lightseekorg / tokenspeed

TokenSpeed is an LLM inference engine built for agentic workloads, with a C++ scheduler control plane and Python execution plane.