# vLLM-omni: Multi-modal Extension for vLLM

vLLM-omni is designed to extend vLLM capabilities to support multi-modality model inference and serving, particularly focusing on non-autoregressive structures and non-textual outputs.