Systems and Model Packages

MonteCarloX keeps systems external on purpose. The framework provides algorithms; companion packages provide concrete models.

Why this split helps

• algorithm implementations stay generic
• model packages can evolve independently
• the same sampler can be reused across many domains

Example: SpinSystems with Ising

using Random
using MonteCarloX
using SpinSystems

rng = MersenneTwister(123)
sys = Ising([16, 16], J=1.0, periodic=true)
init!(sys, :random, rng=rng)
alg = Metropolis(rng; β=0.44)

for _ in 1:100_000
    spin_flip!(sys, alg)
end

println("E  = ", energy(sys))
println("|M| = ", magnetization(sys))

What a custom model package should implement

At minimum:

1. a concrete system type (AbstractSystem subtype)
2. observable functions for analysis
3. update methods that call MonteCarloX primitives (accept!, etc.)
4. initialization utilities (init! pattern)

This is usually enough to plug your model into measurement and algorithm workflows immediately.