Getting started (effective use)

Related: Privacy tuning playbook (technical/privacy_tuning_playbook.md), Federation strategies (technical/federation_strategy_guide.md), and API reference (technical/api_reference.md).

Before you begin - Install Docker and docker compose - Optional: set GRAFANA_ADMIN_USER and write Grafana password to deploy/grafana/credentials/admin_password.txt

Start the stack - Start services: - docker compose -f deploy/docker-compose.yml up -d - Open API docs: http://localhost:8000/docs - Open Grafana: http://localhost:3000 (use the admin password at deploy/grafana/credentials/admin_password.txt)

Run a privacy‑preserving training 1) Register participants (or use simulated ones) 2) Pick a privacy level (epsilon); start with a balanced default 3) Start a federated session; monitor accuracy and privacy budget 4) Export a compliance report for stakeholders

Tips for success - Start with higher epsilon (weaker privacy) to validate utility, then tighten - Watch epsilon consumption and learning curves in Grafana - Use Krum if you expect bad actors; Trimmed Mean for robust averages - Keep audit logs; include report in review packets

What next - See technical/guide.md for configuration details and Kubernetes pointers. - Ready to train? See technical/train_model.md for a step‑by‑step walkthrough.