Workspaces

graal offers two workspaces: Jupyter and VS Code. They share the job scheduling mechanism: a workspace is a pod started in the execution namespace, under the same security constraints and the same project quota.

Jupyter

The notebook a data science team asks for first. The image ships the libraries needed day to day: data manipulation, object storage access, a Trino client to query the warehouse, and an MLflow client to track experiments.

VS Code

code-server, for the data engineer who edits code rather than cells. Same base image, same storage, same identity.

A single door

Both workspaces are served behind a single gateway. The practical consequence for your network team: one entry point to open, and one authentication — the console’s. No password per tool, no second directory.

The project’s data, without copying a credential

A workspace opens inside a project: it sees that project’s tables, buckets and secrets, with the rights of the person who opens it. A notebook that must run on its own becomes a job, schedulable and monitored like the others (see jobs and workflows).

Visualisation stays yours

graal does not replace your dashboarding tool: it plugs into it. Trino exposes a standard JDBC access that your BI tool queries directly.