| layout | page |
|---|---|
| title | AI in Jupyter |
| tagline | Learn about AI-related work from within the Jupyter project |
| permalink | /ai |
Artificial Intelligence (AI) and Large Language Models (LLMs) play an increasingly important role in the industry, and within the Jupyter community. This ecosystem is also relatively young and evolving rapidly. The goal of this page is to point you in the right direction to learn more about various initiatives within the Jupyter ecosystem related to AI.
Jupyter AI is an extension that connects AI agents to computational notebooks in JupyterLab. It is currently incubating as a part of the Jupyter-Frontends subproject.
jupyter-ai-contrib is a community organization hosting extensions, integrations, and experiments that compose or build on Jupyter AI. It is a combination of earlier-stage projects that undergo more rapid development and prototyping, and community-maintained projects.
jupyterlite/ai provides AI code completions and chat for JupyterLab, Notebook 7 and JupyterLite. It is maintained within the JupyterLite organization.
For asynchronous chat, the JupyterLab channel in Zulip is your best bet. Synchronous conversations and demos happen during the weekly Jupyter AI meeting (see calendar) and community workshops. Because the Jupyter AI project is incubating within the Jupyter Frontends subproject, this channel is the best place to have developer-related conversations about AI.
Both projects integrate AI features into JupyterLab and Notebook v7+ interfaces. jupyterlite/ai does not depend on a server component which makes it possible to use it with JupyterLite, the in-browser Jupyter distribution. jupyter-ai enables agentic workflows that can continue to run on the server even when the browser is closed or disconnected. Server dependency distinction aside, the set of features overlaps, with both packages implementing agentic workflows and tool calling. Notably, jupyterlite/ai also implements AI code completion, and jupyter-ai implements multi-agent workflow.
Not yet - Jupyter is primarily organized into semi-independent subprojects, each of which has their own practices and policies for contributing. There is no Jupyter-wide policy regarding AI-assisted contributions. Many sub-projects are currently in the process of defining the policies that work for their needs and practices.
No, Jupyter has a number of initiatives related to AI as part of subprojects (linked above), but there is no one AI subproject that encompasses everything. We're encouraging subprojects to experiment and develop their own AI-related work, and to integrate and connect with other pieces of the ecosystem.