I tidied up my Obsidian (again)
| 8 min read
I've reworked my Obsidian folder structure again, aiming to get it down to a more flat flat-file system. The two latest things I've done are:
- Changed how I save web content to my vault. I've got an n8n flow running on a Raspberry Pi that takes tagged content from my Readwise account and saves it to my vault. It also saves any highlights and notes that I make about content as a separate file, but linked via the Readwise ID (more on that below). Why not use Obsidian Web Clipper? Great question - I'll look into it.
- Made better use of Bases. I've read a few articles now on the major benefits of Bases - a part of Obsidian I hadn't really used much before. It gives me the ability to create lists of content that make more sense to me. For instance, instead of a Dataview query that tries to understand what day it is, and make sense of my complicated Daily Notes folder structure, instead I just create a Base that filters all my content excluding certain folders, and then just displays it in date order.
Now my folder structure looks like this:
_attachments: folder containing non-standard / non-markdown files that are referenced in other files. e.g. images that are pasted into documents._assets: this contains all the "stuff" that runs the vault, including atemplatessubfolder, and a folder to contain all my Bases. It also contains folders for each of the "required metadata parameters" that all notes going forwards will have. Again, I'll talk about this more in a second._webcontent: this folder stores all my content from Readwise. The name is saved as the readwise ID number for the content, and then as a metadata parameter the title is stored, so that I can surface the title in a Base view.inbox: I'm still keeping an inbox folder. This mostly now acts as a place where files generated from orphaned links appear, so that I can quickly find them as part of some light housekeeping/: the rest of my files live in the root of my Vault. This still feels icky to me, and I viscerally hate it. But, I do see the logic and the use of doing this, and so far I've stuck with it.
Why the change?
I've been using Obsidian for years, but it's only recently occurred to me to check out how the guy who runs the show sets up his vault.
After reading that, I fell into something of a rabbit hole of looking through other people's setups, and also came across this post about modular note taking:
Given that my Obsidian is always open, I'd always been a big fan of the Daily Note approach. Everything goes into one note for the day, and then I can separate out bits as needed. Except I never did the separating out bit, and searching for something specific in a note that then has to be tagged in about 30 different ways ended up making no sense.
Modular note taking cuts through all of that. But there's some process prep work involved.
Required metadata
The one consistent bit of advice from people advocating for everything in the root of the vault, is that you absolutely must have good metadata on all your files. So, I've got three parameters that every file now must have:
contenttype: I went back and forth on what this should be, but ultimately settled on "contenttype" as the primary way of filtering content. This gives a quick heads up on what the file is for: is it a blog, is notes from a meeting, is it a quicknote, or is it content that I've clipped from the web?hats: I struggled with the idea of losing the folder structure. It means that my work with one organisation is bundled alongside work I do with another, and also bundled alongside stuff that is just me as an individual human person. The "hats" property means that I can still isolate content based on the role. But as an added benefit, I can classify it against multiple bits of work if there's overlap - in other words, the literal reason for abandoning a rigid folder structure.tags: Essentially, what is this content about? My initial thought was that this would quickly become unwieldy. Not at an individual file level, but for example using an #ai tag would mean that specific tag would be so full as to be pointless. It's why I used to separate things out with nested tags. The thing is, part of the point of a tool like Obsidian is to find links between apparently disparate content. And, between thecontenttypeandhatsproperties, large tag groups can be split into more manageable sizes, whilst retaining the benefit of having everything at once.
Of course, most files will need metadata in addition to these three. But these are the ones that must be there to make my new system work. This however, posed a new problem...
Restructuring old data: enter VSCode
There was a fair bit of backward incompatibility with my old content. A lot of people online recommend wiping the slate and starting the new system with a new Vault. For a lot of others, using AI (Claude in particular) seemed like the way to go.
I didn't want to do either of those things. I've been using Obsidian solidly since 2023, and I didn't want to lose that data. Equally, I don't really like using AI when I know other genuine alternatives exist. For the most part all I needed to do, really, was add things to frontmatter and tidy up some tags. Basically, I wanted a find and replace. While Obsidian can do this, it's not great.
Luckily, one of the benefits of a Markdown based file system is it doesn't really care what you open it with. So I fired up VSCode, opened up my vault folder, and between a quick batch script and a couple of well-thought-out find/replaces, I'd updated somewhere around 1000 files into my new 3-property system in less than 15 minutes.
How has this changed how I use Obsidian?
Even since I last stripped back Obsidian, there's been quite a big change in how I work. All my notes are now created via the "unique note" feature, which has really cut down on the amount of faffing I do deciding whether the thought I'm having should be a note in its own right, whether it should be part of my Daily Notes (which I no longer use), or if it needs to be tacked onto an existing note that I'll need to find. Now I just Alt+Tab over to Obsidian, hit the "Create new unique note" button I've popped on the left hand side, and get to typing.
I also barely use the Files tree on the left anymore - something almost everyone who has a similar system says, but I didn't believe. Instead, I have a pinned Base that shows all the files in my root folder, sorted by Date Modified. The columns show the contenttype, title (which some but not all my notes have), modified time, and tags fields. Turns out, that's enough for me to find the exact note I'm looking for at a glance.
I also have a separate pinned Base that shows all of the files containing highlights or notes from any web content I've saved. It makes it really easy to find content that I want to reference.
The new system has allowed me to retire some more plugins. I now only have the Tasks and Git plugins - saying goodbye to Periodic Notes and Dataview. I really thought I'd miss Dataview, but Bases has come a really long way since it launched. While Dataview is certainly more powerful, for my needs Bases is fine.
And that's it! I'd make a self-deprecating joke at this point along the lines of "tune in next week to see how I've completely overhauled this system", but the truth is I've been running this for a few months now and it kind of just works. Who knows, maybe this will be the last post I ever write about how I write things.
Haha, good one.
Want to chat about anything in this post? I'd love to hear from you!
Feel free to drop me an email
Or find me on Mastodon