nobook 0.0.3

Plain .py files as notebooks

nobook

Plain .py files as notebooks. No .ipynb files ever.

Write Python with # @block=name markers. Each block becomes a Jupyter cell. The .py file is the notebook.

File format

# @block=setup
import math
x = 42

# @block=compute
result = math.sqrt(x)
print(f"sqrt({x}) = {result:.4f}")

A block starts at # @block=name and runs until the next # @block= or end of file. That's it.

Quick start

JupyterLab / Jupyter Notebook

# 1. Create a .py file with block markers
cat > demo.py << 'EOF'
# @block=hello
print("hello from nobook")

# @block=math
import math
print(f"pi = {math.pi:.4f}")
EOF

# 2. Launch Jupyter (no install needed)
uvx nobook

Jupyter opens. Right-click demo.py in the file browser and select Open as Nobook. It renders as a notebook with two cells (hello and math). Run them, edit them, save -- the file stays .py.

VS Code

Install the Nobook VS Code extension (coming soon).

Then:

1. Open any .py file with # @block= markers
2. Right-click and select "Open as Nobook"
3. Or use the command palette: "Nobook: Create New Notebook"

As a project dependency

uv add nobook
uv run nobook

Opening .py files as notebooks

In JupyterLab / Jupyter Notebook, right-click any .py file and select Open as Nobook:

.py files containing # @block= markers will also auto-open as notebooks when double-clicked.

CLI commands

Launch Jupyter

uv run nobook           # launches Jupyter Notebook (default)
uv run nobook lab       # launches JupyterLab
uv run nobook jupyter   # same as default, launches Jupyter Notebook

All launch commands accept extra arguments that are passed through to Jupyter:

uv run nobook --port=9999 --no-browser
uv run nobook lab --port=9999

Run blocks without Jupyter

uv run nobook run example.py                 # execute all blocks, write example.out.py
uv run nobook run example.py --block=setup   # run up to and including "setup"
uv run nobook list example.py                # print block names

Output goes to .out.py with results inlined as comments:

# @block=setup
import math
x = 42
# >>>

# @block=compute
result = math.sqrt(x)
print(f"sqrt({x}) = {result:.4f}")
# >>> sqrt(42) = 6.4807

Errors show as # !!! ... lines. Since output and errors are plain comments, .out.py files are valid Python -- you can run them directly with python example.out.py.

See examples/ for sample input and output files.

Manual launch (without the CLI wrapper)

uv run jupyter notebook --ServerApp.contents_manager_class=nobook.jupyter.contentsmanager.NobookContentsManager

Or for JupyterLab:

uv run jupyter lab --ServerApp.contents_manager_class=nobook.jupyter.contentsmanager.NobookContentsManager

How it works

Jupyter UI (standard, unmodified)
        |
IPython Kernel (standard, unmodified)
        |
NobookContentsManager          <-- intercepts file I/O
        |
.py files with # @block markers

The NobookContentsManager subclasses Jupyter's LargeFileManager. On get(), it parses # @block= markers and returns a notebook model (blocks as code cells). On save(), it converts the notebook model back to .py format. The kernel and UI are completely stock.

.py files without # @block= markers are served normally (as plain text files).

JupyterLab extension

Nobook includes a JupyterLab extension that adds:

• Launcher card -- click "Nobook" in the launcher to create a new .py notebook
• "Open as Nobook" context menu -- right-click any .py file to open it as a notebook
• Block name labels -- editable labels on each cell showing the @block name

The extension is bundled with the package and installs automatically.

VS Code extension

The VS Code extension provides:

• Open .py files with # @block= markers as notebooks
• Block name display on each cell
• "Open as Nobook" context menu
• "Create New Notebook" command

Install from VS Code Marketplace (coming soon).

Development

To test nobook from any project without installing it:

uv run --project <path-to-nobook> python -m nobook

Full development install

git clone <repo-url> && cd nobook
uv sync
make setup-all      # Set up both JupyterLab and VS Code extensions
make build-all      # Build both extensions

JupyterLab extension development

cd jupyterlab-ext
jlpm install
jlpm run watch

VS Code extension development

cd vscode-ext
npm install
npm run watch
# Press F5 in VS Code to launch Extension Development Host

Project structure

nobook/
├── nobook/              # Python package
├── jupyterlab-ext/      # JupyterLab extension (TypeScript)
├── vscode-ext/          # VS Code extension (TypeScript)
├── tests/               # Python tests
└── examples/            # Example notebooks