@sonetto/openai

OpenAI-compatible model provider for @sonetto/atom.

It adapts Chat Completions and Responses APIs into Atom model events, including text, reasoning, tool calls, and tool-result continuation.

Install

pnpm add @sonetto/openai @sonetto/atom

Usage

import { atom } from '@sonetto/atom'
import { openai } from '@sonetto/openai'

const run = atom({
  instructions: 'Answer concisely.',
  model: openai('gpt-4.1-mini', {
    apiKey: process.env.OPENAI_API_KEY,
  }),
})

const result = await run('Say hello.').result
console.log(result.text)

Custom OpenAI-compatible endpoints can set baseURL:

openai('model-name', {
  apiKey: process.env.API_KEY,
  baseURL: 'https://provider.example/v1',
})

Use the Responses API with one option:

openai('gpt-5', {
  apiKey: process.env.OPENAI_API_KEY,
  endpoint: 'responses',
})

License

Copyright © 2026 Yu (Alpha You)

Open sourced under MIT license.