Adaptive teaching platform designed to train teachers on student mental health topics using RAG-powered quizzes with personalized difficulty scaling.

Follow these steps to set up the project on a new machine:

git clone https://github.com/dimitramav/wla.git
cd wla
python3 -m venv .venv
source .venv/bin/activate

# Install CPU-only PyTorch
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu

# Install remaining requirements
pip install -r services/requirements.txt

# Install API dependencies
cd api && npm install

# Install Web dependencies
cd ../web && npm install

You can start all services at once using the provided scripts:

# Start all services (MongoDB, Express, FastAPI, Vite)
./scripts/start.sh

# Start services without the frontend
./scripts/start.sh --no-frontend

# Stop all services
./scripts/stop.sh

The start.sh script handles health checks and environment validation automatically. Logs for each service are piped to /tmp/wla-express.log, /tmp/wla-fastapi.log, and /tmp/wla-vite.log.

If you prefer to start services individually in their own terminals:

1. MongoDB: sudo systemctl start mongod
2. Ollama: Ensure Ollama is running and gemma2:9b-instruct-q4_0 is pulled.
3. Express API: cd api && npm run dev (http://localhost:3001)
4. FastAPI RAG Service:
   • cd services && ../.venv/bin/python -m uvicorn api.main:app --factory --port 8000
   • Or run all cells in services/server.ipynb
5. React Web Frontend: cd web && npm run dev (http://localhost:5173)

The project ships with two benchmarks (RAG retrieval and LLM generation) plus offline rescoring and investigation scripts. See services/benchmarks/README.md for prerequisites, how to run each benchmark, and output layout.

• What is measured and why: services/benchmarks/methodologies/
• Results and findings: services/benchmarks/reports/

This project requires the following to be installed separately:

• MongoDB: Installation Guide
• Ollama: Download Ollama