The built-in assistant

What the assistant is

A chat icon (💬) floats in the bottom-left corner of the panel view. One click opens the assistant as a window above the interface – no ChatGPT account or Alexa link required. It always works with the permissions of the signed-in user on exactly the panel you have open: it can do exactly what you may do in the interface yourself – never more.

Examples:

"Are all windows closed?" "Switch on the hallway light." "Raise all shutters on the ground floor." "Create a rule that switches off the outdoor light at 10 pm." "Mute the outdoor siren for fire alarms."

What the assistant can do

The assistant uses the same functions as the ChatGPT and Claude integrations – at full scope:

• Read: system status, devices, history, statistics, signal strength, system info, rules, scenes, settings.
• Control: switch devices, lights (brightness/colour), shutters, thermostats, apply scenes.
• Set up: create automation rules and scenes, maintain rule groups and profiles (Lupus), pair detectors, change settings (e.g. siren behaviour, e-mail/SMS accounts, reporting).
• Arm/disarm: only with the system's operating code – the assistant asks for it and never guesses.
• Open views: on request it shows the matching page of the interface behind the chat ("Show me the siren settings").

The assistant acts first and only explains when you explicitly ask for instructions or something cannot be done with its tools. Before irreversible steps – deleting rules, activating a profile, arming – it summarises what it is about to do and waits for your confirmation in the chat. While working, the chat shows the current step (e.g. "switching the device").

Permissions and limits

Every action passes the same permission check as the web interface: if your account lacks a permission on this panel (say, "arming"), the tool refuses and the assistant explains. It also only answers topics related to the Alarmanlagenmanager and your system.

Quota and logging

• A daily quota of messages applies per user (shown below the input field). It protects against abuse and caps the costs.
• Inputs, answers and executed actions are logged; the operator can review them for abuse prevention. Details – including the transfer to Anthropic and the reliability of AI-generated answers – are in the privacy policy.

Conversation and history

The assistant remembers the conversation of the current session, so you can refer back to earlier messages. "New conversation" resets the history. After a longer pause or a server maintenance the conversation starts fresh automatically.

How it differs from ChatGPT, Claude and Alexa

The built-in assistant complements the existing integrations, it does not replace them: ChatGPT and Claude control the system from within their apps, Alexa by voice around the house. The assistant here needs none of that – just your normal sign-in.