每次打开网页后，都卡在第二个look中，报错不是视觉模型

[11268211]

Bug description

Bug solved method

Environment information

• System version:windows11
• Python version:3.12
• OpenManus version or branch:main
• Installation method (e.g., pip install -r requirements.txt or pip install -e .):

Screenshots or logs
2025-03-19 15:23:19.435 | INFO | app.agent.toolcall:think:78 - 🛠️ Manus selected 1 tools to use
2025-03-19 15:23:19.435 | INFO | app.agent.toolcall:think:82 - 🧰 Tools being prepared: ['browser_use']
2025-03-19 15:23:19.436 | INFO | app.agent.toolcall:think:85 - 🔧 Tool arguments: {"action": "open_tab", "url": "https://www.baidu.com"}
2025-03-19 15:23:19.438 | INFO | app.agent.toolcall:execute_tool:177 - 🔧 Activating tool: 'browser_use'...
2025-03-19 15:23:25.594 | INFO | app.agent.toolcall:act:147 - 🎯 Tool 'browser_use' completed its mission! Result: Observed output of cmd browser_use executed:
Opened new tab with https://www.baidu.com
2025-03-19 15:23:25.595 | INFO | app.agent.base:run:140 - Executing step 2/20
2025-03-19 15:23:27.971 | ERROR | app.llm:ask_tool:725 - OpenAI API error: Error code: 400 - {'code': 20041, 'message': 'The model is not a VLM (Vision Language Model). Please use text-only prompts.', 'data': None}
2025-03-19 15:23:27.976 | ERROR | app.llm:ask_tool:731 - API error: Error code: 400 - {'code': 20041, 'message': 'The model is not a VLM (Vision Language Model). Please use text-only prompts.', 'data': None}
2025-03-19 15:23:29.137 | ERROR | app.llm:ask_tool:725 - OpenAI API error: Error code: 400 - {'code': 20041, 'message': 'The model is not a VLM (Vision Language Model). Please use text-only prompts.', 'data': None}
2025-03-19 15:23:29.138 | ERROR | app.llm:ask_tool:731 - API error: Error code: 400 - {'code': 20041, 'message': 'The model is not a VLM (Vision Language Model). Please use text-only prompts.', 'data': None}
2025-03-19 15:23:30.529 | ERROR | app.llm:ask_tool:725 - OpenAI API error: Error code: 400 - {'code': 20041, 'message': 'The model is not a VLM (Vision Language Model). Please use text-only prompts.', 'data': None}
2025-03-19 15:23:30.529 | ERROR | app.llm:ask_tool:731 - API error: Error code: 400 - {'code': 20041, 'message': 'The model is not a VLM (Vision Language Model). Please use text-only prompts.', 'data': None}
2025-03-19 15:23:33.123 | ERROR | app.llm:ask_tool:725 - OpenAI API error: Error code: 400 - {'code': 20041, 'message': 'The model is not a VLM (Vision Language Model). Please use text-only prompts.', 'data': None}
2025-03-19 15:23:33.123 | ERROR | app.llm:ask_tool:731 - API error: Error code: 400 - {'code': 20041, 'message': 'The model is not a VLM (Vision Language Model). Please use text-only prompts.', 'data': None}
2025-03-19 15:23:40.960 | ERROR | app.llm:ask_tool:725 - OpenAI API error: Error code: 400 - {'code': 20041, 'message': 'The model is not a VLM (Vision Language Model). Please use text-only prompts.', 'data': None}
2025-03-19 15:23:40.961 | ERROR | app.llm:ask_tool:731 - API error: Error code: 400 - {'code': 20041, 'message': 'The model is not a VLM (Vision Language Model). Please use text-only prompts.', 'data': None}

[fred913]

请提供下你的 config.toml 中 llm.vision 部分填写的是什么 model

[ECat4]

请提供下你的 config.toml 中 llm.vision 部分填写的是什么 model

我也一样

[11268211]

[llm.vision]
model = "Qwen/Qwen2.5-VL-72B-Instruct" # The vision model to use
base_url = "https://api.siliconflow.cn/v1/" # API endpoint URL for vision model

其中browser use工具提示参数错误的问题也总是发生
llm用的Qwen/Qwen2.5-72B-Instruct-128K
不知道是不是模型的原因

[lyy1988323]

请提供下你的 config.toml 中 llm.vision 部分填写的是什么 model

我也一样

我好像找到原因了，要把下面的max_token和temperture删掉

[ECat4]

请提供下你的 config.toml 中 llm.vision 部分填写的是什么 model

我也一样

我好像找到原因了，要把下面的max_token和temperture删掉

跟这个没关系，我发现不同的时候就去掉了，还是一样，不知道是代码里有bug没引用到，还是因为模型的问题

[AbdullaevG]

@ECat4 were you able to solve the problem? My vlm model doesn't work at all, requests go only to llm although the task is related to pictures

[ECat4]

@ECat4 were you able to solve the problem? My vlm model doesn't work at all, requests go only to llm although the task is related to pictures

no, i give up this shit

[hczs]

The api of siliconflow has request restrictions. Just replace it with, for example, the official api of DeepSeek

[github-actions]

This issue has been inactive for 30 days. Please comment if you have updates.