Share serverless compute and environment in Git Folders

The Git Folder Serverless feature allows you to use a single serverless compute resource and a unified Python environment across all notebooks and files within the same Git folder. Dependencies are controlled by a pyproject.toml file at the project root.

Use cases

Imagine you are developing a project with multiple notebooks and Python scripts that rely on the same libraries. With this feature, you avoid setting up the environment notebook by notebook.

• Share the same serverless compute across multiple repository files.
• Manage dependencies centrally via pyproject.toml.
• Run web terminal commands on the same shared compute.

How to use

Open the Git folder editor in the workspace, select Git Folder Serverless from the compute dropdown menu, and add dependencies using notebook commands.

%uv add cowsay

What happens in this code:

1. The command uses the uv package manager to add the dependency to the root pyproject.toml file.

%uv sync

What happens in this code:

1. The command applies the updated dependencies to the shared environment and updates the lock file.

Limitations and requirements

• This feature is in Beta.
• Your project must be in a Git folder and opened in the Git folder editor.
• The environment version in pyproject.toml must be set to 5 or later.
• Only assets inside the associated Git folder can attach to the compute resource.
• Compute sharing is scoped to a serverless usage policy; folders using different policies get separate compute resources.