View Source AWS.SageMakerFeatureStoreRuntime (aws-elixir v0.14.1)

Contains all data plane API operations and data types for the Amazon SageMaker Feature Store.

Use this API to put, delete, and retrieve (get) features from a feature store.

Use the following operations to configure your OnlineStore and OfflineStore features, and to create and manage feature groups:

CreateFeatureGroup DeleteFeatureGroup

DescribeFeatureGroup ListFeatureGroups

Summary

Functions

Retrieves a batch of Records from a FeatureGroup.

Deletes a Record from a FeatureGroup in the OnlineStore.

The PutRecord API is used to ingest a list of Records into your feature group.

Functions

Link to this function

batch_get_record(client, input, options \\ [])

View Source

Retrieves a batch of Records from a FeatureGroup.

Link to this function

delete_record(client, feature_group_name, input, options \\ [])

View Source

Deletes a Record from a FeatureGroup in the OnlineStore.

Feature Store supports both SoftDelete and HardDelete. For SoftDelete (default), feature columns are set to null and the record is no longer retrievable by GetRecord or BatchGetRecord. For HardDelete, the complete Record is removed from the OnlineStore. In both cases, Feature Store appends the deleted record marker to the OfflineStore. The deleted record marker is a record with the same RecordIdentifer as the original, but with is_deleted value set to True, EventTime set to the delete input EventTime, and other feature values set to null.

Note that the EventTime specified in DeleteRecord should be set later than the EventTime of the existing record in the OnlineStore for that RecordIdentifer. If it is not, the deletion does not occur:

  • For SoftDelete, the existing (not deleted) record remains in the OnlineStore, though the delete record marker is still written to the OfflineStore.

  • HardDelete returns EventTime: 400 ValidationException to indicate that the delete operation failed. No delete record marker is written to the OfflineStore.

When a record is deleted from the OnlineStore, the deleted record marker is appended to the OfflineStore. If you have the Iceberg table format enabled for your OfflineStore, you can remove all history of a record from the OfflineStore using Amazon Athena or Apache Spark. For information on how to hard delete a record from the OfflineStore with the Iceberg table format enabled, see Delete records from the offline store.

Link to this function

get_record(client, feature_group_name, expiration_time_response \\ nil, feature_names \\ nil, record_identifier_value_as_string, options \\ [])

View Source

Use for OnlineStore serving from a FeatureStore.

Only the latest records stored in the OnlineStore can be retrieved. If no Record with RecordIdentifierValue is found, then an empty result is returned.

Link to this function

put_record(client, feature_group_name, input, options \\ [])

View Source

The PutRecord API is used to ingest a list of Records into your feature group.

If a new record’s EventTime is greater, the new record is written to both the OnlineStore and OfflineStore. Otherwise, the record is a historic record and it is written only to the OfflineStore.

You can specify the ingestion to be applied to the OnlineStore, OfflineStore, or both by using the TargetStores request parameter.

You can set the ingested record to expire at a given time to live (TTL) duration after the record’s event time, ExpiresAt = EventTime + TtlDuration, by specifying the TtlDuration parameter. A record level TtlDuration is set when specifying the TtlDuration parameter using the PutRecord API call. If the input TtlDuration is null or unspecified, TtlDuration is set to the default feature group level TtlDuration. A record level TtlDuration supersedes the group level TtlDuration.