The local QWEN3 model reports an error when encountering an Invalid URL (POST /v1/responses)

### What is your question?

I used the locally deployed qwen3 model, but an error occurred during operation. The error message is as follows:

`Error 3: Error code: 404 - {'error': {'message': 'Invalid URL (POST /v1/responses)', 'type': 'invalid_request_error', 'param': '', 'code': ''}}`

How to solve it

### Additional information that you believe is relevant to this question

trae_config.json

`{
"default_provider": "openai",
"max_steps": 20,
"enable_lakeview": true,
"model_providers": {
"openai": {
"api_key": "sk-po0xFHDmaSHJdJQO61978",
"base_url": "https://mgallery.xxx.net/v1",
"model": "qwen3-235b-a22b",
"max_tokens": 128000,
"temperature": 0.5,
"top_p": 1,
"max_retries": 10
}
},
"lakeview_config": {
"model_provider": "openai",
"model_name": "qwen3-235b-a22b"
}
}`

Start with the trae_config.json provider settings and reproduce the local QWEN3 request that returns 404 for POST /v1/responses. Trace how the configured base_url and model provider determine that request path; done means the configured local model accepts the request without the Invalid URL error.

Written by the indexing model from the issue text.