comokit4py: Python package brings COMOKIT COVID-19 simulations to HPC

Agent-based models (ABMs) simulate autonomous agents interacting in virtual environments, commonly used for urban segregation, opinion dynamics, and epidemiological crises. During the COVID-19 pandemic, models like COMOKIT—designed to simulate everyday life in Vietnamese cities and test non-pharmaceutical interventions—proved impactful on global policy. However, running such models at scale demands enormous computational resources to replicate simulations across large parameter spaces.

comokit4py, presented by Arthur Brugière and Kévin Chapuis in arXiv:2605.23948, addresses this bottleneck. The Python package provides a streamlined interface to generate, explore, and analyze COMOKIT experiments on HPC infrastructure. By abstracting cluster complexities, it enables researchers to run thousands of simulation replicates efficiently, accelerating insight into epidemic spread and intervention effectiveness. The paper includes 6 pages and 2 figures, submitted to IEEE, highlighting the tool's potential for scalable social simulation.