ᓚᘏᗢ Bengal

The documentation generator built on Python's free-threading future

Every layer — markdown parsing, syntax highlighting, templates, dev server — is pure Python, scales with your cores, and ships zero npm dependencies.

pip install bengal
bengal new site --name mysite && cd mysite && bengal serve

---

Why Bengal?

Bengal is a static site generator where the entire stack is Python you can read, debug, and extend. No JavaScript toolchains. No compiled C extensions in the critical path. Every library in the pipeline — Patitas (markdown), Rosettes (syntax highlighting), Kida (templates), Pounce (dev server) — is purpose-built for Python 3.14+ free-threading.

What this gets you:

• Scales with cores — Parallel builds with true thread parallelism on Python 3.14t. No GIL contention.
• AI-native output — Generates llms.txt, agent manifests, per-page JSON, and Content Signals so AI agents can discover, navigate, and respect your content policies.
• Sub-second rebuilds — Provenance-based incremental builds with content-addressed hashing. Only changed pages rebuild.
• Python-native workflows — Jupyter .ipynb rendering, autodoc for Python/CLI/OpenAPI, pip install and go.
• Batteries included — Sitemap, RSS, social cards, search indexes, broken link detection, content validation.
• Extensible — Pluggable engines for templates, Markdown, and syntax highlighting. 9 plugin extension points.

Use Bengal For

• Documentation sites — Versioned docs, API reference, search, and internal linking
• Blogs and journals — Tags, categories, feeds, related content, and social sharing
• Knowledge bases — Markdown-first publishing with validation and JSON search indexes
• Product and marketing sites — Landing pages, content collections, and social cards

---

Quick Commands

Command	Description
bengal build	Production build
bengal serve	Dev server with live reload
bengal check	Health checks and validation
bengal fix	Auto-fix common issues
bengal inspect graph	Site structure analysis

Aliases: b (build), s (serve), v (check)

---

Site Scaffolding

Interactive Wizard — Guided setup with presets

Run without arguments for a guided experience:

bengal new site

The wizard prompts for site name, base URL, and presents preset options:

🎯 What kind of site are you building?
  📝 Blog            - Personal or professional blog
  📚 Documentation   - Technical docs or guides
  💼 Portfolio       - Showcase your work
  🛒 Product         - Product site with listings and features
  📄 Resume          - Professional resume/CV site
  📦 Blank           - Empty site, no initial structure
  ⚙️  Custom         - Define your own structure

Each preset creates a complete site with appropriate sections, sample content, and configuration.

Direct Template Selection — Skip prompts with explicit options

Create sites non-interactively with --template:

bengal new site --name my-docs --template docs
bengal new site --name my-blog --template blog
bengal new site --name portfolio --template portfolio

Available templates:

Template	Description	Sections Created
default	Basic site structure	Home page only
blog	Personal/professional blog	blog, about
docs	Technical documentation	getting-started, guides, reference
portfolio	Showcase work	about, projects, blog, contact
product	Product site with listings	products, features, pricing, contact
resume	Professional CV	Single resume page
landing	Single-page landing	Home, privacy, terms
changelog	Release notes timeline	Changelog with versions

Add Pages and Templates — Expand without recreating

Add content and template files to an existing Bengal site:

# Add a page under a section
bengal new page --name getting-started --section docs

# Add template files
bengal new layout --name docs-page
bengal new partial --name docs-sidebar

# Add a content type strategy scaffold
bengal new content-type --name tutorial

Use frontmatter, content type strategies, and section index pages to shape how new content is grouped and rendered.

Content Type Scaffolds — Python-defined content behavior

Generate a starter ContentTypeStrategy when a section needs custom matching, template selection, summaries, or pagination behavior:

bengal new content-type --name changelog

The generated file contains TODO markers for matching, template selection, excerpt generation, and pagination so custom content behavior stays in Python.

---

Features

Feature	Description	Docs
SEO & Discovery	Sitemap, RSS, canonical URLs, Open Graph, social cards, search, JSON/LLM output, llms.txt, Content Signals	SEO & Discovery →
Directives	Tabs, admonitions, cards, dropdowns, code blocks	Content →
Notebooks	Native Jupyter .ipynb rendering, Binder/Colab links	Notebooks →
Autodoc	Generate API docs from Python, CLI, OpenAPI	Autodoc →
Remote Sources	Pull content from GitHub, Notion, REST APIs	Sources →
Image Processing	Resize, crop, format conversion (WebP/AVIF), srcset generation	Images →
Content Collections	Type-safe frontmatter with dataclass/Pydantic schemas	Collections →
Theming	Dark mode, responsive, syntax highlighting, search	Theming →
Validation	Health checks, broken link detection, auto-fix	Building →
Performance	Parallel builds, incremental rebuilds, streaming	Large Sites →
Zero-Config Deploy	Auto-detects GitHub Pages, Netlify, Vercel	Deployment →

📚 Full documentation: lbliii.github.io/bengal

---

Configuration

Single-file — Simple projects

# bengal.toml
[site]
title = "My Site"
baseurl = "https://example.com"

Directory-based — Multi-environment projects

config/
├── _default/           # Base configuration
│   ├── site.yaml
│   └── build.yaml
├── environments/       # Environment overrides
│   └── production.yaml
└── profiles/           # Build profiles
    └── dev.yaml

bengal build --environment production    # Production environment
bengal build --profile dev    # Development profile

📖 Configuration guide: Configuration →

---

Project Structure

mysite/
├── content/          # Markdown pages
├── templates/        # Custom templates (optional)
├── assets/           # Static files (CSS, JS, images)
├── data/             # YAML/JSON data files
├── config/           # Configuration directory
└── public/           # Build output

---

Theming

Bengal ships with a modern, accessible default theme:

• Dark mode with system preference detection
• Responsive design with mobile navigation
• Syntax highlighting with copy buttons
• Table of contents with scroll spy
• Full-text search (Lunr.js)

Customize templates:

{# templates/page.html #}
{% extends "base.html" %}

{% block content %}
<article class="prose">
  <h1>{{ page.title }}</h1>
  {{ content | safe }}
</article>
{% endblock %}

---

Pluggable Engines

Bengal follows a "bring your own" pattern — swap engines without changing your content.

Template Engine

Engine	Description	Install
Kida	Bengal's native engine. AST-native, free-threading safe, Jinja2-compatible syntax	Built-in

Kida is the built-in template engine. It uses Jinja2-compatible syntax, so existing Jinja2 templates generally work without changes. See the Kida migration guide if migrating from Jinja2.

Markdown Parsers

Parser	Description	Install
Patitas (default)	Bengal's native parser. Typed AST, O(n) parsing, thread-safe	Built-in
Mistune	Fast, modern parser	Built-in
Python-Markdown	Full-featured, extensive extensions	Built-in

# config/_default/content.yaml
markdown:
  parser: patitas  # default, or mistune (legacy), python-markdown

Syntax Highlighters

Backend	Description	Install
Rosettes (default)	Lock-free, 55+ languages, O(n) guaranteed	pip install rosettes

# config/_default/theme.yaml
highlighting:
  backend: rosettes

Custom backends can be registered via register_backend().

---

Requirements

• Python 3.14+ (uses free-threading and PEP 784 compression)
• Linux, macOS, Windows

---

Philosophy

Bengal prioritizes correctness and clarity over backwards compatibility.

Each release represents the best solution we know how to deliver. When existing behavior no longer reflects the best design, it changes. Upgrades may require reading release notes and making adjustments.

• Fail loudly — Breaking changes produce clear errors
• User control — You choose when to upgrade; we choose what changes
• No hidden layers — No compatibility shims or deprecated code paths

If you need multi-year stability, pin your version.

---

Documentation

📚 lbliii.github.io/bengal

---

Development

git clone https://github.com/lbliii/bengal.git
cd bengal
uv sync --group dev
pytest

Bengal is a standalone uv project. Run uv sync and uv run ... from this checkout; sibling repositories are not required for normal development.

---

The Bengal Ecosystem

A structured reactive stack — every layer written in pure Python for 3.14t free-threading.

ᓚᘏᗢ	Bengal	Static site generator ← You are here	Docs
∿∿	Purr	Content runtime	—
⌁⌁	Chirp	Web framework	Docs
=^..^=	Pounce	ASGI server	Docs
)彡	Kida	Template engine	Docs
ฅᨐฅ	Patitas	Markdown parser	Docs
⌾⌾⌾	Rosettes	Syntax highlighter	Docs

Python-native. Free-threading ready. No npm required.

---

License

MIT