GaussianAPI: RESTful web service for fitting repulsive potentials in density-functional tight-binding with Gaussian process regression

GaussianAPI is a RESTful web service that automates the construction of repulsive potentials for density-functional tight-binding (DFTB) simulations using Gaussian process regression (GPR). The repulsive interaction term in DFTB cannot be derived directly from first principles and requires parameterization based on reference data, which is traditionally a complex and manual process. GaussianAPI addresses this challenge by providing a data-driven and reproducible workflow that transforms reference datasets obtained from quantum chemistry calculations into Slater–Koster (SK) parameter files. The service accepts datasets containing atomic structures, DFT reference forces, and DFT-DFTB force residuals through HTTP requests, either via direct upload or remote access, and performs the fitting procedure using GPR. The generated .skf files encode only the fitted repulsive interaction potential and must be combined with the same existing Slater–Koster electronic parameter set used to compute the DFTB forces in the training data, for complete use in DFTB simulations.