Wrapped up the first iteration of a lightweight blog engine inside this Franklin.jl site. Leaving notes here so I remember how it’s wired together when I inevitably forget six months from now.
All posts live in blog/ as plain Markdown files. Franklin turns each filename into a /blog/<slug>/ URL. The slug is just the filename minus the .md. Example: 2025-11-04-my-first-post.md → /blog/2025-11-04-my-first-post/. Straightforward and easy to manage with grep or shell scripts.
Each post carries minimal metadata so the macros know what to do:
@def title = "First Post"
@def published = Date(2025, 11, 4)
@def tags = ["blog", "personal"]published drives the timeline and defaults to the ISO date if missing. tags are for filtering on the index page.
The index page (menu3.md) calls {{blog_index}}. Behind that:
compute_blog_posts() walks the blog/ folder and grabs metadata via pagevar.
It caches the results as a vector of named tuples, sorted by date.
hfun_blog_index renders article cards (title, date, snippet, tags).
A short JS snippet toggles cards by tag.
No databases, no build chain—just Franklin and a few helper functions.
I wrote a small {{blog_nav}} macro, that expands into < Previous | All posts | Next > for quick post-to-post navigation.
The macro looks up the current post’s identifiers (slug, path, URL) in the cached list, so it keeps working even if Franklin adjusts URLs or folders move.
Tag buttons now show usage counts, remember the active view in the URL hash, and update their pressed state for screen readers.
Each card displays an estimated reading time and a clearer “Read more” call-to-action.
The post footer navigation is now a three-card layout with directional hints instead of the old pipe-separated text links.
Keep the setup minimal. Markdown in, static HTML out. If the blog ever grows, consider adding pagination or RSS, but not yet. The current setup is small, predictable, and completely under control.