CodeQuery allows users to search California building codes with ease using ChatGPT-4. In addition to providing summaries of complex codes, CodeQuery responses include relevant sections of the code for the user to reference. At the core, our product raises awareness of safety issues in our community while also reducing barriers to access and understanding complex legal standards that govern our community spaces. Users are able to research and understand building codes as needed. Users served by our product include:
Renters Homeowners Inspectors & Government Officials Architects Construction companies Real-estate investors or Agents For example, homeowners interested in renovations may need to know about the permitting and inspection process. Renters may want to inquire about the safety of appliances in their apartment. Real estate investors may want to inquire about states with similar codes for future investment projects. The use cases are many -- CodeQuery is the ultimate AI assistant to simplify building codes.
We built CodeQuery using Python scripts to parse a 4000+ page PDF document. Embeddings were created using HuggingFace (Roberta) which were given to Pinecone for vectorization and memory storage. ChatGPT-4 interacts with Pinecone based on user prompts to yield accurate summaries, relevant additional information and specific sections of interest. Our front-end is hosted locally using Django.