Node.js AI API quickstart (₹ pricing, India)

Node.js + unoblox: LLM calls billed in rupees

Node.js developers can call unoblox's /v1 endpoint with the official OpenAI SDK or plain fetch — no custom client needed. Swap the base URL and key, deploy to Vercel, Lambda, or your own server, and get one GST-compliant ₹ invoice every month.

Get started in 2 minutes

npm install openai

import OpenAI from "openai";

const openai = new OpenAI({
  baseURL: "https://api.unoblox.ai/v1",
  apiKey: process.env.UNOBLOX_API_KEY  // ub-gw-... from unoblox.ai/sign-in
});

const response = await openai.chat.completions.create({
  model: "qwen/qwen3-235b-a22b-instruct-2507",
  messages: [
    { role: "user", content: "Explain GST in 50 words" }
  ]
});

console.log(response.choices[0].message.content);

Streaming (real-time tokens)

const stream = await openai.chat.completions.create({
  model: "qwen/qwen3-235b-a22b-instruct-2507",
  messages: [{ role: "user", content: "..." }],
  stream: true
});

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content || "");
}

Handling errors and retries

import OpenAI, { APIError } from "openai";

try {
  const res = await openai.chat.completions.create({ model: "openai/gpt-4o", messages });
} catch (err) {
  if (err instanceof APIError && err.status === 429) {
    // back off and retry — unoblox returns standard OpenAI error shapes
  }
  throw err;
}

Pricing per 1M tokens (input / output)

Model	₹ cost	Best for
Qwen3.8-27B	₹16.32 / ₹48.96	fast, vision
DeepSeek V3.2	₹26.21 / ₹38.30	deep reasoning
Qwen3 1.7B	₹0 / ₹0	prototyping, rate-limited
Claude Sonnet	₹201.6 / ₹1008	production quality

Deployment notes

• Vercel Edge: works with runtime: "edge" — keep the OpenAI client's default fetch transport
• AWS Lambda: pass the key via environment variables; set the function timeout to at least 30s for longer generations
• Express.js: a POST /api/chat route that forwards the body to unoblox and pipes back JSON or a stream
• Discord or Slack bots: run on your own server, call unoblox per message, log tokens for cost tracking

FAQ

Do I need to buffer the whole response before returning it? No. Forward the stream directly to your HTTP response as chunks arrive — Node's stream APIs handle the backpressure.

Can I use TypeScript types? Yes. The OpenAI SDK ships full types — ChatCompletion, ChatCompletionMessageParam, and so on — with editor autocomplete out of the box.

How do I retry failed requests? Catch APIError, check err.status, and retry with exponential backoff. The SDK doesn't auto-retry by default.

Can I run this on Cloudflare Workers? Yes — plain fetch against https://api.unoblox.ai/v1/chat/completions is the safest option in Workers' restricted runtime.

What's a reasonable timeout? Most unoblox responses complete well under 30s; configure your HTTP client's timeout a little above that for longer generations.

Can I run multiple models in parallel? Yes. Fire several create() calls with Promise.all() — each is billed separately, by its own model's ₹ rate.

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