Machine Learning

APIs related to ML services offered by Amorphic.

📄️ Download SageMaker transform job logs for a model run

Returns a presigned S3 URL plus log preview for CloudWatch logs of a SageMaker batch transform job tied to a dataset file model run. Required query params: servicename=sagemaker and serviceloggroup=TransformJobs (else IPV-1053 when missing, LOG-1001 when values differ). Caller must have access to the dataset. Safe to retry (read-only URL/preview mint).

📄️ List SageMaker ML models accessible to the user

Returns registered ML models the caller can access, with optional filter, sort, pagination, and field projection. These are SageMaker models, not Bedrock foundation models.

📄️ Register a SageMaker ML model

Creates an Amorphic ML model from either ArtifactsLocation (custom SageMaker model) or ExistingModelResource (tag an existing SageMaker model). Provide exactly one of those two fields.

📄️ List SageMaker models not registered as custom Amorphic models

Returns SageMaker model names that are not already custom Amorphic models (artifact-created ModelIds). Query parameter fields is accepted but ignored.

📄️ Generate a PUT URL for an ML model artifact

Returns a presigned PUT URL and uploadPath in the ML temp bucket under ml-models/{uuid}/{FileName}. Use uploadPath as ArtifactsLocation when creating a custom model.

📄️ Get registered ML model details

Returns the DynamoDB record for a registered SageMaker ML model. Requires resource access. Optional projectionExpression limits returned attributes.

📄️ Delete a registered ML model

Deletes the Amorphic ML model. Custom models (IsSubscribedModel no) also delete the SageMaker model. Subscribed models are untagged instead. Owner access is required. Blocked when data pipelines depend on the model.

📄️ Update registered ML model metadata

Updates model metadata in DynamoDB. Does not change SageMaker artifacts or AlgorithmUsed. Requires owner or editor access. AlgorithmUsed must be present in the body even though it is not applied.