Portal

Your app's numbers and logs

Every app's page in the portal has a How it is doing section. Your coding agent reads the same figures with get_metrics and get_logs (MCP tools), and the CLI with kaviris logs.

You see your own apps only. Nobody in another organisation can read them, and Kaviris never shows you anyone else's.

The numbers

Pick a window: the last hour, 24 hours, 7 days or 30 days. Metrics are kept for 30 days, so that is the longest window.

Figure What it counts
Visits Requests that reached your app through its address
Errors (5xx) Of those, the answers with a status from 500 to 599, and their share
Typical response, slow The median and the 95th percentile of the time your app took to answer (a wake-up makes a slow one)
Wake-ups How often the app woke from sleep, how many hours it was awake, and how long a wake typically took (the median over the window)
Memory What your app uses now, as seen inside its machine (blank while it sleeps)
Restarts How often the app's process restarted

Two charts show requests and errors over the window, in 48 bars; the strip under them shows when the app was awake (green) and asleep (ember). Point at a bar for its figure, or open The charts as a table.

Free allowance this month compares this app with the free plan's monthly allowance for awake time, data sent and storage (databases and backups), next to your other apps' share. The Usage page has the whole organisation. These figures are measured hourly.

The logs

The section ends with the last 10 lines your app wrote in the past hour. All logs, search and follow shows up to 200 lines from the last 15 minutes up to 14 days (logs are kept for 14 days), only the lines containing a text if you give one, and Follow reloads the page every 10 seconds.

The lines are what your app and its database write to their output (stdout and stderr), shown as plain text. Kaviris adds nothing to them: no visitor addresses or browsers. If your app writes those into its own log, they appear there, and they are yours to handle.

Limits

The numbers are cached for 30 seconds and log reads for 5 seconds. Each organisation can make 60 fresh reads a minute and follow 3 logs at once; past that the API answers 429 rate_limited and the portal says to try again in a minute.

This page as Markdown