Train a model to classify document relevance
Generate text responses based on prompts
Generate images using text and reference images
