Remote Jupyter kernel default for new DataSpell notebooks

Direct answer

Based on current DataSpell behavior, the dependable answer is no: there is no reliable project-level switch that forces every newly created .ipynb to use one remote Jupyter server kernel. The supported behavior is per-notebook kernel and server selection. If your Managed Servers or Jupyter settings expose a default-server option, new notebooks created afterward may start on that remote server, but this is not a guaranteed project default and may apply IDE-wide rather than project-scoped.

Likely explanation vs. confirmed behavior

Likely explanation: DataSpell treats the project interpreter and the Jupyter server as separate layers. The project SDK supplies local kernels, while a remote Jupyter server supplies its own kernel list. Setting a remote Python interpreter does not by itself make the notebook kernel remote. This separation explains why a project interpreter setting does not propagate into notebook metadata as a remote kernelspec.

Confirmed behavior: Existing notebooks are not rewritten when defaults change. Their kernelspec metadata remains in the .ipynb file, so a notebook that referenced a local kernel will continue to do so until you reselect the remote kernel and save.

Steps for this case

1. Open Settings/Preferences | Tools | Jupyter (or Managed Servers, depending on build) and add the remote server via SSH tunnel or Jupyter server URL.
2. If the dialog offers a default-server option, enable it. If not, treat remote use as a manual per-notebook selection.
3. Create a new notebook and inspect the kernel selector. Save the file and open it to check metadata.kernelspec.
4. For existing notebooks, reselect the remote kernel from the notebook toolbar and save; the change is stored in the file.

Limitations and verification

• Connection interruption: The notebook usually keeps the server assignment, but execution fails until the server or SSH tunnel is reachable again. Server-side keep-alive and idle timeout determine whether the kernel process survives.
• Project switching: Configured servers may be IDE-wide while interpreters are project-scoped. A new project can inherit server definitions but not another project's kernel choice. Verify this per project.
• Version sensitivity: JetBrains moves Jupyter settings between Tools | Jupyter and language/framework pages. Behavior is build-sensitive.

Missing diagnostic detail

To confirm whether the default-server toggle exists in your environment, what is the exact DataSpell version and build shown in Help | About? This changes whether step 2 is available.