Plot Panel

A plots pane, a variables explorer and an interactive console for VS Code, so notebook work feels less like scrolling through a log.

If you use notebooks for data work, you know the pattern: you plot something, run three more cells, and now the figure you wanted is somewhere above, between two stack traces. Plot Panel keeps every figure your kernel produces in one pane, shows your variables in another, and gives you a real Python or R prompt at the bottom of the window.

The plots pane

Figures land there by themselves, from notebooks and from the Interactive Window. The current one is shown large, the whole session sits underneath as thumbnails, and the strip reads like a log: every run adds its figures again, even unchanged ones, under a marker, "Run all 3" for a whole notebook or "Run 12" for one cell (12 being its execution count), so after several runs you still know which figure is new.

The history survives a restart, and each project keeps its own. Open another repository and you get a clean strip; come back and your figures are still there.

The title bar carries the usual things: previous and next, save, copy, zoom presets, and a menu to open a plot in an editor tab or a floating window. There is a dark filter for figures that are too bright against a dark theme, and a Plot Code menu that copies the code behind a figure, jumps to the cell that made it, or runs it again.

The variables view

After a cell finishes, the view lists what the kernel holds, in four columns: name, value, type and size. Values are shown for constants (alpha 2.0, best_D 8 for an np.int64); a container shows its type and its size in numpy's shape notation, (10,), (100, 100), (2, 3, 10). Two dividers can be dragged, name|value and value|type; double-click the second to fit the longest type, and hover a type for its full name.

Rows open. A DataFrame opens into its columns, a column into its values, a list into its items, a dict into its keys. Anything a data viewer can display gets a grid button, which hands it to Data Wrangler if you have that installed.

You can group by kind or by size, sort by name, size or recency, and filter by name. Variables that changed during the last run are marked, so a long Run All does not leave you comparing lists by eye.

The console

A Python prompt in the bottom panel, next to the terminal. Enter runs, Shift and Enter add a line, Tab indents, and the arrows walk back through what you typed. Completions come from the session itself, so df. lists the columns that frame actually has rather than a guess from static analysis.

If IPython is available, magics work, including %run on a notebook file. You can open several sessions, restart one, interrupt a runaway loop, and choose which interpreter to start: any environment the Python extension knows about, or R if Rscript is on your path.

Running something in a console makes the variables view follow it, and clicking back into a notebook hands the view back.

Requirements

VS Code 1.106 or later. The Jupyter extension, for notebooks and their kernels. The Python extension is optional and used to find your interpreters. R sessions need Rscript. IPython, if present in the environment you start, is what makes magics work.

Nothing is bundled: the extension has no runtime dependencies, makes no network requests, and collects nothing.

Getting the side-by-side layout

The two views ship in one container, Variables above Plots, that opens in the secondary sidebar as a tab next to Chat (Ctrl+Alt+B, or View, Appearance, Secondary Side Bar, if it is hidden). This needs VS Code 1.106 or later. You can still drag the container anywhere else; VS Code remembers where you put it.

What it cannot do

Worth knowing before you rely on it.

The console runs its own interpreter, so it does not see the variables in your notebook's kernel. Running code inside a notebook kernel and reading the result back is reserved by the Jupyter extension to a short list of publishers. That restriction is also why expansion in a notebook is limited: the view only gets truncated text descriptions, which stop nesting after two levels and cut every nested item at 128 characters. A DataFrame inside a list shows its shape but usually will not open further. A console session has no such ceiling, because the process belongs to the extension.

For the same reason, functions and classes in a notebook are read from the source of the cells you ran rather than from live objects, so what you see is their signature. Imported classes do not show up at all. A console lists them properly.

Figures drawn by plotly, bokeh or ipywidgets never reach the notebook as an image, so there is nothing to capture and the pane says so instead of going quiet. Identical figures are stored once: re-running code that produces exactly the same image selects the existing entry rather than adding a copy.

Variables cannot appear while a cell is still running. The introspection that lists them runs on the kernel, so it waits its turn behind your code.

Settings

With jupyter.generateSVGPlots enabled, kernels also emit SVG and the pane prefers it: crisper when zoomed, usually smaller on disk.

If your figures come out with a dark background, that is jupyter.themeMatplotlibPlots, a Jupyter setting that repaints matplotlib to match your theme. Set it to false for white figures.

Commands

Everything is on the title bars, and all of it is in the command palette under Plot Panel. The ones worth a keybinding are Send Selection to Console (Ctrl+Enter in a Python file), Previous and Next Plot (left and right arrows when the pane has focus), and Refresh Variables.

Clearing has no key by default, so none of yours is taken over. To give it one, open Keyboard Shortcuts (Ctrl+K Ctrl+S) and bind Plot Panel: Clear Plot History and Plot Panel: Clear Plot History and Variables (the latter asks before restarting a notebook's kernel), for instance in keybindings.json:

{ "key": "ctrl+l", "command": "plotPanel.clearHistory", "when": "!terminalFocus" },
{ "key": "ctrl+shift+l", "command": "plotPanel.clearAll", "when": "!terminalFocus" }

Trademarks

An independent project, not affiliated with, endorsed by or sponsored by Posit Software, PBC or Microsoft. Positron and RStudio are trademarks of Posit Software, PBC, named here only to describe the layout this extension reproduces. No Positron source code was consulted or used: the implementation is original, and the behaviour it matches was derived from public behaviour and from the VS Code and Jupyter extension APIs.