With Client SDKs you can call custom or public AI model with just one line of code!

​
Prerequisite

  • Select a model from Model Yard
  • Create an API Key

You can use three ways to call a model on 1019

Client SDK

Run with just a single line. magicpoint.run()

Magicpoint CLI

Run with just a command. magicpoint run

Console

Run from UI. Easy and effective for testing!

​
Client SDK

With Client SDK you need to initialize our SDK in your langauage and then you may able to use custom methods and calls to 1019 backend servers.

​
Initialize SDK

// Please make sure that you run 'npm install magicpoint@latest'
import Magicpoint from 'magicpoint';

const mp = Magicpoint.init({
    API_KEY: process.env.MP_API_KEY
})

​
Run Inference

// Please make sure that you run 'npm install magicpoint@latest'
import Magicpoint from 'magicpoint';

const mp = Magicpoint.init({
    API_KEY: process.env.MP_API_KEY
})

const out = mp.run("MODEL_NAME:MODEL_VERSION", {
    input: {
      width: 512,
      height: 512,
      prompt: "a spaceship into the ocean, dramatic lights, hd",
      scheduler: "K_EULER",
      num_outputs: 1,
      guidance_scale: 7.5,
      num_inference_steps: 50
    }
})

console.log(out)

​
Fine-tune models with your own data

You can improve open-source models with your own data to create new models that are better suited to specific tasks.

Image models like SDXL can generate images of a particular person, object, or style.

// Please make sure that you run 'npm install magicpoint@latest'
import Magicpoint from 'magicpoint';

const mp = Magicpoint.init({
    API_KEY: process.env.MP_API_KEY
})

const training = mp.training.create("sdxl-training:13853038482995843", {
    input: {
        "input_images": "https://example.url/pictures.zip
    },
    destination: "eftal/sdxl-finetuned-with-pictures"
})

console.log(training)

// When Traning completed you can call your custom inference api 
const output = mp.run("eftal/sdxl-finetuned-with-pictures", {
    input: {
        width: 512,
        height: 512,
        prompt: "a spaceship into the ocean, dramatic lights, hd",
        scheduler: "K_EULER",
        num_outputs: 1,
        guidance_scale: 7.5,
        num_inference_steps: 50
    },
})

console.log(output)

​
Magicpoint CLI

With Magicpoint CLI you can call custom or public AI/ML model with just a command.

magicpoint-cli
magicpoint run modelId:modelVersion -input `{"width": 512,"height": 512,"prompt": "a spaceship into the ocean, dramatic lights, hd","scheduler": "K_EULER","num_outputs:" 1,"guidance_scale": 7.5,"num_inference_steps": 50}`

That’s it!

​
Console

The easiest way! You can call public or custom model via console. Select your model and you can use ▶ Demo tab on the page.

Autoscaling and Concurrency