FedMed Flower SuperNode

This plugin packages a Flower SuperNode inside a ChRIS-friendly container. Run the commands below from inside plugins/supernode_plugin/. The actual training loop only knows how to fit a local logistic regression model; Flower merely orchestrates the parameter exchange.

Build

docker build -t fedmed/pl-supernode .

Run (example)

docker run --rm --name fedmed-supernode --network fedmed-net \
  -v $(pwd)/demo/client-in:/incoming:ro \
  -v $(pwd)/demo/client-out:/outgoing:rw \
  fedmed/pl-supernode \
    fedmed-pl-supernode \
      --cid 0 \
      --total-clients 3 \
      --superlink-host ${FEDMED_SUPERLINK_IP} \
      --data-seed 13 \
      /incoming /outgoing

Make sure ${FEDMED_SUPERLINK_IP} is populated via docker inspect fedmed-superlink (see the SuperLink README). After the Flower round completes, the plugin writes client_metrics.json into /outgoing so downstream ChRIS components (or you) can inspect the results.

Accepted parameters

• --cid (default: 0)
• --total-clients (default: 3)
• --superlink-host (default: fedmed-pl-superlink)
• --data-seed (default: 13)

All other options (SuperLink port 9092, ClientApp host 0.0.0.0/port 9094, metrics filename, transport, FLWR state path) are now fixed to defaults and no longer exposed as CLI flags.