I'm deploying a single Quarto dashboard, at the repository root. It uses
the knitr engine and server: shiny. It is R only--There is no Python in the
project and no Python dependency.
THE PROBLEM
Every build classifies the content as quarto-static, and my manifest.json is
overruled. The build log states this directly:
"The publish request indicated this is quarto-static content, but the
manifest.json file indicated it is quarto-shiny. Connect Cloud will
process it as quarto-static."
The build then fails at the final step:
"Failed to publish content: A Python requirements.txt file is required
in your project."
Both failures follow from the content type. When I previously added an empty
requirements.txt to satisfy that check, the build got further and then failed
with a different error, again caused by the static path:
"ERROR: pres-dashboard.qmd is a knitr engine document that uses server: shiny
so cannot be included in a project with an output-dir (shiny document output
must be rendered alongside its source document)."
Is there a selection in the publish flow that I am missing which produces a
quarto-shiny item? The documentation lists a single "Quarto" content type.
Happy to grant access to the repository or re-run the build if that helps.
This behavior definitely sounds odd. We would need to look at the specifics of this project to figure out what is happening. Would you feel comfortable sharing your content id here (its the UUID in the url when you click on the Content)
I couldn't paste the full link because our organization name has a word with the same letters as a blocked word! But the front half is "housingforwardva" (instead of va, type out the full state name)
Yup, that's perfect, thank you! It looks like you have found a bug in our system. Actually, two bugs: the important one is that we have an unexpected limit on how large a single R code chunk can be in a quarto document. The second is that when we hit the first bug we show the wrong error message.
The error message about python and requirements.txt can be ignored. The real issue is the R code chunk size. I've created an issue so we can get it fixed. Fortunately, you should be able to get around it by modifying your quarto document.
If you refactor the large chunk starting with ACCESS FORM then things should work. Either split it into smaller pieces, or pull it out into a separate file (similar to what you do with access_log.R)
Let me know if this gets you unblocked and in the meantime we will try to fix our side so this wouldn't be necessary.
