SUPERWISE Platform Policies

SUPERWISE Platform Policies provides automated governance capabilities for AI model monitoring and management. The product enables users to define monitoring policies, set alert thresholds, and automate response actions for AI systems.

The platform supports multiple policy types including static thresholds for fixed boundaries, moving average thresholds that adapt based on historical patterns, distribution comparison using statistical distance functions, and scheduled evaluation with custom cron expressions. Policies can be created through both SDK and UI interfaces.

The system evaluates policies on user-defined schedules and changes status indicators (healthy, unhealthy, pending, no_data) based on threshold violations. When violations are detected, the platform generates alerts with full incident context and routes them to configured channels.

Alert routing supports multi-channel notifications including Slack, email, PagerDuty, OpsGenie, webhooks, and Datadog. Users can configure different alert channels based on severity levels.

The product integrates with other SUPERWISE Platform capabilities including Guardrails for runtime protection and Observability for metrics evaluation. Policies can monitor metrics such as model accuracy, data drift, and other operational health indicators.

Configuration options include setting metrics to monitor, defining static or dynamic thresholds, specifying evaluation schedules, and configuring alert destinations. The platform maintains historical context for incident investigation and provides consistent policy application across multiple models.