Skip to main content
Embeddings models convert text into numerical vectors — a representation that enables semantic search. In Studio they primarily power Knowledge Base — finding relevant document passages to answer questions (RAG). Long-term conversation memory is configured under the agent’s Memory step and uses Neuralyzer or CoD Summarizer at runtime, not an embeddings picker from this screen.

​
Adding an embeddings model

Click Add Embeddings Model:
FieldDescription
NameDisplay name used when selecting embeddings in agent creation
Class PathThe LangChain class the Runtime uses to call the embeddings service
Model URLThe endpoint URL for the embeddings service

​
Common class paths

ProviderClass Path
Ollama (local)langchain_community.embeddings.OllamaEmbeddings
OpenAIlangchain_openai.OpenAIEmbeddings
Coherelangchain_cohere.CohereEmbeddings
HuggingFacelangchain_community.embeddings.HuggingFaceEmbeddings

​
Ollama embeddings (local)

If you started the stack with --profile with-local-models and included an embed model in OLLAMA_MODELS, it’s already registered: 
OLLAMA_MODELS=qwen2.5:0.5b nomic-embed-text
The nomic-embed-text model is registered automatically as an embeddings model. The others are registered as chat models. For manually added Ollama embeddings:
  • Class Path: langchain_community.embeddings.OllamaEmbeddings
  • Model URL: http://ollama:11434 (inside Docker network) or http://localhost:11434 (from host)
nomic-embed-text is an excellent default for local development. For production, OpenAI’s text-embedding-3-small or Cohere’s embed-english-v3.0 offer better quality.

​
Required for Knowledge Base

An agent cannot use the Knowledge Base feature without an embeddings model configured here and selected for that agent. If Knowledge Base in agent creation shows no embeddings model selector, return here and add one first. For long-term memory, configure Memory instead.

​
Next steps

Knowledge Base

Configure document retrieval (RAG) for an agent.

Memory

Configure short-term or long-term conversation memory (Neuralyzer or CoD Summarizer).