Kaleta 3.17.0

One inbox for all your sites, customers, history and alerts, and a monthly report per customer.

Inbox across all sites

• The fleet console opens on an Inbox: everything that waits on any site, in one list – requests to Claude with their status and age, new enquiries (a count), drafts and files waiting for approval, bookings to confirm, failed updates, sign-in attacks, certificates and domains about to expire, health errors, missing backups, audit findings, and sites that are down or silent.
• Sorted by urgency (urgent / waiting / when convenient), oldest first. Filter by kind, site, customer or only urgent. Cards on a phone.
• Every row is a plain link to the right screen of that site's own admin – a new tab, no sign-in, no token. The console never reaches into a site; it only reads the hourly heartbeat each site sends.
• Rows of a site that has been silent for 26 hours are marked as possibly out of date. Sites still on 3.15 show what their heartbeat has.

Customers

• Give each site a customer – on the site's page (with a note only you see on the console) or in bulk from the list of sites. Renaming a customer renames it for all its sites; renaming to an existing name merges the two.
• The Customers tab shows each customer's sites, what waits and their latest fleet report, and lets you allow a new version for all of one customer's sites at once. A shared-kit version can be meant for one customer only.

History and trends

• Every site keeps a daily history on the console for 400 days: uptime and response time from the console's own 5-minute check, the version, and what needed attention.
• A site's page shows uptime and average response this month and last month, the last 90 days as bars (with a table for screen readers), outages and version changes. Claude gets the same through get_site.

Alerts and a weekly digest

• Alert rules (Fleet → Alerts and digest): site down, stopped reporting, certificate, domain, update failed, failed sign-ins, requests waiting – each on or off, with a threshold, recipients and quiet hours. The defaults behave exactly as before 3.17.
• One set of thresholds drives the site's attention, the alerts and the inbox's urgency; change a threshold once and all three follow.
• An optional weekly digest on Monday morning: attention, what waits, outages and uptime of the week, versions, with plain links. Request titles only if you switch them on.

Monthly fleet report per customer

• On the 1st the console stores a report for each customer: per site uptime, outages, versions, days that needed attention and why, and the site's own headline counts. Kept for 24 months; view and print on the same A4 page as the 3.16 site reports, download, make now or rebuild.
• A summary, not an invoice: no hours, no prices, no personal data. A renamed customer keeps its archive.
• Claude can read it with fleet_report, and the whole inbox with fleet_inbox (read-only).

Audited before release; the findings were fixed first. Migrations 0100 and 0101 add customers, the daily history, outages and the fleet report archive.

Full Changelog: v3.16.0...v3.17.0