n8n-nodes-litellm

An n8n community node package for integrating with a LiteLLM proxy. Provides a standalone node for Chat Completions, Responses API calls, and Embeddings, plus a LangChain-compatible Chat Model node for use inside AI Agent and Chain workflows.

Nodes

LiteLLM

A general-purpose node with three operations:

| Operation | Endpoint | Description | |---|---|---| | Chat Completion | POST /v1/chat/completions | Send messages to any chat model and receive a response | | Responses | POST /v1/responses | Call OpenAI-compatible responses models such as Copilot Codex / GPT-5 | | Embeddings | POST /v1/embeddings | Generate vector embeddings for one or more text inputs |

Chat Completion parameters

| Parameter | Description | |---|---| | Model | Loaded dynamically from your proxy (filtered to chat-capable models) | | System Prompt | Convenience field — prepended as a system message | | Messages | Ordered list of role + content pairs (user / assistant / system / tool) | | Reasoning Effort | low / medium / high — for extended-thinking models (o1, o3, claude-3-7, etc.) | | Tools | JSON array of function/tool definitions (OpenAI format) | | Tool Choice | auto / none / required |

Embeddings parameters

| Parameter | Description | |---|---| | Model | Loaded dynamically (filtered to embedding models) | | Input | Plain string or JSON array of strings for batch embedding | | Encoding Format | float (default) or base64 |

Responses parameters

| Parameter | Description | |---|---| | Model | Loaded dynamically from your proxy (filtered to mode: responses models) | | Input | Prompt or input text sent to /v1/responses | | Instructions | Optional system-style instructions | | Reasoning Effort | low / medium / high for supported reasoning models | | Tools | JSON array of Responses API tool definitions, including built-in tools and function tools | | Tool Choice | auto / none / required |

LiteLLM Chat Model

A LangChain-compatible node that exposes the LiteLLM proxy as an AI Language Model. Connect it to the Model input of an AI Agent, Chain, or any other LangChain node.

| Parameter | Description | |---|---| | Model | Chat model loaded dynamically from your proxy | | Temperature | Sampling temperature 0–2 | | Max Tokens | Maximum tokens to generate | | Reasoning Effort | Extended thinking level for supported models | | Max Retries | How many times to retry failed API calls |

Credentials

LiteLlmApi

| Field | Required | Description | |---|---|---| | Base URL | ✅ | URL of your LiteLLM proxy, e.g. http://localhost:4000 | | API Key | ❌ | Bearer token — leave blank for open (unauthenticated) deployments |

Tip: The credential is entirely optional on both nodes. If you don't attach one, the node will attempt to connect to http://localhost:4000 without authentication.

Dynamic model discovery

The model dropdowns call GET /model/info at node-configuration time and filter results by the mode field reported by LiteLLM (chat for Chat Completion, responses for the Responses API, embedding for Embeddings). If /model/info is unavailable the nodes fall back to GET /v1/models. You can also type any model name manually.

Installation

In n8n (community nodes)

1. Open Settings → Community Nodes
2. Enter @dkhalife/n8n-nodes-litellm and click Install
3. Restart n8n if prompted

Manual / self-hosted

# Inside your n8n custom nodes directory (or the n8n data volume):
npm install @dkhalife/n8n-nodes-litellm

Development

git clone https://github.com/dkhalife/n8n-nodes-litellm.git
cd n8n-nodes-litellm
npm install
npm run build

Link into a local n8n instance:

# From this repo root:
npm link

# From your n8n installation:
npm link n8n-nodes-litellm

Releasing

New versions are published to npm automatically when a version tag is pushed.

npm run release

This will lint, build, prompt for a version bump, update the changelog, commit, tag, and push — which triggers the publish workflow to publish to npm with provenance.

One-time setup: Configure OIDC trusted publishing on npmjs.com for this package. See the publish workflow header for detailed instructions.

Requirements

• n8n ≥ 1.0
• Node.js ≥ 18.17

License

MIT