Skip to content

colab-assist

Google Colab workflow utilities

Test PyPI Documentation Issues

CI workflow status Release workflow status Test coverage

About

  • colab-assist is a small package that shares the utility functions that I find useful for my development workflows on Google Colab.

  • Actually, this is also a semi-mock project that I use to learn Python open-source development. Feedbacks, pointers, and feature suggestions are much appreciated!

Usage

Experimenting your private Python package on Colab

  1. Develop your package any way you like and push it to your private GitHub repo.

  2. Make a repo-specific personal access token (PAT).

  3. Store the PAT as a Colab Secret:

    Colab Secrets demo

  4. On Colab: 

    import colab_assist as A

    • Install → experiment → push → resintall: 

      # Install your private package
A.install("$my_token@me/my_pkg@feat/foo")

# Experiment
from my_pkg import foo
foo()

# Update the repo

# Reinstall updated package
A.install("$my_token@me/my_pkg@feat/foo")

# Reimport updated functions/classes without needing to restart Colab session
foo = A.reload(foo)
foo()

    • Or clone → experiment → push → pull: 

      # Clone your private package and automatically add it to `sys.path`
A.clone("$my_token@me/my_pkg@feat/foo", x="p")

# Experiment
from my_pkg import foo
foo()

# Update the repo

# Pull the update
A.pull("my_pkg")

# Reimport updated functions/classes without needing to restart Colab session
foo = A.reload(foo)
foo()

# Or restart the Colab session with `sys.path` automatically recovered
A.restart()

import colab_assist as A  # Import `colab_assist` recovers previously extended `sys.path`
from my_pkg import foo  # The clone is now again importable
foo()

# Terminate the Colab runtime with your clones automatically cleaned up
A.end()

Text file creation and editing

# Create `foo.txt` at working directory and call `google.colab.files.view()` to edit it
A.edit("foo.txt", x="c")

Dependencies & Installation

  • Although currently colab-assist lists no dependencies, it is intended to only be installed and used in a Google Colab environment. The reason not to explicitly list dependencies for now is that at least one depedency (google-colab) is bespoke for Colab and not hosted on PyPI. However, colab-assist is designed to install and run just fine on a fresh Colab instance.

  • You can install colab-assist very quickly with pre-installed uv on Colab: 

    !uv pip install --system -q colab-assist
    Or with pip: 
    %pip install -q colab-assist

License

Acknowledgements