<![CDATA[Abstrak - Sortasi biji kopi pasca-sangrai merupakan tahapan krusial dalam quality control untuk menjamin konsistensi produk. Namun, proses manual rentan terhadap subjektivitas dan ketidakkonsistenan, khususnya bagi UMKM kopi dengan keterbatasan infrastruktur. Penelitian ini mengembangkan PiKopi, sistem monitoring sortasi terintegrasi dengan dashboard web untuk visualisasi dan pengambilan keputusan berbasis data. Sistem mengintegrasikan tiga komponen utama: (1) modul deteksi cacat berbasis Convolutional Neural Network (CNN) pada perangkat edge computing Raspberry Pi untuk identifikasi real-time, (2) backend API RESTful berbasis Flask dengan database PostgreSQL untuk manajemen data klasifikasi, dan (3) dashboard web responsif berbasis React.js untuk visualisasi data real-time. Arsitektur API-driven memisahkan frontend dan backend, memungkinkan akses monitoring dari berbagai perangkat tanpa instalasi aplikasi. Dashboard menampilkan statistik sortasi real-time, grafik distribusi kualitas, riwayat batch, dan log aktivitas dengan interface yang user-friendly. Model CNN mencapai akurasi deteksi 92% dengan sensitivity dan specificity yang seimbang. Hasil pengujian fungsional menunjukkan sistem mampu melakukan klasifikasi otomatis dan menyajikan data sortasi secara real-time kepada pengguna. Sistem ini memfasilitasi pengambilan keputusan berbasis data, meningkatkan konsistensi mutu produk, serta mengurangi ketergantungan pada inspeksi visual manual bagi pelaku UMKM kopi.Kata kunci: dashboard web; monitoring real-time; sortasi kopi pasca-sangrai; quality control; arsitektur API-driven; Abstract - Post-roasting coffee bean sorting is a crucial stage in quality control to ensure product consistency. However, manual sorting processes are prone to operator subjectivity and inconsistency, particularly for small and medium enterprises (SMEs) with limited infrastructure. This research develops PiKopi, an integrated coffee sorting monitoring system with a web dashboard for visualization and data-driven decision making. The system integrates three main components: (1) a defect detection module based on Convolutional Neural Network (CNN) on Raspberry Pi edge computing devices for real-time identification, (2) a RESTful API backend based on Flask with PostgreSQL database for classification data management, and (3) a responsive web dashboard based on React.js for real-time data visualization. The API-driven architecture separates frontend and backend, enabling monitoring access from various devices without application installation. The dashboard displays real-time sorting statistics, quality distribution charts, batch history, and activity logs with a user-friendly interface. The CNN model achieves detection accuracy of 92% with balanced sensitivity and specificity. Functional testing results demonstrate that the system successfully performs automated classification and presents real-time sorting data to users. This system facilitates data-driven decision making, improves product quality consistency, and reduces reliance on manual visual inspection for coffee SMEs. Keywords: web dashboard; real-time monitoring; post-roast coffee sorting; quality control; API-driven architecture;]]>
