TanStack AI - SUPA Docs

Use TanStack AI with the @tanstack/ai-openai adapter and SUPA’s OpenAI-compatible API. TanStack’s OpenAI adapter builds on the official openai package. Set baseURL to SUPA and pass your SUPA model id (for example google/gemma-3-27b-it).

TanStack AI currently types chat model names as OpenAI model ids. Use a type assertion for custom SUPA model strings until your project exposes them in typings. The runtime request still sends your SUPA model name to the API.

Prerequisites

A SUPA API key from your profile page
Node.js 18+ (TanStack AI recommends current Node LTS; see their docs for any stricter requirement)

Install

npm install @tanstack/ai @tanstack/ai-openai zod

Configure the adapter

Use createOpenaiChat with apiKey, baseURL, and your model id.

import { createOpenaiChat } from '@tanstack/ai-openai';
import type { OpenAIChatModel } from '@tanstack/ai-openai';

const adapter = createOpenaiChat(
  'google/gemma-3-27b-it' as OpenAIChatModel,
  process.env.SUPA_API_KEY!,
  { baseURL: 'https://api.supa.works/openai/v1' },
);

Chat (non-streaming)

The TanStack chat activity accepts your adapter and messages. Set stream: false to collect the full reply as a string.

import { chat } from '@tanstack/ai';

const text = await chat({
  adapter,
  stream: false,
  messages: [
    {
      role: 'user',
      content: [{ type: 'text', content: 'What is a banana?' }],
    },
  ],
});

console.log(text);

Streaming

Omit stream or set stream: true to receive an async iterable of stream chunks (see TanStack AI docs for consuming chunks).

Next steps

Tool calls with TanStack AI shows toolDefinition and .server() with SUPA.
Image attachments with TanStack AI shows multimodal content parts.
Vercel AI SDK if you prefer @ai-sdk/openai and generateText / streamText.