AI allows technological systems to perceive the environment in which they operate, process those perceptions and resolve issues, acting in order to reach a preset objective.

This research aims to show that a non-cloud AI voice assistant in a non-major language can fulfill more customizable functions than typical Big Tech AI solutions, such as enabling a context-sensitive voice alert system in case of a disaster, such as a flood.

The system receives data (already prepared or collected through its own sensor network), processes them and reacts.

A case study was conducted, in which a local instance of a Home Assistant server was connected to flood sensors, presence sensors, cameras and two-way voice systems with a local AI backend, which could work together to identify whether people were present in a certain area, deliver a warning to them in Romanian and notify relevant authorities of their presence.

The system of presence sensors and cameras correctly identified that movement in an area was due to the presence of people, counted them, warned them of an impending flood and sent its findings to a notification system that could be accessed by authorities.

Using affordable, off-the-shelf hardware and open-source software, it is now possible to create a customized disaster alerting system for essentially any language, which can automatically function both as a warning for potentially affected people and a notification for authorities of the scale of the problem.