Writing data
Your own data is a first-class citizen on autario. A dataset you create is private until you say otherwise, it is queryable through exactly the same endpoints as the public catalog, and an app of yours can read it through the same bridge. This page covers the write half of the API.
Create a dataset
A dataset starts empty, with a title and a description. Keep it private unless you have decided otherwise, which is the default, because a public dataset carries a URL contract: once other people cite it, the address has to keep working.
POST /api/v1/datasets
{ "title": "Weekly signups", "description": "One row per week.", "is_public": false }
The response returns the id you will use everywhere else. Titles are stable identifiers in practice, so choose one you can live with; the slug that a public dataset gets is derived from it.
Append rows
Rows go in as JSON objects and the schema is inferred from the first batch. Send at most ten thousand rows per request and page through larger loads; the response tells you how many were inserted and the running total.
POST /api/v1/datasets/<id>/rows
{ "rows": [ { "week": "2026-09-01", "signups": 41 } ] }
Two habits save pain later. Send a real date string rather than a formatted label, because the time column is what every chart and every join uses. And send a column with a value in the first batch rather than a column of nulls, because a column that has never carried a value cannot have its type inferred and will be widened to text.
Clear, delete and replace
Clearing removes the rows and keeps the schema, which is what you want for a full replace. Deleting removes the dataset itself. Both are owner-scoped: a dataset you do not own answers 404, never a permission error, because the existence of somebody else's data is not something an error message should disclose.
DELETE /api/v1/datasets/<id>/rows # keep the shape, drop the rows
DELETE /api/v1/datasets/<id> # drop the dataset
If you are refreshing on a schedule, prefer clear and re-append over deleting and recreating: the id stays stable, so charts, apps and links that point at the dataset keep working.
What a write costs you
Rows you store count toward the storage allowance of your plan, and so do app bundles. The limits page carries the current numbers, generated from the same catalog the server reads, so it cannot be out of date. When you cross a limit the write is refused with a payment-required response that names which limit you hit and what lifts it; it is never silently truncated.
Quality rules that apply to you too
autario refuses to publish a number it cannot stand behind, and the same instincts are worth borrowing for your own data. Do not edit a provider's value to make a chart nicer; flag it instead. Do not mix a fetch timestamp into a time column. Do not load a truncated page of a paginated API and call it a dataset. If a public autario dataset ever looks wrong, the report_data_issue tool exists precisely so that a reader can say so, and safe categories re-ingest automatically.