Setup github action to generate preview automatically

[findmyway]

Ideally, we should use https://github.com/JuliaPluto/static-export-template in the future. (This is not related to the css issue though)

[kir0ul]

Was starting to look into this, and it seems automatically exporting the notebooks to HTML using ExportPluto.yaml looks pretty straightforward :slightly_smiling_face:

What I'm less sure about is what would happen to the notebooks on the blog, like https://juliareinforcementlearning.org/blog/notebooks_for_reinforcement_learning_an_introduction/Chapter01_Tic_Tac_Toe.jl/ (which look manually copied inside https://github.com/JuliaReinforcementLearning/JuliaReinforcementLearning.github.io/tree/master/blog/notebooks_for_reinforcement_learning_an_introduction)? :thinking:

Possible solutions (maybe there are others?):

• Automatically export the notebooks to HTML with the Pluto GH Action, which would look something like this: https://JuliaReinforcementLearning.github.io/ReinforcementLearningAnIntroduction.jl/notebooks/Chapter01_Tic_Tac_Toe.html, and include those HTML pages as an <iframe> in the blog.
• Include https://github.com/JuliaReinforcementLearning/ReinforcementLearningAnIntroduction.jl as a submodule in https://github.com/JuliaReinforcementLearning/JuliaReinforcementLearning.github.io/tree/master/blog/notebooks_for_reinforcement_learning_an_introduction with the notebooks already exported as HTML and maybe setup a CI job that fails if a notebook has changed but has not been exported again to HTML.

Any thoughts on this?

[findmyway]

Thanks for looking into it!

I prefer to simply set up a CI to push these generated htmls into the gh-pages branch in this repo. And update the link in that blog.

(We don't need to keep this repo updated to the master branch of RL.jl, cause generating all those htmls would consume a lot of resources I think. We just need to update the dependency once there are new minor releases 😃 )

[kir0ul]

This is the page I'm referring to: https://juliareinforcementlearning.org/blog/ where there are all the links to the notebooks.

So you mean instead of having a link on the blog that points to https://juliareinforcementlearning.org/blog/notebooks_for_reinforcement_learning_an_introduction/Chapter01_Tic_Tac_Toe.jl/, instead it would point to the gh-pages URL: https://JuliaReinforcementLearning.github.io/ReinforcementLearningAnIntroduction.jl/notebooks/Chapter01_Tic_Tac_Toe.html?

If I understand correctly, I believe there's no way for the gh-pages URL to redirect to this kind of domain: https://juliareinforcementlearning.org/blog/notebooks_for_reinforcement_learning_an_introduction/, it has to be a top level domain.

[findmyway]

So you mean instead of having a link on the blog that points to https://juliareinforcementlearning.org/blog/notebooks_for_reinforcement_learning_an_introduction/Chapter01_Tic_Tac_Toe.jl/, instead it would point to the gh-pages URL: https://JuliaReinforcementLearning.github.io/ReinforcementLearningAnIntroduction.jl/notebooks/Chapter01_Tic_Tac_Toe.html?

Exactly!

In theory, we could still upload generated html pages under https://juliareinforcementlearning.org/blog/notebooks_for_reinforcement_learning_an_introduction . All we need to do is to clone those generated html pages from the gh-pages branch after the following step:

https://github.com/JuliaReinforcementLearning/ReinforcementLearning.jl/blob/9185c8548197dd4a6ef0cd7c84c3531c491e6447/.github/workflows/ci.yml#L103

But I'm afraid that's not necessary.

[kir0ul]

So it's been published correctly here: https://juliareinforcementlearning.org/ReinforcementLearningAnIntroduction.jl/, but there's an error when running using ReinforcementLearning:

ArgumentError: Package ReinforcementLearning is required but does not seem to be installed

My guess is it's missing the following at the beginning of each notebook:

Pkg.add([
        Pkg.PackageSpec(name="ReinforcementLearning"),
    ])
    using ReinforcementLearning

[kir0ul]

• [x] Also don't forget to update the URLs in the README

[findmyway]

Nice work!

It seems I missed the install step:

https://juliareinforcementlearning.org/ReinforcementLearningAnIntroduction.jl/notebooks/Chapter01_Tic_Tac_Toe.html

I'll have a try later today.

[findmyway]

Should be fixed in https://github.com/JuliaReinforcementLearning/ReinforcementLearningAnIntroduction.jl/pull/72

🎉

[kir0ul]

Cool! 😎