Skip to main content

​
UtopikAI’s approach to embedding models

Embedding models convert documents and queries into vectors so UtopikAI can retrieve the right knowledge. They are used for knowledge search and indexing connected sources. They are not the same as LLMs, which generate answers in chat, assistants, agents, and workflows. Which options you see depends on your tenant.

​
Where embeddings are used

Each assistant and each agent has its own knowledge (document sets). The embedding model used to index that knowledge is set at the workspace / search level.

​
Where models are configured

  • Tenant admin — Control which embedding models are available, including filtering the catalog.
  • Knowledge search — Select the embedding model used to index and retrieve connected sources.
Embedding models are configured at the workspace / knowledge-search level. They are not chosen per chat message the way an assistant or agent LLM is. For LLMs used in chat, assistants, agents, and workflows, see LLM Providers and the UtopikAI LLM models page.