This library provides convenient access to the OpenAI REST API from TypeScript or JavaScript.

It is generated from our OpenAPI specification with Stainless.

To learn how to use the OpenAI API, check out our API Reference and Documentation.

npm install openai

deno add jsr:@openai/openai
npx jsr add @openai/openai

These commands will make the module importable from the @openai/openai scope:

You can also import directly from JSR without an install step if you're using the Deno JavaScript runtime:

import OpenAI from 'jsr:@openai/openai';

The full API of this library can be found in api.md file along with many code examples. The code below shows how to get started using the chat completions API.

import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: process.env['OPENAI_API_KEY'], // This is the default and can be omitted
});

async function main() {
  const chatCompletion = await client.chat.completions.create({
    messages: [{ role: 'user', content: 'Say this is a test' }],
    model: 'gpt-4o',
  });
}

main();

We provide support for streaming responses using Server Sent Events (SSE).

import OpenAI from 'openai';

const client = new OpenAI();

async function main() {
  const stream = await client.chat.completions.create({
    model: 'gpt-4o',
    messages: [{ role: 'user', content: 'Say this is a test' }],
    stream: true,
  });
  for await (const chunk of stream) {
    process.stdout.write(chunk.choices[0]?.delta?.content || '');
  }
}

main();

If you need to cancel a stream, you can break from the loop or call stream.controller.abort().

This library includes TypeScript definitions for all request params and response fields. You may import and use them like so:

import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: process.env['OPENAI_API_KEY'], // This is the default and can be omitted
});

async function main() {
  const params: OpenAI.Chat.ChatCompletionCreateParams = {
    messages: [{ role: 'user', content: 'Say this is a test' }],
    model: 'gpt-4o',
  };
  const chatCompletion: OpenAI.Chat.ChatCompletion = await client.chat.completions.create(params);
}

main();

Documentation for each method, request param, and response field are available in docstrings and will appear on hover in most modern editors.

When interacting with the API some actions such as starting a Run and adding files to vector stores are asynchronous and take time to complete. The SDK includes helper functions which will poll the status until it reaches a terminal state and then return the resulting object. If an API method results in an action which could benefit from polling there will be a corresponding version of the method ending in 'AndPoll'.

For instance to create a Run and poll until it reaches a terminal state you can run:

const run = await openai.beta.threads.runs.createAndPoll(thread.id, {
  assistant_id: assistantId,
});

More information on the lifecycle of a Run can be found in the Run Lifecycle Documentation

When creating and interacting with vector stores, you can use the polling helpers to monitor the status of operations. For convenience, we also provide a bulk upload helper to allow you to simultaneously upload several files at once.

const fileList = [
  createReadStream('/home/data/example.pdf'),
  ...
];

const batch = await openai.vectorStores.fileBatches.uploadAndPoll(vectorStore.id, {files: fileList});

The SDK also includes helpers to process streams and handle the incoming events.

const run = openai.beta.threads.runs
  .stream(thread.id, {
    assistant_id: assistant.id,
  })
  .on('textCreated', (text) => process.stdout.write('\nassistant > '))
  .on('textDelta', (textDelta, snapshot) => process.stdout.write(textDelta.value))
  .on('toolCallCreated', (toolCall) => process.stdout.write(`\nassistant > ${toolCall.type}\n\n`))
  .on('toolCallDelta', (toolCallDelta, snapshot) => {
    if (toolCallDelta.type === 'code_interpreter') {
      if (toolCallDelta.code_interpreter.input) {
        process.stdout.write(toolCallDelta.code_interpreter.input);
      }
      if (toolCallDelta.code_interpreter.outputs) {
        process.stdout.write('\noutput >\n');
        toolCallDelta.code_interpreter.outputs.forEach((output) => {
          if (output.type === 'logs') {
            process.stdout.write(`\n${output.logs}\n`);
          }
        });
      }
    }
  });

More information on streaming helpers can be found in the dedicated documentation: helpers.md

This library provides several conveniences for streaming chat completions, for example:

import OpenAI from 'openai';

const openai = new OpenAI();

async function main() {
  const stream = await openai.beta.chat.completions.stream({
    model: 'gpt-4o',
    messages: [{ role: 'user', content: 'Say this is a test' }],
    stream: true,
  });

  stream.on('content', (delta, snapshot) => {
    process.stdout.write(delta);
  });

  // or, equivalently:
  for await (const chunk of stream) {
    process.stdout.write(chunk.choices[0]?.delta?.content || '');
  }

  const chatCompletion = await stream.finalChatCompletion();
  console.log(chatCompletion); // {id: "…", choices: […], …}
}

main();

Streaming with openai.beta.chat.completions.stream({…}) exposes various helpers for your convenience including event handlers and promises.

Alternatively, you can use openai.chat.completions.create({ stream: true, … }) which only returns an async iterable of the chunks in the stream and thus uses less memory (it does not build up a final chat completion object for you).

If you need to cancel a stream, you can break from a for await loop or call stream.abort().

We provide the openai.beta.chat.completions.runTools({…}) convenience helper for using function tool calls with the /chat/completions endpoint which automatically call the JavaScript functions you provide and sends their results back to the /chat/completions endpoint, looping as long as the model requests tool calls.

If you pass a parse function, it will automatically parse the arguments for you and returns any parsing errors to the model to attempt auto-recovery. Otherwise, the args will be passed to the function you provide as a string.

If you pass tool_choice: {function: {name: …}} instead of auto, it returns immediately after calling that function (and only loops to auto-recover parsing errors).

import OpenAI from 'openai';

const client = new OpenAI();

async function main() {
  const runner = client.beta.chat.completions
    .runTools({
      model: 'gpt-4o',
      messages: [{ role: 'user', content: 'How is the weather this week?' }],
      tools: [
        {
          type: 'function',
          function: {
            function: getCurrentLocation,
            parameters: { type: 'object', properties: {} },
          },
        },
        {
          type: 'function',
          function: {
            function: getWeather,
            parse: JSON.parse, // or use a validation library like zod for typesafe parsing.
            parameters: {
              type: 'object',
              properties: {
                location: { type: 'string' },
              },
            },
          },
        },
      ],
    })
    .on('message', (message) => console.log(message));

  const finalContent = await runner.finalContent();
  console.log();
  console.log('Final content:', finalContent);
}

async function getCurrentLocation() {
  return 'Boston'; // Simulate lookup
}

async function getWeather(args: { location: string }) {
  const { location } = args;
  // … do lookup …
  return { temperature, precipitation };
}

main();

// {role: "user",      content: "How's the weather this week?"}
// {role: "assistant", tool_calls: [{type: "function", function: {name: "getCurrentLocation", arguments: "{}"}, id: "123"}
// {role: "tool",      name: "getCurrentLocation", content: "Boston", tool_call_id: "123"}
// {role: "assistant", tool_calls: [{type: "function", function: {name: "getWeather", arguments: '{"location": "Boston"}'}, id: "1234"}]}
// {role: "tool",      name: "getWeather", content: '{"temperature": "50degF", "preciptation": "high"}', tool_call_id: "1234"}
// {role: "assistant", content: "It's looking cold and rainy - you might want to wear a jacket!"}
//
// Final content: "It's looking cold and rainy - you might want to wear a jacket!"

Like with .stream(), we provide a variety of helpers and events.

Read more about various examples such as with integrating with zod, next.js, and proxying a stream to the browser.

Request parameters that correspond to file uploads can be passed in many different forms:

• File (or an object with the same structure)
• a fetch Response (or an object with the same structure)
• an fs.ReadStream
• the return value of our toFile helper

import fs from 'fs';
import OpenAI, { toFile } from 'openai';

const client = new OpenAI();

// If you have access to Node `fs` we recommend using `fs.createReadStream()`:
await client.files.create({ file: fs.createReadStream('input.jsonl'), purpose: 'fine-tune' });

// Or if you have the web `File` API you can pass a `File` instance:
await client.files.create({ file: new File(['my bytes'], 'input.jsonl'), purpose: 'fine-tune' });

// You can also pass a `fetch` `Response`:
await client.files.create({ file: await fetch('https://somesite/input.jsonl'), purpose: 'fine-tune' });

// Finally, if none of the above are convenient, you can use our `toFile` helper:
await client.files.create({
  file: await toFile(Buffer.from('my bytes'), 'input.jsonl'),
  purpose: 'fine-tune',
});
await client.files.create({
  file: await toFile(new Uint8Array([0, 1, 2]), 'input.jsonl'),
  purpose: 'fine-tune',
});

When the library is unable to connect to the API, or if the API returns a non-success status code (i.e., 4xx or 5xx response), a subclass of APIError will be thrown:

async function main() {
  const job = await client.fineTuning.jobs
    .create({ model: 'gpt-4o', training_file: 'file-abc123' })
    .catch(async (err) => {
      if (err instanceof OpenAI.APIError) {
        console.log(err.status); // 400
        console.log(err.name); // BadRequestError
        console.log(err.headers); // {server: 'nginx', ...}
      } else {
        throw err;
      }
    });
}

main();

Error codes are as followed:

For more information on debugging requests, see these docs

All object responses in the SDK provide a _request_id property which is added from the x-request-id response header so that you can quickly log failing requests and report them back to OpenAI.

const completion = await client.chat.completions.create({ messages: [{ role: 'user', content: 'Say this is a test' }], model: 'gpt-4o' });
console.log(completion._request_id) // req_123

You can also access the Request ID using the .withResponse() method:

const { data: stream, request_id } = await openai.chat.completions
  .create({
    model: 'gpt-4',
    messages: [{ role: 'user', content: 'Say this is a test' }],
    stream: true,
  })
  .withResponse();

The Realtime API enables you to build low-latency, multi-modal conversational experiences. It currently supports text and audio as both input and output, as well as function calling through a WebSocket connection.

import { OpenAIRealtimeWebSocket } from 'openai/beta/realtime/websocket';

const rt = new OpenAIRealtimeWebSocket({ model: 'gpt-4o-realtime-preview-2024-12-17' });

rt.on('response.text.delta', (event) => process.stdout.write(event.delta));

For more information see realtime.md.

To use this library with Azure OpenAI, use the AzureOpenAI class instead of the OpenAI class.

import { AzureOpenAI } from 'openai';
import { getBearerTokenProvider, DefaultAzureCredential } from '@azure/identity';

const credential = new DefaultAzureCredential();
const scope = 'https://cognitiveservices.azure.com/.default';
const azureADTokenProvider = getBearerTokenProvider(credential, scope);

const openai = new AzureOpenAI({ azureADTokenProvider });

const result = await openai.chat.completions.create({
  model: 'gpt-4o',
  messages: [{ role: 'user', content: 'Say hello!' }],
});

console.log(result.choices[0]!.message?.content);

Certain errors will be automatically retried 2 times by default, with a short exponential backoff. Connection errors (for example, due to a network connectivity problem), 408 Request Timeout, 409 Conflict, 429 Rate Limit, and >=500 Internal errors will all be retried by default.

You can use the maxRetries option to configure or disable this:

// Configure the default for all requests:
const client = new OpenAI({
  maxRetries: 0, // default is 2
});

// Or, configure per-request:
await client.chat.completions.create({ messages: [{ role: 'user', content: 'How can I get the name of the current day in JavaScript?' }], model: 'gpt-4o' }, {
  maxRetries: 5,
});

Requests time out after 10 minutes by default. You can configure this with a timeout option:

// Configure the default for all requests:
const client = new OpenAI({
  timeout: 20 * 1000, // 20 seconds (default is 10 minutes)
});

// Override per-request:
await client.chat.completions.create({ messages: [{ role: 'user', content: 'How can I list all files in a directory using Python?' }], model: 'gpt-4o' }, {
  timeout: 5 * 1000,
});

On timeout, an APIConnectionTimeoutError is thrown.

Note that requests which time out will be retried twice by default.

List methods in the OpenAI API are paginated. You can use the for await … of syntax to iterate through items across all pages:

async function fetchAllFineTuningJobs(params) {
  const allFineTuningJobs = [];
  // Automatically fetches more pages as needed.
  for await (const fineTuningJob of client.fineTuning.jobs.list({ limit: 20 })) {
    allFineTuningJobs.push(fineTuningJob);
  }
  return allFineTuningJobs;
}

Alternatively, you can request a single page at a time:

let page = await client.fineTuning.jobs.list({ limit: 20 });
for (const fineTuningJob of page.data) {
  console.log(fineTuningJob);
}

// Convenience methods are provided for manually paginating:
while (page.hasNextPage()) {
  page = await page.getNextPage();
  // ...
}

The "raw" Response returned by fetch() can be accessed through the .asResponse() method on the APIPromise type that all methods return.

You can also use the .withResponse() method to get the raw Response along with the parsed data.

const client = new OpenAI();

const response = await client.chat.completions
  .create({ messages: [{ role: 'user', content: 'Say this is a test' }], model: 'gpt-4o' })
  .asResponse();
console.log(response.headers.get('X-My-Header'));
console.log(response.statusText); // access the underlying Response object

const { data: chatCompletion, response: raw } = await client.chat.completions
  .create({ messages: [{ role: 'user', content: 'Say this is a test' }], model: 'gpt-4o' })
  .withResponse();
console.log(raw.headers.get('X-My-Header'));
console.log(chatCompletion);

The log level can be configured in two ways:

1. Via the OPENAI_LOG environment variable
2. Using the logLevel client option (overrides the environment variable if set)

import OpenAI from 'openai';

const client = new OpenAI({
  logLevel: 'debug', // Show all log messages
});

Available log levels, from most to least verbose:

• 'debug' - Show debug messages, info, warnings, and errors
• 'info' - Show info messages, warnings, and errors
• 'warn' - Show warnings and errors (default)
• 'error' - Show only errors
• 'off' - Disable all logging

At the 'debug' level, all HTTP requests and responses are logged, including headers and bodies. Some authentication-related headers are redacted, but sensitive data in request and response bodies may still be visible.

By default, this library logs to globalThis.console. You can also provide a custom logger. Most logging libraries are supported, including pino, winston, bunyan, consola, signale, and @std/log. If your logger doesn't work, please open an issue.

When providing a custom logger, the logLevel option still controls which messages are emitted, messages below the configured level will not be sent to your logger.

import OpenAI from 'openai';
import pino from 'pino';

const logger = pino();

const client = new OpenAI({
  logger: logger.child({ name: 'OpenAI' }),
  logLevel: 'debug', // Send all messages to pino, allowing it to filter
});

This library is typed for convenient access to the documented API. If you need to access undocumented endpoints, params, or response properties, the library can still be used.

To make requests to undocumented endpoints, you can use client.get, client.post, and other HTTP verbs. Options on the client, such as retries, will be respected when making these requests.

await client.post('/some/path', {
  body: { some_prop: 'foo' },
  query: { some_query_arg: 'bar' },
});

To make requests using undocumented parameters, you may use // @ts-expect-error on the undocumented parameter. This library doesn't validate at runtime that the request matches the type, so any extra values you send will be sent as-is.

client.foo.create({
  foo: 'my_param',
  bar: 12,
  // @ts-expect-error baz is not yet public
  baz: 'undocumented option',
});

For requests with the GET verb, any extra params will be in the query, all other requests will send the extra param in the body.

If you want to explicitly send an extra argument, you can do so with the query, body, and headers request options.

To access undocumented response properties, you may access the response object with // @ts-expect-error on the response object, or cast the response object to the requisite type. Like the request params, we do not validate or strip extra properties from the response from the API.

By default, this library expects a global fetch function is defined.

If you want to use a different fetch function, you can either polyfill the global:

import fetch from 'my-fetch';

globalThis.fetch = fetch;

Or pass it to the client:

import OpenAI from 'openai';
import fetch from 'my-fetch';

const client = new OpenAI({ fetch });

If you want to set custom fetch options without overriding the fetch function, you can provide a fetchOptions object when instantiating the client or making a request. (Request-specific options override client options.)

import OpenAI from 'openai';

const client = new OpenAI({
  fetchOptions: {
    // `RequestInit` options
  },
});

To modify proxy behavior, you can provide custom fetchOptions that add runtime-specific proxy options to requests:

Node ^[docs]

import OpenAI from 'openai';
import * as undici from 'undici';

const proxyAgent = new undici.ProxyAgent('http://localhost:8888/');
const client = new OpenAI({
  fetchOptions: {
    dispatcher: proxyAgent,
  },
});

Bun ^[docs]

import OpenAI from 'openai';

const client = new OpenAI({
  fetchOptions: {
    proxy: 'http://localhost:8888/',
  },
});

Deno ^[docs]

import OpenAI from 'npm:openai';

const httpClient = Deno.createHttpClient({ proxy: { url: 'http://localhost:8888/' } });
const client = new OpenAI({
  fetchOptions: {
    client: httpClient,
  },
});

This package generally follows SemVer conventions, though certain backwards-incompatible changes may be released as minor versions:

1. Changes that only affect static types, without breaking runtime behavior.
2. Changes to library internals which are technically public but not intended or documented for external use. (Please open a GitHub issue to let us know if you are relying on such internals.)
3. Changes that we do not expect to impact the vast majority of users in practice.

We take backwards-compatibility seriously and work hard to ensure you can rely on a smooth upgrade experience.

We are keen for your feedback; please open an issue with questions, bugs, or suggestions.

TypeScript >= 4.9 is supported.

The following runtimes are supported:

• Node.js 18 LTS or later (non-EOL) versions.

• Deno v1.28.0 or higher.

• Bun 1.0 or later.

• Cloudflare Workers.

• Vercel Edge Runtime.

• Jest 28 or greater with the "node" environment ("jsdom" is not supported at this time).

• Nitro v2.6 or greater.

• Web browsers: disabled by default to avoid exposing your secret API credentials. Enable browser support by explicitly setting dangerouslyAllowBrowser to true'.

  More explanation

  Why is this dangerous?

  Enabling the dangerouslyAllowBrowser option can be dangerous because it exposes your secret API credentials in the client-side code. Web browsers are inherently less secure than server environments, any user with access to the browser can potentially inspect, extract, and misuse these credentials. This could lead to unauthorized access using your credentials and potentially compromise sensitive data or functionality.

  When might this not be dangerous?

  In certain scenarios where enabling browser support might not pose significant risks:

  • Internal Tools: If the application is used solely within a controlled internal environment where the users are trusted, the risk of credential exposure can be mitigated.
  • Public APIs with Limited Scope: If your API has very limited scope and the exposed credentials do not grant access to sensitive data or critical operations, the potential impact of exposure is reduced.
  • Development or debugging purpose: Enabling this feature temporarily might be acceptable, provided the credentials are short-lived, aren't also used in production environments, or are frequently rotated.

Note that React Native is not supported at this time.

If you are interested in other runtime environments, please open or upvote an issue on GitHub.

See the contributing documentation.