Metadata-Version: 2.4
Name: kushim
Version: 0.0.2
Summary: A framework for building high-quality, verifiable evaluation datasets for LLMs.
Requires-Python: >=3.9
Requires-Dist: datasets>=3.6.0
Requires-Dist: dspy>=2.4.5
Requires-Dist: huggingface-hub[cli]>=0.33.0
Requires-Dist: llama-index
Requires-Dist: polars>=0.20.0
Requires-Dist: python-dotenv>=1.0.0
Requires-Dist: wikipedia>=1.4.0
Description-Content-Type: text/markdown

# Kushim: A Framework for Verifiable LLM Evaluation Datasets

Kushim is a framework for generating high-quality, verifiable Question & Answer datasets from Wikipedia articles. It provides a complete, end-to-end workflow from data generation to packaging and sharing on the Hugging Face Hub.


