notebookflow-app 0.1.0

Visual pipeline editor and local execution engine for JupyterLab

NotebookFlow for JupyterLab

NotebookFlow adds a visual pipeline editor and local execution engine to JupyterLab.

Install

pip install notebookflow-app
jupyter lab

Open a notebook and launch NotebookFlow from the command palette. The Python package contains both the prebuilt JupyterLab extension and the engine; a Node installation and a repository checkout are not required. The engine starts automatically through jupyter-server-proxy when the extension first connects.

NotebookFlow supports Python 3.11–3.13 and JupyterLab 4.6 or newer.

The PyPI distribution is named notebookflow-app; the notebookflow command and Python import package keep their shorter names.

Engine

The engine is a FastAPI and WebSocket server that builds DAGs from notebooks, runs them through a data bus, and exposes pipelines to platform adapters (JupyterLab, VS Code, and the web app).

Layout

notebookflow/
├── core/          # DAG, executor, data bus, triggers
├── nodes/         # Built-in node implementations
├── protocol/      # Extension manifest, registry, loader
├── llm/           # Pipeline author, code synthesis, explainer
└── server.py      # FastAPI + WebSocket entry point

Development

cd engine
uv sync --all-extras
uv run pytest
uv run ruff check .
uv run ty check

Build the prebuilt extension from the repository root:

pnpm --filter @notebookflow/jupyterlab-extension build:lab
uv build engine

The complete Trusted Publishing, upgrade, and rollback procedure is in docs/releases/jupyterlab.md.

Run the engine directly

uv run notebookflow
# Or directly:
uv run uvicorn notebookflow.server:app --reload