Files
foxhunt/crates
jgrusewski bb133d6619 feat(monitoring): add monitoring service for live training metrics over gRPC
New monitoring_service bridges Prometheus training/GPU metrics to gRPC,
enabling `fxt train monitor` TUI and web-dashboard live updates without
direct Prometheus dependency on clients.

Components:
- 6 new hyperopt Prometheus metrics (common::metrics::training_metrics)
- monitoring_service: Prometheus-to-gRPC bridge (port 50057)
- fxt train monitor: live TUI + --once snapshot mode
- web-gateway: REST endpoint + WS training_progress poller
- K8s deployment manifest + CI integration

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-02 22:07:42 +01:00
..