Alarms on Amazon Bedrock model token use in this account and Region, per model and/or
in total, notifying one SNS target. Token metrics are account-wide: they count every
caller in the account, not only workloads deployed by MDAA. Inference through the
OpenAI-compatible bedrock-mantle endpoint publishes to a different namespace and is
not counted.
Alarms on Amazon Bedrock model token use in this account and Region, per model and/or in total, notifying one SNS target. Token metrics are account-wide: they count every caller in the account, not only workloads deployed by MDAA. Inference through the OpenAI-compatible
bedrock-mantleendpoint publishes to a different namespace and is not counted.Use cases: token-spend spike alerting, detecting runaway agents, usage-pattern deviation alarms
AWS: CloudWatch alarms on AWS/Bedrock InputTokenCount / OutputTokenCount + SNS notification
Validation: requires exactly one notification target and at least one threshold