---
title: Observability
category: Observability
description: Metrics, request logs, and runtime logs for every project.
slug: observability
---

Each project has an **Observability** tab (metrics) and a **Logs** tab
(request + runtime logs). The dashboard also links out to the full
Grafana instance at `/grafana`.

![Project observability](/screenshots/observability.png)

## Metrics

The Observability tab charts the last 6 hours from Prometheus (fed by
cAdvisor): memory, CPU %, network in/out per container, plus an edge
request-load chart from the API's own Loki-backed request matrix. Cards
refresh every 15 seconds. An empty state means nothing is deployed yet
— deploy to render telemetry.

Behind the scenes, `GET /projects/:id/status` aggregates health over a
configurable window, and `GET /prometheus/query...` proxies arbitrary
PromQL for custom dashboards.

## Request logs

The Logs tab's **Request Monitoring** view tails the Caddy access log
for the project domain via Loki: method, path, status, latency, and a
30×1-minute histogram. Filter by date range (pauses the live tail),
full-text search, and level checkboxes, or click a row for the
slide-over detail. A live/paused toggle controls the SSE stream.

## Runtime logs

The **Runtime Logs** view follows the running container's stdout
(`docker logs -f`) over SSE — what your app actually printed, as
opposed to what the edge proxy saw. Same search and filter controls.

## Grafana

**Open Grafana** (sidebar, under Observability) jumps to the platform
Grafana with Prometheus and Loki data sources pre-wired, for ad-hoc
queries across all projects and infrastructure.

## Next steps

- Read [Alerts](/docs/alerts) to get notified on CPU, memory, or downtime.
- Read [Deployments](/docs/deployments) for build logs (as opposed to runtime logs).