Amazon’s fully managed service, Bedrock, facilitates the building and scaling of generative AI applications with foundation models from leading providers. Integrated with Amazon CloudWatch, it provides near real-time monitoring and usage analytics. Bedrock also introduces logging in preview, enabling collection of metadata, requests, and responses for all model invocations. Bedrock’s real-time metrics can be utilized to set alarms, detect anomalies, compare latency between models, measure token count, alert on throttling, and various other use cases.