jupyter-vibe-coding

A JupyterLab 4.x extension that adds AI-powered buttons to notebook cells

• 💡 Explain – when a cell throws an error, click Explain to get a plain-language description of what went wrong, powered by an LLM via OpenAI's API.
• 🔧 Fix – click Fix to automatically replace the erroneous cell code with an LLM-corrected version.
• 🪄 Generate Code – a command (jupyter-vibe-coding:generate-code) is contributed to the Cell toolbar via jupyter.lab.toolbars; it opens a prompt dialog and replaces the current code cell with generated code.

Installation

pip install jupyter-vibe-coding

This installs both the Python server extension and the pre-built JupyterLab frontend extension.

You also need an OpenAI API key exported as an environment variable:

export OPENAI_API_KEY=sk-...

Alternatively, you can use environment variables to configure a different OpenAI-API compatible LLM server, including local installations using ollama:

# Preferred API key variable (falls back to OPENAI_API_KEY)
export JUPYTER_VIBE_CODING_API_KEY=sk-key...

# Optional custom OpenAI-compatible endpoint
export JUPYTER_VIBE_CODING_BASE_URL=http://localhost:11434/v1

# Optional model override (default: gpt-4o-mini)
export JUPYTER_VIBE_CODING_MODEL=gpt-oss:20b

Development install

The development install works on Windows, macOS, and Linux without requiring Administrator/Developer Mode privileges.

git clone https://github.com/haesleinhuepf/jupyter-vibe-coding.git
cd jupyter-vibe-coding

# 1. Install JupyterLab and build tools
pip install jupyterlab

# 2. Install JS dependencies
jlpm install

# 3. Build the frontend extension
jlpm run build

# 4. Install the Python package in editable mode
#    This also registers the server extension automatically.
pip install -e ".[test]"

After these steps, start JupyterLab normally:

jupyter lab

Verify both extensions are detected:

jupyter server extension list
jupyter labextension list

You should see jupyter_vibe_coding in the server list and jupyter-vibe-coding in the labextension list.

To pick up TypeScript source changes during development, rebuild and restart:

jlpm run build
jupyter lab

Usage

1. Start JupyterLab: jupyter lab
2. Open or create a Python notebook.
3. Run a cell that contains an error – 💡 Explain and 🔧 Fix buttons appear below the output.
4. Click 🪄 Generate Code in the Cell toolbar to open the prompt dialog.

Architecture

jupyter-vibe-coding/
├── src/
│   ├── index.ts      # JupyterLab plugin (frontend)
│   └── handler.ts    # Helper for calling the Jupyter server REST API
├── style/
│   └── index.css     # Button styles
├── jupyter_vibe_coding/
│   ├── __init__.py   # Server extension registration
│   ├── handlers.py   # Tornado handlers + LLM helper functions
│   └── tests/
│       └── test_handlers.py
├── package.json
├── tsconfig.json
└── pyproject.toml

API endpoints

Running the tests

pip install -e ".[test]"
pytest jupyter_vibe_coding/tests/ -v

License

BSD 3-Clause – see LICENSE.