This study presents a real-time air quality prediction framework for the Mekong Delta (MK), utilizing Graph Attention Network (GAT) and Long Short-Term Memory (LSTM) to address spatial-temporal dependencies in air pollution data. With rapid urbanization exacerbating air quality issues that impact public health, we collected a dataset capturing dynamic changes in air quality across time and space. The GAT model captures spatial relationships, while LSTM handles temporal dynamics for accurate predictions. Our GAT-LSTM model achieved a MAE of 2.64 and MSE of 15.2, with predictions made at the district level and providing hourly forecasts for the next 7 days. A web-based interface has been developed for real-time monitoring, and both the dataset and implementation are publicly available on GitHub to support further research and applications in air quality forecasting.