Vercel AI SDK with unoblox

Vercel AI SDK is the default choice for streaming AI apps in Next.js. unoblox integrates with a one-line provider swap: keep the SDK's hooks and streaming helpers, and unlock Claude, GPT-4o, Qwen, and DeepSeek behind a single endpoint billed in rupees.

Why Vercel AI SDK + unoblox

• Next.js native: streaming, SSR, and TypeScript support out of the box
• One custom provider: configure createOpenAI once, reuse it everywhere
• 40+ models: all through one OpenAI-compatible endpoint
• ₹ billing: monthly GST invoice, input-tax-credit eligible
• Deploy anywhere: Vercel, or any Node.js host — unoblox doesn't require Vercel specifically

Two-line integration

// lib/unoblox.ts
import { createOpenAI } from '@ai-sdk/openai';

export const unoblox = createOpenAI({
  baseURL: 'https://api.unoblox.ai/v1',
  apiKey: process.env.UNOBLOX_API_KEY,
});

// Anywhere in your app
import { generateText } from 'ai';
import { unoblox } from '@/lib/unoblox';

const { text } = await generateText({
  model: unoblox('openai/gpt-4o'),
  prompt: 'Build me a React component that shows a pricing table.',
});

Streaming API route (App Router)

// app/api/chat/route.ts
import { streamText } from 'ai';
import { unoblox } from '@/lib/unoblox';

export async function POST(req: Request) {
  const { messages } = await req.json();

  const result = streamText({
    model: unoblox('anthropic/claude-sonnet-5'),
    messages,
  });

  return result.toTextStreamResponse();
}

Quick React example

import { useCompletion } from 'ai/react';

export default function Chat() {
  const { completion, input, handleInputChange, handleSubmit } = useCompletion({
    api: '/api/chat',
  });

  return (
    <form onSubmit={handleSubmit}>
      <input value={input} onChange={handleInputChange} placeholder="Ask me anything..." />
      <button type="submit">Send</button>
      <div>{completion}</div>
    </form>
  );
}

Environment setup

# .env.local
UNOBLOX_API_KEY=ub-gw-YOUR_KEY

Model selection

Model	₹ per 1M input	Best for
Qwen3 1.7B	FREE	chat responses, ideation
DeepSeek V4 Flash	₹9.07	fast coding, translation
GPT-4o	₹252	complex analysis
Claude Opus	₹504	long-context reasoning

Frequently asked questions

Do I need a Vercel account? No. The AI SDK runs on any Node.js host — Vercel just makes edge streaming convenient.

Can I use streaming, JSON mode, and tool use together? Yes. The Vercel AI SDK exposes all three, and unoblox passes them through unchanged — check /models for per-model tool and vision support.

How do I handle rate limiting? Catch the error from generateText or streamText, check for a 429, and retry with backoff — or add your own circuit-breaker middleware.

Can I switch models per request? Yes. Pick the provider call dynamically, e.g. unoblox(isComplex ? 'anthropic/claude-opus-5' : 'deepseek-ai/deepseek-v4-flash').

What if unoblox has downtime? unoblox targets 99.95% uptime. For extra resilience, wrap calls in your own fallback logic that retries against a second model.

Do tokens count the same way as raw OpenAI calls? Yes — the SDK's usage reporting reflects the same input and output token counts unoblox bills on, visible per-call on your dashboard.

Get started in rupees → https://unoblox.ai/sign-in