[Kkkermit]

What this changes

• A Refresh button on the owner console's Logs tab (/owner?tab=logs), beside Pause and Clear. It reads the run shown and the run list again at once. That covers the run going now, live or paused, and a past run, which otherwise keeps what it read for a minute. While it reads, the button says "Refreshing…" with a spinning icon.
• It says how it went, in a status line beside the buttons (RefreshLogs):
  • Success: "Refreshed at 15:18:12." in green, clearing after five seconds.
  • Failure: "Couldn’t refresh the logs. The log store is busy. The lines shown were read at 15:18:12." in red, with the server's own reason (describe from ErrorState). It stays until a later read of the lines works.
• Lines already read stay on screen when a later read fails. The tab used to swap them for an error page on any failed read, the five-second timer's included. The error page now shows only when nothing has been read yet, for the lines and for the run list alike. A failure on the timer is said in the same red line, so a stale list never looks current.
• Five new strings in all eight languages, and a sentence in AGENTS.md §24 under the log history.

Why

A paused feed and a past run held what they first read, with no way to bring them up to date short of switching tabs or reloading. A failed read also hid the lines already on screen, on the one screen meant for when things are going wrong.

How to verify

• npm run check passes locally with 4,048 tests; the 30 database suites are skipped there. CI was still running on the branch when this was opened.
• New tests in OwnerAnalytics.test.tsx:
  • Refresh brings in a line the first read did not have, and shows the success note.
  • A refresh the API answers with a 500 shows the reason and when the lines were read, and keeps the lines.
  • Three mutations each turned a test red: Refresh not reading, the failure never shown, and the error page replacing the lines.
• In Chromium against the Vite dev server with fixture data, at 1280px and 390px: the success and failure notes both showed, the lines stayed after a failure, nothing scrolled sideways and the page threw nothing. The failure note appears after about 3.4 seconds, because the query client retries a 5xx twice before giving up.
• On a running bot: open /owner?tab=logs, press Pause, press Refresh, and the newest lines appear with "Refreshed at …". Not tried against a live bot.

Checklist

• The branch is cut from develop, and this pull request targets develop
• npm run check passes
• Tests cover the new behaviour, and were shown to fail without it
• No read-modify-save() on numeric fields
• Embeds come from the embed() factory
• No raw console.*
• Every promise is awaited or handled
• No new timers registered with the bot (the five-second note is a browser timeout, cleared on unmount)
• No new custom IDs
• No new environment variables
• No command changed, so docs/commands.md is unchanged
• Every new string is in all eight dictionaries
• Documentation updated: AGENTS.md §24

---

Generated by Claude Code