Aegis in 5 minutes (Quick tour)

What you’ll see - A running privacy-preserving ML stack - One federated training run - Live dashboards and a compliance report

1) Start Aegis - Make sure Docker is running - Start services: docker compose -f deploy/docker-compose.yml up -d (or docker-compose -f deploy/docker-compose.yml up -d) - Open API docs: http://localhost:8000/docs

2) Register two participants - http POST :8000/participants X-Role:admin client_id=c1 key_hex=aa - http POST :8000/participants X-Role:admin client_id=c2 key_hex=bb

3) Set a balanced privacy level - http POST :8000/dp/config X-Role:operator clipping_norm:=1.0 noise_multiplier:=1.0 sample_rate:=0.01 delta:=1e-5 accountant=rdp

4) Choose a federated strategy - http POST :8000/strategy X-Role:operator strategy=trimmed_mean

5) Start training (3 rounds) - http POST :8000/training/start X-Role:operator session_id=tour rounds:=3

6) See it live - Grafana: http://localhost:3000 (default dashboard is provisioned)

7) Create a report - http GET :8000/compliance/report X-Role:viewer | jq -r .markdown > tour_report.md - Alternative with curl: - curl -fsS -H 'X-Role: viewer' http://localhost:8000/compliance/report | jq -r .markdown > tour_report.md

Next: basics/privacy_explainer.md for an approachable overview of DP and FL.