Features

• Telemetry: Added support for reporting model thinking text inside a dedicated <trace_think> block, making it easier to distinguish reasoning content from normal response text while keeping the existing trace format compatible.
• Sessions: Added Redis Cluster-backed session and memory services, allowing deployments that use Redis Cluster to persist conversation sessions and long-term memory through the same service interfaces as existing storage backends.
• Storage: Formatted Redis Cluster session and memory service code to keep the new implementation consistent with the repository style.
• Examples/Docs: Added a script for running with uv on macOS and updated the related installation docs and README instructions, making the macOS setup path easier to follow.