prodigy-recipes

Why we included this project

Anyone setting up annotation for NLP or computer vision will find this repository a useful way to build a labeling pipeline they actually control. It packages Prodigy recipes as commented scripts for tasks like named entity recognition, text classification, terminology bootstrapping, and image annotation with bounding boxes or polygons. The scripts are simplified and explained, so you can trace what each one does, adjust the active-learning selection, swap models, or add filters and callbacks without untangling production code. That dual role, a quick start for common commands like ner.teach and textcat.correct, plus a learning resource for model-in-the-loop design, makes it practical even beyond its own projects. One caveat: the recipes run on Prodigy, so you need a license for the tool itself.