nb2wb

Write in notebooks. Publish anywhere.

nb2wb turns notebook-native writing into paste-ready HTML for editors that do not understand Jupyter, Quarto, or LaTeX. Start from a notebook, a Markdown article, or a Quarto document, then target a neutral preview wrapper or a profile tuned for Substack, Medium, X Articles, LinkedIn, Dev.to, Hashnode, Ghost, or WordPress.

The project also ships a reverse scaffold. Use wb2nb or nb2wb.revert() to turn published HTML back into a Jupyter notebook when you need to recover prose, code blocks, and image-derived content.

Why People Reach for nb2wb

• Render code cells as syntax-highlighted images so theme and spacing survive copy and paste.
• Render display math as images and inline math as readable Unicode text.
• Keep tables as native HTML or convert them to images when a platform editor is unreliable.
• Wrap output for different publishing targets without rewriting article content.
• Serve extracted images through --serve when editors reject embedded data URIs.
• Reverse HTML posts back into notebook scaffolds, with OCR as an opt-in upgrade.
• Apply safety limits and HTML/SVG sanitization by default for server-side use.

Supported Inputs

• Jupyter notebooks: .ipynb
• Quarto documents: .qmd
• Markdown documents: .md
• In-memory notebook payloads: dict / nbformat.NotebookNode
• In-memory text payloads: raw strings or {"format": "...", "content": "..."}
• Reverse conversion inputs: .html, .htm, and in-memory HTML payloads

Installation

Install the base package:

pip install nb2wb

Install extras only when you need them:

pip install "nb2wb[ocr]"     # local OCR for reverse conversion
pip install "nb2wb[openai]"  # OpenAI-backed OCR pipeline
pip install "nb2wb[gemini]"  # Google Gemini-backed OCR pipeline

For development:

git clone https://github.com/the-palindrome/nb2wb.git
cd nb2wb
pip install -e ".[dev]"

Quick Start

Convert a notebook to the default preview wrapper:

nb2wb notebook.ipynb

Target a platform profile:

nb2wb notebook.ipynb -t medium
nb2wb notebook.ipynb -t x
nb2wb notebook.ipynb -t linkedin

Use execution, raw mode, and wrapper overrides when you need them:

nb2wb report.qmd --execute
nb2wb report.ipynb --warnings
nb2wb report.ipynb --raw -o article_raw.html
nb2wb report.ipynb -t ghost --image-strategy embed --article-width 900
nb2wb report.ipynb --serve
nb2wb report.ipynb --verbose

Reverse an HTML article back into a notebook scaffold:

wb2nb article.html
wb2nb article.html -o recovered.ipynb
wb2nb article.html --ocr-pipeline local
OPENAI_API_KEY=... wb2nb article.html --ocr-pipeline openai --model your-model-name
GEMINI_API_KEY=... wb2nb article.html --ocr-pipeline gemini --model gemini-2.0-flash
GEMINI_API_KEY=... wb2nb article.html --ocr-pipeline gemini --model gemini-2.5-flash --verbose

OpenAIOCRPipeline and GeminiOCRPipeline fetch public remote http/https image URLs.

Python API

nb2wb.convert() is content-only. Load files with helpers first, then pass the in-memory payload into the converter.

import nb2wb

payload = nb2wb.load_input_payload("notebook.ipynb")
html = nb2wb.convert(
    payload,
    target="substack",
    config={"latex": {"try_usetex": True}},
    verbose=True,
)

You can also enable package logging explicitly:

import nb2wb

nb2wb.configure_logging(verbose=True)

You can also pass text or notebook payloads directly:

import nb2wb

html = nb2wb.convert(
    {
        "format": "md",
        "content": "# Shipping Notes\n\n`nb2wb` handles this in memory.",
    },
    target="medium",
    raw_mode=True,
)

Reverse conversion follows the same pattern:

import nb2wb

payload = nb2wb.load_html_payload("article.html")
notebook = nb2wb.revert(payload)

Add OCR only when you want image-derived notebook cells:

from nb2wb.ocr.openai import OpenAIOCRPipeline

ocr_notebook = nb2wb.revert(
    payload,
    ocr_pipeline=OpenAIOCRPipeline(model="your-model-name"),
)

Examples

The examples/ directory now covers forward conversion, reverse conversion, API usage, Markdown directives, Quarto {output} chunks, visibility tags, rich HTML/SVG outputs, and target-specific publishing flows.

Useful entry points:

• examples/notebook.ipynb
• examples/markdown.md
• examples/quarto.qmd
• examples/reverse_article.html
• examples/convert_notebook_api.py
• examples/revert_html_api.py

Security Model

nb2wb keeps the safe path on by default:

• HTML and SVG fragments are sanitized.
• CSS URLs are filtered.
• Remote image fetching is SSRF-safe.
• Local image handling blocks traversal and escape paths.
• Notebook payloads are constrained by configurable size and workload limits.

Execution is different. If you enable --execute or execute=True, treat the notebook as untrusted code and isolate that runtime yourself.

Documentation

• Read the full docs: https://nb2wb.readthedocs.io/
• Start in the repo: docs/index.md
• Follow the quick path: docs/getting-started.md
• Explore all features: docs/feature-tour.md
• Integrate the API: docs/python-api.md
• Recover notebooks from HTML: docs/reverse-conversion.md

Development

Run the test suite:

pytest

Build docs locally:

pip install -e ".[docs]"
sphinx-build -b html docs docs/_build/html

The detailed test guide lives in tests/README.md.

License

MIT