Dataset API

Deliver observed and scoped synthetic datasets through one stable API.

Publish observed releases as managed data products, then scope realistic synthetic delivery for an agreed use case with lineage and review under one customer contract.

Start a pilot See pricing

Evidence-backed results Workspace-level access

Release workflow

Finnish vehicle market

Immutable release · dsv-preview-20260813

Completesource coverage verified

VVehicle listings4 source rows in preview

18,420

v0.2

Published

QualitycompleteLineageIncluded

2.0 release schemaBuilder artifacts are validated before publication

1.0 stable API contractBackend changes stay behind the adapter seam

Capabilities

A product surface built for dataset releases

The customer sees a small, predictable API. The builder can gain new recipes, fields, and quality checks without leaking internal churn.

Explore the product

Catalog and releases

Discover datasets, pin an immutable release, and inspect the lineage behind each row.

Schema and definition versions
Parent release and changes

Bounded queries

Query only the rows and fields your application needs, with quality metadata in the response.

Field selection and limits
Coverage and quality status

Synthetic datasets

Scope realistic additional rows from an immutable observed release for a defined use case.

Pilot and custom delivery
Fidelity and privacy review before publish

Process

From source capture to a dependable API row

Every stage has a clear owner and a version boundary.

Capture

Replay Apify JSONL or another source adapter into a source snapshot with coverage evidence.

Source adapter

Build

Normalize fields, reconcile temporal state, and publish a content-addressed release.

dataset_builder

Scope synthetic delivery

Agree the intended use, volume, fields, and review thresholds; publish only after explicit acceptance.

Scoped pilot

Query

Consume the stable Dataset API contract with lineage and quality metadata.

API v1

Next step

Define the synthetic-data acceptance test first.

Start with one observed release, a specific use case, and review criteria that match the decision the data must support.

Scope a synthetic-data pilot Read the product page