Datavo¶
Datavo is a dataset store: import sources into stores, attach derived keys under producers, select worksets, and freeze them into named, reproducible datasets you can split, share and stream.
Install¶
That installs the datavo command-line tool and the datavo_sdk Python client
together (with Entra sign-in). For cloud downloads, add the storage extra you use,
e.g. pip install 'datavo-sdk[azure]'.
Start here¶
- User guide — a fifteen-chapter path from an empty store to two training runs scored on the same held-out split. Begin with Concepts and Setup.
- Just want data on disk? — find a dataset and pull it with the CLI, no build.
- Reference — deep pages on the CLI, shard cache, idempotency, store statistics, and local shards.
- Examples — a runnable script for every chapter.