Migrate to OpenUI Gateway

Move an existing OpenAI Chat Completions integration to OpenUI Gateway by changing the API key, base URL, and model identifier. Your existing messages, streaming, and function-tool workflows can remain unchanged.

Before you begin

• Create an inference API key in the Thesys Console.
• Choose a supported model from Models.
• Decide whether to use managed inference or bring your own provider key.

Update the client

Replace the OpenAI API key and add the Gateway base URL:

import OpenAI from "openai";

export const client = new OpenAI({
-  apiKey: process.env.OPENAI_API_KEY,
+  apiKey: process.env.THESYS_API_KEY,
+  baseURL: "https://api.thesys.dev/v1/embed",
});

Keep the inference API key on the server. See Authentication for credential requirements.

Update the model identifier

Gateway model identifiers use the {provider}/{model} format. Add the provider to the model used by your existing request:

// Before
model: "gpt-5";

// After
model: "openai/gpt-5";

The Chat Completions request and response shapes remain compatible with the OpenAI API. Existing message history, streaming handlers, and function-tool loops do not need to be rewritten.

Add Generative UI

Migrating the client routes your existing model traffic through Gateway. To generate interfaces, add the OpenUI system prompt and connect the response stream to a renderer.

Follow the Generative UI Quickstart to configure a component library and generate your first interface.

Verify the migration

• Confirm requests authenticate with THESYS_API_KEY.
• Confirm the selected model uses the {provider}/{model} identifier.
• Verify streaming responses still reach the application incrementally.
• Run an existing function-tool flow and confirm tool calls are unchanged.
• If Generative UI is enabled, confirm the generated interface renders correctly.