Skip to content

Commit

Permalink
Add Google Vertex AI Feature Store - Feature View Sync Operators, Sen…
Browse files Browse the repository at this point in the history
…sor (#44891)
  • Loading branch information
CYarros10 authored Dec 16, 2024
1 parent 6b55430 commit 16022e0
Show file tree
Hide file tree
Showing 11 changed files with 946 additions and 2 deletions.
32 changes: 32 additions & 0 deletions docs/apache-airflow-providers-google/operators/cloud/vertex_ai.rst
Original file line number Diff line number Diff line change
Expand Up @@ -644,6 +644,38 @@ The operator returns the cached content response in :ref:`XCom <concepts:xcom>`
:start-after: [START how_to_cloud_vertex_ai_generate_from_cached_content_operator]
:end-before: [END how_to_cloud_vertex_ai_generate_from_cached_content_operator]

Interacting with Vertex AI Feature Store
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

To get a feature view sync job you can use
:class:`~airflow.providers.google.cloud.operators.vertex_ai.feature_store.GetFeatureViewSyncOperator`.
The operator returns sync job results in :ref:`XCom <concepts:xcom>` under ``return_value`` key.

.. exampleinclude:: /../../providers/tests/system/google/cloud/vertex_ai/example_vertex_ai_feature_store.py
:language: python
:dedent: 4
:start-after: [START how_to_cloud_vertex_ai_feature_store_get_feature_view_sync_operator]
:end-before: [END how_to_cloud_vertex_ai_feature_store_get_feature_view_sync_operator]

To sync a feature view you can use
:class:`~airflow.providers.google.cloud.operators.vertex_ai.feature_store.SyncFeatureViewOperator`.
The operator returns the sync job name in :ref:`XCom <concepts:xcom>` under ``return_value`` key.

.. exampleinclude:: /../../providers/tests/system/google/cloud/vertex_ai/example_vertex_ai_feature_store.py
:language: python
:dedent: 4
:start-after: [START how_to_cloud_vertex_ai_feature_store_sync_feature_view_operator]
:end-before: [END how_to_cloud_vertex_ai_feature_store_sync_feature_view_operator]

To check if Feature View Sync succeeded you can use
:class:`~airflow.providers.google.cloud.sensors.vertex_ai.FeatureViewSyncSensor`.

.. exampleinclude:: /../../providers/tests/system/google/cloud/vertex_ai/example_vertex_ai_feature_store.py
:language: python
:dedent: 4
:start-after: [START how_to_cloud_vertex_ai_feature_store_feature_view_sync_sensor]
:end-before: [END how_to_cloud_vertex_ai_feature_store_feature_view_sync_sensor]

Reference
^^^^^^^^^

Expand Down
2 changes: 1 addition & 1 deletion generated/provider_dependencies.json
Original file line number Diff line number Diff line change
Expand Up @@ -646,7 +646,7 @@
"google-api-python-client>=2.0.2",
"google-auth-httplib2>=0.0.1",
"google-auth>=2.29.0",
"google-cloud-aiplatform>=1.70.0",
"google-cloud-aiplatform>=1.73.0",
"google-cloud-automl>=2.12.0",
"google-cloud-batch>=0.13.0",
"google-cloud-bigquery-datatransfer>=3.13.0",
Expand Down
Original file line number Diff line number Diff line change
@@ -0,0 +1,147 @@
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership. The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License. You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied. See the License for the
# specific language governing permissions and limitations
# under the License.
"""This module contains a Google Cloud Vertex AI Feature Store hook."""

from __future__ import annotations

from google.api_core.client_options import ClientOptions
from google.cloud.aiplatform_v1beta1 import (
FeatureOnlineStoreAdminServiceClient,
)

from airflow.exceptions import AirflowException
from airflow.providers.google.common.consts import CLIENT_INFO
from airflow.providers.google.common.hooks.base_google import PROVIDE_PROJECT_ID, GoogleBaseHook


class FeatureStoreHook(GoogleBaseHook):
"""
Hook for interacting with Google Cloud Vertex AI Feature Store.
This hook provides an interface to manage Feature Store resources in Vertex AI,
including feature views and their synchronization operations. It handles authentication
and provides methods for common Feature Store operations.
:param gcp_conn_id: The connection ID to use for connecting to Google Cloud Platform.
Defaults to 'google_cloud_default'.
:param impersonation_chain: Optional service account to impersonate using short-term
credentials. Can be either a single account or a chain of accounts required to
get the access_token of the last account in the list, which will be impersonated
in the request. If set as a string, the account must grant the originating account
the Service Account Token Creator IAM role. If set as a sequence, the identities
from the list must grant Service Account Token Creator IAM role to the directly
preceding identity, with first account from the list granting this role to the
originating account.
"""

def get_feature_online_store_admin_service_client(
self,
location: str | None = None,
) -> FeatureOnlineStoreAdminServiceClient:
"""
Create and returns a FeatureOnlineStoreAdminServiceClient object.
This method initializes a client for interacting with the Feature Store API,
handling proper endpoint configuration based on the specified location.
:param location: Optional. The Google Cloud region where the service is located.
If provided and not 'global', the client will be configured to use the
region-specific API endpoint.
"""
if location and location != "global":
client_options = ClientOptions(api_endpoint=f"{location}-aiplatform.googleapis.com:443")
else:
client_options = ClientOptions()
return FeatureOnlineStoreAdminServiceClient(
credentials=self.get_credentials(), client_info=CLIENT_INFO, client_options=client_options
)

def get_feature_view_sync(
self,
location: str,
feature_view_sync_name: str,
) -> dict:
"""
Retrieve the status and details of a Feature View synchronization operation.
This method fetches information about a specific feature view sync operation,
including its current status, timing information, and synchronization metrics.
:param location: The Google Cloud region where the feature store is located
(e.g., 'us-central1', 'us-east1').
:param feature_view_sync_name: The full resource name of the feature view
sync operation to retrieve.
"""
client = self.get_feature_online_store_admin_service_client(location)

try:
response = client.get_feature_view_sync(name=feature_view_sync_name)

report = {
"name": feature_view_sync_name,
"start_time": int(response.run_time.start_time.seconds),
}

if hasattr(response.run_time, "end_time") and response.run_time.end_time.seconds:
report["end_time"] = int(response.run_time.end_time.seconds)
report["sync_summary"] = {
"row_synced": int(response.sync_summary.row_synced),
"total_slot": int(response.sync_summary.total_slot),
}

return report

except Exception as e:
self.log.error("Failed to get feature view sync: %s", str(e))
raise AirflowException(str(e))

@GoogleBaseHook.fallback_to_default_project_id
def sync_feature_view(
self,
location: str,
feature_online_store_id: str,
feature_view_id: str,
project_id: str = PROVIDE_PROJECT_ID,
) -> str:
"""
Initiate a synchronization operation for a Feature View.
This method triggers a sync operation that updates the online serving data
for a feature view based on the latest data in the underlying batch source.
The sync operation ensures that the online feature values are up-to-date
for real-time serving.
:param location: The Google Cloud region where the feature store is located
(e.g., 'us-central1', 'us-east1').
:param feature_online_store_id: The ID of the online feature store that
contains the feature view to be synchronized.
:param feature_view_id: The ID of the feature view to synchronize.
:param project_id: The ID of the Google Cloud project that contains the
feature store. If not provided, will attempt to determine from the
environment.
"""
client = self.get_feature_online_store_admin_service_client(location)
feature_view = f"projects/{project_id}/locations/{location}/featureOnlineStores/{feature_online_store_id}/featureViews/{feature_view_id}"

try:
response = client.sync_feature_view(feature_view=feature_view)

return str(response.feature_view_sync)

except Exception as e:
self.log.error("Failed to sync feature view: %s", str(e))
raise AirflowException(str(e))
Original file line number Diff line number Diff line change
@@ -0,0 +1,163 @@
#
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership. The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License. You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied. See the License for the
# specific language governing permissions and limitations
# under the License.
"""This module contains Google Vertex AI Feature Store operators."""

from __future__ import annotations

from collections.abc import Sequence
from typing import TYPE_CHECKING, Any

from airflow.providers.google.cloud.hooks.vertex_ai.feature_store import FeatureStoreHook
from airflow.providers.google.cloud.operators.cloud_base import GoogleCloudBaseOperator

if TYPE_CHECKING:
from airflow.utils.context import Context


class SyncFeatureViewOperator(GoogleCloudBaseOperator):
"""
Initiate a synchronization operation for a Feature View in Vertex AI Feature Store.
This operator triggers a sync operation that updates the online serving data for a feature view
based on the latest data in the underlying batch source. The sync operation ensures that
the online feature values are up-to-date for real-time serving.
:param project_id: Required. The ID of the Google Cloud project that contains the feature store.
This is used to identify which project's resources to interact with.
:param location: Required. The location of the feature store (e.g., 'us-central1', 'us-east1').
This specifies the Google Cloud region where the feature store resources are located.
:param feature_online_store_id: Required. The ID of the online feature store that contains
the feature view to be synchronized. This store serves as the online serving layer.
:param feature_view_id: Required. The ID of the feature view to synchronize. This identifies
the specific view that needs to have its online values updated from the batch source.
:param gcp_conn_id: The connection ID to use for connecting to Google Cloud Platform.
Defaults to 'google_cloud_default'.
:param impersonation_chain: Optional service account to impersonate using short-term
credentials. Can be either a single account or a chain of accounts required to
get the access_token of the last account in the list, which will be impersonated
in the request. If set as a string, the account must grant the originating account
the Service Account Token Creator IAM role. If set as a sequence, the identities
from the list must grant Service Account Token Creator IAM role to the directly
preceding identity, with first account from the list granting this role to the
originating account.
"""

template_fields: Sequence[str] = (
"project_id",
"location",
"feature_online_store_id",
"feature_view_id",
)

def __init__(
self,
*,
project_id: str,
location: str,
feature_online_store_id: str,
feature_view_id: str,
gcp_conn_id: str = "google_cloud_default",
impersonation_chain: str | Sequence[str] | None = None,
**kwargs,
) -> None:
super().__init__(**kwargs)
self.project_id = project_id
self.location = location
self.feature_online_store_id = feature_online_store_id
self.feature_view_id = feature_view_id
self.gcp_conn_id = gcp_conn_id
self.impersonation_chain = impersonation_chain

def execute(self, context: Context) -> str:
"""Execute the feature view sync operation."""
self.hook = FeatureStoreHook(
gcp_conn_id=self.gcp_conn_id,
impersonation_chain=self.impersonation_chain,
)
self.log.info("Submitting Feature View sync job now...")
response = self.hook.sync_feature_view(
project_id=self.project_id,
location=self.location,
feature_online_store_id=self.feature_online_store_id,
feature_view_id=self.feature_view_id,
)
self.log.info("Retrieved Feature View sync: %s", response)

return response


class GetFeatureViewSyncOperator(GoogleCloudBaseOperator):
"""
Retrieve the status and details of a Feature View synchronization operation.
This operator fetches information about a specific feature view sync operation,
including its current status, timing information, and synchronization metrics.
It's typically used to monitor the progress of a sync operation initiated by
the SyncFeatureViewOperator.
:param location: Required. The location of the feature store (e.g., 'us-central1', 'us-east1').
This specifies the Google Cloud region where the feature store resources are located.
:param feature_view_sync_name: Required. The full resource name of the feature view
sync operation to retrieve. This is typically the return value from a
SyncFeatureViewOperator execution.
:param gcp_conn_id: The connection ID to use for connecting to Google Cloud Platform.
Defaults to 'google_cloud_default'.
:param impersonation_chain: Optional service account to impersonate using short-term
credentials. Can be either a single account or a chain of accounts required to
get the access_token of the last account in the list, which will be impersonated
in the request. If set as a string, the account must grant the originating account
the Service Account Token Creator IAM role. If set as a sequence, the identities
from the list must grant Service Account Token Creator IAM role to the directly
preceding identity, with first account from the list granting this role to the
originating account.
"""

template_fields: Sequence[str] = (
"location",
"feature_view_sync_name",
)

def __init__(
self,
*,
location: str,
feature_view_sync_name: str,
gcp_conn_id: str = "google_cloud_default",
impersonation_chain: str | Sequence[str] | None = None,
**kwargs,
) -> None:
super().__init__(**kwargs)
self.location = location
self.feature_view_sync_name = feature_view_sync_name
self.gcp_conn_id = gcp_conn_id
self.impersonation_chain = impersonation_chain

def execute(self, context: Context) -> dict[str, Any]:
"""Execute the get feature view sync operation."""
self.hook = FeatureStoreHook(
gcp_conn_id=self.gcp_conn_id,
impersonation_chain=self.impersonation_chain,
)
self.log.info("Retrieving Feature View sync job now...")
response = self.hook.get_feature_view_sync(
location=self.location, feature_view_sync_name=self.feature_view_sync_name
)
self.log.info("Retrieved Feature View sync: %s", self.feature_view_sync_name)
self.log.info(response)

return response
Original file line number Diff line number Diff line change
@@ -0,0 +1,16 @@
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership. The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License. You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied. See the License for the
# specific language governing permissions and limitations
# under the License.
Loading

0 comments on commit 16022e0

Please sign in to comment.