Supplementary Materials for "Semantic Graph as Primary Program Database: A Graph-First Architecture for Agent-Oriented Programming"

This repository contains the analysis scripts and generated data for the paper "Semantic Graph as Primary Program Database: A Graph-First Architecture for Agent-Oriented Programming." The materials include: - ablation_study.py: Script to reproduce the graph-first vs. text-first agent tool-call reduction experiment (2.1× improvement reported in the paper)- comparison_framework.py: Code implementing the 12-dimensional comparison framework against 11 program representation approaches- complete_proofs.py: Implementation and validation of the round-trip semantic equivalence proofs- formal_model.py: Formal graph algebra specification (G = (V, E, Σ, λ, μ))- generate_figures.py: Script to regenerate all four figures from the paper- generate_summary.py: Statistical summary generation for the 247-program dataset- statistical_analysis.py: Statistical tests (Shapiro-Wilk, Kruskal-Wallis, Wilcoxon signed-rank) reported in the paper- Figure outputs (PDF): Architecture diagram, information fidelity distribution, radar comparison chart, and encoding efficiency density plot All data is derived from the ZeroLang semantic graph ecosystem. No raw user programs are included.