n8n-nodes-nvidia-nim

Connect any free NVIDIA NIM language model to your n8n workflows — in one node.

Overview

n8n-nodes-nvidia-nim is an n8n community node that gives you access to NVIDIA NIM's free hosted language models directly inside your workflows.
It plugs into n8n's AI Agent as a drop-in language model — just like the built-in OpenAI or Anthropic nodes.

Why use this?

• ✅ Free tier models via build.nvidia.com
• ✅ Works as a proper AI sub-node (connect to any AI Agent)
• ✅ Supports 17+ models including Llama 4, Mistral, Qwen3, Gemma, and more
• ✅ Drop-in OpenAI-compatible API — no extra setup

Supported Models

| Model | Value | |---|---| | Llama 4 Maverick 17B 128E Instruct (default) | meta/llama-4-maverick-17b-128e-instruct | | Mistral Large 3 675B Instruct | mistralai/mistral-large-3-675b-instruct-2512 | | Qwen3 Coder 480B A35B Instruct | qwen/qwen3-coder-480b-a35b-instruct | | Mistral Nemotron (Function Calling) | mistralai/mistral-nemotron | | Mistral Medium 3 Instruct | mistralai/mistral-medium-3-instruct | | Magistral Small 2506 | mistralai/magistral-small-2506 | | Step 3.5 Flash (Agentic) | stepfun-ai/step-3.5-flash | | Seed OSS 36B Instruct (Agentic) | bytedance/seed-oss-36b-instruct | | GLM-4.7 (Tool Calling) | thudm/glm-4.7 | | MiniMax M2.7 | minimax/minimax-m2.7 | | Phi-4 Multimodal Instruct | microsoft/phi-4-multimodal-instruct | | Gemma 3N E4B IT | google/gemma-3n-e4b-it | | Gemma 3N E2B IT | google/gemma-3n-e2b-it | | Gemma 2 2B IT | google/gemma-2-2b-it | | Dracarys Llama 3.1 70B Instruct | abacusai/dracarys-llama-3.1-70b-instruct | | Nemotron Mini 4B Instruct | nvidia/nemotron-mini-4b-instruct | | Solar 10.7B Instruct | upstage/solar-10.7b-instruct |

Installation

Via n8n Community Nodes (Recommended)

1. Open your n8n instance
2. Go to Settings → Community Nodes
3. Click Install
4. Enter n8n-nodes-nvidia-nim and confirm

Manual Installation

# Navigate to your n8n custom nodes directory
cd ~/.n8n/custom/

# Install the package
npm install n8n-nodes-nvidia-nim

Then restart n8n.

Getting Your API Key

1. Visit build.nvidia.com
2. Sign in or create a free account
3. Navigate to API Keys in your dashboard
4. Generate a new key and copy it

Note: The free tier gives you a generous number of inference credits — no credit card required.

Usage

1. Add Credentials

In n8n, go to Credentials → New → NVIDIA NIM API and paste your API key.

2. Add the Node to a Workflow

1. Add an AI Agent node to your workflow
2. In the Agent's Model slot, click the + button
3. Search for NVIDIA NIM and select it
4. Configure the model, temperature, and max tokens
5. Run your workflow

The node acts as a language model supplier — it connects to AI Agents, chains, and any other LangChain-compatible n8n node.

Configuration Options

| Parameter | Type | Default | Description | |---|---|---|---| | Model | Dropdown | llama-4-maverick-17b-128e-instruct | The NIM model to use | | Temperature | Number (0–2) | 0.7 | Controls output randomness. Lower = more deterministic | | Max Tokens | Number (1–4096) | 1000 | Maximum tokens in the response |

Known Issue — Routing Light Not Animating

⚠️ This section documents a bug found in earlier versions of this node and how it was fixed.
If you installed version 1.0.0 or earlier, please update to the latest version.

What was happening

When users ran a workflow containing this node, the routing light (the animated glow/pulse around a node that shows it is actively executing) never appeared — even though the node was working correctly behind the scenes.

Every other node in the workflow showed the animation, but the NVIDIA NIM node stayed visually "dark" throughout execution.

Root Cause

This node uses n8n's supplyData() method instead of execute() — which is the correct pattern for AI sub-nodes (language model supply nodes that connect to AI Agents). However, supplyData has a different execution lifecycle than execute, and n8n tracks visual status through that lifecycle.

Three things were missing:

1. No closeFunction returned

supplyData must return a closeFunction alongside the model response. n8n uses this callback as the signal that the node has finished its work — which is what triggers the routing light to complete its animation and mark the node as done.

// ❌ Before — n8n never knew the node finished
return {
  response: chatModel,
};

// ✅ After — n8n now correctly tracks completion
return {
  response: chatModel,
  closeFunction: async () => {},
};

2. Errors were not re-thrown as NodeOperationError

When the node threw an error (bad API key, network failure, etc.), n8n had no way to update the node's visual state to "failed". The routing light would simply never appear.

// ❌ Before — silent failure, n8n UI stays blank
} catch (error) {
  console.error(error);
}

// ✅ After — n8n marks the node as failed with an error badge
} catch (error) {
  throw new NodeOperationError(this.getNode(), error as Error);
}

3. Missing outputConnectionTypes in node descriptor

Without declaring outputConnectionTypes, n8n's UI couldn't properly register the node as a typed AI supply node, which prevented the execution tracker from attaching to it.

// ✅ Added to NvidiaNim.node.json
"outputConnectionTypes": {
  "ai_languageModel": [{ "type": "ai_languageModel" }]
}

The Fix (v1.1.0+)

All three issues were resolved in v1.1.0. The node now correctly shows the routing light animation during execution, displays an error badge on failure, and cleans up properly after each run.

Development

# Install dependencies
npm install

# TypeScript watch mode (auto-recompile on save)
npm run dev

# Build distributable files
npm run build

# Lint code
npm run lint

# Auto-fix lint issues
npm run lintfix

# Format with Prettier
npm run format

Project Structure

n8n-nodes-nvidia-nim/
├── credentials/
│   └── NvidiaNimApi.credentials.ts   # API key credential definition
├── nodes/
│   └── NvidiaNim/
│       ├── NvidiaNim.node.ts          # Node logic (supplyData)
│       ├── NvidiaNim.node.json        # Node descriptor & UI properties
│       └── nvidia-nim.png             # Node icon
├── dist/                              # Compiled output (generated)
├── package.json
└── README.md

Contributing

Pull requests are welcome! If you'd like to add a new model, fix a bug, or improve documentation:

1. Fork the repo
2. Create a feature branch: git checkout -b feat/your-feature
3. Commit your changes: git commit -m 'feat: add your feature'
4. Push and open a PR

License

MIT

Made with ☕ for the n8n community · Powered by NVIDIA NIM