CODE-COMPANION: A Cross-Platform Computational Framework for Real-Time BIM Compliance Checking and Digital Building Permit Workflows
TL;DR
CODE-COMPANION is presented, an AI-assisted computational framework that transforms natural-language regulations into structured, machine-readable rules and evaluates them deterministically within Building Information Modeling (BIM) environments and transforms compliance checking from a post-design verification task into a proactive design-support process, enabling architects to resolve issues early.
Abstract
Building permit approval remains one of the most time-consuming and inconsistent processes in the Architecture, Engineering, and Construction (AEC) industry, requiring designers and municipal reviewers to manually interpret planning regulations and building codes. This paper presents CODE-COMPANION, an AI-assisted computational framework that transforms natural-language regulations into structured, machine-readable rules and evaluates them deterministically within Building Information Modeling (BIM) environments. The framework formalizes selected regulatory requirements as deterministic computable rules and connects these rules to geometric and parametric information derived from the architectural model. The checking workflow was initially prototyped in Rhino 8 SR29 (v8.29.26063.11001) using Grasshopper (build 8.29.26063.11000) and subsequently implemented in Autodesk Revit 2026.4 using Python 3.13.12 and pyRevit 6.1.0 to provide interactive design-time compliance feedback. Validation was conducted using a real-world case study in Kadikoy, Istanbul, where two alternative building designs were assessed under identical parcel-specific regulations. The prototype successfully identified compliance violations, highlighted non-compliant elements within the 3D model, and generated explainable pass/fail results reporting the evaluated rule, measured value, and governing threshold. By providing continuous, real-time regulatory feedback during design, the framework transforms compliance checking from a post-design verification task into a proactive design-support process, enabling architects to resolve issues early, evaluate design alternatives more efficiently, reduce redesign iterations, shorten permitting time, and make more informed decisions. The results demonstrate that a substantial proportion of deterministic permit regulations can be evaluated automatically while discretionary regulations remain under human judgement.