The whole AEC industry is challenged by the high CO2 footprint caused by the increasing population and construction. Introducing regulations to the construction industry is a solution to reduce CO2 footprint. The workflow behind Leed mini consultant goes as follows: starting with gathering the regulation in a dataset and setting it up, training a rag with it, then feeding images of leed buildings to image to text models, and lastly comparing text with leed and going again to image.

Starting with the first step, we used LM studio to train our models and dataset in pdf format. In details, We loaded the model in LM studio then load the embedding model and then go to vs code, there we input our pdf and text files, transform them to vector encoding  train the rag with LM Studio and get in vector encoding the leed consultant.

For the dataset we used 3 pdf official files – the LEED Certification guidebook – the LEED Building Design and Construction AND the LEED Rating System for Green Buildings.

After that we analyse our pdfs to convert them into txt format. And here we can see that from this process the text that came out needed some cleaning and transformation…like removing unnecessary spaces before feeding it into the RAG.

Text in chunks and then converted to vectors.

In the third step we will explore the image to text method to extract LEED specific sustainable characteristics from building images.