Embeddings

Dense vectors represent text, images, or code chunks so nearest-neighbor search ranks semantic similarity. Embedding models run at index time and sometimes query time inside RAG stacks. Normalization and distance metrics affect recall and latency trade-offs.