JupyterHub-Pilot

One JupyterHub for many teams. Each team gets its own worker VM, and every notebook has an AI assistant.

• A worker VM per team. The Hub starts each user's notebook server over SSH on their team's worker, so one heavy team cannot slow down another.
• A Linux account per user, with cgroup memory and CPU limits on every notebook server.
• AI magics in every notebook: %do, %fix, %review, %rework (Ollama, OpenAI, Claude or Gemini through LiteLLM, or an MCP server).
• Live monitoring at /hub/monitoring: each worker's CPU, memory, disk and network, and each user's usage against their limits.
• Google login, SQLite state that survives Hub restarts, optional Vault secrets, and a built-in load test.

Install

Ubuntu 24.04 or 26.04: one Hub VM plus one worker VM per team, on the same private network.

Hub

sudo apt update && sudo apt install -y python3-pip
sudo pip3 install --break-system-packages --ignore-installed "jupyterhub-pilot[hub]"
sudo jupyterhub-pilot install-hub

It asks for the Hub's private IP, your admin username and the first worker's IP, and prints an SSH key and an agent token for the workers.

Each worker

sudo apt update && sudo apt install -y python3-pip
sudo pip3 install --break-system-packages --ignore-installed "jupyterhub-pilot[worker,ai]"
sudo jupyterhub-pilot install-worker

Then, on the Hub

sudo jupyterhub-pilot add-worker team_beta <worker-private-ip>   # one per extra team
sudo nano /etc/jupyterhub-pilot/hub.env                          # Google client, or test passwords
sudo systemctl enable --now jupyterhub

Put Caddy in front for HTTPS, create a group per team in Admin → Groups, and users can start their servers.

In a notebook

%load_ext jupyterhub_pilot.extension
%do load data/sales.csv and plot monthly revenue

%do writes the code into the next cell for you to check and run.

Learn more

• Full guide: HTTPS, test login, teams, upgrading, troubleshooting
• AWS EC2 walkthrough
• AI magics · Monitoring · Security
• Releases · Issues

MIT licensed.