A planned feature is to introduce a unified interface for handling image inputs across different LLM providers. This would abstract away the complexities of dealing with various data formats (e.g., Base64, URLs, local file paths) and model-specific requirements, providing a simple and consistent API for multimodal applications.