hydrogen-next

Run code interactively with Jupyter kernels. Supports Python, R, JavaScript, and other languages with rich output including plots, images, HTML, and LaTeX.

Features

• Interactive execution: Run lines, selections, or code blocks with inline results.
• Rich media output: Displays plots, images, video, HTML, LaTeX, and more. Ctrl+Click to open images in image-editor.
• Watch expressions: Auto-run and track variables.
• Kernel completions: Autocomplete powered by the running kernel.
• Code introspection: Inline inspection of objects from the kernel.
• Shared namespace: One kernel per language across files.
• Smart code detection: Intelligently detects Python blocks, brackets, and folds.
• Variable explorer: Browse Python variables in a dedicated panel.
• Exec panel: Command history with re-execution support.
• Multi-cursor support: Run with multiple cursors and selections.
• Custom connections: Connect to remote kernels (e.g., Docker).
• Navigation panel: Cell markers via navigation-panel.
• Scrollmap: Cell markers via scrollmap.
• AI integration: Attach code input and output to claude-chat for AI-assisted analysis.

Installation

To install hydrogen-next search for hydrogen-next in the Install pane of the Pulsar settings or run ppm install hydrogen-next. Alternatively, you can run ppm install asiloisad/pulsar-hydrogen-next to install a package directly from the GitHub repository.

Kernel installation

hydrogen-next requires Jupyter kernels to be installed on your system. A kernel is a language-specific backend that executes your code. You can install kernels for many languages — see the full list of available kernels on the Jupyter wiki.

Python (IPython)

Python is the most common kernel. Install it with:

pip install ipykernel
python -m ipykernel install --user

To register a kernel from a virtual environment, activate it first and install with a display name:

source myenv/bin/activate        # Linux/macOS
myenv\Scripts\activate           # Windows
pip install ipykernel
python -m ipykernel install --user --name myenv --display-name "Python (myenv)"

The --name flag sets the kernel directory name (used in magic comments), and --display-name sets the label shown in the kernel picker. Once registered, the kernel remains available even when the venv is not activated — it points directly to the venv's Python interpreter.

To remove a kernel you no longer need:

jupyter kernelspec uninstall myenv

See the IPython kernel documentation for more details.

R (IRkernel)

Install the IRkernel package from R and register it with Jupyter:

R -e "install.packages('IRkernel'); IRkernel::installspec()"

See the IRkernel documentation for more details.

JavaScript / TypeScript

Several JavaScript runtimes provide Jupyter kernels. For IJavascript:

npm install -g ijavascript
ijsinstall

For Deno, the kernel is built-in:

deno jupyter --install

Julia (IJulia)

Install the IJulia package from the Julia REPL:

using Pkg; Pkg.add("IJulia")

Verifying installation

To list all installed kernels:

jupyter kernelspec list

For general information on installing and managing kernels, see the Jupyter documentation.

Kernel selection

When multiple kernels are available for a language, you can specify which kernel to use with a magic comment <comment>:: kernelname on the first line. The comment character is automatically detected based on the language:

#:: python3
import numpy as np

//:: deno
console.log("Hello from Deno");

Matching rules:

• Case-sensitive: Must match exactly (e.g., python3 not Python3)
• Kernel name: The directory name from jupyter kernelspec list (e.g., python3)
• Display name: The human-readable name (e.g., Python 3.13)

If no match is found, falls back to normal behavior (picker or auto-select).

In the kernel picker, press Ctrl+Enter to insert the selected kernel as a magic comment instead of starting it.

Kernel gateways

Connect to remote or local Jupyter servers by configuring kernel gateways in settings.

Example of local jupyter server:

jupyter server --ServerApp.token='test123'

In the hydrogen-next settings, add gateway entries as JSON:

[
  {
    "name": "Local Jupyter",
    "options": {
      "baseUrl": "http://localhost:8888"
    }
  }
]

Use Hydrogen: Connect to Remote Kernel command to select a gateway and kernel. After selecting a gateway, you'll be prompted to choose an authentication method:

• No credentials — for servers without authentication
• Authenticate with a token — prompts for the server token
• Authenticate with a cookie — prompts for a cookie value

If you prefer to skip the prompt, you can include the token directly in the gateway config:

[
  {
    "name": "Local Jupyter",
    "options": {
      "baseUrl": "http://localhost:8888",
      "token": "your-server-token-here"
    }
  }
]

Code block detection

When you run code without a selection, hydrogen-next intelligently detects what to execute based on cursor position.

Priority order

1. Selection - If text is selected, execute exactly that
2. Language Specials - Python compound statements (see below)
3. Brackets - Multi-line bracket expressions (), [], {}
4. Folds - Foldable language constructs
5. Single Line - Current line as fallback

Python support

Cursor Position	What Gets Executed
On def/class line	Entire function/class (with decorators)
On @decorator line	Decorated function/class
On if/elif/else line	Entire if-elif-else chain
On try/except/finally line	Entire try block
On for/while line	Loop with optional else
On with/match line	Entire block
Inside body	Single line only

Bracket expressions

Cursor Position	What Gets Executed
On line ending with [, (, {	Entire bracket block
On line starting with ], ), }	Entire bracket block
Inside bracket block	Single line only

Examples

# Cursor on "if" → executes entire if-elif-else
if x > 0:
    print("positive")
elif x < 0:
    print("negative")
else:
    print("zero")

# Cursor on "print" inside body → executes only that line
if x > 0:
    print("positive")  # ← cursor here = single line

# Cursor on "[" → executes entire list
data = [
    1,
    2,
    3,
]

# Cursor on "2," inside list → executes only "2,"
data = [
    1,
    2,  # ← cursor here = single line
    3,
]

This allows you to execute entire blocks from control lines, while still being able to inspect individual lines inside bodies.

Output interactions

Click on output results to interact with them:

Action	Effect
Click	Copy to clipboard (image or text)
Ctrl+Click (Cmd+Click on macOS)	Open in editor (images open in image-editor)

Images opened via Ctrl+Click are displayed in the image-editor package with full editing capabilities (zoom, pan, filters, save-as).

Provided Service hydrogen.provider

Allows other packages to interact with Jupyter kernels — execute code, get completions, inspect objects, and monitor kernel state.

In your package.json:

{
  "consumedServices": {
    "hydrogen.provider": {
      "versions": {
        "^1.3.0": "consumeHydrogen"
      }
    }
  }
}

In your main module:

module.exports = {
  consumeHydrogen(hydrogen) {
    this.hydrogen = hydrogen;
  },

  async example() {
    const kernel = this.hydrogen.getActiveKernel();
    const result = await kernel.execute("print('Hello')");
    console.log(result.status); // 'ok' or 'error'
  }
};

HydrogenProvider methods

Method	Description
getActiveKernel()	Get the kernel for the active editor
onDidChangeKernel(callback)	Subscribe to kernel changes
getCellRange(editor)	Get the current cell range

HydrogenKernel API

Execution

Method	Description
execute(code)	Execute code, returns Promise<{status, outputs, error}>
executeWithCallback(code, callback)	Execute with streaming callback

State & Control

Property/Method	Description
executionState	Current state: 'idle', 'busy', 'starting'
executionCount	Current execution count
lastExecutionTime	Last execution time string (e.g., "1.23s")
onDidChangeExecutionState(callback)	Subscribe to state changes, returns Disposable
interrupt()	Interrupt running execution
restart([callback])	Restart the kernel
shutdown()	Shutdown the kernel

Introspection

Method	Description
complete(code)	Get completions, returns Promise<{matches, ...}>
inspect(code, cursorPos)	Get documentation, returns Promise<{data, found}>

Kernel info

Property/Method	Description
language	Kernel language (e.g., "python")
displayName	Kernel display name (e.g., "Python 3")
kernelSpec	Full kernel spec object
getConnectionFile()	Path to kernel connection file

Events & Middleware

Method	Description
onDidDestroy(callback)	Called when kernel is destroyed
addMiddleware(middleware)	Add execution middleware

Example: Execute and Handle Results

async function runCode(hydrogen) {
  const kernel = hydrogen.getActiveKernel();

  // Simple execution
  const result = await kernel.execute("x = 42\nprint(x)");

  if (result.status === "ok") {
    console.log("Outputs:", result.outputs);
  } else {
    console.error(`${result.error.ename}: ${result.error.evalue}`);
  }

  // Monitor state
  const disposable = kernel.onDidChangeExecutionState((state) => {
    console.log("Kernel state:", state);
  });

  // Get completions
  const completions = await kernel.complete("import nu");
  console.log(completions.matches); // ['numpy', 'numbers', ...]

  // Cleanup
  disposable.dispose();
}

Shared Module API

hydrogen-next exposes a shared module that allows dependent packages (like jupyter-next) to reuse kernel management and rendering components.

Kernel Registration

External packages can register their kernels with hydrogen-next, making them visible to tools like Variable Explorer, Kernel Monitor, and Inspector:

const hydrogen = atom.packages.getLoadedPackage('hydrogen-next');
const shared = require(path.join(hydrogen.path, 'lib', 'shared'));

// Register a kernel
shared.registerKernel(kernel, filePath, editor, grammar);

// Set as current kernel (for Variable Explorer, etc.)
shared.setCurrentKernel(kernel);

// Unregister on shutdown
shared.unregisterKernel(kernel);

Available Exports

Category	Exports
Kernel	KernelManager, registerKernel, unregisterKernel, setCurrentKernel, getStore
Output Rendering	Display, Output, RichMedia, StreamText, KernelOutputError
Output Utilities	reduceOutputs, normalizeOutput, getBestMimeType, truncateOutput
ANSI Handling	AnsiText, escapeCarriageReturn
Timing	formatExecutionTime, createExecutionTimeTracker
Utilities	kernelSpecProvidesGrammar, tildify, sanitizeHtml

Contributing

Got ideas to make this package better, found a bug, or want to help add new features? Just drop your thoughts on GitHub — any feedback's welcome!