LLM
QuranVerif – AI Quran Search
AI-powered Quran verse search with RAG pipeline, citation verification to prevent hallucination, multi-source tafsir, and dual Simple/Academic mode.
Next.jsFastAPIPostgreSQLpgvectorGeminiPython
System Architecture
[SYS] ─── quranverif-rag.arch
Overview
AI-powered Quran verse search with RAG pipeline, semantic embeddings, and citation verification.
Problem Statement
Traditional Quran search relies on keyword matching which fails for conceptual queries like 'verses about patience'.
Solution & Approach
RAG pipeline: sentence-transformers embeddings → pgvector semantic search → Gemini generation → citation verification.
Key Features
- Semantic search with bilingual query expansion (ID/AR/EN)
- Citation verification: every AI reference cross-checked against actual Quran data
- Multi-source tafsir: Ibn Kathir, Jalalayn, Al-Qurtubi, At-Tabari
- Dual-mode: Simple for general users, Academic for scholars
- i18n: Indonesian and English
Results & Impact
Semantic search outperforms keyword matching for theological queries. Anti-hallucination citation checker ensures accuracy.
Lessons Learned
Citation verification is critical for religious content — AI hallucination is unacceptable in this domain.