AgentiteG

Native performance primitives for Godot, designed for AI-assisted development.

AgentiteG provides high-performance, GDExtension-based operations for common game development hot paths. It's designed to be easy to use with AI coding assistants like Claude Code while delivering native C++ performance.

Features

Spatial Structures

• SpatialHash2D / SpatialHash3D - O(1) spatial queries using hash grids
• KDTree2D / KDTree3D - O(log n) nearest neighbor queries
• QuadTree / Octree - Adaptive spatial subdivision

Array & Math Operations

• ArrayOps - Filter, sort, reduce, select on PackedArrays
• MathOps - Batch vector/matrix operations, distance matrices
• BatchOps - Steering behaviors, flocking, velocity updates

Procedural Generation

• RandomOps - Bulk random generation, Poisson disk sampling, weighted choice
• NoiseOps - Perlin, simplex, Worley noise with FBM, ridged, turbulence variants

Grid & Pathfinding

• GridOps - Coordinate conversion, flood fill, FOV shadowcasting, distance fields
• PathfindingOps - A*, Dijkstra maps, Jump Point Search, flow fields

Collision & Geometry

• CollisionOps - Batch point-in-shape, circle/sphere collisions, ray casting
• GeometryOps - Convex hull, Delaunay triangulation, Voronoi, polygon operations

Interpolation & Statistics

• InterpolationOps - 30+ easing functions, bezier curves, Catmull-Rom splines
• StatOps - Mean, median, std dev, percentiles, histograms, outlier detection

Installation

From Release

1. Download the latest release for your platform
2. Copy the addons/agentiteg folder to your project's addons/ directory
3. Enable the plugin in Project Settings

From Source

# Clone with submodules
git clone --recursive https://github.com/yourusername/AgentiteG.git
cd AgentiteG

# Build
scons platform=<windows|linux|macos> -j4

# Output will be in project/addons/agentiteg/bin/

Quick Start

extends Node2D

var spatial: SpatialHash2D
var enemy_positions: PackedVector2Array

func _ready():
    # Create spatial hash with 64-unit cells
    spatial = SpatialHash2D.new()
    spatial.cell_size = 64.0

func _physics_process(delta):
    # Rebuild spatial hash (or update incrementally)
    spatial.build(enemy_positions)

    # Find all enemies within 200 units of player
    var nearby = spatial.query_radius(player.position, 200.0)

    # nearby is PackedInt32Array of indices into enemy_positions
    for i in nearby:
        var enemy = enemies[i]
        enemy.alert()

Documentation

• Getting Started
• API Reference
• Claude Code Guide

Performance

Operation	Items	GDScript	AgentiteG	Speedup
Radius query	10,000	~20ms	~0.15ms	130x
K-nearest neighbors	10,000	~25ms	~0.2ms	125x
Sort floats	10,000	~5ms	~0.3ms	15x
Filter values	10,000	~1.5ms	~0.08ms	18x
Batch normalize	10,000	~3ms	~0.1ms	30x
A* pathfinding	100x100	~15ms	~0.8ms	18x
Noise grid	256x256	~50ms	~2ms	25x

Interactive Benchmark Demo

See AgentiteG's performance in action with the interactive benchmark demo:

cd project
godot demos/interactive_benchmark/interactive_benchmark.tscn

Features:

• Real-time GDScript vs AgentiteG timing comparison
• 7 scenarios: Spatial Query, Array Filter, Distance, Steering, Flocking, Collision, Flow Field
• Adjustable entity count (100-10,000)
• Live visualization with FPS counter and frame time graph
• Keyboard shortcuts: 1-7 for scenarios, +/- for entity count

Philosophy

1. Zero game-specific logic - Pure algorithms, no architecture opinions
2. Arrays in, arrays out - Work with PackedArrays, return indices
3. Claude-friendly API - Simple signatures, clear documentation
4. Opt-in complexity - Simple defaults, advanced options available
5. Benchmark everything - Performance claims are tested

License

MIT License - see LICENSE

Contributing

See CONTRIBUTING.md for guidelines.