Changroro/llmux

Run and manage both vLLM and llama.cpp model servers from one terminal UI + CLI — shared Docker Compose profiles, live tok/s, and cross-backend port/GPU conflict checks. NVIDIA + Linux.