A framework for generating realistic LLM serving workloads - View it on GitHub