The original title is "Grafana Alerting Checklist: Wiring AI Anomaly Scores Correctly"

Short summary

Static alert thresholds miss gradual issues like slow memory leaks. This 13-point checklist covers wiring ML anomaly models (Prophet, PyOD) into Grafana to detect statistical deviations, from scrape interval matching through synthetic testing. Includes docker-compose stack and alert rules as code.

• •Static thresholds fail at catching slow-burn multivariate drift; anomaly models learn 'normal' and flag deviation
• •13-point checklist: scrape parity, Prometheus retention, label matching, score normalization, contact point testing, synthetic injection
• •Working docker-compose stack and Grafana alert rule template provided for production deployment

Generated with AI, which can make mistakes.