Sheikh Osana
An Arabic-first, voice-enabled Islamic AI assistant powered by IBM Watsonx. Ask in Arabic by text or voice and get streamed, sourced, context-aware answers, plus a Telegram bot for always-on access.
An Islamic AI companion that hears your question in Arabic — typed or spoken — and streams a considered, sourced answer drawn from Islamic scholarship, with Arabic RTL treated as first-class, not an afterthought. A Telegram bot serves those who prefer to chat.
What it solves
- General AI chatbots ignore Arabic nuance, misinterpret Islamic context, and produce unreliable answers.
- Users want sourced, scholarly answers, not hallucinated opinions on Islamic matters.
- Arabic RTL and voice input are often missing or broken in mainstream AI products.
Impact
Native script, voice, and streaming
Grounded in Islamic texts
Web + Telegram bot 24/7
Architecture
Data flow
- User (Arabic text or voice)
- Web or Telegram
- Query to Watsonx with retrieval context
- Islamic scholarship corpus (Quran, Hadith, Fiqh)
- Sourced, grounded answer
- Stream tokens to client
RTL in Arabic, LTR in English
Engineering decisions
IBM Watsonx for Islamic knowledge grounding
Watsonx is fine-tuned on Islamic scholarship corpus (Quran, Hadith, Fiqh texts); retrieval augmentation anchors every answer to primary sources, reducing hallucination and ensuring respectful, accurate responses.
Streaming responses for perceived responsiveness
Tokens stream token-by-token to the client; users see answers appear in real-time instead of waiting for the full response, improving UX for long-form Islamic explanations.
next-intl for bidirectional (RTL/LTR) support
Full Arabic RTL layout, English LTR option, and locale-aware formatting; Tailwind RTL is configured to flip padding, margins, and borders automatically.
Omni-channel: Web + Telegram bot
Web app (Next.js) for full voice + text experience; Telegram bot mirrors the Q&A for users who live in chat, both backed by the same Watsonx knowledge pipeline.
Gallery
