What it means
An embedding converts text into a point in a high-dimensional space, arranged so that distance corresponds to similarity of meaning. "How do I reset my password?" and "I forgot my login" end up close together despite sharing barely any words.
This is what makes meaning computable. Once text is numbers, you can measure similarity, cluster related documents, detect duplicates, and — most commonly — retrieve passages relevant to a question. Embeddings are the mechanism underneath semantic search and therefore underneath most retrieval systems.
Embeddings are produced by their own models, separate from the generative one. Different embedding models produce incompatible spaces, so vectors from one cannot be compared against vectors from another.
Why it matters
Embeddings are the bridge between unstructured text and anything computational. They are also a quiet lock-in point: changing your embedding model means re-embedding your entire corpus, because old and new vectors are not comparable.
In practice
Choose an embedding model deliberately and expect to live with it, or budget for a full re-index. Store the model identity alongside the vectors so a future migration knows what it is looking at.
Where this shows up
Tools and models in our catalog.
PineconeThe leading managed vector database for AI applications. Serverless pricing, 99.99% SLA, and billions of vectors at millisecond query speeds. Widely used in production RAG systems.
Supabase VectorPostgreSQL-based vector storage using the pgvector extension. Seamlessly combines traditional relational data with vector search in a single database.
Hugging FaceThe hub of open machine learning — hundreds of thousands of models and datasets, the Transformers library, and Spaces for demos. The default platform for finding, sharing, and running AI models.