Project Structure

Slothub Project Structure

The project is structured as a mono-repository containing 4 main independent microservices. Each service can be developed and scaled individually. Below is an overview of the directories:

Slothub/
├── frontend/
│   ├── src/                 # React components and pages
│   ├── public/              # Static assets
│   ├── package.json         # Node.js dependencies
│   └── .env                 # API Base URLs
├── backend/
│   ├── src/main/java        # Spring Boot Java source code
│   ├── src/main/resources   # application.yml and static assets
│   ├── pom.xml              # Maven dependencies
│   └── .env                 # Database credentials & JWT config
├── AI/
│   ├── main.py              # FastAPI application entrypoint
│   ├── services/            # OpenAI and S3 handling scripts
│   ├── requirements.txt     # Python dependencies
│   └── .env                 # OpenAI API key, S3 Bucket info
├── agent-core/
│   ├── app.py               # LangGraph state definition & Agent logic
│   ├── tools/               # Bedrock Agent tools (RAG, roadmap, quiz)
│   ├── requirements.txt     # Python dependencies
│   └── .env                 # AWS credentials and Local Dev flags
└── docker-compose.yml       # Shared PostgreSQL database with pgAdmin

Understanding the Components

  • frontend/: The user interface built with React. It dynamically renders conversational widgets sent back from the agent using a specific protocol (>>>[UI_WIDGET:TYPE|args]<<<).
  • backend/: The heart of logical operations. It handles User and Class hierarchy and issues JWT tokens via OAuth2 resource server architecture.
  • AI/: The content factory. When teachers upload PDFs, this service analyzes them using OpenAI and generates study materials and questionnaires. The resulting files are safely pushed to AWS S3.
  • agent-core/: The cognitive center (Slozy AI). Running with LangGraph framework, it processes student queries in real-time. By connecting to the shared database, it retrieves educational context through pgvector extension to implement RAG workflows.

In the next section, we will start by initializing our shared infrastructure using Docker and AWS.