domino-model-monitoring

Detect production model data drift by comparing input feature distributions against baseline training data to identify changes and trigger alerts. This requires configuring drift tests on your deployed Model API using training data baselines.

Model quality tracking monitors deployed API performance by calculating metrics like accuracy, precision, recall, F1, AUC, and regression error over time. It compares live predictions against actual outcomes to track degradation.

Yes, you can configure alerting rules to notify stakeholders and drive remediation workflows when drift or quality metric thresholds are breached. Alert triggers are defined during the monitoring setup on your deployed Model API.

Yes, production model monitoring requires a deployed Model API and your original training data to establish baselines. You must also configure specific drift tests, quality metrics, and alert rules to begin tracking.

Capture live prediction streams by enabling prediction capture and dataset logging within your Domino Model API monitoring configuration. This logs incoming data and outputs to ensure reproducibility and support investigations.