openclaw - 💡(How to fix) Fix Image model cannot recognize LM Studio deployed models (qwen3-vl) [3 comments, 3 participants]

1. Error: Unknown model anthropic/qwen3-vl-30b-a3b-instruct-mlx
2. System falls back to default provider anthropic, causing the error

Fix Plan

The fix involves updating the resolveAutoImageModel function to correctly handle LM Studio models.

• Update the model catalog to include LM Studio models
• Modify the resolveAutoImageModel function to check for LM Studio models before falling back to the default provider

Example code:

def resolveAutoImageModel(provider, model):
    # Check if provider is lmstudio2
    if provider == 'lmstudio2':
        # Check if model is in LM Studio catalog
        if model in lm_studio_catalog:
            return model
    # Fall back to default provider if not found
    return resolveImageModelFromAgentDefaults(provider, model)

# Update model catalog to include LM Studio models
lm_studio_catalog = {
    'qwen3-vl-30b-a3b-instruct-mlx': 'lmstudio2/qwen3-vl-30b-a3b-instruct-mlx',
    # Add other LM Studio models here
}

Verification

To verify the fix, follow these steps:

• Configure agents.defaults.imageModel.primary to use the LM Studio model
• Try to use the image analysis tool
• Check that the error "Unknown model anthropic/qwen3-vl-30b-a3b-instruct-mlx" is no longer thrown

Extra Tips

• Make sure to update the lm_studio_catalog dictionary with all available LM Studio models
• Consider adding error handling to the resolveAutoImageModel function to handle cases where the model is not found in the catalog.