Claude Code – Disabling telemetry also disables 1-hour prompt cache TTL (github.com)

🤖 AI Summary
A bug report has highlighted an issue with Claude Code regarding prompt cache Time-to-Live (TTL) settings when telemetry is disabled. Users have discovered that when they opt out of telemetry—using parameters like DISABLE_TELEMETRY=1—sessions that should normally receive a 1-hour prompt cache TTL instead default to a 5-minute TTL. This inconsistency impacts the performance and efficiency of sessions for developers relying on optimal caching configurations. The significance of this bug lies in the potential disruption it causes for AI/ML practitioners who depend on Claude Code for high-performance applications. The ability to maintain a longer cache TTL is crucial for reducing latency and enhancing responsiveness in AI applications. Fixing this issue will ensure that cache tier selection operates independently from telemetry settings, thus providing users the benefits of the 1-hour prompt cache without telemetry interference. This enhancement could lead to improved user experiences and more effective resource management in AI-powered solutions.
Loading comments...
loading comments...