What is a vector embedding?
A vector embedding is a list of numbers (often 768 or 1024 dimensions) that represents a piece of text in a way machines can compare. Two texts with similar meaning produce vectors that are close in the high-dimensional space. This is the foundation of semantic search and RAG.
How embeddings are made
A neural network trained on language is run over the input text. The final hidden state (or a pooled version of it) becomes the embedding. The training objective taught the network to place semantically related text close together, so the resulting vectors carry meaning.
Why they beat keyword search
Keyword search misses paraphrase: a query for "reset password" cannot find a doc titled "forgot my login". Embeddings handle paraphrase natively because the underlying meaning maps to nearby vectors regardless of word choice.
We run a self-hosted, open-weight embeddings model on hardware we own. All embedding compute stays on our infrastructure: customer content is never sent to a third-party model. See the RAG concept for the rest of the pipeline.
See it working on message.com.
Unified inbox, grounded AI, flat pricing. 14 days free, no credit card.
Start free trial →