argilla

Why we included this project

Fine-tuning a language model usually hits the same bottleneck: labeled data, not compute. Argilla is a shared workspace where AI engineers and domain experts review and annotate text and multimodal datasets together, with a Python SDK driving the workflow and a web UI handling the human side of review. The same setup covers classification and NER feedback as well as preference pairs for RLHF and RAG evaluation, and it deploys on Hugging Face Spaces or Docker. Because the process is scriptable, it works for continuous evaluation: you keep correcting data as the model improves rather than running one labeling sprint up front. The codebase is mature and in maintenance mode, patched but not gaining features, so it is a reasonable choice for teams that want a proven labeling layer now.