Embeddings
Embeddings are numerical representations of text, images, or other data — lists of numbers positioned so that similar meanings sit close together. They let software compare things by meaning instead of exact words, and they are what a vector database stores and searches.
An embedding model turns a piece of content into a fixed-length vector. Content with similar meaning lands near other similar content in that space, so 'cancel my subscription' sits close to 'end my plan' even with no shared words. This is the quiet foundation under semantic search, RAG, recommendations, clustering, and deduplication — anywhere you need a machine to judge how alike two things are.
When building a product, the choice of embedding model is one of the highest-leverage and most overlooked decisions: it sets the ceiling on retrieval quality, so a weak model means RAG fails no matter how good the prompt is. Practical details matter too — different content types may need different models, multilingual content needs a multilingual model, and you must re-embed everything if you change models. We pick the embedding model deliberately and measure it, because it quietly determines whether everything downstream works.
// faq
Frequently asked questions
- What is the difference between embeddings and a vector database?
- Embeddings are the numbers that represent meaning; a vector database is the system that stores those numbers and searches them quickly. You generate embeddings with an embedding model, then put them in a vector database to retrieve the closest matches. One is the representation, the other is the storage-and-search engine around it.