easy-dataset

Why we included this project

Anyone who has assembled a fine-tuning or RAG dataset by hand knows the real work is parsing documents, splitting text into sensible chunks, and writing question-answer pairs. Easy Dataset wraps that whole pipeline in a guided desktop and web application. You feed it PDF, Markdown, DOCX, or EPUB files, it applies a few chunking strategies, and an LLM API generates questions, answers, and chain-of-thought content you can review and clean before exporting. Output covers single-turn, multi-turn, and image QA formats, with exports to Alpaca, ShareGPT, and JSONL, plus one-click configuration for LLaMA Factory and direct upload to Hugging Face. If you also need to verify the data, the built-in evaluation side turns documents into test sets and runs automated judging alongside a double-blind human arena. For small teams that want a visual, low-code route from raw domain documents to usable training and evaluation data, this is a practical pick.