Visual Debugger for Jupyter Lab/IPython Notebooks | Installation, Code Examples & Debugging

A new visual debugger for JupyterLab/IPython notebooks adds interactive, breakpoint-based debugging directly in the notebook UI—complete with variable inspection in table or tree views—so developers can step through code without leaving the notebook environment. The catch: it doesn’t work with the standard IPython kernel, forcing users to install and select an experimental kernel (referred to as a “sales”/“X” Python kernel) before debugging features appear.

Setup starts with creating a dedicated Conda environment pinned to Python 3.8. After activating that environment, the workflow installs core notebook tooling (including JupyterLab and the notebook package) from Conda Forge. With the environment ready, the next step is installing the JupyterLab extension for the visual debugger by pointing JupyterLab’s extension installer at the GitHub repo for the “Jupyter web debugger.” Once the extension is installed, launching the Jupyter web instance brings up the debugger panel in the interface.

In the browser UI, the debugger is enabled via a toggle. A simple demo shows how breakpoints work: after creating a list and iterating over it, the user clicks in the gutter to set a breakpoint, runs the cell, and then watches execution pause at the marked line. While stopped, variables appear in a dedicated “variables” area. Toggling the view formats those variables into a more readable table, and a tree view option helps inspect nested structures. From there, execution can be continued or stepped through line by line, with the debugger updating values such as the loop variable as it advances through the list.

A more practical example demonstrates debugging during data filtering. The notebook defines a contrived list of YouTube creators—some labeled with “Dr.” in their names—and assigns each creator a subscriber count. The code builds a dictionary of “doctors” by iterating through names and subscriber counts and using string matching: if a name contains “doctor,” the entry is added to the doctors dictionary. Breakpoints placed inside the loop let the user observe the control flow in real time: the debugger halts at the conditional, shows when the doctors dictionary is empty at the start, and then reveals it growing as qualifying “Dr.” entries are encountered. Stepping through successive iterations makes it clear which creators match the condition and which do not.

Finally, the notebook computes aggregate metrics: it sums the subscriber counts stored in the doctors dictionary, calculates the total subscribers across all creators, and derives the fraction of subscribers belonging to the “doctors” subset. Using the debugger’s variable inspection at intermediate steps, the fraction comes out to roughly 13.1%. The overall takeaway is straightforward: once the extension and experimental kernel are in place, the visual debugger turns notebook execution into an interactive, inspectable process—breakpoints, stepping, and structured variable views included—making it easier to validate logic and catch mistakes during development.