Ashton Sidhu developed a safety enhancement feature for the BerriAI/litellm repository, focusing on integrating HiddenLayer Guardrails to improve the security of large language model interactions. Using Python and FastAPI, Ashton implemented guardrail hooks and configurable options that allow the LiteLLM framework to block or redact unsafe content, addressing compliance and user trust concerns. The work included comprehensive user documentation and setup guidance to ensure correct adoption and usage. Although the contribution was limited to a single feature over one month, Ashton’s approach established a technical foundation for future customization and monitoring of safety policies within the backend system.