OpenAI Responses API

The OpenAI Responses endpoint exposes the OpenAI Responses API through Neon AI Gateway. Use it with the OpenAI SDK's responses.create() method by changing only the baseURL.

Base URL: https://<branch-host>/openai/v1

This endpoint is also reachable at the longer /ai-gateway/openai/v1/responses path. Both behave identically and neither is deprecated. See Shorter paths for the full list of aliases.

If you're using an OpenAI-compatible client that accepts a base URL, set it to either https://<branch-host>/openai/v1 or https://<branch-host>/ai-gateway/openai/v1. The request and response shapes are the standard OpenAI Responses API shape.

Setup

Set these environment variables. See Get started for how to obtain them.

NEON_AI_GATEWAY_TOKEN=nt_live_...
NEON_AI_GATEWAY_BASE_URL=https://br-winter-pond-aptw82ef-api.ai.c-2.us-east-2.aws.neon.tech

Supported models

This endpoint accepts OpenAI models only:

Sending a non-OpenAI model ID returns 400 model "<model-id>" is not available on the openai_responses endpoint.

Basic request

import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: process.env.NEON_AI_GATEWAY_TOKEN,
  baseURL: `${process.env.NEON_AI_GATEWAY_BASE_URL}/openai/v1`,
});

const response = await client.responses.create({
  model: 'gpt-5-4',
  input: [{ role: 'user', content: 'What is Neon?' }],
});

console.log(response.output_text);

Streaming

const stream = await client.responses.create({
  model: 'gpt-5-4',
  input: [{ role: 'user', content: 'Explain database branching.' }],
  stream: true,
});

for await (const event of stream) {
  if (event.type === 'response.output_text.delta') {
    process.stdout.write(event.delta);
  }
}

Image generation with Vercel AI SDK

The @neon/ai-sdk-provider package re-exports the OpenAI provider's Responses image-generation tool as neon.tools.imageGeneration(). Use it with OpenAI-routed models such as gpt-5-mini.

Use streamText, not generateText: image results are returned as tool-result parts, and a full base64 image reliably runs into size limits on a non-streaming response.

import { neon } from '@neon/ai-sdk-provider';
import { streamText } from 'ai';

const result = streamText({
  model: neon('gpt-5-mini'),
  messages: [{ role: 'user', content: 'Create a simple Neon database mascot.' }],
  tools: {
    image: neon.tools.imageGeneration({
      outputFormat: 'jpeg',
      size: '1024x1024',
      partialImages: 3,
    }),
  },
});

return result.toUIMessageStreamResponse();

AI SDK generateImage() is not supported by AI Gateway; image generation is available only through this Responses tool.

Error handling

For authentication, quota, and upstream errors, see Troubleshooting.

Next steps

• Models: full model catalog and which models require this endpoint
• Chat completions: use any model via the unified OpenAI-compatible endpoint
• Authentication: credential scopes and branch binding