AWS introduces Amazon Bedrock Ops Alert, an automated monitoring solution that detects operational issues, adjusts alarm thresholds dynamically, and creates context-aware support cases for generative AI workloads.

Amazon Bedrock Ops Alert is a CloudFormation-based solution that implements three complementary detection layers: Layer 1 monitors throttles, client errors, and server errors for immediate alerting; Layer 2 compares RPM (requests per minute), TPM (tokens per minute), and latency against dynamically calculated thresholds; Layer 3 uses CloudWatch machine learning to identify unusual patterns across metrics.

The solution automatically calculates alarm thresholds by querying the Service Quotas API for current quota values and applying configured percentages, storing thresholds in AWS Systems Manager Parameter Store rather than requiring manual recalculation when quota increases are approved.

Organizations using the solution can reduce operational overhead through duplicate case prevention, context-aware support case automation that equips AWS support engineers with needed information, and contextualized notifications to AI SRE teams—addressing reactive operations, case context enrichment, and manual threshold updates across multiple foundation models.