Visualization of the PyHealth codebase

[ivanmilevtues]

I have generated high-level diagrams for the PyHealth project. The idea is to help people get up-to-speed with the codebase (as contributors), you can check it here:
https://github.com/CodeBoarding/GeneratedOnBoardings/blob/main/PyHealth/on_boarding.md

These diagrams are designed to help people quickly understand a codebase. We believe that, especially considerting that a lot of scientists interact with this codebase, they can benefit a lot from visualization like the one we have, instead of having to read the full codebase. Also considering the 400+ forks, there are a lot of people who edit the codebase, I suppose that usually they are interested in just one component, with the diagram they can immediatly navigate to it and have the context on how it interacts with the rest of the codebase!

Any feedback is more than welcome! We also just released a free GitHub Action that can automatically update the diagrams so they are always up-to-date.

Full disclosure: we're trying to turn this into a startup, but we're still in a very early stage and figuring out what will actually be useful for people.

[jhnwu3]

These are excellent points. There's honestly a whole bunch that are deprecated that we do need to fix in terms of documentation. That said there's really three major stacks to Pyhealth (not listing exact apis yet, mostly because I'm writing this comment on phone lol)

• Pyhealth models (afaik, it's just a bunch of models with different schemas linked by the tasks class)
• Pyhealth data (data -> dataset -> tasks)
• Pyhealth evaluation (metrics, uncertainty quantification, and a hidden Interpretability module)

Im going to hopefully get around to the tutorials this weekend where I can draft this diagram. It's just we've been battling technical debt for the past couple months.

[ivanmilevtues]

Thanks for the answer @jhnwu3!

Having up-to-date docs is a burdon that is why we are trying to automate this part. I'd love to hear your opinion on the generation we have for your repo. If you think it can help you I can setup a PR where the diagrams are added and we add our new FREE github action to keep them up-to-date with new PRs.

The diagram captures all 3 of the components you've listed. (Maybe you didn't see it as the github app doesn't render the mermaid diagrams, unlike a browser, which does)

Let me know when you have more time I'd love to hear what you have to say about the diagram and if we can do something to help you <3

[jhnwu3]

Ah I see the diagram now. Yeah, that is helpful! Not sure if you're able to join our Pyhealth discord. Would love to chat more on there 😂. Its a bit more mobile friendly. I'd love to collaborate and discuss how this might integrate with our current readthedocs setup this weekend.

I'm unfortunately at work right now. I need to update some of the basic tutorials that are outdated now since we do want to show programmatically how everything plays together.

[ivanmilevtues]

Thank you for the response @jhnwu3, I will join the discord tomorrow and engage in contact there. See how and if we can help and do something toghether!

See you there!

[ivanmilevtues]

Hey all, a quick update if someone is looking at the thread. We are now open-source: https://github.com/CodeBoarding/CodeBoarding.
Didn't want to post at discord as we are not yet integrated so thought it would be out-of-place :)

[jhnwu3]

Silly question, are you guys able to generate this in a .rst format? I think we're stuck with using the .readthedocs formatting. Not sure, if it's possible, our docs/ directory is all .rst.

Edit: also, sorry for the late response, I've been swamped at work and with other obligations. I'm finally hitting the tutorials and docs again.