Release 5.12.3

5.12.3 speeds up large statistics tables with server-side pagination, adds a configurable urgent-message banner that can be turned on across all instances at once, starts flagging the old operator panel design as unsupported, fixes a chat error in the Instance Agent, and improves the reliability of automatically scheduled actions.

Highlights

Faster statistics tables — the dialogs table now loads only the requested page instead of the entire history, keeping large accounts fast and responsive.
Urgent message banner — support can display a short, dismissible announcement across the operator panel, either for one instance or the whole fleet, without touching individual instance settings.
Old design deprecation notice — the legacy operator panel now shows a banner pointing to the new design when deprecation is enabled for an instance.
Instance Agent chat fix — a rare "agent already has active run" error is now recovered automatically instead of appearing as a raw error in the chat.
Internal reliability improvements to statistics filters and to the automatic execution of scheduled actions.

Statistics

Faster loading for large dialog tables

Before: the dialogs table in Statistics loaded the entire matching history into memory before showing a page, which could slow down or briefly freeze the account on accounts with a long history or a wide date range.

Now: the table loads only the requested page from the database, with the total row count shown separately. Loading time no longer depends on how much history exists.

Why it matters: statistics stay responsive regardless of how much data has accumulated, and one broad, dateless report can no longer slow the account down for other users.

More: View a dashboard

Before: in some statistics tables, filter dropdown lists (for example, available statuses or channels) could accumulate stale values over time and no longer reflect what was actually available in the selected period.

Now: filter dropdowns always show the values relevant to the current table and period.

More: Apply report filters

Operator Panel

Urgent message banner

Before: there was no built-in way to broadcast a short announcement (for example, about planned maintenance) to operators inside the panel — support had to reach out through other channels.

Now: an urgent message banner can be shown at the top of the operator panel, in both the old and the new design. It can be turned on with a short text for a single instance or globally for all instances, is dismissible by each operator for an hour, and reappears automatically if the text is changed.

Why it matters: critical announcements reach operators directly where they work, without requiring a config change on every affected instance.

Old design deprecation notice

Before: instances still using the legacy operator panel design had no in-product signal that the design was being phased out.

Now: when deprecation is enabled for an instance, the legacy panel shows a banner linking to the new design, disables some legacy tabs, and displays a placeholder instead of the dialog list. This only takes effect once a link to the new design is configured, so operators are never left without a way forward.

Why it matters: gives instances still on the legacy design clear notice and a direct path to move to the current interface.

Instance Agent (Cursor)

Fixed "agent already has active run" chat error

Before: in some cases — most often after clicking Stop while the agent was still processing an earlier message — the next message in the chat could fail with a raw technical error ("agent already has active run") instead of a normal response.

Now: the Instance Agent automatically detects and clears the stuck run and retries the message once. If a genuine conflict remains (for example, another run is truly still active), the chat shows a clear, localized message instead of the raw error.

Why it matters: removes a confusing technical error from a common interaction (sending a follow-up message right after Stop) and keeps the conversation moving without manual retries.

More: Instance Agent overview

Automations

Improved reliability of scheduled action execution

Before: in rare cases, after a temporary connectivity issue with the automation infrastructure, a scheduled automatic action (for example, an "Automatic Check") could silently stop running on an instance — the schedule still showed as active, but nothing executed until someone manually restarted the service.

Now: the underlying process automatically detects this stuck state and recovers on its own, without manual intervention.

Why it matters: scheduled automatic actions keep running reliably through temporary infrastructure hiccups, without requiring anyone to notice and restart anything.

More: Configure automatic check parameters

Notes

Internal improvements to how recurring broadcast schedules are processed, reducing unnecessary system load without changing broadcast behavior.
AI-driven statistics requests that don't specify a date range now default to the last 7 days instead of returning unbounded results.