Grafana Dashboards

Archestra publishes five Grafana dashboards for its metrics, traces, and logs.

Dashboard	Shows	Data Sources
Archestra GenAI Observability	LLM requests, tokens, cost, latency, external agent runs, and traces	Prometheus, Tempo
Archestra MCP Monitoring	Tool calls, error rates, duration, MCP server deployments, and traces	Prometheus, Tempo
Archestra Agent Sessions	Recent sessions, and one session's LLM calls, tool calls, and logs	Prometheus, Tempo, Loki
Archestra Application Metrics	HTTP traffic, Node.js runtime, task queue, and PostgreSQL	Prometheus
Archestra / Knowledge Base Operations	Connector syncs, the embedding pipeline, and search performance	Prometheus

Installing the Dashboards

Prerequisites: a Grafana service account token with the Editor role or the narrower fixed:folders:writer role. The Agent Sessions dashboard also needs backend logs in Loki.

Run the install script:

bash

GRAFANA_URL=https://grafana.example.com GRAFANA_TOKEN=glsa_xxx \
  bash <(curl -sL https://raw.githubusercontent.com/archestra-ai/archestra/main/platform/dev/grafana/install-dashboards.sh)

To use basic authentication, set GRAFANA_USER and GRAFANA_PASS in place of GRAFANA_TOKEN.

In Grafana, go to Dashboards → Archestra. The folder holds the five dashboards.
Open a dashboard and pick your Prometheus, Tempo, and Loki data sources in the selectors at the top.

Re-run the script to update the dashboards; it updates existing ones in place. If you cannot run the script, download the JSON files linked in the table and upload each one in Grafana under Dashboards → New → Import.

PostgreSQL Metrics

The PostgreSQL panels on Application Metrics expect metric names from one exporter. Pass --postgres-provider to match yours:

Provider	Metric Prefix	Use When
`helm` (default)	`pg_*`	The bundled Bitnami PostgreSQL chart runs its metrics exporter
`otel`	`postgresql_*`	An OpenTelemetry Collector PostgreSQL receiver reads any PostgreSQL, including RDS, Cloud SQL, and Azure
`cloudsql`	`stackdriver_cloudsql_*`	The Stackdriver exporter reads Google Cloud SQL metrics
`azure`	`azure_*`	Azure Monitor metrics for Azure Database for PostgreSQL

bash

GRAFANA_URL=https://grafana.example.com GRAFANA_TOKEN=glsa_xxx \
  bash <(curl -sL https://raw.githubusercontent.com/archestra-ai/archestra/main/platform/dev/grafana/install-dashboards.sh) \
  --postgres-provider otel

Exemplars

LLM and MCP metrics carry exemplars: the trace ID of a sample request. With exemplars on, a dot on a latency or cost panel opens its trace in Tempo. To turn on exemplars:

Start Prometheus with --enable-feature=exemplar-storage.
In Grafana, go to Connections → Data sources, open your Prometheus data source, and under Exemplars add a link with the label name traceID to your Tempo data source.

If you provision data sources as code, add this to the Prometheus data source:

yaml

jsonData:
  exemplarTraceIdDestinations:
    - name: traceID
      datasourceUid: tempo