A high-performance media generation library for rendering Quranic Arabic text and translations into customizable, professional-grade images.

Requirements: Python 3.13+

📌 Table of Contents

• User Guide
  • Installation
  • Quick Start
  • Presets Reference
  • Built-in Workflows
  • Demo Gallery
• Developer Guide
  • Architecture
  • Core Thesis
  • Core API Reference
  • Advanced Workflows
  • Performance & Parallelism
  • Development Suite
• Community & License

User Guide

Installation

# Clone the repository
git clone https://github.com/AtTheZenith/quranmedialib.git
cd quranmedialib

# Install with uv (recommended)
uv pip install -e .

Quick Start

from quranmedialib import DatabaseManager, VerseWorkflow, LANDSCAPE_PRESET

# Initialize database manager
db = DatabaseManager()

# Load a 1080p Landscape preset
preset = LANDSCAPE_PRESET["default"]["1080p"]
workflow = VerseWorkflow(preset)

# Render Surah 1, Ayah 1
translations = ["In the name of Allah,", "the Entirely Merciful, the Especially Merciful."]
pages = list(workflow.get_iterator(surah=1, ayah=1, translations=translations))

# Save first page
pages[0][0].save("output.png")

db.close()

Presets Reference

The library uses a resolution-independent system. All sizing parameters scale linearly from a 1080p baseline.

Built-in Workflows

Workflows are high-level orchestrators that handle data retrieval, layout, and rendering.

SurahWorkflow

Processes an entire surah page by page.

from quranmedialib import SurahWorkflow, LANDSCAPE_PRESET

preset = LANDSCAPE_PRESET["default"]["1080p"]
workflow = SurahWorkflow(preset)

for page_num, page_images in enumerate(workflow.get_iterator(surah=112), 1):
    for img, suffix in page_images:
        img.save(f"surah112_p{page_num}_{suffix}.png")

VerseRangeWorkflow

Processes a range of verses with support for parallel rendering.

from quranmedialib import VerseRangeWorkflow, SQUARE_PRESET

preset = SQUARE_PRESET["default"]["1080p"]
workflow = VerseRangeWorkflow(preset)

# translations: list[list[str]] -> [verse_index][page_index]
translations = [["Trans V1"], ["Trans V2"]] 
iterator = workflow.get_iterator(surah=1, start_ayah=1, end_ayah=2, translations=translations)

Demo Gallery

Visual examples of generated content:

Preset	Image
Landscape (1080p)
Story (1080p)
Square (1080p)

Developer Guide

Architecture

graph TD
    DB[(SQLite Database)] -->     WF[Workflow Orchestrator]
    WF --> Layout[Layout Engine: VImage]
    Layout --> Pipe[Rendering Pipeline]
    Pipe --> Mask[Mask-First Rendering]
    Mask --> Comp[Composition: Frame]
    Comp --> Out[Final Image]
    
    subgraph Config
        FC[FrameConfig] -.-> Comp
        VC[VerseConfig] -.-> Layout
        WC[WordConfig] -.-> Pipe
        TC[TextConfig] -.-> Pipe
    end

Core Thesis

QuranMediaLib is built on the "Boring Code" philosophy: prioritizing linear, obvious logic over clever abstractions to ensure long-term maintainability.

Key Engineering Pillars:

• Resolution Independence: Layouts are defined relative to a 1080p height and scaled linearly.
• Memory Efficiency: Use of __slots__ for high-frequency objects and lru_cache for database queries.
• Performance: Mask-first rendering minimizes expensive RGBA operations.

Core API Reference

DatabaseManager

A thread-safe singleton managing SQLite connections to Quranic databases.

• get_verse(surah, ayah): Retrieves verse text.
• get_verses_from_surah(surah): Retrieves all verses in a surah.
• minimize_caches(): Explicitly clears internal LRU caches.

Configuration

• Preset: Unified config container (preset.frame, preset.word, preset.verse, preset.text).
• FrameConfig: Controls frame dimensions, background, and alignment.
• VerseConfig: Controls word spacing, row spacing, max rows, and wrapping.
• WordConfig: Defines Arabic font, size, colors, and spacing.
• TextConfig: Defines translation font, size, and rich-text formatting.

Resource Management

Custom assets can be loaded via:

• FontResource.from_path("path/to/font.ttf")
• DatabaseConfig.from_path("path/to/db.sqlite")

Advanced Workflows

IsolateWordsWorkflow

Used to isolate individual words within their layout context, useful for word-by-word study tools.

from quranmedialib import IsolateWordsWorkflow, LANDSCAPE_PRESET

preset = LANDSCAPE_PRESET["default"]["1080p"]
workflow = IsolateWordsWorkflow(preset)

# Isolates each word of the verse
iterator = workflow.get_iterator(
    surah=1, 
    verse_words=["الله", "الرحمن", "الرحيم"], 
    translations=["Allah", "The Merciful", "The Compassionate"]
)

Performance & Parallelism

For bulk rendering, the library provides ParallelRenderer, which distributes tasks across CPU cores.

• Execution Modes: ExecutionMode.PROCESS (recommended for CPU-heavy tasks) or ExecutionMode.THREAD.
• Memory Guard: Per-process RSS enforcement via worker_heartbeat() every 10 verses. Workers crash immediately if they exceed 256MB. Aggregate RSS ~700MB during parallel renders is safe — no aggregate monitor needed.

Development Suite

quranmedialib.check — Validation, Benchmarking & Reference Management

The check module is the canonical entrypoint for all regression testing. It wraps pixel validation, performance benchmarks, and unit tests into a single command.

# Install dev dependencies
uv pip install -e ".[dev]"

# Full suite (pixel validation + benchmarks + unit tests)
uv run -m quranmedialib.check test

# Quick pixel validation only (no benchmarks)
uv run -m quranmedialib.check test --no-benchmark

# Unit tests only (skip pixel validation and benchmarks)
uv run -m quranmedialib.check test --unit

# (Re)generate reference images for v4.1.0
uv run -m quranmedialib.check update --version v4.1.0

# Cross-version pixel comparison
uv run -m quranmedialib.check compare v4.0.0 v4.1.0

# Standalone benchmark run
uv run -m quranmedialib.check benchmark

# List all canonical validation scenarios
uv run -m quranmedialib.check list

Reference pipeline: The update command renders all canonical scenarios to PNG files under src/quranmedialib/check/references/<version>/. Each reference set includes scenarios.json (metadata), sha256sums (integrity hashes), and a perf.json benchmark artifact. The compare command performs pixel-level diffs between versions.

Performance benchmarks: The benchmark path runs file-based rendering (saves PNGs to disk via parallel async I/O, counts pages) to avoid deserializing ~3.8GB of RGBA image bytes over IPC. Memory is checked every 10 verses via worker_heartbeat() (per-process 256MB limit). Batch sizes use natural chunking (ceil(tasks / workers)) with adaptive down-capping in bytes IPC mode via _bytes_mode_max_batch(). No aggregate memory monitor — per-process enforcement catches leaks before they cascade.

Lint and Format

uv run -m ruff check .
uv run -m ruff format .

Community & License

We welcome contributions from developers who value engineering rigor and performance.

• Contribute: See CONTRIBUTING.md for technical standards.
• Conduct: See CODE_OF_CONDUCT.md.
• License: Apache License 2.0 - see LICENSE.