The setup package bundles helpers that prepare a private GitHub project to run smoothly inside Google Colab. These routines remove repetitive manual steps around authentication, repository synchronization, dependency installation, and notebook execution.

• setup/__init__.py re-exports the public helper functions for concise imports, for example from setup import clone_or_update_repo.
• setup/project.py contains the implementations described in this document.

• store_credentials_from_env: Reads GitHub credentials from environment variables (defaults GITHUB_USERNAME, GITHUB_TOKEN), writes them to ~/.netrc, and applies strict file permissions so authenticated git operations can proceed non-interactively.
• ensure_project_name: Retrieves the project name from the PROJECT_NAME environment variable or prompts for it once, then caches the result back into the environment for consistency across helpers.
• clone_or_update_repo: Clones https://github.com/<owner>/<project>.git when absent or fast-forwards an existing checkout using git pull --ff-only, returning the resolved project path.
• change_directory: Switches the current working directory to the provided path and returns the resolved location, useful right before running project-specific commands.
• setup_google_drive_project: Mounts Google Drive through google.colab.drive, ensures a project directory under MyDrive/projects, manages a /project symlink, and exports the resulting location via OUT_DIR.
• setup_vscode_tunnel: Installs vscode-colab on demand, authenticates, and starts a VS Code tunnel named after the project so you can connect from the desktop client.
• install_project_requirements: Detects the installed PyTorch/CUDA versions, builds the matching PyTorch wheel index URL, upgrades pip, and installs dependencies from requrment.txt, returning the wheel index used.
• execute_notebook: Executes a notebook (default eval.ipynb) headlessly via nbconvert, applies a configurable timeout, and writes the executed notebook to eval_output.ipynb (or a provided path).

Together these utilities let you authenticate, mount storage, sync code, configure dependencies, and run evaluation notebooks with minimal manual intervention in Colab.