ATLAS: A software for interactive modeling and analysis of homogeneous and heterogeneous graphs

Despite the increasing adoption of graph-based learning methods, the integration between interactive graph modelling and programmatic graph analytics remains limited. We present ATLAS, a general-purpose software that bridges interactive graph modelling with modern graph analytics and graph-learning workflows. The software is implemented in Python and supports the interactive construction of both homogeneous and heterogeneous graphs through a graphical interface. It enables the definition and customisation of nodes and edges, the realtime computation of structural metrics, and the automatic generation of executable Python code compatible with major graph libraries, including NetworkX, igraph, PyVis, DGL, and PyTorch Geometric. In addition, existing code can be imported to reconstruct graph structures, supporting bidirectional integration between visual modelling and programmatic workflows. The software also facilitates the visual inspection of complex graph structures and graph-based explanations produced by Explainable Artificial Intelligence (XAI) methods. Two real-world application scenarios, based on the Cora and MovieLens datasets, demonstrate these functionalities for homogeneous and heterogeneous graphs, respectively. A quantitative evaluation of parsing time, rendering time, and memory consumption further characterises the computational behaviour of ATLAS as graph size and complexity increase. Our source code is publicly available on GitHub (https://github.com/cilabuniba/atlas)