@rbxts/a-plant-generator

Deterministic, renderer-independent procedural botany for roblox-ts. It combines bounded L-system derivation, stable 2D/3D turtles, plant topology, mesh builders, phyllotaxis, developmental animation, cellular map systems, IFS utilities, and optional Roblox renderers without depending on a game framework.

Status: 0.1.2. Core APIs are usable; cellular maps and EditableMesh adapters are explicitly experimental while their mechanisms are functional and tested.

Preset architectures	Golden-angle phyllotaxis
Growth progression	Inflorescence architectures

Install

npm install @rbxts/a-plant-generator
Copy

30-second quick start

import { PLANT_PRESETS, PlantCompiler, PlantGenerator } from "@rbxts/a-plant-generator";

const model = PlantCompiler.compile(PLANT_PRESETS["broad-canopy-tree"]!);
const plant = PlantGenerator.generate(model, {
	seed: 12345,
	iterations: 5,
	limits: { maxSymbols: 50_000, maxSegments: 5_000 },
});

print(plant.statistics.segments, plant.descriptor.modelHash);
Copy

The same model, seed, parameters, time, and limits produce the same topology and compact descriptor. Derivation never creates Instances.

Growth and rendering

import { PartPlantRenderer, PlantCompiler, PlantGenerator, PLANT_PRESETS } from "@rbxts/a-plant-generator";

const model = PlantCompiler.compile(PLANT_PRESETS["timed-tree"]!);
const result = PlantGenerator.generate(model, { seed: 9, iterations: 4 });

// Parent ownership is explicit; the renderer never assumes Workspace.
const renderer = new PartPlantRenderer({ parent: script, maxInstances: 2_000 });
const handle = renderer.render(result);
handle.setGrowth(0.65);
handle.setLod("medium");
// Later: handle.destroy(); renderer.destroy();
Copy

For frame-budgeted generation:

const session = PlantGenerator.createSession(model, {
	seed: 9,
	iterations: 6,
	limits: { maxWorkUnits: 250_000 },
});
while (!session.isComplete()) session.step(500);
const result = session.getResult();
Copy

Capabilities

• deterministic context-free, stochastic, context-sensitive, parametric, and bracketed rewriting;
• edge/node replacement graphs with explicit contact points;
• stable 2D/3D turtles, polygon capture, custom commands, tropism, and axial branch context;
• renderer-neutral branch graphs, buds, attachments, pruning, bounds, biomass, traversal, and hashes;
• tubes with parallel-transport frames, mesh validation/merge/transform, blades, petals, ribbons, disks, compound leaves, and flowers;
• planar/cylindrical phyllotaxis, deterministic jitter, spacing relaxation, and parastichy analysis;
• timed symbols, absolute-time scrubbing, event enumeration, and built-in, piecewise, keyframed, or registered growth functions;
• experimental planar/spherical/3D cellular structures and weighted/controlled IFS tools;
• versioned model specs, migrations, named behavior registries, immutable preset extension, compact network descriptors, LRU caches, and strict budgets;
• pooled/batched Part rendering with default or custom organ Parts, plus an injected, size-guarded EditableMesh capability boundary.

Architecture and performance

The pipeline is model spec -> compiled production index -> bounded derivation -> turtle -> branch graph -> optional mesh -> optional renderer. Core modules use plain data and never access Roblox services. See Architecture, Concepts, Book coverage, and Performance.

Exponential derivation is always a risk. Set symbol, segment, vertex, triangle, stack, iteration, Instance, and total work limits for untrusted or user-authored models. Prefer incremental sessions and distance LOD for interactive worlds.

Compatibility

• roblox-ts 3.0.0;
• TypeScript 5.5.3, matching the compiler's pinned TypeScript line;
• @rbxts/types 1.0.938 or compatible later API dumps;
• Node 20.19 or newer for development tools;
• EditableMesh requires the Roblox capability/security settings described in Roblox rendering.

Guides

Getting started · Examples · L-systems · Turtles · Tree modeling · Phyllotaxis · Animation · Networking · Extending · Security · Troubleshooting

API reference and all guides are published at mateusdcc.github.io/plant-generator.

Contributing and license

See CONTRIBUTING.md. Package code is MIT licensed. The Algorithmic Beauty of Plants and its text, figures, photographs, and illustrations are not included and are not licensed by this repository; see NOTICE.md and Academic references.