Hujjah

Privacy-first Islamic research engine with Quran, hadith, and AI chat.

Impact Highlights

Made 111 Quran translations searchable by indexing the full corpus into a SQLite database with FTS5 and Arabic diacritic normalization.
Eliminated hallucination risk in Islamic Q&A with a RAG pipeline using BGE-M3 embeddings, grounding every response in sourced verses and hadith.
Made hadith chain scholarship accessible with an interactive sanad explorer — BFS graph traversal, React Flow infinite canvas, chain quality indicators, narrator enrichment via Wikidata/Wikipedia.
Solved the Bengali Islamic content gap with a three-tier hadith corpus, Sunnah.com import + github-classic Bengali fallback (re-linking 24K translations), and bilingual AI chat.
Delivered uninterrupted Quran audio by proxying the Islamic Network CDN with verse highlight/scroll, stay-in-surah auto-play, and resume support.
Deployed a production web application with Docker Compose and Cloudflare Tunnels — providing a seamless experience with privacy-first features.

Next.jsSQLiteFTS5Groq AITransformers.jsBGE-M3

Next.js architecture with SQLite for fast database reads.
Arabic FTS5 search with diacritic stripping and multi-language query normalization.
AI chatbot with RAG implementation, extractive Bengali fallback, and language auto-detection.
Sunnah.com hadith import pipeline; unified three-tier corpus with grading and translator switcher.
Sanad explorer: BFS level assignment, expandable nodes, hadith counts, React Flow draggable nodes, infinite canvas Map view.
RadialSanad and ChainPreview views with SVG connectors and GitHub-style edge routing.
Narrator search with FTS5 prefix matching, Arabic normalization, multi-language fallback.
Quran audio proxy with real progress bar, reciter selection, per-verse controls.
Settings page: appearance, AI model, Arabic font switcher, database sections.
Dark/light mode, Bengali/Arabic bilingual UI, mobile nav, collapsible chain sections.
Docker Compose and Cloudflare Tunnel deploy pipeline.