Dashboards

A dashboard is one saved query rendered as a chart over a time window. Pick a small predicate (severity:error, dataset:player-actions), optionally slice by dataset, event, or severity, choose the time window and chart type, and the dashboard renders a time-bucketed chart of event counts.

Use dashboards for the ops views you want at a glance: errors per minute, kicks per hour, transactions per day. Each dashboard is one chart. To build a multi-pane view, create several dashboards and visit the index page when you need them.

The index page

Logs ▸ Dashboards. Card grid, one card per saved dashboard. Filter chips on the left flip between:

• ALL - every active (non-archived) dashboard in the project
• MINE - just the ones you created
• ARCHIVED - everything you have archived (defer-cleanup view)

On the right of the strip, the DASHBOARDS · N / M chip shows your slot usage against your plan. Click + NEW DASHBOARD to open the create form.

The predicate language

The predicate is what filters the chart. It's deliberately small: dashboards read from a rolled-up storage table that only carries dataset, event, and severity columns, so allowing arbitrary payload filters would silently match nothing.

Grammar:

predicate := clause (AND clause)*
clause    := key ':' value (',' value)*
key       := dataset | event | severity
value     := lowercase identifier (a-z, 0-9, _, -)

Examples:

severity:error
severity:warn,error,critical
dataset:player-actions AND severity:error
event:login,logout AND dataset:auth

The empty string is a valid predicate (no filter, count everything).

If you want richer filters (payload fields, free-text matching, etc.), use the search page. Dashboards are intentionally fast and simple.

Group by

Optional. Slices the chart into multiple series:

• NONE - one series, total event count per bucket
• dataset - one series per dataset that appears in the window
• event - one series per distinct event name
• severity - one series per severity level (uses the canonical ordering: trace, debug, info, warn, error, fatal)

Only one group-by field at a time in v1.

Time windows and bucket sizes

Pick a window; we pick the bucket size:

Tuned for ~60-180 data points per chart - enough density to read trends, sparse enough to read individual buckets.

Chart types

• BAR - vertical bars per bucket, stacked when grouped
• LINE - one line per series
• STACKED - filled area chart, series stacked
• TABLE - sum per series, sorted by total descending. Useful for "top N" views without rendering a chart.

Refresh

Dashboards do not auto-poll in v1. Click REFRESH at the top of the chart to re-run the query. Auto-refresh is on the roadmap; tell us if you need it sooner.

Save as dashboard from search

The search page has a ◰ SAVE AS DASHBOARD button next to the regular save button. Click it to open the new-dashboard form with the current query pre-populated. If your search query is full SOQL with payload predicates or OR / NOT clauses, the form will tell you which parts aren't supported as a dashboard predicate.

Archive is reversible

Archive hides a dashboard from the default list. To bring it back, open the ARCHIVED filter, click the row, and use the UNARCHIVE button on the Settings tab. Nothing is deleted; the row sits in storage with archived_at stamped.

This means dashboards behave the same way datasets do: archive is for cleanup, not destruction.

Common error responses

If you hit one of these on save:

What this page does NOT do

• It does not let you compose multi-pane dashboards. One query, one chart. v2.
• It does not let you build dashboards on payload fields. They are not in the rollup; use search instead.
• It does not auto-refresh. Click REFRESH.
• It does not share dashboards publicly. Org members with LOGS READ can view; outside that, they cannot.
• It does not export the chart as PNG / CSV. On the roadmap.