Skip to documentation content
Dataset APIDocs
Back to website
FIEN

Generated SDKs

Use contract-generated clients against production, sandbox, or deterministic local fake data.

9 min Updated 2026-08-15Docs version 1.0
In this guide
  1. 122 generated language packages
  2. 2OpenAPI is the source of truth
  3. 3TypeScript and Python fake-data quick starts
  4. 4Contract and native E2E verification
In this guide
  1. 122 generated language packages
  2. 2OpenAPI is the source of truth
  3. 3TypeScript and Python fake-data quick starts
  4. 4Contract and native E2E verification

One contract, generated packages

Every Dataset API SDK is generated from openapi/dataset-api-v1.json with the framework’s pinned OpenAPI Generator. Endpoint methods, models, authentication, base URLs, and package versions therefore move with the canonical API contract instead of a hand-maintained copy.

The package pages below are read from the generated Dataset API manifest. A stale OpenAPI snapshot or generated file causes the SDK checks to fail.

ArtifactGuarantee
GENERATION.jsonBinds the package to the OpenAPI version, specification SHA-256, and generator version
SDK_VERIFICATION_PLAN.jsonLists contract-derived operations and required checks
FAKE_API.mdDocuments the shared product-aware fake endpoint and demo credential
manifest.jsonRecords every generated file hash and all 22 language targets

TypeScript against fake data

Install the locally generated package
npm install ./sdks/generated/dataset-api/typescript
TypeScript
import { Configuration, DatasetsApi } from "@datasetapi/sdk";

const config = new Configuration({
  basePath: "http://localhost:3000/api/fake/dataset-api/v1",
  accessToken: process.env.DATASET_API_KEY,
});

const catalog = await new DatasetsApi(config).listDatasets({ xAPIVersion: "1" });
console.log(catalog.data[0].latest_release);

Python against fake data

Install a locally generated package
python -m pip install ./sdks/generated/dataset-api/python
Python
import asyncio
import os
import dataset_api

async def main():
    config = dataset_api.Configuration(
        host="http://localhost:3000/api/fake/dataset-api/v1",
        access_token=os.environ["DATASET_API_KEY"],
    )
    async with dataset_api.ApiClient(config) as client:
        catalog = await dataset_api.DatasetsApi(client).list_datasets(x_api_version="1")
        print(catalog.data[0].latest_release)

asyncio.run(main())

Deterministic fake API

Run the Dataset API site locally, set DATASET_API_KEY to a disposable local test credential, then use http://localhost:3000/api/fake/dataset-api/v1. The HTTP fake is backed by the same Dataset Builder adapter as the preview UI and exposes catalog, release, query, build, cancellation, download, and health operations with stable fixtures.

The fake is for integration development, examples, and SDK verification. It is not a substitute for sandbox performance or production data-quality testing.

Reset deterministic fixtures
npm run sdk:fake:reset -- --product dataset-api

Regenerate and verify

  1. 1
    Change the canonical OpenAPI contract
  2. 2
    Run npm run sdk:generate to regenerate every configured product and language package
  3. 3
    Run npm run sdk:check to prove committed output is current
  4. 4
    Run npm run sdk:e2e -- --product dataset-api --language typescript and repeat for Python
  5. 5
    Run the operation E2E suite before publishing a package
Was this page helpful?