n8n Embeddings Ollama nodeConfigure Embeddings Ollama in n8n.
Embeddings are the layer that lets an AI compare meaning instead of matching words. The n8n Embeddings Ollama node builds them with a model served by Ollama, exposes a single parameter, and plugs into the embeddings port of a root node. Built for anyone assembling retrieval on their own hardware.
Verified Trustpilot reviews · AI, automation & growth agency
What does the n8n Embeddings Ollama node do?
The Embeddings Ollama node turns text into embeddings, the numerical vectors an AI uses to place pieces of content close to each other when they mean the same thing. It calls a model served by Ollama to produce those vectors. It is a sub-node, meaning it has no main input of its own and never runs alone: it attaches to the embeddings port of a root node, and that root node decides when it fires.
The first scenario is an internal knowledge base you do not want to send anywhere. A document loader splits your handbook into chunks, this node vectorizes each chunk through the model picked in Model, and a vector store, a database built to hold and search vectors, keeps the result. The three published templates that use this node all follow that shape, starting with the local retrieval chatbot built around Ollama and Qdrant.
The second one is question answering on top of that store. The same node has to run again at query time, because a question can only be compared to your chunks if both were embedded by the same model. Wire it to the store used by an AI Agent, and the agent retrieves passages instead of guessing from its training data.
The third is a workflow split across machines. Retrieval stays local through Ollama while the conversation itself runs on a hosted chat model, the pattern used by the template that pairs this node with an OpenAI chat model. Vectors never leave your infrastructure, only prompts go out.
Pick a different embeddings sub-node when you have no Ollama server to point at. This node only speaks to Ollama, so a provider that bills its own API calls is the alternative to weigh against running and maintaining the model yourself. Our n8n review covers the self-hosting side of that choice, and the n8n training walks through building the store itself.
Two limits are worth knowing before you build. The node exposes one parameter and nothing else, so batch size, timeouts and retries are not yours to tune here. And the model choice is sticky: change it later and every vector already stored was produced with different dimensions, which means reindexing the whole collection.
What do you set up before the first run?
- 01
Pull the model into Ollama
The node does not download anything for you. Get the embedding model onto your Ollama server first, from the Ollama models library that the node description points to. Until the model is there, the Model parameter has nothing real behind it and the run fails on the first chunk you try to vectorize.
- 02
Create the Ollama credential in n8n
This node authenticates against your Ollama server, so it needs a credential. Create it once from the Credentials menu in n8n, following the Ollama credential page in the n8n documentation for the exact values to enter. Once saved, it is reusable: every workflow that touches Ollama points at that same credential instead of redeclaring it.
- 03
Attach the credential and wire the port
Drop the node on the canvas, select the credential you just created, then choose the model in Model. Last step is the connection itself: drag the
ai_embeddingoutput up to the embeddings slot of a root node. A sub-node with no parent stays inert, whatever its credential says.
Where does this sub-node plug in?
n8n splits AI work between a root node that receives the workflow items and sub-nodes attached to its ports, one type of sub-node per port. This node is a sub-node, so it only has an output.
Output (what it plugs into)
- Embeddings
ai_embedding
01Embeddings
This is the port the node offers to a root node that needs text turned into vectors. Drag the connection up to the embeddings slot of the parent, and the parent calls it whenever it has content to index or a query to match. Nothing else on the canvas can reach this node.
Key parameters
- Required: a sub-node never executes by itself, so an Embeddings Ollama node left unconnected stays inert and no run will ever touch it.
Which parameter do you actually set?
The Embeddings Ollama node has one parameter. For each one: the node as you configure it in n8n, what the parameter changes, and our field notes.
Model
modelWhat you see in n8n
Notes & use cases
The node has exactly one setting: which Ollama model generates the embedding. The choice decides the shape of every vector you produce, so it is worth making before you index anything.
Key parameters
- Model: required, picked from a list.
all-minilmreturns vectors of 384 dimensions, the lighter option.nomic-embed-textreturns 768 dimensions.
all-minilm stays small and quick to search; a knowledge base where wording matters gets more room to work with nomic-embed-text. Models are pulled from the Ollama models library first, otherwise the node has nothing to call.Need help automating Embeddings Ollama with n8n?
A person reads every message.
Questions people ask next
01Is the Embeddings Ollama node included in n8n?
02What credentials does the n8n Embeddings Ollama node need?
03What are the limits of this node?
04What do you need to connect at minimum, and which node should you start with?
05n8n or Make for a local embeddings pipeline?
Get our weekly integration tips.
No spam. Unsubscribe anytime.
