Worked on the spring-ai repository to enhance the reliability of usage calculations in the AnthropicChatModel by addressing a critical bug related to token handling. Applied Java development skills to implement robust error handling and API integration, introducing null checks for both input and output tokens. This approach ensured that missing token values defaulted to zero, preventing potential NullPointerExceptions during runtime and maintaining accurate usage metrics for downstream services and billing. The fix improved production stability by safeguarding against runtime errors, ultimately supporting more reliable billing processes and reinforcing defensive programming practices within the codebase over the course of the month.