Verso.Python 1.2.2

Verso.Python

Python language kernel extension for Verso notebooks.

Overview

Runs the Python already installed on the machine as a separate process, so a notebook sees the same interpreter, virtual environment, and packages a terminal would. Nothing is embedded and no Python is bundled. Requires Python 3.8 or newer, including 3.13 and 3.14.

The interpreter comes from an explicit setting, the VERSO_PYTHON variable, a #!python selection, an active virtual environment, an active conda environment, a .venv or venv directory beside the notebook or above it, PATH, or a well-known install location, in that order. The first candidate that reports a supported CPython version wins.

Features

• Python execution with persistent state across cells, output streamed as it is produced, and full traceback formatting
• Interrupt and restart, including a restart that clears imported module state
• Crash isolation: a native crash takes down the interpreter, reports itself, and the next cell runs in a fresh one
• IntelliSense answered inside the running interpreter, so it sees the packages that are actually installed and the names earlier cells defined. Uses jedi when available, with standard library fallbacks for completions and syntax diagnostics
• Bidirectional variable sharing between Python and every other kernel in the notebook
• Matplotlib integration with automatic figure capture, including figures produced part-way through a cell
• Interactive widgets built on ipywidgets or anywidget, live in both editors: moving a control reaches the interpreter, the author's callbacks run, and the notebook saves the state a reader last saw. A file without a kernel behind it draws the same widget from that state and says so
• #!bind to share a widget's trait as a notebook variable, so a C#, F#, or PowerShell cell reads the control's value and writing it moves the control
• display() function supporting _repr_html_(), _repr_png_(), _repr_svg_(), with an optional MIME type hint (for example display(obj, "application/json"))
• input() support, prompting in the notebook
• Package management through #!pip, the %pip and ! shell escapes, consent-gated install on import, inline script metadata, and a per-notebook requirement list, all uv-accelerated when uv is present
• Interpreter selection with #!python, including #!python --list to see what was discovered

Installation

dotnet add package Verso.Python

This package ships with Verso and is loaded automatically, so a notebook needs no reference to it. Add it directly when embedding the engine in your own application.

It depends on Verso.Abstractions and nothing else. No Python is bundled and no interpreter is embedded, so the package carries no native dependency of its own.

Quick Start

import json
data = {"name": "Verso", "version": 1}
print(json.dumps(data, indent=2))

Install packages explicitly:

#!pip pandas matplotlib

Or just import them, and approve the prompt:

import pandas as pd

Check or change the interpreter:

#!python --list
#!python /usr/local/bin/python3.13