Intelligent BIM Assistant by Leveraging Large Language Models

Abstract

This thesis focuses on developing an AI-powered tool that simplifies the exploration of Building Information Modeling (BIM) data for Solibri users by leveraging the Neo4j graph database and OpenAI's large language model (LLM). The system translates natural language prompts into meaningful Cypher queries executed on the Neo4j database, enabling advanced search and validation capabilities without requiring users to have expert knowledge of filters and rules. The project encompasses comprehensive data conversion and cleaning, thoughtful database architecture and modeling, and the creation of a robust data insertion pipeline. The end result is an AI agent that generates accurate Cypher queries from human prompts, streamlining BIM workflows through efficient element discovery and compliance verification. This work underscores the potential of LLMs to transform BIM applications by reducing manual workload and allowing professionals to focus on more complex, human-centric tasks.