Resources · n8n integration

n8n Lemonade Model nodeConfigure Lemonade Model in n8n.

The n8n Lemonade Model node hands a language model to another node instead of running on its own. It exposes 7 parameters, from the model picker down to sampling controls, and plugs into the ai_languageModel port of agents and chains. For builders who want their completions served by a Lemonade server.

Try the Lemonade Model node on n8nSee the 7 parameters

Verified Trustpilot reviews · AI, automation & growth agency

Why automate

What does the n8n Lemonade Model node do in a workflow?

The node supplies a language model to a root node. In n8n, a root node is the one that sits in the main flow and receives items, while a sub-node like this one hangs off one of its ports and only wakes up when the root node asks for a completion. Models here are loaded and managed through the Lemonade server, so the picker lists whatever that server currently serves.

First scenario, a drafting assistant. An AI Agent reads a support ticket and writes a reply, with this node wired to its model port. Sampling Temperature stays low so two similar tickets get two similar answers, and Max Tokens to Generate caps the reply before it turns into an essay.

Second scenario, classification at volume. Rows arrive from Google Sheets, a chain labels each one, and Stop Sequences cuts the output at the end of the label so nothing downstream has to strip commentary. The result lands back in the sheet or in a Slack message.

Third scenario, a tool-using agent. Wire the node to an AI Agent Tool and the model decides which tool to call, while Frequency Penalty and Presence Penalty keep it from looping on the same phrasing.

When to reach for something else: this node talks to a Lemonade server, so a workflow that needs a hosted vendor endpoint goes to OpenAI, Anthropic or Google Gemini instead. Same port, different sub-node, nothing else in the workflow changes. That swap is also the honest test of whether a served model holds up on your own task, and the rest of the build stays where it is.

Known limits. The node is a sub-node with no main input, so it never runs on its own and never appears alone in an execution. It covers text completion, not embeddings or transcription. Every parameter except the model itself is optional, which is convenient until a default surprises you: Max Tokens to Generate ships at -1, meaning no limit at all. Version 1 of the node exposes 7 parameters, so anything beyond that list is simply not on this panel, and older workflows may still show an earlier version with fewer options.

Connect

How do you authenticate the node?

  1. 01

    Create the Lemonade credential

    Open the Credentials menu in n8n and add the credential the node asks for. n8n stores it once and every workflow on the instance can reuse it, on n8n Cloud or on a self-hosted instance, so there is nothing to paste again the next time an agent needs a model.

  2. 02

    Attach it to the node

    Drop the Lemonade Model node onto the canvas, open it, and select the credential in the field at the top of the panel. A credential belongs to the n8n instance, not to a single workflow, so the same entry serves a drafting agent and a classification chain at the same time.

  3. 03

    Pick a model and test

    Open the Model list. The node queries the Lemonade server and shows the models it loads and manages, which is how you confirm the credential works before building anything else. An empty list means the server side is not serving what you expect, not that the node is broken.

Connections

Where does this node plug in?

A port is the small connector under a root node where one type of sub-node attaches. This node has no main input of its own: it only exists at the end of a model port.

Output (what it plugs into)

  • Modelai_languageModel
01ModelThis is the single connection the node offers, and the reason it exists. Drag from the model port of a root node and the completion requests it makes are served by Lemonade.

This is the single connection the node offers, and the reason it exists. Drag from the model port of a root node and the completion requests it makes are served by Lemonade.

Key parameters

  • Required: the root node has no model of its own, so the port has to be filled before it can generate anything.
  • AI Agent: the usual host, where the model picks tools and writes the answer.
  • Basic LLM Chain: prompt in, text out, and the simplest root node to test a served model on.
  • Text Classifier: the model reads and labels instead of writing, alongside Sentiment Analysis and Information Extractor.
Use cases
one node can feed several root nodes in the same workflow, so an agent and a classifier share the same served model.
Parameters

Which parameters does the node expose?

The Lemonade Model node has 7 parameters. For each one: the node as you configure it in n8n, what the parameter changes, and our field notes.

01

Model

model

What you see in n8n

Notes & use cases

Picks the model that generates the completion. The list is not typed by hand: models are loaded and managed through the Lemonade server, and the node reads that list when you open the dropdown.

Key parameters

  • Model: the only required field on the node, filled from the list the server provides.
Use cases
swapping the entry here is how a workflow moves from one served model to another without touching the agent above it.
02

Sampling Temperature

options.temperature

What you see in n8n

Notes & use cases

Controls how random the generated text is. Lower values make the output more focused and deterministic, higher values make it more diverse and random.

Key parameters

  • Sampling Temperature: a number, default 0.7, left alone unless the output is either too repetitive or too loose.
Use cases
pull it down for a classifier that must return the same label twice for the same row, raise it when an agent writes copy and the phrasing keeps coming out identical.
03

Top P

options.topP

What you see in n8n

Notes & use cases

Restricts the pool the model draws from. It chooses from the smallest possible set of tokens whose cumulative probability exceeds the probability top_p, which reduces repetitions and reads as more human.

Key parameters

  • Top P: a number, default 1, meaning nothing is filtered out until you lower it.
Use cases
a summary that keeps wandering into unlikely wording tightens up here, and the trap is turning this down at the same time as temperature, which narrows the output twice over.
04

Frequency Penalty

options.frequencyPenalty

What you see in n8n

Notes & use cases

Weighs tokens by how often they have already appeared in the generated text. Positive values discourage repetition, negative values encourage it.

Key parameters

  • Frequency Penalty: a number, default 0, so no adjustment is applied out of the box.
Use cases
long outbound emails that hammer the same product name benefit from a positive value, while a node that has to echo a fixed term in every sentence is left at the default.
05

Presence Penalty

options.presencePenalty

What you see in n8n

Notes & use cases

Looks at whether a token has appeared at all rather than how often. Positive values penalize tokens already present in the generated text so far, which encourages diversity.

Key parameters

  • Presence Penalty: a number, default 0, adjusted only when an answer circles the same few ideas.
Use cases
brainstorming tasks where the agent should move on to a new angle after each bullet, rather than restating the first one in new words.
06

Max Tokens to Generate

options.maxTokens

What you see in n8n

Notes & use cases

Caps the length of the completion. A token is the unit a model counts in, roughly a word fragment. Set it to -1 for no limit, which is exactly what the node does by default.

Key parameters

  • Max Tokens to Generate: a number, default -1, and worth setting on any node whose output is written somewhere with a length constraint.
Use cases
a Slack notification or a spreadsheet cell. Be cautious with a large value, since it can lead to very long outputs.
07

Stop Sequences

options.stop

What you see in n8n

Notes & use cases

Tells the model where to stop. Give it a comma-separated list of sequences and generation ends as soon as one of them shows up in the text.

Key parameters

  • Stop Sequences: a string, empty by default, so nothing stops the model early until you fill it.
Use cases
a model that answers the question and then keeps going with an explanation nobody asked for gets cut at the marker that opens the explanation.
Need help

Need help automating Lemonade Model with n8n?

A person reads every message.

FAQ

Lemonade Model in n8n, answered

01Is the Lemonade Model node included in n8n?
Yes. It belongs to the n8n AI package, shipped with n8n, so there is nothing to install and nothing extra to pay on the n8n side. That holds on n8n Cloud and on a self-hosted instance running the Community Edition under the Sustainable Use license, and a workflow behaves the same either way. What sits behind the node is a separate matter: the provider serving the model bills its own API calls under its own terms, and n8n adds nothing on top of that.
02What do you need for the n8n Lemonade Model node to work?
Two things. A credential, created once from the Credentials menu and reusable across every workflow on the instance, and a root node to attach to, since this one is a sub-node with no main input and never runs on its own. Once both are in place, open the Model list and pick an entry. Models are loaded and managed through the Lemonade server, so the picker reflects the server state rather than anything stored in n8n. The credential itself lives on the n8n instance, which means a second workflow needing the same model asks for nothing new.
03What are the limits of the node?
It does one job: text completion, served through a model port. There is no main input, so it cannot sit in the middle of a flow and transform items. Version 1 of the node exposes 7 parameters and no others, which means anything outside that list is not available here. Older workflows may show an earlier version with fewer options. Two defaults deserve attention, Max Tokens to Generate at -1, which sets no limit at all, and Stop Sequences empty, which lets the model run to its natural end.
04What do you connect at a minimum, and which sub-node should a beginner start with?
At a minimum, one root node and this sub-node on its model port. The easiest starting point is Basic LLM Chain, which takes a prompt and returns text without the extra machinery. Once that runs, AI Agent is the natural next step, because it can call tools and keep a conversation going. If the model has to be served elsewhere, the same port accepts OpenAI, Anthropic or Google Gemini sub-nodes, and the rest of the workflow is untouched. Nothing else has to be rebuilt for that swap, which is the point of the sub-node model in the first place.
05n8n or Make for this kind of AI workflow?
It depends on where the workload should live. Make is hosted, with no self-hosting option, and is billed per operation, which is predictable and needs no infrastructure. n8n runs on n8n Cloud or on your own server, which matters when a model is served from a machine you control and the prompts carry data you would rather keep in house. Cost models differ too, per operation on one side, per instance on the other. Both build workflows visually, so the choice is about hosting, data control and billing rather than the editor.
Hack'celeration Lab

Get our weekly integration tips.

No spam. Unsubscribe anytime.

Same platform

Other docs you might need on this stack.

Browse all integrations