n8n-nodes-custom-llm

Custom LLM node for n8n - Use any LLM API via custom HTTP request with curl import support.

Features

• Manual Request Mode: Manually configure HTTP requests with full control over URL, headers, body, and query parameters
• cURL Import Mode: Import HTTP request options directly from curl commands
• Dynamic Prompt Injection: Inject prompts into any location in the JSON body using dot notation (e.g., messages[0].content)
• Custom Parameters: Configure any LLM parameters like max_tokens, temperature, top_k, etc. with Liquid template support
• Authentication Support: Built-in support for Header Auth and Basic Auth
• Liquid Templates: Full support for n8n Liquid expressions in all fields

Installation

Global Installation (Development)

1. Clone or download this repository:

git clone <repository-url>
cd n8n-nodes-custom-llm

2. Build the project:

npm install
npm run build

3. Link the module globally:

npm link

4. Navigate to your n8n directory:

cd ~/.n8n

5. Link the nodes package:

npm link n8n-nodes-custom-llm

6. Restart n8n

Installation in n8n Directory

1. Copy the entire project to your n8n directory:

cp -r n8n-nodes-custom-llm ~/.n8n/nodes/

2. Navigate to the node directory:

cd ~/.n8n/nodes/n8n-nodes-custom-llm

3. Install and build:

npm install
npm run build

4. Restart n8n

Docker Installation

Add the following to your n8n docker-compose.yml:

volumes:
  - ./n8n-nodes-custom-llm:/nodes/n8n-nodes-custom-llm

Then rebuild and restart n8n.

Usage

Mode 1: Manual Configuration

1. Add the "Custom LLM" node to your workflow
2. Select "Manual" request mode
3. Fill in:
   • URL: Your LLM API endpoint (e.g., https://api.openai.com/v1/chat/completions)
   • Method: POST (most LLM APIs use POST)
   • Authentication: Choose auth type (Header Auth, etc.)
   • JSON Body: Base request template
   • Prompt Field: Where to inject the prompt (e.g., messages[0].content)
   • Prompt: Your prompt text (supports Liquid templates)
   • Custom Parameters: Add parameters like max_tokens, temperature, top_k

Mode 2: Import from cURL

1. Get a curl command from your LLM provider documentation
2. Select "Import from cURL" request mode
3. Paste the curl command
4. Use the Prompt Field and Custom Parameters sections to override values dynamically

Example Configurations

OpenAI API Setup

• URL: https://api.openai.com/v1/chat/completions
• Method: POST
• JSON Body:

  {
    "model": "gpt-4",
    "messages": [
      {
        "role": "user",
        "content": ""
      }
    ]
  }

• Prompt Field: messages[0].content
• Custom Parameters:
  • max_tokens: 1000
  • temperature: 0.7

Anthropic Claude API Setup

• URL: https://api.anthropic.com/v1/messages
• Method: POST
• JSON Body:

  {
    "model": "claude-3-opus-20240229",
    "messages": [
      {
        "role": "user",
        "content": ""
      }
    ]
  }

• Prompt Field: messages[0].content
• Custom Parameters:
  • max_tokens: 1000
  • top_k: 250

Liquid Template Examples

Using Previous Node Data

• Prompt: {{ $json.userInput }}
• Custom Parameter: {{ $json.tokenCount }}

Conditional Values

{% if $json.priority == 'high' %}
  0.1
{% else %}
  0.7
{% endif %}

Complex Prompt

You are a {{ $json.role }}. 
Please answer the following question: {{ $json.question }}

Requirements

• n8n 1.0.0 or higher

License

MIT