Docs · TypeScript SDK
TypeScript SDK
@synthforgeio/sdk wraps the REST API with typed methods and polling helpers.
Before you start: you need a personal API key. Create one in the app, then come back.
@synthforgeio/sdk is a thin typed client for schemas, AI schema generation, and datasets.
On npm: npm install @synthforgeio/sdk. MIT licensed, zero runtime dependencies.
The helpers create the job, wait for it to finish, and throw a typed error if it fails.
typescript
// npm install @synthforgeio/sdk
import { SynthForge } from '@synthforgeio/sdk';
const sf = new SynthForge({
apiKey: process.env.SF_API_KEY!, // sfk_…
});
// One call: a schema from the description, then a dataset from the schema.
const result = await sf.descriptionToDataset({
description: 'customers with orders; each customer places 1 to 5 orders',
default_rows: 500,
formats: ['csv'],
includeDownloadUrl: true,
});
console.log(result.downloadUrl);
// Or two steps, to read the schema before any rows are generated.
const drafted = await sf.generateSchema({
description: 'a lending library: members, books, and loans',
});
console.log(drafted.schema.schema_data);
const { downloadUrl } = await sf.generateDataset({
schema_id: drafted.schemaId,
default_rows: 200,
includeDownloadUrl: true,
});Each resource is also there on its own, when you want to drive the steps yourself:
typescript
const { items } = await sf.schemas.list({ limit: 20 });
const schema = await sf.schemas.create({
name: 'orders',
schema_data: {
tables: [
{
name: 'orders',
columns: [
{ name: 'id', field_type: 'uuid', primary_key: true },
{ name: 'label', field_type: 'word' },
],
},
],
},
});
const dataset = await sf.datasets.create({
schema_id: schema.id,
name: 'demo',
row_counts: { orders: 100 },
formats: ['csv'],
seed: 42,
});
const status = await sf.datasets.wait(dataset.id);
if (status.status === 'completed') {
const { url } = await sf.datasets.getDownloadUrl(dataset.id);
// fetch(url) → ZIP
}Next