The SEMAR platform gained a conversational guide, tested on three smart systems and a university class.
Setting up an Internet of Things project is rarely as simple as plugging a sensor into a board. You have to pick the right pins, choose a protocol, and get the network settings right before the device can talk to a server. SEMAR, an integrated server for collecting and visualizing IoT data, was built to make that last part easier. Its authors found that the first part still stops novices cold.
Their answer is an assistant that runs in a browser and holds a conversation. A user describes the project, the hardware on hand, and where it will run. The service replies with step by step instructions for wiring sensors, configuring the board, setting up the network, and choosing a protocol.
The service has three parts: a setup guidance function, a response function, and a conversational interface. It runs on GPT-4o, and it grounds answers with retrieval augmented generation over the SEMAR setup guide.
The authors narrowed the requirement scope to four domains: processing boards (Raspberry Pi or Arduino), sensor connectivity (UART, GPIO, I2C), network connectivity (WiFi, cellular, LAN), and communication protocols (MQTT, HTTP, WebSocket).
The team ran two evaluations. First they used the service on three real systems: a smart light, a smart water heater, and a room temperature monitor. It produced working configurations for each, including one where the user swapped boards mid conversation.
Second, they handed it to 22 undergraduates at Institut Bisnis dan Teknologi in Indonesia, working in five groups. Of the 19 students with no IoT experience, 16 finished the assignment, and all three experienced students did too. Usability scores were high for ease of use, usefulness, learning curve, efficiency, and error handling.
The assistant was not flawless. It once generated a pin constant that would not compile, and it recommended a package that fails on the newest Raspberry Pi. Students fixed both problems by asking follow up questions, and several asked for the ability to edit messages they had already sent.
Kotama et al. (2025). Implementation of Sensor Input Setup Assistance Service Using Generative AI for SEMAR IoT Application Server Platform. Information.
