Retrieval
Retrieval for AI applications
Index content once and use it for semantic search, recommendations and RAG with embeddings generated on QDivZero.
How it fits together
Create an embedding deployment, create a Retrieval database that references its serving name, then index documents with content and metadata.
One index, several experiences
Use the same database for text, image or multimodal search when the embedding model supports it. Record view signals to bias results toward a user's interests.
Independent lifecycle
The index remains a managed QDivZero resource. The embedding deployment must be running when new content or queries need embeddings.