Skip to content

Getting started

Subscribe and deploy your regional package, then use the public notebook to invoke an existing endpoint with a fictional identifier.

Subscribe and deploy

  1. Review the Marketplace offer, version 0.2.0, region, supported instance, software terms and AWS charges.
  2. Subscribe and select Continue to configuration. Copy the model-package ARN for your chosen region.
  3. Create the SageMaker model in your account with that ARN, your execution role and EnableNetworkIsolation=True. Use a supported ml.m5.large or ml.c5.xlarge endpoint and wait for InService.
  4. Download the public invocation notebook. Set your region and existing endpoint name. Use an approved AWS session with permission to invoke that endpoint; do not paste credentials into the notebook.

The notebook invokes an existing endpoint only. It does not subscribe, deploy, delete resources or retrieve secrets. See the shared SageMaker guide for the difference between caller permissions and the execution role.

First request

{"text":"RFC LUNA910210AB1; RUT 12.345.678-5","mode":"replace","country_hint":"auto"}

The values are fictional. Inspect version, capabilities, warnings and validation fields in the result. person_address_ner must be false. Failed checksums do not prevent recognized candidates from being redacted.

Batch and cleanup

Use the same bounded request contract for real-time and Batch Transform. Configure line splitting when supplying JSON Lines. SageMaker payload settings do not increase the application's 256 KiB body limit. Keep S3 input/output private and authorized to your execution role.

Inspect your exact endpoint and batch job when finished. Delete only the temporary hosting resources you created; the invocation notebook does not clean them up. Keep or remove test objects according to your retention policy. Closing a notebook does not stop compute charges.

Last updated November 15, 2026 · Suggest an improvement