GonkaGate Vercel AI SDK Setup

Connect an existing Vercel AI SDK app to GonkaGate by creating one OpenAI-compatible provider, using a current model ID, and returning result.toTextStreamResponse() from your server route. Keep GONKAGATE_API_KEY in the same server runtime that handles the route.

Install the provider package

Installation

npm install @ai-sdk/openai-compatible

Create one shared provider

Create one provider and reuse it across your routes:

Create one shared provider

import { createOpenAICompatible } from "@ai-sdk/openai-compatible";

const apiKey = process.env.GONKAGATE_API_KEY;

if (!apiKey) {
  throw new Error("Set GONKAGATE_API_KEY");
}

export const gonkagate = createOpenAICompatible({
  name: "gonkagate",
  apiKey,
  baseURL: "https://api.gonkagate.com/v1",
});

Use it in a server route

Use the provider in a server route that returns streamed text:

Use it in a server route

import { streamText } from "ai";

import { gonkagate } from "@/lib/gonkagate";

export async function POST(request: Request) {
  const { prompt } = (await request.json()) as { prompt: string };

  const result = streamText({
    model: gonkagate("<model-id-from-get-v1-models>"),
    prompt,
  });

  return result.toTextStreamResponse();
}

Replace <model-id-from-get-v1-models> with a current value from GET /v1/models.

Verify the stream

Send one short request through the route:

Request Example

curl http://localhost:3000/api/chat \
  -H "Content-Type: application/json" \
  -d '{"prompt":"Return exactly: AI SDK connected"}'

Expected result: the route returns streamed text that includes AI SDK connected.

Common failures

Response or symptom	What it usually means	What to do
`GONKAGATE_API_KEY` is missing	The variable is not loaded in the server runtime that handles the route	Load it in the same runtime where the route runs
`401 invalid_api_key`	The key value or key state is wrong	Recheck Authentication and API Keys
`404 model_not_found`	The model ID is stale or unsupported	Refresh it from GET /v1/models
`429 insufficient_quota`	The prepaid USD balance is too low for the request	Top up before retrying
`429 rate_limit_exceeded`	You hit a runtime limit	Honor `Retry-After` and add bounded backoff
The route responds but the client still handles the stream incorrectly	The route and caller no longer agree on a text-stream response	Keep `result.toTextStreamResponse()` on the route and make sure the caller expects streamed text

Vercel AI SDK Setup

Install the provider package

Create one shared provider

Use it in a server route

Verify the stream

Common failures

See also

On this page