observability-alert-noise-analyzer

What problem does it solve? On-call engineers drowning in alert fatigue need a systematic way to identify which CloudWatch alarms are noise—duplicates, flapping thresholds, and correlated clusters—so they can reduce alert volume without missing real incidents. ## Core Features & Use Cases - Duplicate Detection: Identifies exact duplicate alarms monitoring the same metric, namespace, and dimensions, distinguishing them from legitimate tiered warning/critical pairs. - Flapping & Correlation Analysis: Detects alarms with more than 6 state transitions in 24 hours and clusters alarms that fire within 5 minutes of each other, including cross-domain incidents. - Noise Score & Recommendations: Produces a quantified noise score (LOW/MEDIUM/HIGH) with dedup candidates, snooze window suggestions, and escalation reports. - Use Case: After an incident where an on-call engineer received 50+ alerts, run this analysis to find that three alarms monitor the same RDS metric and two others flap nightly, then hand the cleanup list to the observability team. ## Quick Start Ask the assistant to analyze CloudWatch alarm noise in your AWS region over the last 24 hours and report duplicates, flapping alarms, and a noise score.