Build Your Own AI Chat Interface
Your Students. Your Pedagogy. Your Code.
Why I Made This Site
In late 2024, I spent five intense days working to build a customized chat application for my students that I call Babbleborg. I’m an educator who knows just enough about code to be dangerous, so I was shocked by the quality of what AI helped me create. The application worked. My students liked it. It’s still running at my school today.
I wanted to share what I built, but sharing code I can’t maintain feels irresponsible. So I’m trying something different: what if I share how to build it instead? This site is an experiment, a guide that empowers technically savvy educators to create their own AI chat interfaces, perfectly tailored to their students and pedagogy.
I’ve documented every architectural decision, broken the build into testable stages, and created context files you can feed directly to your coding assistant. The result? You build it yourself, you own the code, and you control exactly how your students interact with AI. I hope you enjoy the process as much as I did!
— Tom Gromak
Why Build Your Own?
Commercial AI tools and default interfaces are designed for general use, not educational contexts. They often promote AI dependence rather than AI literacy, use default system instructions you can’t control, and train on your students’ conversations.
This guide helps you build something better: an AI learning environment with complete pedagogical autonomy.
As educators, you know your students better than any distant corporation. This approach gives you:
Complete Control Over System Instructions
Shape the AI's behavior to support authentic learning, not just provide answers. Design for your student's specific needs, not sycophancy and friction-free task completion.
No Corporate Training on Student Data
API-based architecture means conversations aren't used for model training. Student interactions remain private and are never harvested to improve commercial products. Learn more →
Student Data Ownership
All conversations are stored locally on student devices, not in the cloud. Students control their own data and can export or delete it at any time.
Customizable Learning Supports
Build modular instructional strategies like simplified language for English learners, step-by-step scaffolding, or Socratic questioning that you can mix and match for different contexts.
No Authentication Complexity
Simple, fast deployment without login systems. Each installation serves a single class or use case, eliminating the overhead of user management and permissions.
Dramatic Cost Savings
Pay pennies per conversation instead of $5-15 per student per month for commercial AI tools. Costs scale with actual usage, not seat licenses. Learn more →
Battle-Tested in Real Classrooms
This architecture has been successfully deployed across classes spanning middle school through high school, including:
- Technology and computer science courses
- Global studies and humanities
- Learning support programs
- College counseling
- General academic support
The approach works because it prioritizes educator autonomy and student learning over convenience, shortcuts, and dependence.
Who This Guide Is For
This is not a beginner-friendly tutorial. You should be comfortable with:
- Shared hosting environments (cPanel, Plesk, or similar)
- Basic PHP concepts and file structures
- Setting environment variables
- Reading error logs and troubleshooting
- Working with command-line AI coding assistants
Want to get fully up to speed? Use the Learning Roadmap to fill in any gaps.
Great for: Tech-savvy educators, instructional technologists, ed-tech coordinators, or technically proficient homeschool parents working in educational contexts.