DEV Community: Tim Apple

folderblog 0.2.11: posts ordered by date and time

Tim Apple — Tue, 10 Nov 2026 18:27:39 +0000

folderblog 0.2.11 puts posts in order by date and time, so several posts on the same day finally show up in the right order. The home page, archives, tag pages, feeds and the podcast list are all newest first by it.

How it works

Every post date now carries a time of day. To set the time yourself, write it in date:

date: 2026-11-10 09:28

A full timestamp with an offset works too, like date: 2026-11-10T09:28:00-06:00.

A post with only a day (a 2026-11-10- filename or date: 2026-11-10) gets the time it was first committed to git, or else the file’s modified time. That time is saved in dates.toml the first time the post is published, so editing it later never reshuffles the day. The day, and so every address, stays the same.

You can also set a timezone in blog.toml:

timezone = "America/Chicago"

Dates written without an offset are read in that zone. Without it, folderblog uses the computer’s own zone, as before.

A couple of other details:

Friendlier date typos. An unreadable date is now a warning (and the filename’s date is used) when the filename has one. folderblog check also warns about an unreadable updated.

Read this before you update

timezone needs 0.2.11 or newer. Older versions refuse a blog.toml that has it, so update every machine that builds the blog first.
Expect a one-time change to dates.toml. The first build adds a time for each post dated by filename or a date-only date:. Commit it with your blog.
Feeds now show real times instead of midnight, and a few same-day posts may swap places in the feed.

Grab it

System	Command
Arch / Omarchy	`sudo pacman -U folderblog-bin-0.2.11-1-x86_64.pkg.tar.zst`
Debian / Ubuntu	`sudo apt install ./folderblog_0.2.11-1_amd64.deb`
Fedora / RHEL / openSUSE	`sudo dnf install ./folderblog-0.2.11-1.x86_64.rpm`
autonomix-cli	`autonomix-cli add https://github.com/ninepointlabs/folderblog`

Already installed? Install the new package, then run systemctl --user restart folderblog. The packages and a plain tarball are on the v0.2.11 release, and the source is on GitHub.

Originally posted on blog.ninepointlabs.com.

Dogear 0.1.3: EPUB books, marked up like a PDF

Tim Apple — Mon, 09 Nov 2026 01:31:11 +0000

Dogear is the reader for studying, made for Omarchy. Version 0.1.3 opens EPUB books, so it now reads, highlights and remembers your PDFs and EPUBs.

EPUB books

Open an .epub the same way as a PDF: from the Open dialog, the command line, by dropping it on the window, or from the library. The first time, Dogear sets the book in pages (6 × 9 in, chapters starting on a new page), which takes a second or two for a long book, and keeps those pages. After that everything works as it does for a PDF: highlights in five colors, notes, bookmarks, answer lines, search, the library and the Markdown export. The Contents panel shows the book’s own table of contents.

Your marks belong to the book

The library lists the EPUB, and the EPUB is what Dogear copies into Library/. The laid-out pages live in your library folder, so they sync with your marks and another computer shows the same pages. They’re never laid out again, so your marks stay on the words you marked. Another copy of the same book (same identifier, such as its ISBN) is offered your marks, just as with PDFs.

Good to know

Books get one plain, readable style. Bold, italics, headings, lists, tables and images are kept; the publisher’s styling, links inside the book and most SVG drawings aren’t.
DRM-protected EPUBs can’t be opened.
There’s no choice of page size or font yet.
Fix: the Arch package now depends on qt6-svg. Without it, Dogear’s icon (and SVG images in books) couldn’t be shown. Most desktops already have it, so few noticed.

Get it

autonomix-cli add https://github.com/ninepointlabs/dogear
autonomix-cli install dogear
# or, if you already track it:
autonomix-cli update

On Arch or Omarchy you can also grab dogear-0.1.3-1-any.pkg.tar.zst and run sudo pacman -U on it; on any other Linux, the AppImage runs as is. Both are on the v0.1.3 release, and the source is on GitHub. Questions go to help@ninepointlabs.com.

Originally posted on blog.ninepointlabs.com.

omaforge 0.2.9: star addons to find them later

Tim Apple — Thu, 05 Nov 2026 02:21:50 +0000

omaforge is the World of Warcraft addon manager for Linux and Omarchy. Version 0.2.9 lets you star addons, so you can bookmark one you haven’t installed and come back to it later.

What’s new

Star any addon. Click the ☆ on an addon in Explore, Search or Installed, or in its details view. Stars are saved between sessions, are shared by all your game versions, and stay when you remove the addon. An addon with no known source can’t be starred.

A new Starred tab. It sits between Search and Backups and lists everything you’ve starred, with a filter, a sort (Recently starred or Name) and a Not installed toggle. Install adds the addon to the game version selected in the sidebar, and if an addon has no build for that version you get the usual “no stable release” message.

Shortcut change. Ctrl+4 now opens Starred, and Backups moved to Ctrl+5.

Get it

Arch / Omarchy: sudo pacman -U omaforge-0.2.9-1-any.pkg.tar.zst, then omaforge setup omarchy
Debian 13+ / Ubuntu 26.04+: sudo apt install ./omaforge_0.2.9-1_all.deb
Fedora: sudo dnf install ./omaforge-0.2.9-1.noarch.rpm
Any other Linux: the AppImage, chmod +x omaforge-0.2.9-x86_64.AppImage && ./omaforge-0.2.9-x86_64.AppImage. Add --install-desktop-entry to put it in your app menu.
With autonomix-cli: autonomix-cli add https://github.com/ninepointlabs/omaforge

The packages are on the v0.2.9 release, and the source is on GitHub.

Originally posted on blog.ninepointlabs.com.

Persistwise: writing that outlasts the platform

Tim Apple — Fri, 30 Oct 2026 18:32:31 +0000

I have a bad habit with blogs. I start one somewhere that feels right, write for a while, then the platform changes, the export is a mess, and the old posts sit there like a burned CD nobody can play. Geocities. LiveJournal. A host that quietly folded. Same movie, different sequel.

So I built Persistwise, and today I’d like to invite you to try it.

persistwise.com · Start a free trial

What it does

Persistwise is not another place to write. You keep blogging wherever you already do: HEY World, Substack, your own site, a newsletter, even a social feed. You give Persistwise the feed, and it does the quiet filing in the background.

It checks your feeds every 15 minutes. When it finds a new post, it keeps three copies:

A Wayback Machine snapshot. Persistwise asks the Internet Archive to capture the post. That snapshot lives on archive.org, it’s public, and anyone can open it without Persistwise.
An IPFS pin. The saved copy is pinned through Pinata onto the public IPFS network, with a content ID that points at that exact copy of your post.
A saved copy we keep. The post’s HTML is stored on our own storage, so there is something to fall back on when an outside service is slow or down. Only you can open it.

Your Archive page shows one line per post, with a link to each copy and a plain status for each one. If the Wayback Machine is busy, Persistwise waits and tries again later, so you don’t have to babysit anything.

RSS, Atom and JSON Feed all work. The first check takes in every post your feed lists at that moment, then new posts get picked up as they appear.

Why I wanted it

Most of us don’t lose writing because we’re careless. We lose it because the web is a rental. The landlord changes the locks, the feed dies, the pretty URL turns into a 404. The Wayback Machine is wonderful, but it only saves what someone asks it to save. I wanted a boring, stubborn librarian whose only job is to ask every time.

I also didn’t want to build another editor. I already have places I like to write, and so do you. The whole point is that you don’t have to move.

Keep: 30 days free, then $40 a year

There is one plan, called Keep. It’s 30 days free, then $40 a year, and you don’t need a card to start. Create an account, press “Start free trial,” and add your feed.

About three days before the trial ends, we’ll email you if you haven’t added a card. If you decide not to keep going, Keep pauses and nothing is deleted. Your Archive page stays readable, and your Wayback snapshots and IPFS pins stay where they are. The Wayback snapshots are kept by the Internet Archive, not by us, so they don’t depend on Persistwise at all. We don’t delete people for leaving.

A yearly price felt right for something meant to last. Longevity and a monthly habit you cancel in February don’t get along very well.

Honest notes

It’s early, and I’d rather tell you the edges up front. Each post is captured once, when Persistwise first sees it, so later edits aren’t saved. The copies are the page’s HTML only, without images or stylesheets, so they may look plainer than the original. There’s no export button yet. And Wayback and IPFS copies are public, so please only add feeds for writing you’re happy to keep public.

Give it a try

If you have a feed you care about, I’d be grateful if you’d let Persistwise keep it. Go to persistwise.com, read how it works if you’re curious, and start your free trial. If anything doesn’t work the way you expect, please write to help@ninepointlabs.com. I’d really like to hear from you.

Thanks for reading. Now go write the next one, and let the librarian do the filing.

Sunsetting Rune, and why I'm writing in WordCraft now

Tim Apple — Thu, 29 Oct 2026 18:56:19 +0000

Earlier this month I started building Rune, a word processor for Linux and Omarchy with AI built right into the writing surface. Today I’m archiving it, and I’d like to tell you why.

Why I built Rune

Word processors on Linux felt like a choice between heavyweight ports and plain text editors pretending to be word processors. And every one of them treated AI as a sidebar plugin. I wanted the opposite: a fast, native app where the model is part of the page, where you can ask it to fix a paragraph and the fix lands right in your document as one undo step.

Rune got further than I expected. It went from an idea to version 0.8.7 in about two weeks, with packages for Arch, Debian and Fedora, an AppImage for most other distros, headers and footers, tables, and sign-in for ChatGPT and OpenRouter.

What I learned

Quite a lot, honestly. Rune started on LibreOfficeKit and ended up as a single Qt process that reads and writes ODF itself, because the first design had a performance ceiling I couldn’t get past. Along the way I learned to test real files instead of trusting documentation. That habit caught a bug in Qt’s own ODF writer, where a table right after a list silently disappeared when LibreOffice opened the file. By the end, Rune was checked by 504 automated tests. I’ll carry that way of working into everything I build next.

Then I found WordCraft

Recently I came across WordCraft, a fast, open-source word processor written in Rust by the ArtCraft team. I tried it, and it’s really good.

It does what Rune set out to do, and more. It has a ribbon and styles, works with .docx natively along with .odt, Markdown and more, exports PDF, and has track changes and comments, offline spelling and grammar, footnotes, a table of contents, and citations in APA, MLA, Chicago and IEEE. It runs on Linux as a Flatpak, plus macOS, Windows, FreeBSD and the web.

The part that really got me is how it treats AI. Its commands are available through a command-line tool, a JSON control channel and an MCP server, so AI assistants can work inside your documents. That’s the idea I was chasing with Rune, done in a more open way: you bring whichever assistant you like.

It’s still in development, but it’s already well past where Rune is.

Why archive instead of compete

I could keep going. But I’m one person, word processors are a lot of work to maintain, and the world doesn’t need two small open-source editors chasing the same idea when one of them is already further along. I’d rather spend my time on things only Nine Point Labs can build, and send you to the better tool.

So Rune is archived. The code stays up under the MIT license, along with the existing releases, if you’d like to learn from it or fork it. The README now points to WordCraft.

Give WordCraft a try

If you write on Linux, or anywhere really, please give WordCraft a try. The source is on GitHub if you’d like to dig in or help out.

Thank you to everyone who tried Rune or sent a kind word about it. It meant a lot.

Originally published on The Word from Nine Point Labs.

StoryTracker takes its own screenshots now

Tim Apple — Wed, 28 Oct 2026 15:29:07 +0000

StoryTracker is the World of Warcraft addon that keeps a record of your character’s adventures: quests, zones, levels, deaths, bosses, loot and more. It comes in two versions, one for WoW Forever and one for Retail. Yesterday’s releases (Forever 1.0.4 and 1.0.5, Retail 1.1.3 to 1.1.5) add automatic screenshots at the moments that make a good story, and make the addon easier to install with an addon manager.

Screenshots at story moments

StoryTracker now takes a screenshot about three quarters of a second after:

a quest turn-in
a level up
an achievement (on Forever, only if the client fires the achievement event)
a boss kill
your first visit to a zone
other story moments, such as entering a dungeon or battleground, a battleground result, rare or elite kills, a new mount, epic or better loot, duels, joining a guild, a reputation standing going up, and deaths

There’s also an optional timed shot every 15 minutes, taken only when you’re out of combat, not on a loading screen and not AFK.

It takes at most one shot every 20 seconds. When two moments land together, like a turn-in that also dings, the shot is labelled with the bigger one. If you turn on “hide UI”, the interface is hidden for the shot while you’re out of combat, and it’s never touched in combat.

Each shot is logged in the addon’s saved data with its time, the file name WoW gives it, what happened, the zone and subzone, and your character’s name, realm and level. Log entries older than 30 days are dropped. Addons can’t delete files, so StoryTracker never removes screenshots itself; clean up the game’s Screenshots folder outside the game.

Turning it on and off

Screenshots are on by default, and everything is under /storytracker shots:

/storytracker shots on|off turns automatic screenshots on or off.
/storytracker shots <event> on|off controls each kind: quest, levelup, achievement, boss, newzone and story.
/storytracker shots timer on|off|<minutes> sets the timed shot (1 to 120 minutes, 15 by default).
/storytracker shots hideui on|off hides the interface for out-of-combat shots (off by default).
/storytracker shots log shows the last few logged shots.

Easier installs with an addon manager

Each release is now built with the BigWigs packager and comes with a small release.json file that says which game it’s for. Addon managers such as OmaForge use it to install the right StoryTracker for each client, so the Forever build goes to Forever and the Retail build goes to Retail. Installed copies also link back to their GitHub page now. Retail 1.1.5 updates the addon for the 12.1.5 client so it isn’t flagged as out of date.

Why it matters

StoryTracker’s notes are what turn a play session into a post on My Azeroth Life. Words were only half of it. Now each story moment comes with a picture taken right when it happened, and the log says exactly which shot goes with which moment, so a post can show the boss kill or the first look at a new zone instead of a stock image.

Privacy

Nothing changes here. StoryTracker has no network access, since WoW addons can’t use the internet. Screenshots are saved by WoW to its own Screenshots folder on your computer, and the log only records when and where each one was taken.

Download the latest Forever release or Retail release, or install it with OmaForge. The code is on GitHub for Forever and Retail.

Originally posted on blog.ninepointlabs.com.

Omostrich 0.3.3: tighter signing grants and safer profile pictures

Tim Apple — Tue, 27 Oct 2026 14:39:21 +0000

Omostrich keeps your Nostr key on your own computer and signs for your apps from the Omarchy bar. Two releases landed this morning. Version 0.3.2 finishes the profile picture hardening that 0.3.1 started, and version 0.3.3 makes your signing permissions narrower and your approval prompts clearer, especially around direct messages.

What changed in 0.3.3

"Always allow" now covers one kind of event, not everything. Before, remembering a signing request for an app meant that app could sign any kind of Nostr event without asking again. Now the button names the kind it covers, for example "Always allow sign kind 4 (DM)", and Authorized apps lists each app's grants kind by kind.

Encrypt and decrypt prompts show who's on the other end. Each request shows the other person as a shortened npub (marked "(you)" when it's yourself) and whether you're encrypting to them or decrypting from them. Encrypt requests also show a short preview of the plain text. A kind 4 signing request shows its recipient, and DM ciphertext is no longer previewed. None of this goes into the log.

The signer keeps running when notifications can't. If notify-send is missing or fails, Omostrich logs it once, with an error code only, and carries on instead of crashing.

Smaller fixes:

Signing requests for kind 14 and kind 1059 (private messages and gift wraps) show a warning and can only be approved once, never remembered.
Your public key is handed out without a prompt only to an app that connected this session or that you've remembered. Anyone else gets asked.
Gift-wrap notifications look back two days, and unlocking no longer sets off a burst of old ones.

What changed in 0.3.2

0.3.1 put limits on how big your cached profile picture could be, but the file could still carry compressed metadata that expands while the image is being read. A small, hostile picture could push the panel to several gigabytes of memory. Now the daemon rewrites the downloaded picture down to just the parts needed to draw it, with no re-encoding and no new dependency:

PNG keeps only the image data and the colour chunks it needs. An unknown critical chunk is refused.
JPEG drops comments, EXIF, ICC, XMP and Photoshop data.
GIF keeps the first frame only.
WebP keeps the image data and drops everything else. Animated WebP is refused, so the placeholder shows.

Anything after the end of the image is dropped too, and a picture cached by 0.3.1 is fetched again on startup, so the cache only ever holds the stripped copy. This came out of the Omarchy plugin marketplace review (#12494).

Why it matters

Your signer should do exactly what you agreed to and nothing more. A blanket "always allow" was easy to click and hard to reason about later, and a DM prompt that doesn't say who it's for asks you to approve blind. Per-kind grants and clearer prompts mean you can see what each app is allowed to do and who your messages are going to.

Upgrading

Old blanket signing grants meant "sign any kind", and there's no way to tell which kind you had in mind, so they're removed the first time 0.3.3 starts. The app and its other grants stay, daemon.log names each affected app, and the app asks once more for each kind it needs.

To update an existing install:

cd ~/Projects/omostrich
git pull
npm ci
./install-daemon.sh
./install-plugin.sh
systemctl --user restart omostrich

The restart matters, because the running signer keeps the old code until it's restarted. The full notes are on the 0.3.2 and 0.3.3 release pages, the code is on GitHub, and more about Omostrich is at omostrich.com.

Originally posted on blog.ninepointlabs.com.

omaforge 0.2.8: fixes from a full QA pass

Tim Apple — Mon, 26 Oct 2026 22:21:03 +0000

omaforge is the World of Warcraft addon manager for Linux and Omarchy. Version 0.2.8 is a fixes release from a full QA pass, coming the same day as 0.2.7.

What’s fixed

Safer uninstall. omaforge uninstall now asks before removing an addon it can’t trace to a source, such as one you wrote yourself. Without a terminal it refuses unless you pass --yes.

Damaged files no longer crash omaforge or lose data. A broken config.toml or flavors.toml gives a one-line message naming the file and the problem, and omaforge never writes over it. In the app, a red banner names the file and offers Open file, Rescan, or setting it aside (it’s kept as <name>.broken-<time>). A damaged state.json is moved to state.json.damaged-<time> and a fresh one is started, with a warning in both the CLI and the app.

Fits half a screen. The window now goes down to 640 px wide. Below 1000 px the sidebar narrows, the tabs move under the title, and less-used actions (Import/Export, and each addon’s channel, Pin, Ignore and Remove) move into a ⋯ menu. Alt+↑ and Alt+↓ switch between game versions.

Clearer messages. An unknown addon id, an AddOns folder omaforge can’t write to, an unknown root or provider, and an empty GitHub search now say what went wrong in plain words. CurseForge versions no longer show up as 1.0.1.zip or offer an update that isn’t there, check lists each addon it couldn’t check and why, and Wago says plainly that it isn’t supported yet.

Get it

Arch / Omarchy: sudo pacman -U omaforge-0.2.8-1-any.pkg.tar.zst, then omaforge setup omarchy
Debian 13+ / Ubuntu 26.04+: sudo apt install ./omaforge_0.2.8-1_all.deb
Fedora: sudo dnf install ./omaforge-0.2.8-1.noarch.rpm
Any other Linux: the AppImage, chmod +x omaforge-0.2.8-x86_64.AppImage && ./omaforge-0.2.8-x86_64.AppImage. Add --install-desktop-entry to put it in your app menu.
With autonomix-cli: autonomix-cli add https://github.com/ninepointlabs/omaforge

The packages are on the v0.2.8 release, and the source is on GitHub.

Originally posted on blog.ninepointlabs.com.

omaforge 0.2.7: ready for World of Warcraft: Forever

Tim Apple — Mon, 26 Oct 2026 16:31:25 +0000

omaforge is the World of Warcraft addon manager for Linux and Omarchy. Version 0.2.7 gets it ready for World of Warcraft: Forever, which launches on November 4.

What’s new

omaforge knows the Forever release client. Battle.net installs Forever in its own _forever_ folder (product wowf). omaforge now lists that install as Forever, even before you’ve started the game for the first time and there’s no Interface/AddOns folder yet. If Battle.net hasn’t filled in a version, omaforge shows not installed yet instead of guessing. The first addon you install creates Interface/AddOns for you.

Pick it from the command line with -c forever. The beta client is still -c classic_beta.

Leftover beta folders get a proper name. If you played the beta, you may still have a _classic_beta_ folder that Battle.net no longer lists. It now shows up as Forever Beta rather than an unknown game.

Why it matters

You can have your addons in place before launch day instead of waiting for the game to create its folders. Install Forever through Battle.net, open omaforge, and set it up the same way you already do for Retail.

Get it

Arch / Omarchy: sudo pacman -U omaforge-0.2.7-1-any.pkg.tar.zst, then omaforge setup omarchy
Debian 13+ / Ubuntu 26.04+: sudo apt install ./omaforge_0.2.7-1_all.deb
Fedora: sudo dnf install ./omaforge-0.2.7-1.noarch.rpm
Any other Linux: the AppImage, chmod +x omaforge-0.2.7-x86_64.AppImage && ./omaforge-0.2.7-x86_64.AppImage. Add --install-desktop-entry to put it in your app menu.
With autonomix-cli: autonomix-cli add https://github.com/ninepointlabs/omaforge

The packages are on the v0.2.7 release, and the source is on GitHub.

Originally posted on blog.ninepointlabs.com.

folderblog 0.2.10: podcast download stats with OP3

Tim Apple — Mon, 26 Oct 2026 15:29:14 +0000

folderblog 0.2.10 adds one thing podcasters have been asking about: download numbers. If you publish a show with folderblog, you can now count how often your episodes are downloaded with OP3, the free, open-source podcast stats service.

How it works

Add one line under [podcast] in blog.toml:

analytics_prefix = "https://op3.dev/e/"

Every episode in your feeds then goes through OP3 in the form OP3 documents, including your show’s guid so OP3 can attribute downloads sooner. There’s no account to create. After a few days, your public stats page is at https://op3.dev/show/<your podcast guid>.

A few details:

It’s off by default. Existing feeds don’t change unless you add the line.
Only the feed changes. Uploads, the deploy’s size check and the player on your episode pages keep using your own address. That also means plays on your episode pages aren’t counted, only downloads from podcast apps.
folderblog podcast tells you whether stats are on.
Other prefix services work too, as long as the prefix starts with https:// and ends in /.

The My Azeroth Life podcast is already running on it.

Read this before you update

analytics_prefix needs 0.2.10 or newer. Older versions refuse a blog.toml that has it, so update every machine that builds the blog first.
If you used the old, undocumented op3 = true: it now means the same thing, and your episode addresses gain ,pg=<guid>. Podcast apps keep track of episodes by guid, so nothing gets downloaded twice.

Grab it

System	Command
Arch / Omarchy	`sudo pacman -U folderblog-bin-0.2.10-1-x86_64.pkg.tar.zst`
Debian / Ubuntu	`sudo apt install ./folderblog_0.2.10-1_amd64.deb`
Fedora / RHEL / openSUSE	`sudo dnf install ./folderblog-0.2.10-1.x86_64.rpm`
autonomix-cli	`autonomix-cli add https://github.com/ninepointlabs/folderblog`

Already installed? Install the new package, then run systemctl --user restart folderblog. The packages and a plain tarball are on the v0.2.10 release, and the source is on GitHub.

Originally posted on blog.ninepointlabs.com.

folderblog 0.2.9: Fixes from a full QA pass, and podcast addresses that stay put

Tim Apple — Sun, 25 Oct 2026 01:29:24 +0000

folderblog 0.2.9 is a fix release that came out of a full QA pass over this morning’s 0.2.8. That pass followed the published instructions step by step, the way a new user would, and the rough edges it found are fixed here. Most of it is podcast polish, plus a change that keeps your post addresses the same when you clone or copy a blog.

Read this before you update

Three changes are worth knowing about first:

The podcast feed’s own title changes. If your show has the same name as your blog, the feed itself is now titled “Your Blog (Podcast)”, not just the link to it. Podcast apps will pick up the new name. To keep the old one, give the show its own name in podcast/index.md or with [podcast] title.
There’s a new file to commit: dates.toml. A post without a date of its own gets its address from the day folderblog first saw it. That date used to live only in the ignored .blog/ folder, so a fresh clone could give the post a different address. It’s now kept in dates.toml in the blog folder. After you update, the service writes it from the dates already recorded, so no address changes. If your blog is in git, commit dates.toml.
A new [podcast.storage] prefix setting needs 0.2.9 or newer. It pins the folder your episode files use on the audio host, so renaming the blog folder no longer changes every episode’s address. If you leave it out, nothing changes. podcast setup now writes it, and folderblog check suggests the exact line for existing shows. Older versions refuse a blog.toml that has it, so update every machine that builds the blog first.

Podcasts

Publishing an episode keeps your notes. folderblog podcast publish --episode (and the MCP publish_episode tool) now changes only the draft: line. Before, it rewrote the whole front matter and dropped the comments from the template.
check looks at draft episodes too. Problems like a missing guid show up as warnings, and the suggested guid is the same every time you run it. folderblog podcast now counts episodes left out because of a problem, and a draft no longer needs audio hosting set up.
No more doubled page title. The default podcast page no longer reads “Blog · Blog”.
S3-compatible storage is documented. With type = "s3" and no key, folderblog now tells you how to give it one (FOLDERBLOG_STORAGE_KEY_ID and FOLDERBLOG_STORAGE_SECRET, or a key file) instead of pointing you at the R2-only podcast setup.

Clearer messages everywhere else

Honest retry messages. A held publish only says it’ll be retried when the background service is actually running. Otherwise it tells you what to run.
Machines without a systemd user session. folderblog status says so plainly, and a failed install-service removes the unit file it wrote (or puts the old one back) and tells you what to do instead.
check names pictures nothing shows. Pictures in posts/ and pages/ that no page uses are still published, and now you’ll know which ones they are. It also notes that a robots.txt under a base_url path isn’t read by search engines.
Docs that don’t break. Install links are pinned to the release, building from source uses --tag, and there’s new coverage of template route headers, base_url for other hosts, machines without a systemd user session, the preview address and folderblog data examples.

Grab it

System	Command
Arch / Omarchy	`sudo pacman -U folderblog-bin-0.2.9-1-x86_64.pkg.tar.zst`
Debian / Ubuntu	`sudo apt install ./folderblog_0.2.9-1_amd64.deb`
Fedora / RHEL / openSUSE	`sudo dnf install ./folderblog-0.2.9-1.x86_64.rpm`
autonomix-cli	`autonomix-cli add https://github.com/ninepointlabs/folderblog`

Already installed? Install the new package, then run systemctl --user restart folderblog, and commit the new dates.toml if your blog is in git. The packages and a plain tarball are on the v0.2.9 release, and the source is on GitHub.

Originally posted on blog.ninepointlabs.com.

folderblog 0.2.8: Podcast setup that works on real R2, and a Subscribe box

Tim Apple — Sat, 24 Oct 2026 15:27:45 +0000

folderblog 0.2.8 is a small fix release on top of yesterday’s 0.2.7, the one that turned a blog folder into a podcast. If you tried folderblog podcast setup and it stopped at the “find it again” step, this is the release you want.

Podcast setup works with real Cloudflare R2

After setup uploads a small test file, it checks that the file can be found again. On real Cloudflare R2 that check failed even though the upload had worked. R2 compresses some files on the way out, and when it does, its reply leaves out the file size, so folderblog read the missing size as a missing file.

folderblog now asks R2 for uncompressed replies, so the size comes back. If a reply still has no size, any successful answer counts as “the file is there”, and the size is only compared when R2 sends one. The same fix applies when a deploy checks whether an episode’s audio is already uploaded, so a working upload isn’t treated as missing. The tests now use a fake R2 that compresses the way the real one does, which is how this slipped through before.

This is the fix that got the first real show on folderblog running. My Azeroth Life was set up with it, and the story of how that show is made is here on the blog.

Easier to subscribe

The podcast feed has its own title. When the show and the blog share a name, the podcast feed is now called “Your Blog (Podcast)”, so feed readers and browsers can tell it apart from the blog feed.
A Subscribe box on the show page. The podcast page that folderblog podcast on sets up now has a Subscribe section near the top, with the full feed address ready to select and paste into a podcast app. It’s plain CSS that uses your theme’s colors, with no JavaScript.

Grab it

System	Command
Arch / Omarchy	`sudo pacman -U folderblog-bin-0.2.8-1-x86_64.pkg.tar.zst`
Debian / Ubuntu	`sudo apt install ./folderblog_0.2.8-1_amd64.deb`
Fedora / RHEL / openSUSE	`sudo dnf install ./folderblog-0.2.8-1.x86_64.rpm`
autonomix-cli	`autonomix-cli add https://github.com/ninepointlabs/folderblog`

Already installed? Install the new package, then run systemctl --user restart folderblog. The packages and a plain tarball are on the v0.2.8 release, and the source is on GitHub.

Originally posted on blog.ninepointlabs.com.