Pre-built dashboard templates for visualizing GenAI traces and metrics collected by TraceVerde.
Comprehensive LLM observability dashboard with:
- Request rate by provider (time series)
- Token usage: prompt vs completion (stacked time series)
- Cost tracking over time by provider
- Latency distribution (histogram)
- Error rate by provider and error type
- Top 10 models by usage (table)
- Summary stats: total cost, requests, tokens, avg latency (24h)
Real-time GPU monitoring dashboard with:
- GPU utilization gauge (per GPU)
- Memory utilization gauge (per GPU)
- Temperature gauge with thresholds
- Power consumption stat
- Time series for utilization, memory, temperature, power
- Multi-GPU aggregate metrics (mean utilization, total memory, total power, max temp)
- Open Grafana and navigate to Dashboards > Import
- Click "Upload dashboard JSON file"
- Select the desired JSON file from this directory
- Select your Prometheus datasource when prompted
- Click "Import"
Prerequisites:
- Prometheus must be configured to scrape your OpenTelemetry Collector
- The OTel Collector must be configured with a Prometheus exporter
- TraceVerde must be running with metrics enabled
Example OTel Collector config:
exporters:
prometheus:
endpoint: "0.0.0.0:8889"
service:
pipelines:
metrics:
exporters: [prometheus]Basic saved objects for OpenSearch Dashboards including index pattern and visualizations for GenAI trace data.
- Open OpenSearch Dashboards
- Go to Management > Saved Objects
- Click "Import"
- Select the
.ndjsonfile - Click "Import" and resolve any conflicts