Distributed AI client

Mix.install(
  [
    {:kino_bumblebee, "~> 0.5"},
    {:exla, "~> 0.9", override: true}
  ],
  config: [nx: [default_backend: EXLA.Backend]]
)

Connect to server

server_node_inputs =
  ["A", "B", "C"]
  |> Enum.into(%{}, fn node_id ->
    {
      node_id,
      %{
        node: Kino.Input.text("SERVER_#{node_id}_NODE_NAME"),
        cookie: Kino.Input.text("SERVER_#{node_id}_COOKIE")
      }
    }
  end)

server_node_inputs
|> Enum.map(fn {_, inputs} ->
  [inputs.node, inputs.cookie]
end)
|> List.flatten()
|> Kino.Layout.grid(columns: 2)

server_node_inputs
|> Enum.map(fn {_, inputs} ->
  node_name =
    inputs.node
    |> Kino.Input.read()
    |> String.to_atom()

  cookie =
    inputs.cookie
    |> Kino.Input.read()
    |> String.to_atom()

  Node.set_cookie(node_name, cookie)

  Node.connect(node_name)
end)

Call server AI

image_input = Kino.Input.image("Image", size: {224, 224})
form = Kino.Control.form([image: image_input], submit: "Run")
frame = Kino.Frame.new()

Kino.listen(form, fn %{data: %{image: image}} ->
  if image do
    Kino.Frame.render(frame, Kino.Text.new("Running..."))

    image = image.data |> Nx.from_binary(:u8) |> Nx.reshape({image.height, image.width, 3})

    output = Nx.Serving.batched_run(ViT, image)

    output.predictions
    |> Enum.map(&{&1.label, &1.score})
    |> Kino.Bumblebee.ScoredList.new()
    |> then(&Kino.Frame.render(frame, &1))
  end
end)

Kino.Layout.grid([form, frame], boxed: true, gap: 16)