!Project banner
!Project banner

STACKIT AI Model Serving nodes for n8n

n8n community nodes that connect to STACKIT AI Model Serving’s OpenAI-compatible API. Includes:

- STACKIT Chat Model — Chat-completions compatible LLM node.
- STACKIT Embeddings — Text embeddings generation node.

n8n is a fair-code licensed workflow automation platform.

- Installation
- Credentials
- STACKIT Chat Model
- STACKIT Embeddings
- Compatibility
- Resources

Installation

Follow the official guide to install community nodes: n8n Community Nodes Installation

Once installed, search for “STACKIT Chat Model” or “STACKIT Embeddings” in the node panel.

Credentials

These nodes use a single credential: “STACKIT AI Model Serving API”.

What you need:

- API Key from your STACKIT AI Model Serving project
- API URL (optional): defaults to

How it works:

- Set the API Key and (optionally) API URL in the credential "STACKIT AI Model Serving API".
- All node requests use the credential’s API URL. If not set, the default endpoint above is used.
- The Model picker fetches from ${API_URL}/models using your credential.

Notes:

- These nodes are compatible with OpenAI-style models and endpoints provided by STACKIT

STACKIT Chat Model

Use this node to obtain a ChatCompletions-capable model as an AI Model output for n8n’s AI features or your own custom chains and agents. No LangChain dependency is required. The node accepts a variety of input shapes (including common LangChain-like prompt structures) and normalizes them for the API.

Key options:

- Model: dynamically loaded from your STACKIT project (non-embedding models)
- Temperature, Top P, Penalties, Max tokens, Retries, Timeout
- Response format: text or JSON (when using JSON mode, ensure your prompts instruct the model to return valid JSON)

Typical usage:

1. Add “STACKIT Chat Model” to the canvas
2. Configure the node with your desired parameters (API-Key, Model and Options)
3. Connect to AI nodes (e.g., AI Agent, AI Chain)

Behavior details:

- Input normalization: The model function accepts strings, arrays of OpenAI-style messages, and objects like { input }, { messages }, { prompt }, as well as common LangChain prompt shapes (StringPromptValue/ChatPromptValue). These are normalized to OpenAI-compatible messages internally.
- Error handling: Unknown errors are wrapped in n8n’s NodeOperationError so failures appear clearly readable in the UI.
- Tracing in UI: The node uses addInputData/addOutputData so every call shows messages in and response/error out in the execution view.

STACKIT Embeddings

Use this node to generate vector embeddings from text.

Key options:

- Model: dynamically loaded embeddings models (intfloat/* models)
- Dimensions: choose target dimensionality if supported
- Batch size and timeout controls
- Strip new lines option for cleaner inputs

Typical usage:

1. Add “STACKIT Embeddings” to the canvas (AI > Embeddings)
2. Configure the node with your desired parameters (API-Key, Model and Options)
3. Connect the output to vector stores, RAG flows, or custom logic

Behavior details:

- Batch and timeouts: Inputs are batched (configurable) and timeouts accept seconds for small values (converted to ms). Newline stripping is enabled by default for cleaner inputs.
- Error handling: Unknown errors are wrapped in NodeOperationError for readable failures in the UI.
- Tracing in UI: The node logs inputs and outputs using addInputData/addOutputData so each request is traceable.

Compatibility

Compatible with n8n@1.60.0 or later.

Resources

- Product:
- n8n Community Nodes:
- OpenAI compatibility:
- GitHub repo: