Skip to main content

Embedding Models

An embedding model turns text into a numeric vector, used for similarity search in a knowledge base or in a document index node.

Configuration​

From Manager → Settings → AI Models (Embedding Models section):

  • Name — label shown in selectors.
  • Provider — chosen among the Embedding-type providers configured by the administrator (Ollama, OpenAI, Voyage AI or Mistral).
  • Model name — e.g. nomic-embed-text for Ollama.

A "Test" button computes a sample vector with the entered configuration and returns its dimension, to verify the configuration works before using it on a real knowledge base.

Once configured, an embedding model becomes selectable when creating a knowledge base or in a document index node. If no suitable provider appears in the selector, ask an administrator to configure one from Providers.