June 2026 monthly summary for UKGovernmentBEIS/inspect_ai: Delivered enhancements to the SageMaker streaming handler to capture reasoning tokens from thinking-mode models, improving reliability and accuracy of streamed responses with multi-byte support, robust parsing, and comprehensive testing. The changes enable complete reasoning content to be reconstructed and accurately counted for usage metrics, delivering measurable business value and a foundation for deeper model insights.

May 2026 monthly summary for aws-samples/amazon-nova-samples: Implemented Tau-Bench agentic evaluation notebooks supporting local and SageMaker-based evaluation, with multi-domain support (airline, retail, telecom) and configurable sample and message limits. Documentation and workflow automation enhancements completed to streamline end-to-end model evaluation and improve reproducibility. No major bug fixes documented this month; emphasis on feature delivery and workflow stabilization.