Worked on the ggml-org/llama.cpp repository, focusing on improving server-side grammar processing by addressing a bug in command line argument handling. Implemented a solution in C++ that ensures the server task defaults to standard grammar values when none are specified, reducing ambiguity and preventing misconfiguration across deployments. The approach involved hardening argument parsing logic to enhance workflow stability and minimize the risk of regression. By refining how grammar-related arguments are processed, the update improved reliability for users managing grammar workflows in server environments. The work demonstrated attention to robust server development practices and careful risk management within the C++ codebase.