Abstract: To address the issues of variable selection, of model selection and, of accuracy improvement during the construction of surrogate models in space structure design, a general framework for an optimal predictive adaptive agent model is proposed. This framework is developed on the SiPESC platform, an open and plugin-oriented service software system, by employing established surrogate modeling design architectures and functionalities. In this framework, variable filtering services are deployed to identify variables that are strongly correlated and critically significant for the construction of surrogate models. Accuracy evaluation filtering services are then applied to determine the most suitable surrogate model algorithms. Moreover, adaptive sampling services are refined to augment the accuracy of the surrogate modal post-construction. Additionally, established mainstream variable filtering criteria and efficient adaptive refinement criteria are integrated into the framework, then an automated rapid design software is developed for optimal surrogate modeling in engineering problems. This study demonstrates that: the designed architecture effectively leverages existing surrogate modal algorithms, facilitating automated variable selection and surrogate modal algorithm determination, thereby significantly enhancing the efficiency of the surrogate modal construction process for space structure design.