llmux 0.5.0

Zero-reload model switching for vLLM - manages multiple models on shared GPU
Documentation