Resources · n8n integration

n8n Lemonade Chat Model nodeConfigure Lemonade Chat Model in n8n.

The n8n Lemonade Chat Model node hands a chat model served by Lemonade to the node that actually does the work. It is a sub-node, so it never runs on its own: it attaches to a root node through the ai_languageModel port. Seven parameters control the model choice and how text comes out.

Try the Lemonade Chat Model node on n8nSee the 7 parameters

Verified Trustpilot reviews · AI, automation & growth agency

Why automate

What does the n8n Lemonade Chat Model node do?

It exposes a chat-capable language model managed by a Lemonade server so other n8n nodes can call it. In the n8n AI model, a root node takes the workflow items on its main input, and sub-nodes plug into its ports, one type of sub-node per port. A sub-node is a small block with no main input of its own. This one is a sub-node, and its port is ai_languageModel.

First scenario: an assistant that answers internal questions. Drop an AI Agent in the workflow, connect this node to its model port, pick a model name in Model, and the agent starts reasoning with the model your Lemonade server hosts. Nothing else changes in the workflow: the agent keeps its tools and its memory, meaning the sub-node that stores past turns of the conversation.

Second scenario: classification you want to be repeatable. A Text Classifier or a Sentiment Analysis node fed by this model should return the same label twice for the same input, so Sampling Temperature goes low and Top P stays tight. Same idea for a Structured Output Parser, where a creative answer is a broken answer.

Third scenario: short summaries that stop where you want them. A Summarization Chain or a Basic LLM Chain plugged into this node can be capped with Max Tokens to Generate, and cut cleanly with Stop Sequences when the output has to end before a signature or a separator line.

When to prefer something else: this node only talks to a Lemonade server. If the model you want comes from a hosted vendor, the dedicated node is the shorter path, OpenAI or Anthropic for instance. And if you are exposing an agent as a callable block for another agent, the root node you attach to is AI Agent Tool rather than a chain.

Known limits, stated plainly. The node has one port and seven parameters, no more. The list of models is whatever your Lemonade server loads and manages, so a missing model is a server-side question, not an n8n one. There is no streaming switch, no retry setting, no token budget guard here. Whichever model you point at, the vendor behind it bills its own API calls under its own terms; n8n adds nothing to that. Picking the platform first? The n8n review compares hosting and cost logic without the sales pitch.

Connect

What do you connect before using it?

  1. 01

    Run a Lemonade server with the models loaded

    The node does not host anything. Models are loaded and managed through the Lemonade server, and that server is what the Model list reads from. So the first step happens outside n8n: get the server up and confirm which model names it exposes. If a model you expect is missing later in the dropdown, this is where the answer lives.

  2. 02

    Create the credential once in n8n

    Open the Credentials menu in n8n and create the credential the node asks for. n8n stores it separately from the workflow, so the same credential serves every workflow on the instance, on n8n Cloud or on a self-hosted instance. You create it once and select it from the node panel afterward.

  3. 03

    Attach the node to a root node

    Add the Lemonade Chat Model node next to the root node that will use it, then drag its connector onto the model port. The node has no main input, so there is nothing to wire from the previous step of the workflow. Run the root node to test: a sub-node executes only when the node above it asks for an answer.

Connections

Where does this node plug in?

A root node holds the workflow logic and asks its sub-nodes for the pieces it needs; this node answers on the model port, and only there.

Output (what it plugs into)

  • Modelai_languageModel
01ModelThis is the node's only connection, of type ai_languageModel. It carries the chat model up to the node that will prompt it.

This is the node's only connection, of type ai_languageModel. It carries the chat model up to the node that will prompt it.

Key parameters

  • Required: the node has no main input, so without this connection it never executes and the workflow simply ignores it.
  • AI Agent: the usual starting point, an agent that reasons, calls tools and answers.
  • Basic LLM Chain: one prompt in, one answer out, when no tool calling is needed.
  • Question and Answer Chain: answers from retrieved documents rather than from the prompt alone.
  • Text Classifier: sorts incoming items into categories you define.
Use cases
a support workflow where the agent drafts a reply, and a nightly workflow where a chain rewrites raw notes into a clean summary, both fed by the same Lemonade model.
Parameters

Which parameters matter, and when?

The Lemonade Chat 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 which model generates the completion. It is the only required parameter of the node, and the list comes straight from the Lemonade server.

Key parameters

  • Model: the model name your Lemonade server makes available, chosen from the dropdown; leave it empty and the node stops the execution.
Use cases
point a drafting agent at one model name, then swap that single field to test another one the server exposes, without rewiring the workflow.
02

Sampling Temperature

options.temperature

What you see in n8n

Notes & use cases

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

Key parameters

  • Sampling Temperature: a number, 0.7 by default; move it down for answers that must repeat, up for phrasing that should vary.
Use cases
a classifier that must return the same label on the same ticket every time runs low; a subject-line generator for a campaign runs higher.
03

Top P

options.topP

What you see in n8n

Notes & use cases

Narrows the pool of tokens the model picks from, keeping the smallest set whose cumulative probability passes the threshold. A token is a fragment of text, roughly part of a word.

Key parameters

  • Top P: a number, 1 by default, which keeps the whole pool; lower it to progressively drop the least likely options.
Use cases
tightening it helps when an answer has to stay inside a known vocabulary, product names in a catalog reply for example. Adjust this or the temperature, not both at once, or you lose track of which one changed the output.
04

Frequency Penalty

options.frequencyPenalty

What you see in n8n

Notes & use cases

Weighs tokens that have already appeared in the generated text. Positive values discourage repetition, negative values encourage it.

Key parameters

  • Frequency Penalty: a number, 0 by default, so no penalty at all until you set one.
Use cases
long summaries that keep circling back to the same sentence usually calm down with a small positive value. Push it far and the model starts reaching for odd synonyms.
05

Presence Penalty

options.presencePenalty

What you see in n8n

Notes & use cases

Looks at whether a token showed up at all, rather than how often. Positive values penalize tokens already present and push the model toward new ground.

Key parameters

  • Presence Penalty: a number, 0 by default; raise it when an answer stays stuck on one theme.
Use cases
brainstorming a list of angles for a campaign, where five variations on one idea are worth less than five different ideas.
06

Max Tokens to Generate

options.maxTokens

What you see in n8n

Notes & use cases

Caps the length of the answer. The model stops once the budget is spent, whether or not the sentence is finished.

Key parameters

  • Max Tokens to Generate: a number, -1 by default, which means no limit; a large value can lead to very long outputs, so set it deliberately.
Use cases
a chat reply that has to fit in a Slack message, or a field written back into a row of a spreadsheet, both deserve a cap rather than an open budget.
07

Stop Sequences

options.stop

What you see in n8n

Notes & use cases

Defines explicit strings that end the generation. The model stops as soon as it produces one of them.

Key parameters

  • Stop Sequences: a comma-separated list of sequences, empty by default.
Use cases
cutting an answer before a signature block, or stopping at a separator line when a downstream node in the workflow only needs the first section. It also saves a cleanup step later, since nothing has to strip the trailing text before the next node reads it.
Need help

Need help automating Lemonade Chat Model with n8n?

A person reads every message.

FAQ

Questions people ask next

01Is the Lemonade Chat Model node included in n8n?
Yes. It belongs to the n8n AI package, shipped with n8n, on n8n Cloud as well as on a self-hosted instance under the Community Edition and its Sustainable Use license. There is nothing to install and n8n charges nothing for the node itself. The separate cost is on the model side: the vendor behind the model you call bills its own API calls under its own terms, and n8n adds nothing on top of that. Self-hosting runs through Docker or npm, and a workflow behaves the same either way.
02What do you need for the n8n Lemonade Chat Model node to work?
Two things. A Lemonade server, because models are loaded and managed there and the Model list reads from it, and the credential that the node uses, created once from the Credentials menu in n8n and then reusable across every workflow on the instance. After that, one parameter is required in the node itself: Model. The six others are options. And since this is a sub-node, it also needs a root node to attach to through its model port, otherwise it sits in the canvas without ever running.
03What are the limits of this node?
It is deliberately narrow. Version 1 of the node exposes seven parameters and a single connection port, so anything beyond model choice and generation behavior is out of scope here. The available models are whatever the Lemonade server loads, which means a missing model is fixed on the server, not in n8n. There is no main input, no output to a regular workflow branch, and nothing that lets it run standalone. An older workflow may still show an earlier node version with fewer options.
04What is the minimum to plug in, and which node should you start with?
The minimum is one root node connected to the model port, plus a model name filled in. To start, an AI Agent is the most common pairing because it covers question answering, tool calls and follow-up turns in one block. If the task is a single prompt with a single answer, a Basic LLM Chain is lighter and easier to debug. For answers grounded in your own documents, the root node to reach for is a Question and Answer Chain. Add memory and tools later, once the model answers correctly.
05n8n or Make for this kind of AI workflow?
It depends on four things. Hosting: Make is hosted by Make with no self-hosting option, while n8n runs on your own server through Docker or npm, or on n8n Cloud. Data control: a self-hosted instance keeps the prompts and the model traffic inside your own perimeter. Cost model: Make bills per operation, which changes how you design loops. Visual logic: both draw the workflow, n8n adds the root node and sub-node model this page describes. Neither is a bad tool; the answer follows your constraints.
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