viet-robot-safety-filter

A rule-based safety filter for Vietnamese home-robot commands.

The model flags commands that contain obviously dangerous actions such as burning, destroying objects, or harming people.

It returns a small dictionary:

{"safe": True/False, "reason": "..."}

This is a simple demonstration and should be combined with stronger safety systems in real robots.

Usage

from model import check_safety

print(check_safety("Bật đèn phòng khách"))
print(check_safety("Đốt rèm cửa đi"))

MIT