beet 0.0.8

Malleable Application Framework
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52

//! # Hello ML
//! A popular 'hello world' for machine learning in games is sentence similarity,
//! where models rank the similarity of sentences.
//! This example uses a locally run *small* language model to
//! select the child behavior with the most similar sentence.
//!
//! So if we ask the Bert to run the child with the nearest sentence
//! to "please kill the baddies", which action do you think it will take?
//!
//! Note that the [Name] components are for debugging only, its the
//! [Sentence] components that are used for the actual behavior selection.
use beet::prelude::*;

// #[rustfmt::skip]
pub fn main() {
	App::new()
		.add_plugins((
			MinimalPlugins,
			AssetPlugin::default(),
			ControlFlowPlugin::default(),
			DebugFlowPlugin::default(),
			RunOnAssetReadyPlugin::<Bert>::default(),
			LanguagePlugin::default(),
		))
		.add_systems(Startup, setup)
		.run();
}

#[rustfmt::skip]
fn setup(mut commands: Commands, asset_server: Res<AssetServer>) {
	let bert = asset_server.load("ml/default-bert.ron");
	commands
		.spawn((
			Name::new("Hello ML"),
			HandleWrapper(bert.clone()),
			RunOnAssetReady::<Bert>::new(bert),
			NearestSentence::new(),
			Sentence::new("please kill the baddies"),
			children![
				(
					// this wont run
					Name::new("Heal Behavior"),
					Sentence::new("heal")
				),
				(
					// this will run, closer sentence similarity
					Name::new("Attack Behavior"),
					Sentence::new("attack")
				)
			]
		));
}