Merge pull request #43 from teamdatatonic/feat/extending-model-monito…

…ring-config

Extending model monitoring config

components/src/components/model_batch_predict.py

@@ -38,6 +38,7 @@ def model_batch_predict(
     max_replica_count: int = 1,
     monitoring_training_dataset: dict = None,
     monitoring_alert_email_addresses: List[str] = None,
+    notification_channels: List[str] = [],
     monitoring_skew_config: dict = None,
     instance_config: dict = None,
 ):
@@ -62,6 +63,9 @@ def model_batch_predict(
             https://cloud.google.com/python/docs/reference/aiplatform/latest/google.cloud.aiplatform_v1beta1.types.ModelMonitoringObjectiveConfig.TrainingPredictionSkewDetectionConfig
         monitoring_alert_email_addresses (List[str]):
             Email addresses to send alerts to (optional).
+        notification_channels (List[str]):
+            Notification channels to send alerts to (optional).
+            Format: projects/<project>/notificationChannels/<notification_channel>
         monitoring_training_dataset (dict): Metadata of training dataset. See:
             https://cloud.google.com/python/docs/reference/aiplatform/latest/google.cloud.aiplatform_v1beta1.types.ModelMonitoringObjectiveConfig.TrainingDataset
         instance_config (dict): Configuration defining how to transform batch prediction
@@ -152,7 +156,9 @@ def is_job_successful(job_state: JobState) -> bool:
 
         message["modelMonitoringConfig"] = {
             "alertConfig": {
-                "emailAlertConfig": {"userEmails": monitoring_alert_email_addresses}
+                "emailAlertConfig": {"userEmails": monitoring_alert_email_addresses},
+                "notificationChannels": notification_channels,
+                "enableLogging": True,
             },
             "objectiveConfigs": [
                 {

pipelines/src/pipelines/prediction.py

@@ -24,6 +24,7 @@
 RESOURCE_SUFFIX = env.get("RESOURCE_SUFFIX", "default")
 # set training-serving skew thresholds and emails to receive alerts:
 ALERT_EMAILS = []
+NOTIFICATION_CHANNELS = []
 SKEW_THRESHOLDS = {"defaultSkewThreshold": {"value": 0.001}}
 # or set different thresholds per feature:
 # SKEW_THRESHOLDS = {"skewThresholds": {"payment_type": {"value": 0.001}}, ... }
@@ -109,6 +110,7 @@ def pipeline(
             max_replica_count=max_replicas,
             monitoring_training_dataset=lookup_op.outputs["training_dataset"],
             monitoring_alert_email_addresses=ALERT_EMAILS,
+            notification_channels=NOTIFICATION_CHANNELS,
             monitoring_skew_config=SKEW_THRESHOLDS,
         )
         .after(prep_op)