Code Assist

Project Overview

Code Assist is an online building code companion. It excels at addressing all your building code inquiries. Powered by cutting-edge LLMs and a robust vector database (Pinecone), this app ensures precise and reliable results.

The Challenge

Navigating building codes is a complex and time-consuming task for professionals in the architecture, engineering, and construction (AEC) industry. Traditional methods of code lookup are often inefficient, requiring manual searches through extensive documents. This can lead to delays in project timelines, potential oversights in code compliance, and increased risk of errors. Moreover, interpreting code language and applying it to specific project contexts can be challenging, especially for less experienced professionals.

The Solution

Code Assist addresses these challenges through an innovative AI-powered approach: 1. Natural Language Processing: Utilizes advanced LLMs to understand and interpret user queries in natural language. 2. Vector Database Integration: Employs Pinecone, a robust vector database, to store and quickly retrieve relevant code information. 3. Contextual Understanding: Capable of interpreting building code in the context of specific project requirements. 4. Instant Code Lookup: Provides immediate access to relevant building code sections, eliminating the need for manual searches. 5. Explanation and Clarification: Offers explanations of code requirements in plain language, making interpretation easier. 6. Cross-Reference Capability: Identifies and presents related code sections that may be relevant to the user's query. 7. Regular Updates: Maintains an up-to-date database of building codes to ensure accuracy and compliance with the latest standards. 8. User-Friendly Interface: Designed for ease of use, allowing professionals to quickly get the information they need. By leveraging AI and advanced data storage techniques, Code Assist aims to streamline the process of building code consultation, improve accuracy in code interpretation, and ultimately enhance efficiency and compliance in the AEC industry.

Results

Code Assist has shown promising results in its initial deployment. Users report significant time savings in code lookup processes, with some estimating a 60% reduction in time spent on code-related queries. The accuracy of responses has been particularly praised, with users noting the app's ability to provide relevant code sections and explanations. While long-term impact data is still being collected, early feedback suggests improved confidence in code compliance among users. The app continues to be refined based on user interactions, with ongoing improvements in its language understanding and response accuracy.

Technologies Used

• LLMs
• Pinecone
• Natural Language Processing
• React
• Node.js