# vLLM-Omni CLI Guide

The CLI for vLLM-Omni inherits from vllm with some additional arguments.

## serve

Starts the vLLM-Omni OpenAI Compatible API server.

Start with a model:

```bash
vllm serve Qwen/Qwen2.5-Omni-7B --omni
```

Specify the port:

```bash
vllm serve Qwen/Qwen2.5-Omni-7B --omni --port 8091
```

If you have custom stage configs file, launch the server with command below
```bash
vllm serve Qwen/Qwen2.5-Omni-7B --omni --stage-configs-path /path/to/stage_configs_file
```