Developed multi-completion support for VLLM and TogetherAI integrations within the safety-research/safety-tooling repository, enabling users to request and receive multiple generated responses per API call. Leveraged Python for backend development, introducing an 'n' parameter to VLLM requests and refining validation logic to ensure the correct number of valid choices are returned for each request. Focused on enhancing API integration workflows, the work improved reliability and user experience by supporting richer multi-response scenarios. The implementation addressed both request handling and response validation, contributing to more robust backend processes and laying groundwork for future extensibility in safety tooling environments.