Developed an enhancement for the jeejeelee/vllm repository, enabling the Automatic Speech Recognition engine to abort ongoing requests when cancellations occur. This feature was implemented as a frontend change using Python, with a focus on asynchronous programming and backend development. By allowing in-flight ASR requests to be terminated promptly, the update improved system responsiveness and reduced unnecessary compute resource usage during cancellation scenarios. The work emphasized robust testing and collaborative development, as reflected in signed commits and multiple co-authors. This targeted improvement addressed cancellation latency, resulting in a more efficient and user-friendly ASR workflow within the project’s codebase.