brickkit

Inspiration🌟 Inspiration

Growing up, I always loved LEGO but often found myself limited by the few sets I owned. I realized that there are millions of kids who share that same problem — they have creativity but not the endless supply of bricks to express it. That’s what inspired AI LEGO Builder: a tool that lets anyone describe what they want to build and uses AI to generate a complete, buildable LEGO model using only the pieces they already have. It’s about unlocking creativity, not increasing cost.

⚙️ How I Built It

The core idea was to connect generative AI with the LDraw/LeoCAD ecosystem. The system takes a natural language prompt — like “a red race car with a spoiler” — and maps it to real LEGO parts from the LDraw database. It then structures those parts into a 3D model (.mpd/.ldr) that users can view, rotate, and download directly from the website. I used Python and async I/O for file handling and generation, Playwright + aiohttp for part lookups, and embedded LeoCAD to visualize the model online.

🧠 What I Learned

I learned how to bridge text-to-structure generation — translating human language into geometric constraints. Working with the LDraw format taught me a lot about 3D coordinate systems, matrix transformations, and hierarchical model structures. I also gained a deeper understanding of asynchronous programming and how to optimize AI-agent pipelines for performance.

🚧 Challenges

Getting the AI to generate physically buildable models was tough — especially ensuring proper part connections and orientation matrices. Integrating the generated .mpd files with LeoCAD’s viewer inside the web interface was another big hurdle. Finally, balancing speed and realism (so that builds load quickly while remaining detailed) required a lot of iteration and debugging.

💡 Reflection

I built this in just 24 hours for a hackathon, and it’s already capable of generating and visualizing complex LEGO builds from text. The next step is allowing users to specify their own brick collections — letting the AI design only with what they own. Beyond the tech and monetization potential, this project reminded me how powerful creativity becomes when accessibility meets AI.

Built With

• fastapi
• nextjs
• pydantic
• react