get an easy http api for ONNX models on android,windows,linux,mac,ios with native acceleration.

> moxgo serve

the builtin cli makes it easy to play with popular classification models

> moxgo pull birdnet
> moxgo run birdnet --file what_bird_is_this.mp3

> moxgo pull yolo11n
> moxgo run yolo11n --file is_this_a_dog.png

you can add your own ONNX models without publishing them to the registry. place the model files in ~/.moxgo/models/<name>/ with a manifest.json and a model.onnx.

~/.moxgo/models/my-model/
├── manifest.json
├── model.onnx
└── labels.csv        (optional)

the manifest describes inputs, outputs, and runtime options:

{
  "inputs": {
    "images": {
      "type": "image",
      "width": 640,
      "height": 640
    }
  },
  "outputs": {
    "output0": {
      "labels": "coco80"
    }
  }
}