Datasets
Datasets let you push your own data into A vs B and read it back at the edge, by key, in a single fast lookup. Upload a target list of customer IDs, a table of content keyed by product, or a recommendation feed generated by your own pipeline. Then read it back from the browser snippet, the Node SDK, or directly over HTTP.
Every dataset is versioned. A new upload lands as a new version, and you activate it with one atomic flip. You can roll back instantly by re-activating an older version, and readers never see a half-updated dataset.
In this section
What Are Datasets?The dataset primitive: LIST, TABLE, and FEED types, the version lifecycle, retention, and limits.
Uploading & VersioningThe NDJSON row format, the dashboard upload flow, commit validation, activation, and rollback.
Serving & Reading DatasetsRead datasets from the browser snippet with avsb.dataset() or from your server with the Node SDK.
Datasets API & CLI ReferenceThe REST endpoints, the avsb datasets CLI commands, and the nightly-push automation recipe.
Troubleshooting DatasetsRun the upload self-test, recover a version stuck on Processing, and confirm a delete reclaims its storage.