RAG system with chat
Thomas Lorenz ZT: Local RAG system for regulatory planning compliance
Civil engineers must consider a multitude of legal requirements and guidelines during the planning process . We had the opportunity to develop a local RAG (Reference, Consultation, and Approach) system for Thomas Lorenz ZT, an internationally active civil engineer , which provides quick answers to important planning questions via chat.
Task
Civil engineers must consider a multitude of legal requirements and guidelines during the planning phase . This information is often scattered, difficult to access, and time-consuming to verify. Errors or omissions can lead to delays, additional costs, or legal risks.
The need for reliable, fast and local support in the planning process was therefore particularly high.
Customer & Sector
Thomas Lorenz ZT GmbH
Construction industry
Services
AI Consulting
Conception & Tech Consulting
Web App
Solution
We developed a local Retrieval Augmented Generation (RAG) system for Thomas Lorenz ZT that indexes all relevant documents and guidelines, searches them intelligently, and can be operated via chat.
Duration
Start: 11/2025 – ongoing
Tech stack
Angular
Postgre SQL
Ollama
Rust
The solution
We developed a local Retrieval Augmented Generation (RAG) system that indexes and intelligently searches all relevant documents and guidelines.
The AI provides precise, context-related answers including document sources to planning questions, thus directly supporting civil engineers in their workflow.
Thanks to the local infrastructure, the system remains fully compliant with data protection regulations and independent of external services.
Special challenges
This project required the reliable conversion of technical standards documents containing complex mathematical formulas, special characters, and demanding formatting into machine-readable formats . This necessitated adapting the LLM (Large Language Management) system accordingly.
Furthermore, it was necessary to analyze and summarize extensive bodies of documents thematically.
A particular challenge was handling technical drawings and graphics, as these cannot be meaningfully interpreted by LLMs/OCR. As a solution, the corresponding source documents are referenced for graphics.
Conclusion
With the new RAG system, civil engineers have a powerful tool at their disposal that accelerates planning processes, reduces errors, and increases legal certainty. The solution brings clarity to a complex regulatory environment and ensures noticeable efficiency gains in daily work.
Team/Credits
AI consulting: Robert Reihs, Leonhard Brunner-Plasch
PM: Anita Brunner-Irujo, Zaim Sarajlic
Development: Leonhard Brunner-Plasch, Helmut Hasler
Text
Would you like to know more about this project?
