Q & A

Questions that actually come up, with the short answer first.

Scenarios

Why is my scenario read-only? Someone else holds the edit lock — the banner names them. A studio admin can take over; otherwise the lock expires on its own shortly after their tab closes.

I saved, but the call still runs the old flow. Saving creates a version; publishing puts it live. Check the header — it shows the latest version and the published version side by side.

Can I undo a publish? Open History, restore the version you want (it becomes a new latest version), and publish that. History is never rewritten, so the misstep stays visible in the trail.

v1 or v2 — which editor am I in? The header badge says v2 in the v2 canvas. New scenarios always start in v2; v1 exists so old documents stay readable. Conversion to v2 is explicit and keeps the v1 history.

A node keeps re-asking the caller forever. Its retry lane has no bound. Set Re-prompts allowed (asks) and Times this node may run (the loop guard) — only a rule naming [EXECUTION_LIMIT] catches the guard firing.

Campaigns & calls

Contacts uploaded, nothing dials. In order: is the campaign started, is the schedule window open in the tenant timezone, is the scenario published, and does the callbot have an outbound trunk on the gateway host.

Where is the recording? On the call page, when recording was enabled for that leg. Consent handling can lawfully suppress it — the call's compliance events say so explicitly.

Platform

Which parts need the internet? Only the AI providers you configure. With on-premise ASR/LLM/TTS (ADR-0028) the whole stack runs air-gapped; the telephony path never leaves your network either way.

How do I upgrade? Rehearse the new tag in staging, then set the same TAG in both production environment files and redeploy each stack. Updates are health-gated and roll back automatically on failure, and database migrations run on api start. See Operating.