@two-layer-cake/adapter-openai

OpenAI LLM adapter for the two-layer-cake SDK.

Installation

npm install two-layer-cake @two-layer-cake/adapter-openai openai

Configuration

import { OpenAIAdapter } from '@two-layer-cake/adapter-openai';

const adapter = new OpenAIAdapter({
  apiKey: process.env.OPENAI_API_KEY,
  model: 'gpt-4o',
  maxTokens: 4096,
});

Config Options

| Option | Type | Default | Description | | ----------- | -------- | ------- | --------------------------- | | apiKey | string | — | OpenAI API key (required) | | model | string | — | Model ID (required) | | maxTokens | number | 4096 | Maximum tokens per response |

Usage

import { Orchestrator } from 'two-layer-cake';
import { OpenAIAdapter } from '@two-layer-cake/adapter-openai';

const llm = new OpenAIAdapter({
  apiKey: process.env.OPENAI_API_KEY,
  model: 'gpt-4o',
});

const orchestrator = new Orchestrator({
  planner: { llm, mode: 'single-shot' },
});

const result = await orchestrator.run('Summarize the latest AI research trends');
console.log(result.aggregatedOutput);

Features

• Full LLMAdapter interface: complete(), stream(), completeStructured()
• Token counting via character-based estimation
• Streaming support with usage reporting
• Structured output via OpenAI's zodResponseFormat API

License

MIT