easy-dataset

Easy Dataset is an open-source tool for creating high-quality datasets for LLM fine-tuning, RAG, and evaluation. It addresses common challenges such as poor AI-generated QA pairs from large documents, context window limitations, duplicate questions, and the need for structured dataset management. The tool offers a full pipeline from document parsing to dataset construction, annotation, export, and evaluation. It features a project-based workflow, intelligent document chunking (chapter-aware recursive splitting), AI-assisted domain label generation, batch question and answer generation with COT support, quality check mechanisms, and multi-format export (Alpaca, ShareGPT). It also integrates a dataset plaza aggregating sources like HuggingFace and Kaggle.