Adding searchable metadata to your features

The code in this section uses the UpdateFeatureMetadata operation in the AWS SDK for Python (Boto3) to add searchable metadata to your features for different scenarios. For information about the other languages to submit a query, see See Also in the Amazon SageMaker API Reference.

For more Feature Store examples and resources, see Amazon SageMaker Feature Store resources.

Example code

After you've updated the metadata for a feature, you can use the DescribeFeatureMetadata operation to see the updates that you've made.

The following code goes through an example workflow using the AWS SDK for Python (Boto3). The example code does the following:

For more Feature Store examples and resources, see Amazon SageMaker Feature Store resources.

Step 1: Setup

To start using Feature Store, create SageMaker AI, boto3 and Feature Store sessions. Then set up the S3 bucket you want to use for your features. This is your offline store. The following code uses the SageMaker AI default bucket and adds a custom prefix to it.

# SageMaker Python SDK version 2.x is required
%pip install 'sagemaker>=2.0.0'
import sagemaker
import sys

import boto3
import pandas as pd
import numpy as np
import io
from sagemaker.session import Session
from sagemaker import get_execution_role
from botocore.exceptions import ClientError


prefix = 'sagemaker-featurestore-introduction'
role = get_execution_role()

sagemaker_session = sagemaker.Session()
region = sagemaker_session.boto_region_name
s3_bucket_name = sagemaker_session.default_bucket()
sagemaker_client = boto_session.client(service_name='sagemaker', region_name=region)
            

Step 2: Create a feature group and add features

The following code is an example of creating a feature group with feature definitions.

feature_group_name = "test-for-feature-metadata"
feature_definitions = [
    {"FeatureName": "feature-1", "FeatureType": "String"},
    {"FeatureName": "feature-2", "FeatureType": "String"},
    {"FeatureName": "feature-3", "FeatureType": "String"},
    {"FeatureName": "feature-4", "FeatureType": "String"},
    {"FeatureName": "feature-5", "FeatureType": "String"}
]
try:
    sagemaker_client.create_feature_group(
        FeatureGroupName=feature_group_name,
        RecordIdentifierFeatureName="feature-1",
        EventTimeFeatureName="feature-2",
        FeatureDefinitions=feature_definitions,
        OnlineStoreConfig={"EnableOnlineStore": True}
    )
except ClientError as e:
    if e.response["Error"]["Code"] == "ResourceInUse":
        pass
    else:
        raise e       
     

Step 3: Add metadata

Before you add metadata, use the DescribeFeatureGroup operation to make sure that the status of the feature group is Created.

sagemaker_client.describe_feature_group(
        FeatureGroupName=feature_group_name
    )               
            

Add a description to the feature.

sagemaker_client.update_feature_metadata(
    FeatureGroupName=feature_group_name,
    FeatureName="feature-1",
    Description="new description"
)
            

You can use the DescribeFeatureMetadata operation to see if you successfully updated the description for the feature group.

    sagemaker_client.describe_feature_metadata(
    FeatureGroupName=feature_group_name,
    FeatureName="feature-1"
)

You can also use it to add parameters to the feature group.

sagemaker_client.update_feature_metadata(
    FeatureGroupName=feature_group_name,
    FeatureName="feature-1",
    ParameterAdditions=[
        {"Key": "team", "Value": "featurestore"},
        {"Key": "org", "Value": "sagemaker"},
    ]
)
            

You can use the DescribeFeatureMetadata operation again to see if you have successfully added the parameters.

    sagemaker_client.describe_feature_metadata(
    FeatureGroupName=feature_group_name,
    FeatureName="feature-1"
)