5. Ship an app

Build a marimo dashboard notebook and share it with your team

The agent answers questions and sends a weekly report. Some people want to explore the numbers themselves. In this last part you build a marimo dashboard — a reactive Python notebook — and share it with your workspace.

Before you start

• You completed 4. Build an agent.
• Apps appears on your navigation rail. Apps is enabled per account; if it is missing, contact [e-mail].
• The data for the dashboard is a CSV file in Files.

A running app reads the workspace files you select on its Sources tab, copied into the app at startup, and credentials from its vault. A notebook has no builder chat, so it does not connect knowledge bases. See Build apps for everything a custom app can use.

1. Create the notebook

1. Open Notebooks

   Click Apps on the left navigation rail, then the Notebooks card.

2. Fill in the dialog

   Click New notebook and fill in:

   Field	Value
   Title	A display name.
   Name	Lowercase letters, digits, dashes or underscores. It names the notebook file and cannot be changed later.
   Description	Optional. What the dashboard shows.
   Sandbox environment	Keep Workspace default. This field appears only when Sandbox Environments is enabled.

3. Create

   Click Create. The notebook starts and opens in the marimo editor.

2. Give the notebook its data

1. In the notebook view, click App details, then open the Sources tab.
2. Click the edit (pencil) icon, select the CSV file, and save the selection.
3. Click Rebuild sources, or stop and reopen the notebook, so the running notebook gets a local copy.

The notebook reads a read-only copy of each selected file under the directory in the DATAGEN_SOURCES_DIR environment variable, at the file’s full storage key. A file on the Workspace tab of Files has a key that starts with .raw/.

3. Build the dashboard

Click Open app to return to the notebook, and add the cells below. marimo reruns dependent cells automatically, so the dashboard updates as soon as someone changes a control.

Load the data

import os

import marimo as mo
import pandas as pd

SOURCES = os.environ["DATAGEN_SOURCES_DIR"]
df = pd.read_csv(f"{SOURCES}/.raw/<folder>/<file>.csv")

Filter

segment = mo.ui.dropdown(
    options=sorted(df["<segment_column>"].dropna().unique().tolist()),
    label="Segment",
)
segment

Key numbers

view = df if segment.value is None else df[df["<segment_column>"] == segment.value]

mo.hstack(
    [
        mo.stat(value=f"{len(view):,}", label="Rows"),
        mo.stat(value=f"{view['<metric_column>'].sum():,.0f}", label="<metric name>"),
    ]
)

Detail table

mo.ui.table(view)

If the notebook needs a Python package that is not installed, add it to a Sandbox Environment with a built notebook image and choose that environment for the notebook.

4. Share it

The notebook is already shared: every notebook lives in the workspace’s shared notebooks folder, and workspace members open it from Apps > Notebooks with Open.

• Each member who opens the notebook gets their own sandbox. The notebook’s Instances tab lists who has one running.
• Workspace admins can limit who sees the notebook on its Access tab, when that tab is available in your workspace.
• There is no public link. Only signed-in workspace members can open the notebook.

What you built

Across the five parts you built a complete internal AI workflow on DataGen AI Cloud:

Part	Building block	Result
1	Data	A connection and a published analytic model
2	Files	A shared folder of reference documents
3	Context	A versioned knowledge base built from the documents
4	Agents	An agent that answers questions and posts a weekly report to Slack
5	Apps	A marimo dashboard shared with the workspace

Learn more in Apps overview, Build apps and Publish and share.

Previous: Build an agent

Build the agent and schedule its weekly report.

Next: Apps overview

Notebooks and custom apps, and when to use each.