Introduction

What it looks like

config.yaml

apiKey: 'test-key'
port: 3000
responses:
  - id: 'greeting'
    messages:
      - role: 'user'
        content: 'hello'
        matcher: 'contains'
      - role: 'assistant'
        content: 'Hello! How can I help you today?'

// Your test code
const openai = new OpenAI({
  baseURL: 'http://localhost:3000/v1',
  apiKey: 'test-key',
});

const response = await openai.chat.completions.create({
  model: 'gpt-4',
  messages: [{ role: 'user', content: 'Say hello!' }],
});
// Returns: "Hello! How can I help you today!"

Why use it

Testing LLM applications is hard because:

• Real API responses vary each time
• API calls cost money and have rate limits
• Network issues can break tests
• You can’t test specific edge cases reliably

This mock server solves these issues by giving you complete control over responses during testing.