explosion/prodigy-recipes

Provides annotated recipe implementations for NER, text classification, terminology extraction, and image tasks that demonstrate active learning workflows, custom model integration, and pattern-based pre-labeling with spaCy components. Recipes are designed as customizable Python scripts that plug into Prodigy's streaming pipeline via the `-F` flag, enabling users to modify sorting functions, filtering logic, and model callbacks to tailor annotation workflows. Integrates with spaCy tokenizers and models, transformer-based tokenizers (BERT), fuzzy matching libraries, and supports pluggable ML models for closed-loop active learning scenarios.

505 stars. No commits in the last 6 months.