Embedding Space

Embedding space is the high-dimensional vector space (e.g., hundreds or thousands of dimensions) in which embeddings are represented. Points that are close in this space correspond to text with similar meaning; vector search finds neighbors in embedding space (e.g., via cosine similarity).

In DocLD

DocLD uses Pinecone with integrated embeddings; chunks and queries are embedded into the same space so retrieval returns semantically relevant passages. Dimensionality is determined by the embedding model.

Related Concepts

Embedding space is where embeddings and vectors live. Vector search and semantic search operate in this space; dimensionality and the choice of embedding model affect quality.