Operations & Administration (Control)
Demo readiness: live avatar only for appointments
The live avatar's GPU machines do not run permanently but only for scheduled appointments: they start by themselves ten minutes before and stop afterwards.
Why#
The live avatar needs two GPU machines (renderer and speaker). In continuous operation they cost around a thousand dollars a month, even when nobody is watching. They are only needed for demonstrations, customer meetings and test slots — so that is exactly when they run.
How it works#
Click pathAdministrationOverviewDemo readiness
- 1Open “Demo readiness” in the administration's overview (address /admin/demo-bereitschaft; operator only).
- 2At the top there is a status light per service: off, starting, ready, stopping or error — with the reason in words and the time of the last check.
- 3Under “New appointment” choose title, start, duration (at most 240 minutes) and the services; additional pods can be added by their ID. “Create appointment”.
- 4Ten minutes before the start the machines start; the light only shows “ready” once the service's health check responds. Ten minutes after the end they are stopped.
- 5For spontaneous demonstrations “Make ready now (60 min)”; to end early “Stop now”. “Emergency stop” prevents any start until it is lifted.
Settings and switches#
- Switchable per tenantDefault: onModule: Avatar-Dozent
The controller's rules#
- Lead and trail time — ten minutes each around every appointment.
- Cost brake — no appointment counts longer than four hours, whatever is entered.
- Merging — overlapping time spans result in a single run.
- Never without an appointment — machines start only for an appointment or at the push of a button.
- Stop, never delete — a stopped machine keeps its address; the environment does not need to change for each demonstration.
- The controller runs every minute; immediately after every button press.
Costs#
Each appointment shows an estimate of its costs (lead and trail time included); at the top the month's total so far, the total including planned appointments and, for comparison, continuous operation. The operator enters the price per hour per service.
The actual running time is booked to the minute in cost tracking and appears in the cost dashboard under avatar.
Appointments from a booking page#
An external booking page for test slots can create and cancel appointments itself via a signed interface. It can only order the two fixed services, at most two hours per slot and only in limited numbers; the log shows every entry with the source “Slot booking”.
Who may do what#
Only the operator. Every command is recorded in the page's log and in the administration log.
Tips#
- Overlapping or closely consecutive appointments are merged into one run — the machines do not stop in between.
- If a machine runs without an appointment (for example started by hand), the controller stops it after ten minutes. A machine that is still processing render jobs is stopped only afterwards; the emergency stop stops immediately.
- Website and course use the live avatar only when the light shows “ready”. Otherwise the pre-rendered clips or the still image are shown — nobody waits for a machine that is not coming.
Related features#
- IT control centreThe operational view in real time: database, avatar machines, render pipeline, live avatar and services — refreshed every five seconds, with status in colour, word and symbol.
- Costs, usage & controllingMonthly costs and revenue, usage per student and customer, the derivation of the rendering costs and the controlling cockpit with the state of the university.
- Feature catalogue: switch every feature per tenantEvery feature of the platform can be switched on and off for all customers or for a single customer — off means: not there, not just greyed out.

