Embeddings
Numerical vector representations of text that capture semantic meaning, used for knowledge base retrieval.
Embeddings are a way to represent text as arrays of numbers (vectors) such that semantically similar text produces similar vectors. This mathematical property makes embeddings useful for search: when a caller asks a question, the system converts it to an embedding and finds the closest matching chunks in the knowledge base.
In Stellar, when you upload documents to your knowledge base, the content is split into chunks and each chunk is converted into a 1536-dimensional embedding vector. These vectors are stored in a pgvector database alongside the original text.
During a call, when the AI agent needs to reference your business information, the relevant question is embedded and a similarity search finds the most relevant knowledge base chunks.