@ai-sdk/openai

Vercel AI SDK - OpenAI Provider

The OpenAI provider contains language model support for the OpenAI chat and completion APIs. It creates language model objects that can be used with the generateText, streamText, generateObject, and streamObject AI functions.

Setup

The OpenAI provider is available in the @ai-sdk/openai module. You can install it with

npm i @ai-sdk/openai

Provider Instance

You can import createOpenAI from @ai-sdk/openai and create a provider instance with various settings:

import { createOpenAI } from '@ai-sdk/openai';

const openai = createOpenAI({
  // optional base URL for proxies etc.
  baseURL: '',

  // optional API key, default to env property OPENAI_API_KEY
  apiKey: '',

  // optional OpenAI organization:
  organization: '',

  // optional OpenAI project:
  project: '',

  // optional custom headers:
  headers: {
    'custom-header': 'value',
  },
});

The AI SDK also provides a shorthand openai import with an OpenAI provider instance that uses defaults:

import { openai } from '@ai-sdk/openai';

Models

The OpenAI provider instance is a function that you can invoke to create a model:

const model = openai('gpt-3.5-turbo');

It automatically selects the correct API based on the model id.

You can also provide model-specific parameters or select a model API by using .chat or .completion.

Chat Models

You can create models that call the OpenAI chat API using the .chat() factory method. The first argument is the model id, e.g. gpt-4. The OpenAI chat models support tool calls and some have multi-modal capabilities.

const model = openai.chat('gpt-3.5-turbo');

OpenAI chat models support also some model specific settings that are not part of the standard call settings. You can pass them as an options argument:

const model = openai.chat('gpt-3.5-turbo', {
  logitBias: {
    // optional likelihood for specific tokens
    '50256': -100,
  },
  user: 'test-user', // optional unique user identifier
});

Completion Models

You can create models that call the OpenAI completions API using the .completion() factory method. The first argument is the model id. Currently only gpt-3.5-turbo-instruct is supported.

const model = openai.completion('gpt-3.5-turbo-instruct');

OpenAI completion models support also some model specific settings that are not part of the standard call settings. You can pass them as an options argument:

const model = openai.completion('gpt-3.5-turbo-instruct', {
  echo: true, // optional, echo the prompt in addition to the completion
  logitBias: {
    // optional likelihood for specific tokens
    '50256': -100,
  },
  suffix: 'some text', // optional suffix that comes after a completion of inserted text
  user: 'test-user', // optional unique user identifier
});