Using a small set of ordinary mobile devices together with modern humanoid robots, we built a wireless local area network for robot-assisted teaching in classrooms. All data is kept within the network, so that no sensitive communication is shared with any third party. We employ a webcam application to be processed in OpenCV, a capable large language model supported with context knowledge. This is assisted with current information sources like date and time, and a real-time object detection and segmentation framework put together to build a common Python application. It enables students to interrogate the robot about its general and problem-specific knowledge, including the current spatial and temporal information of learning objects for teaching. Through the Web Real-Time communication interface, it is possible to instruct the robot to navigate through the classroom and perform certain tasks. This lightweight brownfield automation setup proves to be versatile for interactive teaching with robots.