AI Power Grid — JavaScript / TypeScript SDK

Free, decentralized, OpenAI-compatible AI inference.

The Grid API speaks the OpenAI protocol, so this SDK is a thin subclass of the official openai package: it points at the Grid, reads your key from the environment, and adds Grid-specific conveniences. Everything you know from the OpenAI SDK works unchanged.

Install

npm install grid-ai

Quick start

Sign in at the Grid developer console, create an API key, then set it as AIPG_API_KEY:

import { Grid } from 'grid-ai';

const client = new Grid(); // reads AIPG_API_KEY from the environment

const stream = await client.chat.completions.create({
  model: 'gpt-oss-120b',
  messages: [{ role: 'user', content: 'Explain AI Power Grid in one line.' }],
  stream: true,
});

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

See what's online

The Grid's available models change as workers connect and disconnect. Don't hardcode a model blindly — ask which ones are servable right now:

const client = new Grid();
console.log(await client.onlineModels());
// ['gpt-oss-120b', 'qwen3-27b', ...]

An empty array means no workers are connected — requests will 503 until one is.

Video, img2img, ControlNet, LoRAs — the full Grid

The OpenAI-compatible surface covers text and basic txt2img. For the full media parameter surface, use client.grid, which calls the synchronous /v1 image and video endpoints:

const client = new Grid();

// Video:
const result = await client.grid.video('a timelapse of a city at night', {
  model: 'LTX-2.3',
  width: 768,
  height: 512,
  seconds: 4,
  fps: 24,
});

// img2img / ControlNet / LoRAs — anything the workers support:
const img = await client.grid.image('make it watercolor', {
  models: ['FLUX.2 Klein 4B FP8'],
  sourceImage: '<base64>',
  params: { loras: [{ name: 'watercolor', model: 1.0 }] },
});

// Or full control with a raw payload:
const raw = await client.grid.generate({ prompt: '...', models: ['...'], params: {} });

client.grid waits for the synchronous Grid response and returns the finished OpenAI-shaped result. Use timeoutMs for long media jobs; there is no SDK submit/poll mode on this /v1 client.

It's just OpenAI underneath

Grid extends the openai client, so anything the OpenAI SDK does — images, tool calling, structured output, the full .chat / .images / .models surface — works here too. You can also point existing OpenAI code at the Grid by setting baseURL: 'https://api.aipowergrid.io/v1' if you'd rather not switch packages.

Config

new Grid({ apiKey })	Explicit key (overrides env)
AIPG_API_KEY	Env var read when no key is passed
new Grid({ baseURL })	Override the endpoint (default https://api.aipowergrid.io/v1)

All other openai client options are passed straight through.

Links

• Docs
• Get an API key
• Discord

License

MIT