Maktabati/shamela-vectors

🇬🇧 English

The largest open-source vector database of the complete al-Maktaba al-Shamela (المكتبة الشاملة) Islamic text corpus, prepared for semantic search and Retrieval-Augmented Generation (RAG). Includes the full Quran text (6,236 verses, Hafs 'an 'Asim).

Statistics:

• ~8–11M chunks from 8,589 classical Islamic books (final count pending — indexing in progress)
• 6,236 Quran verses (one verse = one chunk)
• 40 categories covering the full breadth of Islamic scholarship
• Period: 1st century AH to 15th century AH
• Language: Arabic (primary)

⚠️ Dataset upload in progress — indexing is currently running on a local machine. The Parquet files will be uploaded once indexing and HNSW build are complete. This README already reflects the final schema and usage instructions.

Schema is fully consistent with Maktabati/openiti-vectors — same field names, types, and conventions. Both datasets can be queried together seamlessly.

🔍 Live Demo

Try the search demo directly in your browser — no setup required:

👉 maktabati.ai · Shamela Search (coming soon)

Keyword search (author, title, full text). For semantic / vector search, import into Qdrant using the instructions below.

---

🇸🇦 العربية

أكبر قاعدة بيانات متجهية مفتوحة المصدر للمكتبة الشاملة، تضمّ ما يزيد على 7 ملايين مقطع نصي مستخرج من 8,589 كتابًا من كتب التراث الإسلامي، مُعدَّة للبحث الدلالي وتطبيقات الذكاء الاصطناعي التوليدي (RAG).

تشمل المجموعة كامل القرآن الكريم برواية حفص عن عاصم (6,236 آية) — آية واحدة لكل مدخل — وهو ما يجعل الاسترجاع القرآني دقيقًا على مستوى الآية.

تم التضمين والفهرسة بالكامل على حاسوب شخصي دون أي خدمات سحابية، باستخدام نموذج intfloat/multilingual-e5-base بأبعاد 768 ومسافة كوساين.

الإحصاءات:

• ~8–11 مليون مقطع نصي من 8,589 كتابًا (الفهرسة جارية حاليًا)
• 6,236 آية قرآنية (آية = مقطع واحد)
• 40 تصنيفًا تغطي علوم الإسلام كاملةً
• النصوص من القرن الأول إلى الخامس عشر الهجري
• منجز على حاسوب شخصي — دون خدمات سحابية

---

🇩🇪 Deutsch

Die größte Open-Source-Vektordatenbank des vollständigen al-Maktaba al-Shamela-Korpus islamischer Texte, vorbereitet für semantische Suche und Retrieval-Augmented Generation (RAG). Enthält den vollständigen Korantext (6.236 Verse, Hafs 'an Asim).

Statistiken:

• ~8–11 Millionen Textabschnitte aus 8.589 klassischen islamischen Büchern
• 6.236 Koranverse (ein Vers = ein Abschnitt)
• 40 Kategorien — das gesamte Spektrum der islamischen Wissenschaften
• Zeitraum: 1. bis 15. Jahrhundert n. H.
• Sprache: Arabisch

Vollständig eingebettet und indiziert auf einem handelsüblichen Heimcomputer ohne Cloud-Dienste. Das Schema ist konsistent mit Maktabati/openiti-vectors.

🇹🇷 Türkçe

al-Maktaba al-Shamela İslami metin külliyatının tamamına ait en büyük açık kaynaklı vektör veritabanı; anlamsal arama ve Üretim Destekli Geri Getirme (RAG) için hazırlanmıştır. Tam Kur'an metnini içermektedir (6.236 ayet, Hafs rivayeti).

İstatistikler:

• 8.589 klasik İslami kitaptan ~8–11 milyon metin parçası
• 6.236 Kur'an ayeti (bir ayet = bir parça)
• 40 kategori — İslam ilimlerinin tüm alanlarını kapsar
• Dönem: Hicri 1. ila 15. yüzyıl
• Dil: Arapça

Bulut hizmeti kullanılmadan, tüketiciye yönelik bir bilgisayarda tamamen gömülü ve dizine eklenmiştir. Şema Maktabati/openiti-vectors ile tam uyumludur.

🇮🇷 فارسی

بزرگ‌ترین پایگاه داده برداری متن‌باز از مجموعه کامل المکتبة الشاملة، آماده‌شده برای جستجوی معنایی و تولید متن مبتنی بر بازیابی (RAG). شامل متن کامل قرآن کریم (۶٬۲۳۶ آیه، روایت حفص از عاصم) می‌باشد.

آمار:

• ~۸ تا ۱۱ میلیون قطعه متنی از ۸٬۵۸۹ کتاب اسلامی کلاسیک
• ۶٬۲۳۶ آیه قرآنی (هر آیه = یک قطعه)
• ۴۰ دسته‌بندی — پوشش کامل علوم اسلامی
• دوره: قرن اول تا پانزدهم هجری
• زبان: عربی

به‌طور کامل بر روی یک رایانه شخصی و بدون استفاده از خدمات ابری پردازش شده است. طرح‌واره با Maktabati/openiti-vectors کاملاً سازگار است.

---

Technical Details

Parameter	Value
Embedding Model	intfloat/multilingual-e5-base
Vector Dimension	768
Distance Metric	Cosine
Chunk Size	512 tokens
Overlap	50 tokens
Tokenizer	XLM-RoBERTa (via multilingual-e5-base)
Passage Prefix	passage: (indexing)
Query Prefix	query: (search)

Indexed and embedded on a consumer-grade home PC:

Component	Spec
CPU	AMD Ryzen 9 5900X
RAM	32GB DDR4
GPU	AMD Radeon RX 6900 XT (16GB VRAM, ROCm)

No cloud services used; fully reproducible on consumer hardware.

---

Dataset Fields

Common Fields (identical with Maktabati/openiti-vectors)

Field	Type	Description
text	string	Original chunk text with full tashkeel
text_norm	string	Normalized text (no tashkeel, unified alef variants)
sha256	string	SHA-256 hash of chunk text (64 chars)
author	string	Arabic author name (e.g. ابن كثير)
title	string	Arabic book title (e.g. تفسير ابن كثير)
death_year	int32	Author death year in AH
language	string	"ar"
page	string	Page reference — books: "V02P045", Quran: "2:255"
char_start	int32	Character offset start within page text
char_end	int32	Character offset end within page text
chunk_no	int32	Chunk index within page
indexed_at	string	ISO 8601 timestamp of indexing

Shamela-specific Fields

Field	Type	Description
source	string	"shamela" or "quran"
book_id	int32	Shamela book ID
page_id	int32	Global page ID
page_num	int32	Page number within volume
part	int32/string	Volume/part number (numeric or Arabic label e.g. "المقدمة")
category_id	int32	Shamela category ID
category_name_ar	string	Category name in Arabic (e.g. كتب السنة)
sequence_num	int32	Reading order sequence number within book
book_type_label	string	Book type (كتاب، رسالة، ...)

Quran-specific Additional Fields

Field	Type	Description
surah_num	int32	Surah number (1–114)
surah_name	string	Surah name in Arabic (e.g. البقرة)
ayah_num	int32	Ayah number within surah
global_id	int32	Sequential Quran verse ID (1–6236)

Note on Quran text_norm: Unlike book entries, text_norm is identical to text for Quran verses — no normalization applied, preserving ة and ى.

---

Source Data

• Shamela corpus: AuthenticIlm/Shamela4_Full_DB
• Quran text: Hafs 'an 'Asim (رواية حفص عن عاصم), Majma' al-Malik Fahd (مجمع الملك فهد لطباعة المصحف الشريف), via Shamela4_Full_DB
• Quran verification: Spot-checked Al-Fatiha, Ayat al-Kursi (2:255), Al-Ikhlas, An-Nas — all correct

---

Usage

Import into Qdrant

import pyarrow.parquet as pq
from pathlib import Path
from qdrant_client import QdrantClient
from qdrant_client.models import (
    VectorParams, Distance, PointStruct,
    HnswConfigDiff, OptimizersConfigDiff
)

client = QdrantClient("http://localhost:6333")

# Create collection (vectors on disk, HNSW in RAM)
client.create_collection(
    collection_name="shamela",
    vectors_config=VectorParams(size=768, distance=Distance.COSINE, on_disk=True),
    hnsw_config=HnswConfigDiff(m=16, ef_construct=100, on_disk=False),
    optimizers_config=OptimizersConfigDiff(indexing_threshold=0),
    on_disk_payload=True
)

# Import Parquet files
for path in sorted(Path("shamela-vectors").glob("*.parquet")):
    table = pq.read_table(path)
    points = [
        PointStruct(
            id=row["id"],
            vector=row["vector"],
            payload={k: row[k] for k in table.schema.names if k not in ("id", "vector")}
        )
        for row in table.to_pylist()
    ]
    client.upsert(collection_name="shamela", points=points, wait=False)

# Enable HNSW indexing after upload
client.update_collection(
    collection_name="shamela",
    optimizers_config=OptimizersConfigDiff(
        indexing_threshold=1,
        max_optimization_threads=4
    )
)
# Wait until indexed_vectors_count == points_count

Semantic Search

from transformers import AutoTokenizer, AutoModel
import torch
from qdrant_client import QdrantClient

model_name = "intfloat/multilingual-e5-base"
tokenizer  = AutoTokenizer.from_pretrained(model_name)
model      = AutoModel.from_pretrained(model_name)
model.eval()

def embed(text: str) -> list:
    # Important: use "query: " prefix for search queries
    encoded = tokenizer(f"query: {text}", return_tensors="pt",
                        truncation=True, max_length=512)
    with torch.no_grad():
        output = model(**encoded)
    emb = output.last_hidden_state.mean(dim=1)
    return torch.nn.functional.normalize(emb, p=2, dim=1)[0].tolist()

client  = QdrantClient("http://localhost:6333")
vector  = embed("ما حكم الصلاة في الأماكن المغصوبة")
results = client.search(
    collection_name="shamela",
    query_vector=vector,
    limit=10,
    with_payload=True
)

for r in results:
    print(f"Score: {r.score:.3f} | {r.payload['author']} | {r.payload['title']}")
    print(r.payload['text'][:200])
    print()

Filter by Source or Category

from qdrant_client.models import Filter, FieldCondition, MatchValue

# Quran only
results = client.search(
    collection_name="shamela",
    query_vector=vector,
    query_filter=Filter(must=[
        FieldCondition(key="source", match=MatchValue(value="quran"))
    ]),
    limit=10
)

# Hadith books only (كتب السنة)
results = client.search(
    collection_name="shamela",
    query_vector=vector,
    query_filter=Filter(must=[
        FieldCondition(key="category_name_ar",
                       match=MatchValue(value="كتب السنة"))
    ]),
    limit=10
)

# Filter by author
results = client.search(
    collection_name="shamela",
    query_vector=vector,
    query_filter=Filter(must=[
        FieldCondition(key="author", match=MatchValue(value="ابن كثير"))
    ]),
    limit=10
)

---

Related Datasets

• Maktabati/openiti-vectors — OpenITI corpus vectors, same schema (Arabic, Persian, Turkish, Urdu)

---

License

The texts are from al-Maktaba al-Shamela and subject to their respective licenses. The embeddings and metadata in this dataset are released under CC BY 4.0.

---

Citation

@dataset{maktabati_shamela_vectors_2026,
  title     = {Shamela Vector Database},
  author    = {maktabati.ai},
  year      = {2026},
  publisher = {HuggingFace},
  url       = {https://huggingface.co/datasets/Maktabati/shamela-vectors},
  note      = {Complete al-Maktaba al-Shamela corpus: 8,589 books + 6,236 Quran verses,
               embedded with intfloat/multilingual-e5-base on consumer hardware}
}