Overview
Step-by-step workflows for the Adaption API and app. Run the notebooks in Google Colab or follow along in the docs.
Runbooks take you through a complete workflow from start to finish. The notebook runbooks come from the adaption-runbooks repo on GitHub. Each one has a page here and a notebook you can open in Google Colab or run locally.
Get started
Section titled “Get started”- Get an API key from the Adaption app. See Create API keys.
- Open a notebook in Colab and add
ADAPTION_API_KEYto Colab secrets (🔑 in the sidebar).
To run the notebooks locally, clone the repo and run:
export ADAPTION_API_KEY=...uv run --with adaption --with jupyterlab jupyter labNotebook runbooks
Section titled “Notebook runbooks”| Runbook | What it does | Notebook |
|---|---|---|
| Invent a Dataset | Generates a dataset from one prompt and scores its diversity | Colab · GitHub |
| Expand a Dataset | Grows 500 GSM8K rows with domain, general, and similar rows | Colab · GitHub |
| Extend Conversation Trajectories | Writes next user turns that need the whole conversation to answer | Colab · GitHub |
| Train a Model with AutoScientist | Trains on Alpaca with data augmentation and downloads the best checkpoint | Colab · GitHub |
App runbooks
Section titled “App runbooks”These walk through the Adaption app, with no code.
- Adaptive Data App Walkthrough: import a dataset, map its columns, run adaptation, and review the results.
- Processing Unstructured Documents: turn PDFs, spreadsheets, and slide decks into a dataset.
- AutoScientist App Walkthrough: pick a dataset, launch training, and read the result.
Contribute
Section titled “Contribute”To add a runbook, see CONTRIBUTING.md in the repo.