Virtual Env

[stefanbringuier]

How does one setup a Python virtual environment? Is it done prior to pip install marimo or does a marimo notebook have a mechanism for creating and manging virtual environments?

[akshayka]

Hi @stefanbringuier, welcome! Thanks for opening a discussion,

Setting up a virtual environment is done prior to pip install marimo; marimo does not yet have a mechanism for managing virtual environments, but we may have one in the future (great idea!).

Instructions:

Create a virtual environment called "marimo-env" (you can name it whatever you like).

1. python -m venv marimo-env

Activate the environment (unix/macOS, let me know if you are running Windows):

2. source marimo-env/bin/activate

Install marimo

3. pip install marimo

When you're done working in the environment

4 deactivate

To learn more:

• Check out this excellent guide on scientific computing with Python: https://learn.scientific-python.org/development/tutorials/dev-environment/
• Read the official Python tutorial: https://docs.python.org/3/tutorial/venv.html
• Or, ask me more questions :)

[akshayka]

I wonder if you have multiple marimo installations, and it's picking up the wrong one.

Can you type which marimo and paste the output here?

EDIT: The output of which marimo should be /path/to/my_project/bin/marimo. If it's not, there's a path issue, in which case I would try

deactivate
pip uninstall marimo
source my_project/bin/activate
marimo edit

In particular, if you explicitly use marimo from your environment, that should (hopefully!) work:

source my_project/bin/activate
my_project/bin/marimo edit 

[stefanbringuier]

I think you are correct and I do have marimo installed at the system level.

$ which marimo
/usr/local/bin/marimo

It doesn't look like there is a my_project/bin/marimo binary that I can add to my path via the my_project/bin/activate script. Any thoughts on a workaround?

[akshayka]

Hmm. I'd uninstall marimo from system level (if you haven't already), then try again from scratch.

$ deactivate
$ pip uninstall marimo
$ which marimo # sanity check, should be empty
$ rm -rf my_project/ # destroy the environment
$ python -m venv my_project # create it again from scratch
$ source my_project/bin/activate
$ pip install marimo
$ which marimo # sanity check, shouldn't be empty
$ pip install numpy
$ pip install matplotlib
$ marimo edit

Sorry, Python environment management can be a bit cumbersome. A good argument for marimo to abstract it away ...!

[stefanbringuier]

Did the trick.

I was hoping to have a "vanilla" marimo setup at the system level for quick stuff, but it's probably safer and cleaner otherwise.

Yeah if marimo can abstract and handle it, that would be awesome. I would probably never touch Jupyter notebooks again.

[akshayka]

Awesome, glad to hear it!

Yeah if marimo can abstract and handle it, that would be awesome. I would probably never touch Jupyter notebooks again.

Out of curiosity, how do you deal with virtual environments for Jupyter today? Through VS Code? Or do you just install your packages system wide

[s-celles]

Hi,

Daytona seems an interesting approach to consider for building development environment.

Daytona.io is an open-source Development Environment Manager designed to simplify and streamline the process of setting up development environments[1][3]. Key features of Daytona include:

1. Single-command setup: Developers can create a full dev environment using just one command: 'daytona create'[1].

2. Flexibility: It works on any machine and with any infrastructure, supporting both local and remote environments[1][3].

3. Automation: Daytona automates the creation of development environments, connects to local instances of VS Code (or browser-based alternatives), creates VPN tunnels, and adds reverse proxies for sharing[1].

4. Standardization: It helps eliminate configuration drift and ensures consistent environments throughout the development lifecycle[3].

5. Scalability: Daytona can scale with development requirements, removing constraints of local machines[3].

6. Security: It offers fine-grained code access controls and comprehensive dependency management[3].

7. Integration: Daytona supports various Git providers (GitHub, GitLab, Bitbucket) and popular IDEs (Visual Studio Code, JetBrains, Vim)[3].

Daytona.io has gained significant traction since its launch, amassing over 6,700 GitHub stars[2]. The company has also secured $7 million in funding to advance its mission of simplifying development environments[2][4].

Daytona (can) rely on Docker as provider.

https://www.youtube.com/watch?v=uL-TaEhvVwk provides an introduction to this tool

https://www.youtube.com/watch?v=DmLj1fTQebw&t=30s provides an example with Jupyter Notebooks in Seconds - No Setup, Remote Access, GPU-Ready... It could probably be Marimo.

I wonder if some of you have ever experiment it and have an (unbiased) opinion about it.

Citations:
[1] https://news.ycombinator.com/item?id=39616709
[2] https://www.youtube.com/watch?v=RNf_vfgc91c
[3] https://www.daytona.io
[4] https://www.linkedin.com/company/daytonaio
[5] https://github.com/daytonaio/daytona/actions
[6] https://twitter.com/daytonaio
[7] https://dribbble.com/SLRNCL/projects/7008631-Daytona-io
[8] https://x.com/daytonaio/highlights

Best regards