hf-dataset-distill

An project to help with distillation of models to get dataset with Batch API (50% discount)

• GitHub: Claude Opus 4.6 • GPT-5.4
• HuggingFace: Claude Opus 4.6 • GPT-5.4

Features

• Easy to use commands
• JS/TS Stack
• Hugging Face compatible

Prerequisites

• Bun installed
• AI Provider API Key
• A money in AI Provider balance
• Source dataset

Commands

Installation

bun install

Running

Development:

bun run dev

Production:

bun run start

Build dataset

bun run export

An file should be ready at dataset/

Dataset stat

bun run stat

Source dataset

PROMPT:

Generate a dataset for different/mixed purposes with focus on programming language.

Most of dataset should be:
- Rust
- Golang
- TypeScript/JavaScript/Node.js
- Also other programming languages

should be included.

Most output should be simpler, not verbose or expanded.
Only last 15% rows should be verbose with expanded details.

Output dataset entries: ~2000 rows

License

Apache-2.0