Compactifications of cluster varieties and convexity

In [GHKK18], Gross-Hacking-Keel-Kontsevich discuss compactifications of cluster varieties from "positive subsets" in the real tropicalization of the mirror. To be more precise, let $\mathfrak{D}$ be the scattering diagram of a cluster variety $V$ (of either type -- $\mathcal{A}$ or $\mathcal{X}$), and let $S$ be a closed subset of $\left(V^\vee\right)^{\text{trop}}(\mathbb{R})$ -- the ambient space of $\mathfrak{D}$. The set $S$ is positive if the theta functions corresponding to the integral points of $S$ and its $\mathbb{N}$-dilations define an $\mathbb{N}$-graded subalgebra of $\Gamma(V, \mathcal{O}_V)[x]$. In particular, a positive set $S$ defines a compactification of $V$ through a Proj construction applied to the corresponding $\mathbb{N}$-graded algebra. In this paper we give a natural convexity notion for subsets of $\mathfrak{D}$, called "broken line convexity", and show that a set is positive if and only if it is broken line convex. The combinatorial criterion of broken line convexity provides a tractable way to construct positive subsets of $\mathfrak{D}$, or to check positivity of a given subset.