np-notebooks-launcher 0.1.17

A tool to launch JupyterLab with a modified version of a notebook, based on user-specified parameters.

np-notebooks-launcher

A GUI tool for launching JupyterLab with a filtered, pre-configured version of a Jupyter notebook. Designed for experiment workflows that use the same notebook across different experiment types (e.g. ephys, hab, training).

What it does

1. Reads the first cell of a notebook to find _-prefixed variables annotated as Literal, int, float or bool types — these are the configurable parameters:

   

2. Shows a GUI with a dropdown for each variable, pre-populated with the options and default from the notebook:

   

3. Generates a filtered copy of the notebook with the selected values injected and irrelevant cells removed.

4. Launches JupyterLab with the filtered copy, with the first cell modified:

   

The launcher also has a Reset & Update button which is hard-coded to reset to the latest commit on github.com/alleninstitute/np_notebooks:main and runs uv sync on the repository.

Usage

uvx np-notebook-launcher path/to/notebook.ipynb

How to annotate a notebook

First cell — configurable variables

The first cell of the notebook is special. Any _-prefixed variable with a Literal type annotation will appear as a dropdown in the launcher GUI:

# variables in this cell with underscore-prefix can be injected by launcher:
_experiment: Literal["pretest", "ephys", "opto", "hab", "training"] = "pretest"

• The variable name must start with _
• The type annotation must be Literal[opt1, opt2, ...]
• A default value must be assigned (= "pretest")
• Supported value types: str, bool, int, float

When launched, the first cell is replaced with a markdown summary of the injected values.

Cell directives — conditional visibility

Add a directive comment as the first line of any cell to control whether it appears in the filtered notebook.

Code cells:

# /// show-if: _experiment=ephys

Markdown cells:

<!-- /// show-if: _experiment=ephys -->

Directives

Directive	Behavior
show-if: <condition>	Cell is included only when condition is true
hide-if: <condition>	Cell is hidden only when condition is true

Cells with no directive are always included.

Condition syntax

Conditions use and, or, not, and names matching the possible option values of the variable. The namespace prefix (e.g. _experiment=) is informational — what follows = is the boolean expression.

# /// show-if: _experiment=ephys
# /// show-if: _experiment=(ephys or opto) and not pretest
# /// hide-if: _experiment=training

For a string variable like _experiment = "ephys", the condition context is {"ephys": True, "pretest": False, "hab": False, ...} — each option is a key and only the selected one is True. Unknown names evaluate to False.

Project structure

src/np_notebooks_launcher/__init__.py   # all logic: parsing, filtering, GUI, CLI
notebooks/dynamic_routing.ipynb         # example annotated notebook
tests/test_launcher.py                  # unit + integration tests
pyproject.toml

Development

Install with uv:

uv sync

Run tests:

uv run task test

The entry points np-notebooks-launcher and launch both map to np_notebooks_launcher:main.

Pushing to main triggers CI/CD with publishing to PyPI.

Key functions

Function	Description
parse_first_cell_variables(cell)	Extracts annotated _vars from cell 0
parse_cell_directive(cell)	Reads show-if/hide-if from a cell's first line
cell_is_visible(cell, ctx)	Evaluates directive against a context dict
filter_notebook(nb, ctx, variable_selections)	Returns a filtered deep copy of the notebook
generate_filtered_notebook(source, ctx, ...)	Filters and writes the notebook to disk
run_launcher(notebook_path)	Opens the GUI