The modernization of power distribution grids, driven by the integration of DER, necessitates advanced modelling capabilities. A critical challenge is the semantic gap between the extensive geospatial asset data and the detailed electrical models required for engineering analysis. This data is often stored in static, unstructured KMZ formats that contain inherent topological errors. To address this, this paper presents a novel low-cost, Python framework that fully automates the conversion of this raw GIS data into solvable mathematical models for computation. This process generates executable files for industry-standard, script-based simulators such as MATPOWER and OpenDSS. The framework's core technical contributions include an XPath and Regex-based engine for metadata extraction. A graph-theory pipeline then utilizes a Minimum Spanning Tree (MST) algorithm to algorithmically heal topological disconnections. A Dijkstra-based method is then used for model abstraction. The core methodology was first validated using a bidirectional, reverse-engineering process on a 4-bus test case. This confirmed a lossless round-trip conversion to and from a data-rich KMZ format. Subsequently, it was applied to a complex, real-world 11 kV Kenyan distribution feeder. The generated model converged in a Newton-Raphson power flow. This demonstrated its utility as a powerful diagnostic instrument by enabling detailed feeder voltage profiling and loss analysis. Results were cross-validated against DIgSILENT PowerFactory, a graphic-based simulator, showing excellent numerical agreement. This study validates a scalable framework that transforms static, error-prone GIS data into dynamic, multi-platform diagnostic models. This approach provides a feasible pathway to accelerate grid modernization.