Penerapan Model AI Gemini pada Pengembangan Chatbot Tanya Fiqih Haji dan Umrah

Ahmad Farisi
Chriestina, Chriestina
Dafid, Dafid
Adit Jansa

Jurnal Sains dan Teknologi
Vol. 4 No. 6 (2025): Desember 2025

This study aims to develop a hybrid mobile application prototype called “Tanya Fiqih Haji dan Umrah” as a supportive consultation medium for Islamic jurisprudence (fiqh) related to Hajj and Umrah. The application utilizes the Gemini 2.5 Flash model with a Retrieval-Augmented Generation (RAG) approach that is restricted to two official references issued by the Indonesian Ministry of Religious Affairs, namely the Hajj Manasik Infographic Book 2025 and the Hajj Supplications and Dhikr Guide 2025. The system is implemented using a modern web-to-mobile hybrid architecture based on React, TypeScript, Tailwind CSS, and Capacitor, enabling deployment as a lightweight Android application with local conversation history storage. System evaluation was conducted through expert validation on ten test questions representing the core themes of Hajj and Umrah fiqh. The results show that 70% of AI responses were categorized as fully aligned with the references, 10% as partially accurate, and 20% as not aligned. These findings indicate that the RAG mechanism helps maintain consistency between AI-generated answers and authoritative literature, potentially reducing hallucination in generative models, although expert clarification remains necessary in certain cases. This study contributes to the application of RAG-based large language models in the domain of Hajj fiqh and offers a technological support tool that may benefit pilgrims, particularly independent pilgrims.