Phoenix supports exporting metrics via a Prometheus endpoint, which Grafana can scrape for visualization. To enable this, set the enablePrometheus option to true in your Phoenix server configuration (e.g., in your Helm values or Docker/Kubernetes deployment), which exposes metrics on port 9090 for Prometheus to collect and Grafana to visualize. See the Helm values.yaml and related Slack discussion for details.Would you like more detail on how to configure Prometheus and Grafana to connect to Phoenix?Sources:
- https://arize-ai.slack.com/archives/C04R3GXC8HK/p1719072670.110699
- Helm values.yaml
💡Hint: Mention RunLLM in the thread for followups.
Hey Daniel just to be clear these are Prometheus metrics of the Phoenix instance itself, not your app.
You can export your traces and metrics to grafana via another exporter.
So phoenix is considered an otel collector and backend so it won’t export but you can either add two exporters in your app or use the OTEL collector to send data to phoenix and grafana
I don’t unfortunately. Grafana docs are probably best and it depends on how you want to setup your otel metrics and logs. I know Kibana has some examples with Phoenix.
The benefits of using both is that you will have everything correlated with the same IDs.
If you just want metrics that are stored in phoenix it does have APIs and that’s probably what I would hook into
