Orchestrating AI Agents for Building Performance: LangGraph, RAG & Real-Time Engineering Calculations

I plan to show an autonomous multi-agent system that automates building energy performance calculations. You will see:

The implementation of a stateful agent network using LangGraph
RAG system used for retrieving data from NASA radiation tables, and engineering codes, standards, and design guides
Integration of Python REPL tools to perform calculations
State management through a supervisor agent that controls the flow of the calculations and communication
MongoDB for persistent storage to save every user’s building information
Fernet encryption to protect the project’s sensitive information

During the live code walkthrough I will show the agent creation, state management, prompt engineering, and tool integration used to make the LLMs perform reliable engineering calculations.

This demo focuses on the core architecture, LangGraph state management, RAG, and engineering calculations with additional features planned for the future.