Frequently Asked Questions

Currently we support CSV, JSON, JSONL, and TXT. Parquet support is planned. Free tier uploads are limited to 50 KB per file.

You can explore the interface without signing in. To run refinements, track jobs, and publish datasets you will need a free account.

The free tier currently allows up to 1,000 rows / month and 50 KB file uploads. Higher limits and larger files will be available on paid plans.

• Google Gemini — best general quality and structured output
• Groq — fastest responses
• Cerebras — high throughput for larger batches
• SambaNova — strong enterprise performance

Start with Gemini if you’re unsure.

Uploaded files are processed to generate the refined dataset. We retain job metadata and the final refined output so you can re-download or publish later. Raw files are not kept longer than necessary for the job. See our Privacy Policy for full details.

Yes. Use the optional Instructions field to tell the model what to do — extract fields, standardize formats, remove PII, classify text, etc. Clear, specific instructions produce better results.

You receive a cleaned, consistently typed dataset with an inferred schema and a quality score. You can download it or push it directly to Hugging Face.

Yes. After validation you can push the refined dataset to your Hugging Face account (requires linking your HF token).

Visit the Contact page and send us a message. We usually respond within 1–2 business days.

An API is on the roadmap. For now the product is available through the web interface. Contact us if you have enterprise integration needs.