n8n-nodes-laikatest

This is an n8n community node for LaikaTest - a prompt management and A/B testing platform for LLM applications.

Features

• Get Prompt: Fetch versioned prompts by name
• Get Experimental Prompt: Fetch A/B tested prompt variants with automatic bucketing
• Push Scores: Send evaluation metrics back to LaikaTest for experiment analysis

Installation

Community Nodes (Recommended)

1. Go to Settings > Community Nodes
2. Select Install
3. Enter n8n-nodes-laikatest
4. Agree to the risks and select Install

Manual Installation

npm install n8n-nodes-laikatest

Credentials

To use this node, you need a LaikaTest API key:

1. Sign up at laikatest.com
2. Navigate to your project settings
3. Copy your API key
4. In n8n, create new credentials of type "LaikaTest API"
5. Paste your API key

Operations

Get Prompt

Fetch a prompt template by name.

Parameters: | Parameter | Required | Description | |-----------|----------|-------------| | Prompt Name | Yes | Name of the prompt template | | Version ID | No | Specific version (e.g., "10" or "v10") | | Variables | No | Key-value pairs to inject into {{placeholder}} syntax |

Output:

{
  "content": "Your prompt content here",
  "type": "text",
  "promptVersionId": "pv-123"
}

Variable Compilation: If your prompt contains placeholders like {{name}}, add variables to replace them:

Variables:
  Key: name     Value: John
  Key: company  Value: Acme

The output will have placeholders replaced with the provided values.

Get Experimental Prompt

Fetch an A/B tested prompt variant. Users are automatically bucketed for consistent experiences.

Parameters: | Parameter | Required | Description | |-----------|----------|-------------| | Experiment Title | Yes | Title of the experiment | | User ID | No* | User identifier for consistent bucketing | | Session ID | No* | Session identifier for consistent bucketing | | Additional Context | No | Extra key-value pairs for bucketing | | Variables | No | Key-value pairs to inject into {{placeholder}} syntax |

*At least one of User ID or Session ID is recommended for consistent bucketing.

Output:

{
  "content": "Variant A prompt content",
  "type": "chat",
  "experimentId": "exp-123",
  "bucketId": "bucket-A",
  "promptVersionId": "pv-456",
  "promptId": "prompt-789"
}

Variable Compilation: Works the same as Get Prompt - add variables to replace {{placeholder}} syntax in the returned content.

Push Scores

Send evaluation metrics for experiment analysis.

Parameters: | Parameter | Required | Description | |-----------|----------|-------------| | Experiment ID | Yes | From Get Experimental Prompt output | | Bucket ID | Yes | From Get Experimental Prompt output | | Prompt Version ID | Yes | From Get Experimental Prompt output | | User ID | No* | User identifier | | Session ID | No* | Session identifier | | Scores | Yes | Array of score metrics |

*At least one of User ID or Session ID is required.

Score Types:

• int - Integer values (e.g., rating: 5)
• float - Decimal values (e.g., accuracy: 0.95)
• bool - Boolean values (e.g., helpful: true)
• string - Text values (e.g., feedback: "Great!")

Output:

{
  "success": true,
  "statusCode": 200,
  "data": {}
}

Example Workflow

1. Get Experimental Prompt - Fetch a prompt variant for user
2. Use the prompt content with your LLM
3. Push Scores - Send metrics using the experiment metadata from step 1

Development

# Install dependencies
npm install

# Build
npm run build

# Run tests
npm test

# Watch mode
npm run dev

License

MIT