Ilm عِلْم

Search the Quran & Sunnah. Deeply.

A semantic search platform for Islamic scholarship — explore the Quran with tafsir, 34K+ hadiths with narrator chains, and interactive isnad graphs. Powered by meaning, not just keywords.

Free & Open Source Semantic Search Bilingual Open Data AI-Powered

View on GitHub Blog

114 Surahs

6,236 Ayahs

34K+ Hadiths

18K+ Narrators

Features

Everything you need to study

Intelligent Search

Find verses and hadiths by what they mean, not just keywords. Hybrid search fuses BM25 full-text with 1024-dimension semantic vectors across the entire corpus.

Quran Reader

114 Surahs with Tajweed-colored Arabic, Sahih International translation, and expandable Tafsir Ibn Kathir commentary — per ayah.

Hadith Explorer

Browse 34K+ hadiths from 926 books across 6 canonical collections, each with full narrator chains and source attribution.

Interactive narrator graph visualization

Narrator Networks

Interactive graph visualization of 18K+ narrators. Trace isnad chains, explore teacher-student relationships, and check Ibn Hajar reliability grades.

Early Quranic manuscript — Berlin, Wetzstein II 1913

Early Manuscripts

View high-resolution images of early Quranic manuscripts per ayah from the Corpus Coranicum archive — Berlin-Brandenburg Academy of Sciences.

Personal Study Notes

Capture your thoughts while studying. Annotate any ayah or hadith, collect evidence by topic, and embed Quran verses and hadiths inline with @mentions.

@Mentions Topic Collections Tags Color Highlights Rich Embeds

Provenance

Data Sources

Built on open scholarly datasets. Every hadith, ayah, and narrator is traceable to its source.

34K

SemanticHadith KG

Hadiths with narrator chains and knowledge graph from 6 canonical collections

6,236

QUL (Tarteel)

QPC Hafs Arabic + Sahih International English from Quranic Universal Library

6,236

Tafsir Ibn Kathir

Classical exegesis per ayah via QUL

33K

Sunnah.com

Human-verified English translations across 6 canonical collections

18K

AR-Sanad

Narrators with Ibn Hajar's reliability classifications

✦

Corpus Coranicum

Early manuscript images per ayah from Berlin-Brandenburg Academy

Get Started

Run it locally

Ilm runs entirely on your machine — no cloud, no accounts, full privacy.

Prerequisites

✓ Rust (latest stable)
✓ Node.js v20+
✓ Python 3 (for data download)
✓ SurrealDB
✓ Ollama (for Ask AI)

Get the Code

            # Clone

            git clone https://github.com/arriqaaq/ilm.git

            cd ilm

            # Download pre-built data (no ingestion needed)

            make download-data

            # Build & run

            make dev

Open Architecture

Under the Hood

A modern stack built for Islamic scholarship.

Architecture

Frontend

SvelteKit

Quran · Hadith · Search · Narrators · Graphs

JSON API

Backend

Rust / Axum

Ingest

RAG

Database

SurrealDB

Graph · HNSW vectors · BM25

Embeddings

FastEmbed

bge-m3 · 1024-dim

Training Pipeline

Raw Data SemanticHadith 34K · QUL · Sunnah.com

Parse & Enrich Join translations + narrator bios + tafsir

Generate QA ChatML pairs matching RAG prompt pattern

LoRA Fine-tune MLX on Phi-4-mini / Command-R

GGUF → Ollama Quantize Q4_K_M · ollama create · serve