Back to GitHub projects

jev-curate: TypeSafe Jev dataset sifter

A high-throughput dataset curation tool that uses the TypeSafe AI Jev System One API to filter and score synthetic and pretraining data in Parquet/JSONL streams with typed judgments.

仓库明确面向 TypeSafe AI 的 Jev / System One 模型构建,提供 Rust CLI 与 Python 绑定,调用 api.typesafe.ai 的客户端、预设评估规则、离线测试和基准数字,属于实际集成项目。

· 32 starsOpen on GitHub