Qurany AI
State-of-the-art Islamic AI Companion prioritizing accuracy with a 'Source-First' multi-agent RAG pipeline, hybrid search, and semantic caching.
Commercial or client work — public links are not available.
Overview
Qurany AI is an advanced Islamic AI Companion built for accuracy, speed, and spiritual depth following a strict 'Source-First' philosophy — every response is grounded in authentic Quranic verses, Sahih Hadiths, and verified classical exegesis (Tafsir). Operating as a stateless Python AI microservice, it utilizes a 3-tier architecture with a LangGraph multi-agent pipeline routing to 6 specialized parallel sub-agents (Quran, Hadith, Tafsir, Topic, Web, Utility). The platform relies on an ultra-fast hybrid RAG search combining OpenAI semantic embeddings mapped with pgvector, BM25 keyword matching, and FlashRank cross-encoder re-ranking. Additionally, it implements an optimized 2-layer semantic cache (L1 in-memory, L2 Supabase) to instantly resolve redundant queries, and an acoustic pronunciation assessment engine running Whisper + Azure Speech SDK in parallel.
What it does
- 6 parallel sub-agents
- 2-layer semantic cache
- Whisper + Azure STT