Skip to content

Analytics Dashboard

RTSM includes a runtime analytics system that tracks per-stage latency, segmentation output, and pipeline throughput. Analytics are available via the REST API and the built-in visualization dashboard.

Enabling Analytics

Analytics are enabled by default. Configure in config/rtsm.yaml:

analytics:
  enable: true
  retention_s: 3600     # Keep 1 hour of data
  buffer_frames: 300    # Ring buffer size (per-frame granularity)

REST API Endpoints

Pipeline latency

curl http://localhost:8002/stats/detailed

Returns per-stage timing breakdown:

Stage Description
segmentation Mask extraction (model inference)
heuristics Mask filtering (area, depth, border, planarity)
scoring Priority scoring and top-K selection
clip CLIP visual encoding (224x224 crops)
association Proximity + embedding matching to existing objects

Each stage reports mean, P50, P95, and max latency.

Segmentation analytics

curl http://localhost:8002/analytics/segmentation

Returns segmentation-specific metrics:

  • Masks per frame (mean, min, max)
  • Dual-confirmation rate (when using dual backend)
  • Label distribution across confirmed objects
  • Drop counters (frames rejected at each gate)

Drop counters

The pipeline tracks where frames are dropped:

Drop point Description
tracking_state Camera tracking quality below threshold
throttle Non-keyframe interval throttling
queue_full Ingest queue overflow
gate_rejection Sweep-based novelty gating

Visualization Dashboard

The analytics dashboard is built into the 3D visualization frontend. When the visualization server is running (ws://localhost:8083/ws), the frontend includes:

  • KPI Scorecards — Input Hz, Processing Hz, Latency, Queue depth, Object count, Association rate
  • Time-series charts — Throughput, per-stage latency breakdown, segmentation rates (powered by uPlot)
  • Config display — Current backend, filtering thresholds, and scoring parameters

The dashboard updates in real-time via WebSocket push from the visualization server.

Prometheus Metrics

RTSM exports Prometheus-compatible metrics at http://localhost:8002/metrics:

curl http://localhost:8002/metrics

Key gauges:

Metric Description
rtsm_live_objects Current number of live objects
rtsm_confirmed_objects Number of confirmed objects
rtsm_upserts_total Total vector DB upserts

These can be scraped by Prometheus and visualized in Grafana for long-term monitoring.

Next Steps