About
Islamic knowledge is one of the most richly interconnected bodies of thought in human history — but today that knowledge is scattered across books, websites, and apps that were never designed to work together.
VisualDhikr is building the semantic data layer for Islamic knowledge online.
Not just the texts — but the meaning-level connections between them. An open knowledge graph that maps how Quranic verses, hadith, classical tafsir commentary, and fiqh rulings from all four Sunni schools relate to one another, and to the people, places, events, and concepts that tie them together. Every connection is sourced. Every fiqh ruling is traceable to a specific page in a classical text.
Long-term, this data will be available through a public API — so developers, researchers, educators, and other Islamic platforms can build on it without starting from scratch.
A note on accuracy
AI identified the connections; primary sources ground them. But AI is fallible — it can misattribute quotations, mislabel relationships, or generate inaccurate descriptions. If you find a mistake, please let us know at info@visualdhikr.org. Every correction makes the platform more trustworthy for everyone.
How it works
VisualDhikr is built on a knowledge graph — a structured database where every verse, hadith, person, place, event, theme, and concept is a node, and the connections between them are edges. The graph currently contains over 8,000 entities and 300,000 relationships, drawn from the Quran, Sahih al-Bukhari, Sahih Muslim, 8 classical tafsir sources, and fiqh rulings across the four Sunni madhabs.
AI-driven extraction:We used Google Gemini to analyse every verse, hadith, and ruling — identifying which entities each text mentions, addresses, or relates to. These AI-generated relationships are marked as “inferred” and are subject to scholarly review. Structural relationships (which verse belongs to which surah, which hadith to which book) come directly from the source data.
Semantic search:Every text in the database has a 3072-dimensional embedding vector (generated by Gemini). When you search or ask the chatbot a question, your query is embedded and matched against these vectors using pgvector, finding semantically relevant sources even when the exact words don't match.
Grounded generation:Topic deep-dives, stories, and Q&A answers are generated by AI, but grounded entirely in the knowledge graph. The AI receives the relevant source texts as context and is instructed to cite only what it's been given. Citations are validated post-generation — any reference that doesn't match the provided sources is stripped.
Tech stack: Next.js, TypeScript, PostgreSQL with pgvector, sigma.js for graph visualisation, Google Gemini 2.5 Flash for generation, deployed on Google Cloud Run.
Where we're headed
VisualDhikr is becoming a uniformly accessible semantic layer for Islamic knowledge — structured, queryable, and open. Every verse, hadith, tafsir commentary, and fiqh ruling linked to every other, traceable to primary sources, and available through a public API that developers, researchers, and educators can build on.
The data layer is the foundation. On top of it, we're building value-added services: live Q&A grounded in the knowledge graph (ask a question, get an answer with citations to specific verses and hadith), deep topic research pages that present the full scholarly conversation across sources on any subject, and semantic search that understands meaning rather than just matching keywords.
More languages are coming (Turkish, French, Indonesian, Bengali). More hadith collections (Abu Dawud, Tirmidhi, Nasa'i, Ibn Majah). Seerah as a structured narrative layer. And ultimately, a platform where qualified scholars can review, correct, and enrich the AI-extracted content — closing the loop between machine-scale extraction and human-quality verification.
Support this work
VisualDhikr is free, independent, and has no advertising. If you find value in it:
Verify content
Help check AI-extracted connections and rulings against primary sources. The most impactful contribution.
Scholarly review
Review fiqh extractions and introductions, especially across the four madhabs.
Sponsor
Fund infrastructure, AI processing, and ongoing development.
Feedback
Use the platform and tell us what's broken, confusing, or missing.
For sponsorship or collaboration: info@visualdhikr.org