Currently, hydrotherapy relies on physiotherapists, which can lead to inconsistencies due to the shortage of professionals and human error. This paper proposed Joint Vector-YOLO NAS (JV-YOLO Nas) to accurately classify hydrotherapy hip abduction range. The methodology involved recording underwater hip abduction exercises with high precision, where YOLO-NAS Pose was used to extract joints frame by frame and turned into joint vectors for precise angle calculation. These angles were categorized into predefined ranges. A dataset was created from these categories to train a custom YOLO-NAS classification model. The model achieved 90.44% mAP and worked at 18-20 frames per second on land, providing real-time feedback on angle ranges. This method allow precise feedback for real time monitoring for hydrotherapy exercise specifically in hip abduction type exercise. This can potentially be adopted to other types of hydrotherapy exercises.