Building SiteGen AI v2: From a Website Generator to an AI Software Engineer
After migrating SiteGen AI to a more flexible AI architecture, I spent some time thinking about the future of the project.
The question I kept asking myself was simple.
"Do I really want to build another AI website generator?"
The answer was no.
There are already many tools capable of generating beautiful landing pages and portfolio websites.
If SiteGen AI was going to continue evolving, it needed a bigger purpose.
A Different Vision
Instead of generating only websites, I wanted SiteGen AI to generate complete software projects.
When developers start a new project, they don't just create an HTML file.
They create an entire application.
That application usually contains:
- Frontend
- Backend
- Database
- Authentication
- REST APIs
- Configuration files
- Documentation
- Deployment setup
Generating only the frontend solves a small part of the development process.
I wanted SiteGen AI to solve much more than that.
Thinking Like a Software Architect
I realized that code generation shouldn't be the first step.
Planning should be.
Before writing a single line of code, every software project begins with understanding the requirements.
What type of application is this?
Which technology stack should be used?
What database is appropriate?
How many user roles are required?
Which APIs should exist?
These questions are normally answered by software architects.
I wanted SiteGen AI to answer them automatically.
The New Architecture
Instead of sending one prompt to an AI model and expecting a complete application, I started designing a workflow where different parts of the project are generated independently.
The process begins by analyzing the user's requirements.
Once the project is understood, separate modules generate different parts of the application.
For example:
- Project Planning
- Frontend Generation
- Backend Generation
- Database Design
- Authentication
- API Development
- Documentation
- Deployment Configuration
Each module focuses on a single responsibility.
This makes the generated projects more organized and easier to maintain.
Why This Matters
As projects become larger, a single AI response becomes difficult to manage.
Breaking the work into smaller tasks improves consistency and allows each part of the application to be generated with more context.
It also makes future improvements much easier.
If I want to improve backend generation, I don't have to modify the frontend generation process.
Each module can evolve independently.
The Long-Term Goal
My goal is no longer to generate simple webpages.
I want SiteGen AI to become a development assistant capable of creating production-ready software projects from natural language.
A user should be able to describe an idea, and the platform should generate a complete project structure that developers can continue building upon.
This is an ambitious goal, but every new feature brings the project one step closer.
What I've Learned
One of the biggest lessons from this journey is that software architecture matters more than writing code quickly.
Good architecture allows a project to grow.
Poor architecture eventually limits what the project can become.
Building SiteGen AI has taught me much more than AI integration.
It has helped me understand project planning, modular design, scalability, and maintainability.
Looking Ahead
SiteGen AI is still evolving.
There are many challenges ahead, and many features I want to build.
In future articles, I'll share the progress of SiteGen AI v2, the technical decisions behind it, and the lessons I learn while building it in public.
This project started as a simple experiment.
Today, it has become one of the most valuable learning experiences of my software development journey.
Thank you for reading.