An Electron desktop application that brings cloud-level agricultural AI completely offline. Farmers upload or capture a leaf photo — the app runs a cascaded ONNX inference pipeline entirely on local hardware: Swin Transformer V2 classifies the disease, U-Net with EfficientNet-B0 calculates exact severity via pixel-wise segmentation. A Laplacian variance blur filter rejects low-quality images before inference. Results are matched against a local SQLite disease treatment database covering 15 classes across tomatoes, potatoes, and bell peppers. Built as a hackathon project with cross-platform installers for Windows (.exe) and macOS (.dmg).