public landing page
login home page
chat UI with civit

Civit – Your Civil Engineering AI Assistant

Inspiration

As a civil engineering student who’s also passionate about AI and full-stack development, I constantly found myself needing instant guidance on technical topics like structural analysis, concrete design, hydraulic flow, and AutoCAD scripting. Traditional resources are either scattered, outdated, or too rigid. Inspired by ChatGPT, I envisioned an AI assistant specifically tailored for civil engineers, capable of understanding context, solving engineering problems, and providing real-time assistance—whether for study, design, or field work. That vision became Civit.

What it does

Civit is an AI-powered assistant that functions like ChatGPT, but with prompt desing specific for civil engineering. It allows users to:

Ask civil engineering questions in natural language (e.g., "What’s the formula for critical depth in open channels?")
Get design guidance on concrete structures, slabs, columns, beams, and more
Solve hydraulic problems like pipe sizing, flow estimation, or Manning’s equation
Receive AutoCAD Lisp code or suggestions for structural drafting
Access summarized code clauses, formulas, and fieldwork best practices

It’s built to support students, professionals, and field engineers who need quick, reliable, and intelligent help.

How we built it

Frontend: Next.js + TailwindCSS + DaisyUI for a clean, ChatGPT-inspired interface
AI Agent: Integrated OpenRouter.
Prompt Engineering: Carefully crafted system prompts to steer GPT toward civil-related reasoning
Auth: JWT authentication using jose (Edge Runtime compatible)
Deployment: Developed inside Bolt IDE, exported to GitHub, deployed to Netlify

The project is modular and easily scalable to add new AI tools like unit converters, live drawing interpreters, or engineering calculators.

Challenges we ran into

Edge Runtime limits: Packages like bcryptjs failed under Edge runtime. I had to refactor logic and switch to Edge-compatible libraries or move code to Node runtime.
Bolt deployment issues: Misconfigured base and publish directories caused Netlify and Bolt IDE to throw "no such file" errors.
Token signing: Needed to switch from jsonwebtoken to jose for Edge compatibility, and rework JWT handling logic
Narrowing AI focus: Prompt engineering was tricky—GPT had to stay on-topic for engineering and not drift into general-purpose answers

Accomplishments that we're proud of

Created a working full-stack AI assistant in a limited hackathon window
Designed a specialized AI UX for engineering rather than general chat
Solved Edge Runtime problems and successfully deployed with authentication
Built an assistant that understands the field I study and can help others like me