In the classical Quranic sciences, naẓm (نَظْم) is the study of coherence — how verses, words, and themes are ordered into a single interconnected whole. Nazm-AI builds software that makes that coherence visible, using AI to surface the connections scholars have always known were there.
An AI-powered Quran knowledge graph. Search any verse, drop it onto an infinite canvas, and let AI surface the connections — shared root words, thematic resonance, theological contrast. The Quran's internal architecture emerges as you explore.
| Feature | Description |
|---|---|
| Canvas | An infinite graph of verse nodes joined by AI-generated edges. Expand any verse by Theme, Root Word, or Contrast — click an edge to read the reasoning behind the connection. |
| Search | Direct reference lookup (2:255), full-text keyword search, and semantic search powered by verse embeddings. One click adds any verse to the canvas. |
| Knowledge graph | Connections are persisted and grow over time — the graph gets richer the more the community explores. |
| Shareable canvases | Every canvas state serialises to a single URL you can copy and share. |
| Audio | Play all verses on a canvas in Quran order with a single button. |
| 99 Divine Names | The complete Asmāʾ al-Ḥusnā — Arabic script, root morphology, a Maturidi taxonomy, a verse feed, and contemplative reflections. |
| Bookmarks | Save verses to your account; synced across devices via the Quran Foundation API. |
| Social | Daily engagement streaks, a friends leaderboard, and head-to-head challenges (24h / 48h / 7d). |
Software that touches sacred text carries responsibility. Open Hikmah is built to a deliberate standard:
- Maturidi / Hanafi tradition — the project's theological frame of reference. Positions are attributed with care; schools are never conflated.
- Strict Tanzīh (transcendence) — divine attributes are never described in ways that imply physical form, location, or resemblance to creation. This constraint is encoded directly into the AI prompts.
- Verified references only — every verse reference is validated against canonical sources before it is shown. The model is never trusted to recall scripture from memory.
- Sacred data, even in tests — real, plausible Arabic throughout — never placeholder text in place of the Word.
Full stack details
| Layer | Choice |
|---|---|
| Framework | Next.js 16 (App Router), React 19, TypeScript |
| Canvas | @xyflow/react |
| State | Zustand |
| Styling | Tailwind CSS v4 |
| AI | Anthropic Claude (adaptive thinking), with Google Gemini as fallback |
| Semantic search | Gemini embeddings + PostgreSQL pgvector |
| Quran data | alquran.cloud + Quran Foundation API |
| Auth | Quran Foundation OAuth2 (PKCE) |
| Database | PostgreSQL + Drizzle ORM |
| Testing | Vitest + Testcontainers |
| CI | GitHub Actions |
We welcome contributors — engineers, designers, and people with knowledge of the Islamic sciences alike.
"We have made it easy to learn lessons from the Quran: will anyone take heed?" — Qur'an 54:17 (trans. Abdel Haleem)
Built with care by the Nazm-AI community.
