Penelitian ini bertujuan untuk mengimplementasikan chatbot asisten informasi pada website Universitas Kuningan menggunakan model Gemini 3 Flash. Permasalahan yang melatarbelakangi penelitian ini adalah adanya kendala pengguna dalam memperoleh informasi akademik secara cepat, seperti informasi PMB, UKT, beasiswa, jadwal, fakultas, program studi, dan pejabat struktural. Metode pengembangan sistem yang digunakan adalah Chatbot Development Life Cycle (CDLC) yang terdiri dari tujuh tahapan, yaitu Problem Identification, Requirement Analysis, Design, Implementation, Testing and Debugging, Integration, dan Monitoring. Sistem dibangun menggunakan React JS, TypeScript, Vite, serta Supabase PostgreSQL sebagai basis data. Sistem menerapkan arsitektur hibrida yang menggabungkan Rule-Based System, Retrieval-Augmented Generation (RAG), dan model Gemini 3 Flash untuk menghasilkan jawaban yang relevan berdasarkan knowledge base. Pengujian sistem dilakukan menggunakan Black Box Testing, White Box Testing, Confusion Matrix, dan User Acceptance Test (UAT). Hasil Black Box Testing menunjukkan bahwa seluruh fitur utama sistem berjalan sesuai rancangan. Pengujian White Box Testing menghasilkan nilai Cyclomatic Complexity sebesar 3, sedangkan evaluasi menggunakan Confusion Matrix pada 30 pertanyaan uji memperoleh nilai accuracy, precision, recall, dan F1-score sebesar 96,67%. Hasil UAT terhadap 55 responden memperoleh persentase sebesar 92,87% dengan kategori sangat layak. Dengan demikian, UNIKU AI dapat membantu layanan informasi akademik secara lebih cepat, terarah, dan real-time.

This study aims to implement an academic assistant chatbot for the Universitas Kuningan website using the Gemini 3 Flash model. The research is motivated by the existing challenges users face in obtaining academic information efficiently, particularly regarding student admission, tuition fees, scholarships, academic schedules, faculties, study programs, and structural officials. The system is developed using the Chatbot Development Life Cycle (CDLC) methodology, which consists of seven stages: Problem Identification, Requirement Analysis, Design, Implementation, Testing and Debugging, Integration, and Monitoring. The chatbot architecture is built using React JS, TypeScript, Vite, and Supabase PostgreSQL as the database. The system applies a hybrid architecture that combines a Rule-Based System, Retrieval-Augmented Generation (RAG), and the Gemini 3 Flash model to generate relevant responses based on the established knowledge base. System testing is conducted using Black Box Testing, White Box Testing, Confusion Matrix evaluation, and a User Acceptance Test (UAT). The Black Box Testing results show that all core system features function as designed. The White Box Testing produces a Cyclomatic Complexity value of 3, while the Confusion Matrix evaluation on 30 test questions achieves 96.67% for accuracy, precision, recall, and F1-score. Furthermore, the UAT results from 55 respondents obtain a percentage score of 92.87%, placing the system in the 'Very Feasible' category. In conclusion, UNIKU AI effectively enhances academic information services by making them faster, more targeted, and available in real time.