Smart Home Energy Scheduling Agent with Tariff and Appliance Tools

Can a small local language model plan a household's appliances from a request in plain words? This project builds a LangGraph agent that reads the request, gets the tariff, the measured base load and the appliance table from a Home-Assistant-style MCP tool server, picks start times, and revises them from a checker's report, at most 3 proposals, before a person approves the schedule on a simulator. The model (qwen2.5:3b) runs in local Ollama on a laptop with RTX 2050 4 GB (Ryzen 7, 8 GB RAM).

NOTE: Research and teaching demo. Every appliance action runs on a simulator; no real device is switched, and a person approves a schedule before it runs on the simulator. The tariff is made up (illustrative units per kWh, not a supplier's price). Savings come from a simulation on one historical household and are not money a real household would save.

Category: Department of Computer Science and Engineering / Agent Development

Agent Development Project AG21 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.