Welcome
Programmatic access to Adaptive Data—ingest, adapt, evaluate, and export training data through the Adaption API and Python SDK.
Adaptive Data is a platform that brings data-optimization techniques typically reserved for frontier labs to everyday teams. The product analyzes your data’s structure and learns how to adapt and optimize it so you can ship model-ready datasets with clearer quality and control.
The API and Python SDK expose the same lifecycle you use in the web app — ingest, adapt, evaluate, and export — so you can run it from a notebook, a service, or CI.
Next steps
Section titled “Next steps”- Create API keys — authenticate the API and SDK for your environments.
- Getting started — install the SDK, ingest data, start an adaptation run, and download results.
Ways to create a dataset
Section titled “Ways to create a dataset”The Python SDK supports three ingestion paths (detailed on the getting started page):
- Local file —
datasets.upload_file()runs the full presigned upload flow (CSV, JSON, JSONL, Parquet). - Hugging Face —
datasets.create_from_huggingface()pulls from a dataset URL; server-side processing is asynchronous—wait before you adapt. - Kaggle —
datasets.create_from_kaggle()imports from a Kaggle dataset; register your Kaggle credentials under API keys settings first.