This blog focuses on Agentic workflow created during the workshop in creating an “IFC Compliance” checking app through different tools like Claude code, Cursor, Github, Gradio, Python and Js.
The objective of creating this app was simple and straight to the eye, Create an app that can read the models and replace the conventional method for planners, agents and government to check the design of buildings manually. One of the key objectives was product management learnt alongside the technical skills.
Project Management through Agents
As conventional project management would have lead to 1 year of intense work in order to create such app, but we learnt how to use agents in PRD and establish skills for agents to learn co-creating.
We created a start-up teams which collaborates together in creating a product. Our start-up named as “Garage coders” who tries to create product from waste. And our team focused on structural compliance check and also wrapping up the entire app into an explainable RAG agent which helps non-technical audience to use the app.
Structural compliance check
One of the major challenge was reading the 3d model and converting it to a standard AI readable file.
After creating an independent gradio app which checks the structural compliance with IFC model, we connected this individual platform to the main product.
Retrieval-Augmented Generation
Now, came across one of the gaps in connecting this app with audiences across industry, this dealt with difficulty of navigation and functions to check compliance. As the user was confused with too many data structures, we came up to create a AI bot which explains everything that happens within the app, but the problem is the AI tokens cost and hallucination that it leads to breaking out of the scope of the app.
In order to address this gap, we created a RAG agent which Retrieval Augmented Generation which picks the specific file for the required explanation rather than surfing all context and wasting all our tokens.
Finally, we connected the RAG agent to the Chatbot interface for easy user interactions.
Github Link: https://github.com/iaac-maai/structure_compliance_check
IFC Comliance check app: https://ifcore-platform.tralala798.workers.dev/projects/c542fd50-7649-4f25-a875-9212a890bc48
