The publishing pipeline for allandeutsch.com

🌱October 4, 2021.
Last tended February 7, 2024.
budding 🌿
1 minute read ⏱

Like any good PM, I opted to start from the ideal user experience. This was unusually easy since I am the only user; talk about putting the personal in personal computing!

My ideal workflow for publishing is that I can edit my notes in Obsidian on any device, and by adding a single line to the frontmatter they show up on my website automatically. Step by step, this looks something like:

  1. I edit my notes locally on any device of my choice using Obsidian
  2. By default notes are private
  3. I can add publish: true to any note’s frontmatter to opt-in to publishing the note
  4. Notes are automatically synced across devices and publishing automatically using git
  5. Git commits to my private notes repo triggers an automatic site build & deployment of the notes opted-in to publishing

Structuring the build pipeline

Because I want my notes to be private by default, they exist in their own private git repo. The SvelteKit code for this site is private, though I hope to open-source it one day. That leaves the site deployment, which is hosted using my GitHub account’s GitHub Pages repo and pushed to automatically by my GitHub Actions workflow.

Here’s an overview of what happens in the CI workflow:

  1. Commits to either the private notes or private site code repos trigger a build
  2. The build action pulls the private notes and processes them for publishing:
    1. A python script deletes all notes not opted-in for publishing
    2. The remaining notes are run through obsidian-export to convert them from Obsidian markdown to Commonmark
  3. The site’s SvelteKit code is pulled from another repo
  4. The processed notes are dumped into a content folder
  5. The SvelteKit site is built using the static adapter
  6. The completed build is pushed to my GitHub Pages repo

This entire process takes around 2 minutes from start to finish. I’ve noticed that sometimes GitHub takes another minute or two to distribute the latest build across the Pages CDN.

Initiating builds

My build pipeline is fully automated, but that’s not the only piece. I generally don’t need to manually author commits, either. While I generally avoid committing broken code, notes don’t really have a broken state that would fail to compile. A half-written note is in a valid state for backing up to git, so notes are a great candidate for automatic git commits.

I achieve automated commits in 2 ways, depending on the device. On a computer, I use the obisidian-git plugin. It’s easy to set up and works great. I have it configured to automatically backup my notes to git every 30 minutes if there are any changes, and it provides a command to manually trigger a backup too.

On mobile & tablet devices, though, it’s not so great. iOS and Android lack a native git client for the plugin to use, so it ships with a JS implementation of git that is very slow. I didn’t like this approach, so opted for a different workflow.

On mobile & tablet, I use a combination of Working Copy as a git client and Apple Shortcuts to operate it. I have 2 shortcuts:

  1. A button shortcut that replaces my Obsidian app icon. When tapped, it:
    1. Pulls the git repo containing my notes, then
    2. Launches obsidian once the pull is completed
  2. An automated shortcut that pushes all changes to my notes repo when I leave the Obsidian app

I’ve found this setup makes it extremely easy to get content published and shared with people on demand.

Specifying publishing details for notes

The last detail worth covering is how I specify metadata for published notes. TL;DR: I use frontmatter.

All the metadata on this and other notes is specified in the frontmatter. Slugs make links readable, so I specify my slugs in the frontmatter. For this note that frontmatter is slug: publishing-this-site. Really simple! I do similar for the created date, last modified date, title, description, etc. I use mdsvex to process my markdown into svelte components, and it makes the frontmatter metadata available so I can easily use it when rendering the page.

Hopefully this overview inspires you to build your own publishing pipeline and put the personal back in personal computing. If you have any questions/comments/feedback, please reach out to me on Twitter!