besides using git-annex to manage images etc content I do not think there are limiting factors forbidding generating previews for PRs, in particular -- they could potentially limit only to the content necessary!
ATM annex info say
size of annexed files in working tree: 135.93 megabytes
so we might potentially start hitting limits on github gh-pages (what are they?)
Optimization
- we could just
get only the talks/slides touched by PR
- we could potentially fetch only the content needed for those particular slides!
We used to have a 'tracking' functionality in datalad to monitor for open files and auto 'annex get' it so then we could have -m http.server and get those . That is gone. But I think it might be feasible to make it work with datalad-fuse if we add mode to simply annex get all files instead of trying to proxy via fsspec etc. Should be the easiest! then we would publish result after going through all the slides in the talk
besides using git-annex to manage images etc content I do not think there are limiting factors forbidding generating previews for PRs, in particular -- they could potentially limit only to the content necessary!
ATM annex info say
so we might potentially start hitting limits on github gh-pages (what are they?)
Optimization
getonly the talks/slides touched by PRWe used to have a 'tracking' functionality in datalad to monitor for open files and auto 'annex get' it so then we could have
-m http.serverand get those . That is gone. But I think it might be feasible to make it work with datalad-fuse if we add mode to simplyannex getall files instead of trying to proxy via fsspec etc. Should be the easiest! then we would publish result after going through all the slides in the talk