Apps from your local Nebi environments

You can deploy a dashboard or web app from the hub’s home page (the Deploy App button) and have it run in a local Nebi environment you built yourself in JupyterLab.

This page is for clusters where your administrator has set this up. If the Software Environment list in the Deploy App form shows environments with names like team/project instead of folder paths, your cluster uses a shared Nebi server and this page does not apply.

1. Build the environment in JupyterLab

Open a terminal in JupyterLab and make a folder for the project. The folder is the environment:

mkdir ~/sales-dashboard && cd ~/sales-dashboard
pixi init
pixi add panel bokeh-root-cmd
nebi init

• pixi init and pixi add create pixi.toml and pixi.lock, the two files that describe the environment.
• nebi init registers the folder with Nebi, which is how the Deploy App form finds it.
• Check that it is there with nebi workspace ls.

The app runs entirely in this environment, so it must contain everything the app imports and the framework you deploy it with. For a Panel app that is panel and bokeh-root-cmd; the hub starts Panel apps through bokeh-root-cmd, and an environment without it cannot start one.

Environments you create in the Nebi tile on the JupyterLab launcher work the same way.

2. Put the app’s code in your home

Save the app’s script anywhere in your home, for example ~/sales-dashboard/app.py:

import panel as pn

pn.extension()

region = pn.widgets.RadioButtonGroup(options=["North", "South", "West"])
pn.Column("# Sales dashboard", region, pn.bind(lambda r: f"Showing **{r}**", region)).servable()

3. Deploy the app

1. On the hub’s home page, click Deploy App and give the app a name.
2. Choose the Framework, for example Panel.
3. Open Software Environment and pick your environment. It is listed by its folder, for example /home/jovyan/sales-dashboard.
4. In File path, enter the path to the script: /home/jovyan/sales-dashboard/app.py.
5. Click Next, pick a server size, and deploy.

The first start takes a few minutes. The app gets its own copy of the environment, built from your pixi.toml and pixi.lock, before it starts. Your JupyterLab can be running or stopped; the app does not need it.

How the environment list stays up to date

The list shows the environments you have in your main JupyterLab:

• While your JupyterLab is running, the list is read from it each time you open the form, so a new environment appears straight away.
• While your JupyterLab is stopped, the list is the one saved when it last stopped.

So create, change and remove environments in your main JupyterLab. A change made anywhere else (an app, or a JupyterLab you deployed as an app) shows up once your main JupyterLab has run again.

Changing an environment later

Add or remove packages in JupyterLab as usual (pixi add ..., pixi remove ... in the environment’s folder). A running app keeps the copy it started with. Restart the app to rebuild its copy from the new pixi.toml and pixi.lock.

Do not delete or move an environment’s folder while an app uses it: the app needs those two files every time it starts.

When something is not right

What you see	What to do
The Software Environment list is empty	Start your JupyterLab, wait for it to open, then open the Deploy App form again. You need to have run JupyterLab at least once, and to have created an environment there.
A new environment is missing from the list	Check nebi workspace ls in a JupyterLab terminal. Only environments that are fully set up are listed, and only those inside your home or created in the Nebi tile.
The app starts but fails on a missing package or framework	Add the package to the environment in JupyterLab, then restart the app.

For administrators: setting this up is covered in Nebi integration.