openmed

Why we included this project

OpenMed is a local-first toolkit for teams that handle clinical text and need to pull out medical entities and strip patient identifiers without sending data anywhere. Instead of wiring up a cloud API, you run its models on your own hardware, from a CPU laptop to Apple Silicon via MLX, Android phones, or a GPU server, which makes it a natural fit for HIPAA-sensitive and air-gapped work. A single analyze_text call returns labeled spans for clinical entities such as disease names and drug names, and it can redact patient identifiers as well. The same model catalog works across Python, Swift, Android, and the browser, so a small team can prototype in Python and later ship the same extraction inside a mobile or on-prem product. It is not a general-purpose NLP framework; it comes with over two thousand biomedical models across 21 languages and expects you to work with those rather than train your own, which keeps it focused and practical for health data.