Part 2: Deploy the dashboard and schedule

The pipeline is deployed and writing to the playground lake. Now we deploy the marimo dashboard alongside it and set up a schedule.

Add the dashboard to the deployment

Import the dashboard module in __deployment__.py and add it to __all__:

from agent_traces_dashboard import app as agent_traces_dashboard

This registers the dashboard as an interactive job. The platform can run pipelines and interactive applications like marimo notebooks or Streamlit apps.

The dashboard also needs to point at the playground destination, not DuckDB.

Update the connection:

dlt.attach("agent_traces", destination="playground", dataset_name="agent_logs")

When deploying notebooks, you must pass destination and dataset_name explicitly to dlt.attach().

Deploy and run:

uv run dlthub deploy
uv run dlthub run

Run mode

Open the notebook in the platform UI. It runs in run mode, not edit mode, so all the code is hidden and you see only the reports and visuals. This is the view you share with your team.

The data lives in the playground destination. It could just as well be MotherDuck, BigQuery, Snowflake, or a vector database like LanceDB. dlt writes to all of them with the same pipeline code.

Share it

Publish the dashboard to get a public URL:

uv run dlthub job publish agent_traces_dashboard

Or share it within the workspace via the platform's Users and Roles.

Scheduling

To keep the data fresh, schedule the pipeline to run on a cron trigger.

Add it to the decorator in __deployment__.py:

from dlt.hub.run import trigger

@run.pipeline("agent_traces", trigger=trigger.schedule("0 12 * * *"))
def ingest_agent_logs(): ...

Confirm the schedule:

uv run dlthub job list

You can also create followup chains: run the ingestion pipeline, and on success, run the dashboard to refresh the report. The platform supports job.success triggers that chain jobs together.

You can manage jobs from the platform UI too - start runs, cancel runs, and manage schedules.

← Deploy to the cloud | Where to go from here →