llmux 0.4.0

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