ConardLi/easy-dataset

A powerful tool for creating datasets for LLM fine-tuning 、RAG and Eval

What it solves

Easy Dataset simplifies the complex process of creating high-quality, structured datasets for Large Language Models (LLMs) from unstructured, domain-specific documents. It eliminates the manual effort of parsing files, splitting text, and generating question-answer pairs required for model fine-tuning, RAG, and performance evaluation.

How it works

The tool provides a visual interface that guides users through a data pipeline: it parses various document formats (PDF, DOCX, etc.), uses intelligent algorithms to split text into meaningful chunks, and leverages LLM APIs to automatically generate questions, comprehensive answers (including Chain of Thought), and domain label trees. It also includes a system for cleaning noise from the data and evaluating the resulting dataset's quality using judge models or human blind tests.

Who it’s for

It is designed for both technical and non-technical users who need to build specialized datasets for fine-tuning LLMs, improving RAG recall rates, or conducting vertical domain model evaluations.

Highlights

  • Comprehensive Document Support: Handles PDF, Markdown, DOCX, TXT, and EPUB with intelligent recognition.
  • Diverse Dataset Types: Supports single-turn QA, multi-turn dialogues, and image-based QA datasets.
  • Integrated Evaluation: Features automated scoring via Judge Models and a double-blind "Arena" for human comparison.
  • Seamless Integration: One-click configuration for LLaMA Factory and direct upload capabilities to Hugging Face Hub.
  • Flexible Model Support: Compatible with any OpenAI-format API, including local models via Ollama.

Related

  • Project
    drmingler/smart-llm-loader

    A Python package that converts documents and scanned images into clean, structured markdown chunks using OCR and LLM-powered semantic chunking for RAG systems.

  • Project
    shcherbak-ai/contextgem

    ContextGem is an open‑source Python library for extracting structured, typed facts from documents using LLMs and low‑cost decision models. You declare aspects (text sections) and concepts (typed values) in plain language; the library auto‑generates prompts, routes each extraction to the appropriate model, validates output, and returns each fact with its source paragraph and optional justification. It supports probability‑based answers from decision models, batch processing, and full documentation.

  • Project
    OpenDCAI/DataFlow

    DataFlow is an open‑source Python framework that lets you build, run, and share low‑code pipelines for generating, cleaning, and evaluating LLM training data. It provides a library of reusable operators, a visual WebUI, an AI‑assistant for auto‑creating pipelines, and a Ray‑based distributed execution layer.

  • Project
    yifanfeng97/Hyper-Extract

    A smart knowledge extraction CLI that transforms unstructured documents into structured knowledge bases, graphs, and lists using LLMs and domain-specific templates.

  • Project
    docling-project/docling

    Docling is an open‑source Python library/CLI that parses many document types (PDF, Word, slides, spreadsheets, HTML, e‑books, audio, video, email, etc.) into a unified, structured representation. It extracts layout, tables, code, formulas, images, and speech‑to‑text, then exports to Markdown, HTML, JSON, DocLang, DocTags, etc. Designed for generative‑AI workflows, it offers plug‑and‑play integrations with LangChain, LlamaIndex, Crew AI, Haystack, and supports local execution for privacy‑sensitive data.