[Bug]: qwen3-tts 在线服务部分按照readme操作之后，调不通服务400 Bad Request

[laohua-wang]

Your current environment

The output of python collect_env.py

Your output of `python collect_env.py` here

Your code version

The commit id or version of vllm

The commit id or version of vllm-omni

🐛 Describe the bug

qwen3-tts部署好服务后，报错：400 Bad Request 请求体按照文档写的

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[Gaohan123]

@linyueqian Please take a look. Thanks!

[linyueqian]

Hi @laohua-wang, I just tested the endpoint locally and it works correctly with the README example. To help debug your issue, could you please provide:

1. The exact curl command or request you're using
2. The full error response body (not just the status code)
3. Your vllm-omni version (pip show vllm-omni)

[kylehh]

I found a bug in the online example script and fixed it in #1036
Or you can specify model name using -m option in the example script

[laohua-wang]

Hi @laohua-wang, I just tested the endpoint locally and it works correctly with the README example. To help debug your issue, could you please provide:

1. The exact curl command or request you're using
2. The full error response body (not just the status code)
3. Your vllm-omni version (pip show vllm-omni)

Hi, thank you for your reply. My vllm-omni version is 0.14.0rc1.

[laohua-wang]

Hi @laohua-wang, I just tested the endpoint locally and it works correctly with the README example. To help debug your issue, could you please provide:

1. The exact curl command or request you're using
2. The full error response body (not just the status code)
3. Your vllm-omni version (pip show vllm-omni)

Thank you for your reply , I have solved my problem

[laohua-wang]

Hi @laohua-wang, I just tested the endpoint locally and it works correctly with the README example. To help debug your issue, could you please provide:

1. The exact curl command or request you're using
2. The full error response body (not just the status code)
3. Your vllm-omni version (pip show vllm-omni)

I'm a bit confused. Why is the "Voice" parameter necessary when using the cloning function? If it's included, it seems that the model's voice cloning function won't be activated.

[linyueqian]

Hi @laohua-wang, I just tested the endpoint locally and it works correctly with the README example. To help debug your issue, could you please provide:

1. The exact curl command or request you're using
2. The full error response body (not just the status code)
3. Your vllm-omni version (pip show vllm-omni)

I'm a bit confused. Why is the "Voice" parameter necessary when using the cloning function? If it's included, it seems that the model's voice cloning function won't be activated.

I will take a look. The validation logic has some glitch.

[linyueqian]

@laohua-wang I just checked again. voice is not a required parameter for base tasks. there is a route bug because we are taking vllm's default endpoint setting. i will patch a fix soon and you can test again.