPI.