[Feature] Add Gradio Demo for Qwen2.5Omni

[SamitHuang]

PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED.

Purpose

Support gradio web service for Qwen2.5-Omni and Qwen3-Omni

Test Plan

Run

 python gradio_demo.py  --model Qwen/Qwen2.5-Omni-7B --port 7861

, then open http://127.0.0.1:7861/ in a browser

Test Result

Future Updates

TODOs (from demo to app level):

• Support camera recording input
• Maybe save history and support multi-turn dialogues
• Support realtime interact

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft.

BEFORE SUBMITTING, PLEASE READ https://github.com/hsliuustc0106/vllm-omni/blob/main/CONTRIBUTING.md (anything written below this line will be removed by GitHub Actions)

[hsliuustc0106]

I think we can make this gradio function as a universal frontend built under vllm-omni folder,

maybe serving like:

vllm serve xxmodel --display Gradio

[SamitHuang]

Good suggestion. But the UI can vary for different models; each model needs its own build_interface function. We need a clean solution to tell model developers how to link their UI and inference input args to this universal --display gradio arg

[hsliuustc0106]

Offline

[hsliuustc0106]

shall use the bash.sh instead of python script to align with the offline inference,

in addition, it seems offline inference is not paired with Gradio Demo, they should not be placed on the same level.

[hsliuustc0106]

please help delete this section

[SamitHuang]

updated based on #64

[hsliuustc0106]

please align with PR #76 to support multimodal inputs with multiple requests

[SamitHuang]

rebase and update to support text + audio + image + video inputs

[hsliuustc0106]

@codex address that feedback

[Gaohan123]

lgtm, thanks!