Skip to main content
Most users should use the Python SDK, which calls these endpoints for you. They are listed here for direct integration. Every request authenticates with your API key:
Authorization: Bearer hk_live_...
The base URL is your environment’s origin (the same value you set as HOLLERITH_BASE_URL), e.g. https://hollerith.monarcha.ai.

Limits

MethodPathPurpose
GET/v1/model-limitsThe published limits (see Model). Unauthenticated.

Fit — prepare a reusable context

Fit is asynchronous: submit returns a queued job plus a fitted-context id; poll until ready.
MethodPathPurpose
POST/v1/fitsSubmit a labeled table; returns a fitted-context id.
GET/v1/fits/{id}Poll the context until status: "ready".

Predict

MethodPathPurpose
POST/v1/predictionsScore rows. Pass fittedContextId to reuse a saved context (sends only the rows to score); omit it to send the training table inline.
GET/v1/predictions/{id}Poll job status.
GET/v1/predictions/{id}/resultRead the scored output once status: "succeeded".

Evaluate

MethodPathPurpose
POST/v1/evaluationsSubmit a labeled set for a held-out / k-fold metric.
GET/v1/evaluations/{id}/resultRead the metric once succeeded.

Forecast

MethodPathPurpose
POST/v1/forecastsSubmit a time series; poll/read via the /v1/predictions/{id} routes.

Uploads

Large datasets are uploaded directly to object storage, not through the API body.
MethodPathPurpose
POST/v1/uploadsRequest a presigned URL (single or multipart) to upload a dataset; the upload key is then referenced when you submit a job. The SDK does this automatically.

Errors

Every error response is a typed envelope with a code, a human-readable problem / cause / fix, and a requestId for support. Quote the requestId if you contact us.