Vision and Multimodal Support

AgentForge supports image and multimodal prompts for APIs that implement vision capabilities.

VisionMixin

• VisionMixin is a mixin class that adds image/multimodal support to API integrations.
• Used by GeminiVision and can be used in custom APIs.

How to Use

• To use image/multimodal support, select a provider class that supports images, such as GeminiVision or LMStudioVision.
• Pass images to the generate method via the images argument.

Example

response = agent.model.generate(model_prompt, images=[image_data], **params)

Configuration Example

model_library:
  gemini_api:
    GeminiVision:
      models:
        gemini_vision:
          identifier: gemini-3.5-flash
          params:
            temperature: 0.7

Notes

• Not all APIs support images. Check the API documentation for support.
• For custom APIs, inherit from VisionMixin and implement required methods.