Amazon Bedrock Ops Alert is a CloudFormation-based automated monitoring solution for generative AI workloads on Amazon Bedrock. It implements three detection layers: critical error detection (throttles, client/server errors), usage rate monitoring against dynamically calculated RPM/TPM thresholds, and CloudWatch ML-based anomaly detection. When alarms trigger, the system classifies them as quota-related or non-quota, checks for duplicate open support cases, and either creates a new AWS Support case pre-filled with usage-validated context or appends to an existing one. Alarm thresholds automatically recalculate via EventBridge when quotas change, eliminating manual threshold maintenance. Email notifications with case IDs are sent to AI SRE teams after case processing. The solution is deployable via a GitHub repository and requires AWS Business or Enterprise Support for automated case creation.
