HiveWatch¶
HiveWatch is a lightweight observability toolkit for federated and distributed machine learning workloads. It helps you capture round-level metrics, client updates, communication costs, and geographic activity on a live map without forcing you into a single training framework.
With HiveWatch you can:
log client and round metrics from custom training loops,
stream and replay runs locally with the built-in map dashboard,
send metrics to Weights & Biases, MLflow, or both, and
integrate observability into APPFL workflows with minimal glue code.
Technical Components¶
HiveWatch is organized around a few clear technical components.
Runtime API
A small instrumentation API for logging round starts, client updates, round summaries, failures, and checkpoints.
SSE Emitter
Streams local events, writes replayable run artifacts, and powers the built-in map dashboard workflow.
WandB Emitter
Sends round metrics, per-client metrics, and alert-style events to Weights & Biases.
MLflow Emitter
Logs tracking metrics, tags, and artifacts to local or remote MLflow servers.
Map Dashboard
Visualizes client geography and run progress from saved .jsonl and
.map.json artifacts.
APPFL Integration
Fits into APPFL server-side orchestration so metrics can be emitted during real federated runs.