Argilla

Argilla is an open source collaboration platform where AI engineers and domain experts build high-quality datasets together. Datasets are defined programmatically through a Python SDK with customizable fields and questions, then labeled and reviewed in a web UI that supports filtering, semantic search, and AI-generated suggestions to speed up annotation. It covers traditional NLP tasks such as text classification and named entity recognition alongside LLM-era workflows like preference data collection for RLHF, prompt and response evaluation, and multimodal review. The server can be self-hosted or launched on Hugging Face Spaces, and the platform integrates closely with the Hugging Face datasets ecosystem. Community projects built with Argilla include cleaned datasets such as UltraFeedback that were used to fine-tune open models, and organizations including the Red Cross and Prolific have used it for annotation work. The code is released under the Apache 2.0 license.