Amazon Nova Launches Multimodal Embeddings Model

Amazon published documentation for the Nova Multimodal Embeddings model, describing how it generates embeddings for text, images, audio, and video to support search, RAG, and recommendation use cases. It explains embeddingPurpose parameter modes—retrieval system (GENERIC_INDEX, GENERIC_RETRIEVAL, TEXT_RETRIEVAL, etc.) and ML task (CLASSIFICATION, CLUSTERING)—and provides architecture and walkthroughs for tuning and deployment.