Worked on enhancing telemetry for the 0xPlaygrounds/rig repository by implementing comprehensive tracking of cached input tokens across multiple AI providers. Leveraging Rust for backend development, the solution introduced new telemetry fields and updated span instrumentation to record token usage at both provider and agent levels. This approach enabled detailed cost tracking, improved cache-hit observability, and supported more accurate billing decisions for AI services. The work focused on robust integration validation and multi-provider compatibility, laying the foundation for ongoing performance monitoring and cost optimization. The technical depth addressed both operational transparency and the evolving needs of AI provider integrations.