Developed and delivered a Reasoning Output Token Metrics feature for the DataDog/dd-trace-js repository, enhancing the OpenAI plugin’s ability to track token usage during reasoning tasks. The work involved integrating new metric instrumentation across multiple code paths and ensuring seamless reporting through existing dashboards. Comprehensive unit and integration tests were added to validate accurate capture and reporting of these metrics, supporting improved observability and cost optimization for LLM-powered workflows. The implementation leveraged JavaScript and Node.js, with a focus on API integration and full stack development practices. This contribution deepened the plugin’s monitoring capabilities without introducing any new bugs.