How to Fix OpenAI Node SDK TypeError: Configuration is not a constructor

Quick Diagnosis

  • ✓ Did you recently upgrade the openai npm package to version 4 or newer?
  • ✓ Does your Node process crash immediately with TypeError: Configuration is not a constructor?
  • ✓ Are you calling openai.createChatCompletion or openai.createCompletion directly on your client instance?

Environment

The OpenAI client wrapper operates server-side across configurations including Express Server Environments, Next.js App Router Route Handlers, Serverless Vercel Functions, and Custom Node API Endpoints.

SDK Version Constructor Class Pattern Completion Method Signature API Invocation Result
SDK v3.x new Configuration({ apiKey }) openai.createChatCompletion({ ... }) Success (Legacy SDK format)
SDK v4.x new Configuration({ apiKey }) openai.createChatCompletion({ ... }) TypeError: Configuration is not a constructor
SDK v4.x new OpenAI({ apiKey }) openai.chat.completions.create({ ... }) Success (v4 Nested resources format)

Minimal Repro

Under the architecture of OpenAI Node SDK v4, the library was completely rewritten to optimize network execution times, introduce native edge compatibility, and simplify the developer API surface. In v3, client initialization required importing and nesting the Configuration and OpenAIApi classes, and resource requests were made using top-level methods like createChatCompletion. Version 4 deprecated and removed these separate helper classes, merging client setup into a single unified OpenAI constructor class. Furthermore, method signatures were refactored into nested resource directories. Attempting to invoke obsolete constructor classes or calling top-level completion methods directly on a v4 client instance causes the JavaScript compiler to throw TypeError: Configuration is not a constructor or TypeError: openai.createChatCompletion is not a function during initialization. Re-aligning imports to instantiate OpenAI and nesting requests under chat.completions.create resolves the mismatch.

// CRASH: Using obsolete v3 configuration classes under SDK v4
const { Configuration, OpenAIApi } = require('openai');

const configuration = new Configuration({
  apiKey: process.env.OPENAI_API_KEY,
});
const openai = new OpenAIApi(configuration);
TypeError: Configuration is not a constructor
    at Object.<anonymous> (server.js:4:23)

Resolution

When resolving OpenAI v4 migration errors, developers can choose between two main structural options depending on their codebase size.

If you are updating your code manual or starting a new integration, instantiating the unified OpenAI client directly is applicable. This configuration prevents class constructor errors and targets v4 nested completion resources.

  1. Import the default OpenAI class directly from openai.
  2. Instantiate using new OpenAI() and supply your apiKey.
  3. Re-route completion methods to openai.chat.completions.create.
  4. Remove the .data property wrap from response choices parsers.
import OpenAI from 'openai';

// Correct client initialization
const openai = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY
});

export async function GET() {
  // Correct method path and payload parsing
  const response = await openai.chat.completions.create({
    model: 'gpt-4',
    messages: [{ role: 'user', content: 'Hello API' }]
  });

  return Response.json({ text: response.choices[0].message.content });
}

Option B: Run the Automated Migration Codemod

When you manage large enterprise repositories where manual refactoring is cost-prohibitive, running OpenAI’s automated migration utility is applicable.

# Execute automated codebase migration codemod
npx openai migrate

When This Fix Won’t Work

If you transition to the OpenAI client class but still attempt to query legacy endpoints (like createCompletion) with chat-based models (such as gpt-4), the API will return HTTP client exceptions.

Operational Runbook

Case 1: Client Setup

  1. Update import statements to pull OpenAI from openai.
  2. Construct the client directly with configuration keys.

Case 2: Completion Methods

  1. Replace openai.createChatCompletion with openai.chat.completions.create.
  2. Remove .data payload wraps.

Rollback Strategy

To roll back this migration, replace the OpenAI client instance with the legacy Configuration and OpenAIApi classes, update method calls back to openai.createChatCompletion(), restore response payload references to read .data wrappers, and downgrade your package dependencies to OpenAI Node SDK v3.x.

Verification

  • Application process boots and executes route handlers without throwing client configuration TypeErrors.
  • OpenAI API responses resolve successfully without returning payload schema compilation warnings.
  • Integration tests verify that model response strings match expected chat completions outcomes.

Error Trigger Point Lifecycle

Load API keys ➔ Import OpenAI client class ➔ Call client constructor [ERROR OCCURS HERE] ➔ Execute chat.completions.create ➔ Resolve message content payload ➔ Handle connection exceptions

References