Incident rules + redesign kicks off

Incident rules

A new per-type incident catalog with a Settings page. Every check kind (uptime / SSL / DNS / headers / robots / domain) gets its own rule set: when does a sample become an open incident, how long until it closes, what counts as a flap.

• Rule-aware ingest — workers route through apply_to_monitor so changes take effect on the next probe.
• Retroactive recompute — change a threshold, the system re-evaluates recent history without re-running probes.

Project-first redesign

The first slice of a project-first redesign (inspired by Oh Dear) lands today:

• Bundles backfill into proper projects under the hood — no migration on your end.
• New Add Project flow at /projects/new with an in-page plan-limit banner instead of a hard redirect.
• Project overview with an Active / Disabled card grid.

Polish

• Sidebar collapses by default, project actions menu (kebab) with cascade-delete confirmation.
• Breadcrumbs with full clickable path on every page.
• Header horizon: status-driven colour with a subtle ambient sweep — at-a-glance whether everything's green.
• App-shell layout fixes — sticky sidebar, header inside the content column, borders all line up now.