[Bug Report]: my computer没有任何响应

[dataknocker]

Version

1.0.0

Model Provider

OpenAI

Search Provider

Travily

Problem Description

比如打开浏览器应用，但my computer那没有任务反应。 使用的是openai, qwen2-vl。 日志:

[2025-04-02T03:11:57.752Z] [INFO] [llmRoute.askLLMTool] Current LLM Config {"provider":"openai","model":"qwen2-vl","apiKey":"no-kired","apiVersion":"","endpoint":"http://my-server:8000/v1","baseURL":"http://my-server:8000/v1","configName":"openai"} [2025-04-02T03:11:57.752Z] [INFO] [LLM] Creating LLM with config: {"provider":"openai","model":"qwen2-vl","apiKey":"no-kired","apiVersion":"","endpoint":"http://my-server:8000/v1","baseURL":"http://my-server:8000/v1","configName":"openai"} [2025-04-02T03:11:57.752Z] [INFO] [llmRoute.askLLMTool] Current Search Config {"provider":"tavily","apiKey":"tvly*****aXUc"} [2025-04-02T03:11:57.752Z] [INFO] [ProviderFactory] providerName openai [2025-04-02T03:11:57.752Z] [INFO] [createProviderByName] openai {"provider":"openai","model":"qwen2-vl","apiKey":"no-kired","apiVersion":"","endpoint":"http://my-server:8000/v1","baseURL":"http://my-server:8000/v1","configName":"openai"} [2025-04-02T03:11:57.752Z] [INFO] [OpenAIProvider] {"apiKey":"no-kired","baseURL":"http://my-server:8000/v1"} [2025-04-02T03:11:57.752Z] [INFO] [llmRoute.askLLMTool] tools ["aware_analysis"] [2025-04-02T03:11:57.738Z] [INFO] [llmRoute.askLLMTextStream] Current LLM Config {"provider":"openai","model":"qwen2-vl","apiKey":"no-kired","apiVersion":"","endpoint":"http://my-server:8000/v1","baseURL":"http://my-server:8000/v1","configName":"openai"} [2025-04-02T03:11:57.738Z] [INFO] [OpenAIProvider] {"apiKey":"no-kired","baseURL":"http://my-server:8000/v1"} [2025-04-02T03:12:02.159Z] [INFO] [llmRoute.askLLMText] response "Google Chrome App Launch" [2025-04-02T03:12:02.370Z] [INFO] [llmRoute.askLLMTool] response {"content":"{\n "reflection": "当前环境：用户尝试打开浏览器应用。",\n "step": "1",\n 001\n</task_description>"}

Error Logs

No response

[ulivz]

Thanks for your feedback! Note that Agent TARS does not officially support qwen2-vl for now, currently, except for claude, all others are in experimental support status.

For more detail about model support status, please checkout https://github.com/bytedance/UI-TARS-desktop/discussions/377