On This Page
Auto-Tag in Spaces
Learn from how you've tagged files in a folder, then suggest and apply tags to untagged files in a Sortio Space.
What Auto-Tag Does
As of Sortio v8.11.0, every Space has an Auto-tag panel. Auto-tag profiles the macOS Finder tags you've already applied inside a folder and uses that profile to suggest tags for any files that aren't tagged yet.
Instead of forcing you to predefine a taxonomy, Auto-tag learns from the work you've already done. If half your "2024 Taxes" folder is tagged Receipt, Invoice, or Statement, Sortio will suggest those same tags for the untagged files in the folder, with a per-file confidence score.
1. Build
Scan a folder you've already tagged and build a tag corpus.
2. Suggest
Get per-file tag suggestions with confidence scores.
3. Apply
Approve the chips you like, then write the tags to disk.
ℹ️ macOS only
Finder tags are a macOS feature, so the Auto-tag panel only appears on Mac builds of Sortio. On Windows and Linux you'll see Spaces without this panel.
The Build / Suggest / Apply Workflow
Open any Space, scroll to the Auto-tag panel, and walk through these three steps. The whole flow stays in the panel: no separate window, no setup wizard.
Step-by-step
1
Pick a "Learn from" folder
By default this is the Space's own folder. You can point it at any other folder where you've tagged files. This is the source of the tag profile, not the destination.
2
Click "Build tag corpus"
Sortio scans the source folder, reads each file's Finder tags, and profiles which tags appear on which kinds of files. The result is a tag corpus stored locally with the Space.
3
Click "Suggest tags for untagged files"
Sortio runs the corpus against the Space and produces a per-file list of suggested tags. Each suggestion shows a confidence score and is rendered as an approvable chip.
4
Approve or unapprove chips
Click a chip to toggle its approved state. You stay in control of every tag that ends up on every file. Sortio never writes tags to disk during the suggestion step.
5
Click "Apply selected tags"
Sortio writes the approved tags to the files using the macOS Finder tag API. From here on you can act on those tags from Finder-tag-aware Automations, the macOS sidebar, Spotlight, or any tool that reads Finder tags.
AI Mode vs. Deterministic Heuristic
The Auto-tag panel has a Use AI toggle, on by default. It controls how Sortio decides which tags to suggest for each file.
Use AI: on (default)
Sortio sends each candidate file's profile and the tag corpus to an AI model and asks for the most likely tags. Quality scales with the model you've selected.
Costs 1 credit per AI suggestion call.
Use AI: off
Sortio uses a deterministic heuristic over filenames, paths, and the corpus. It's free, fast, and entirely local.
No credits used. Works offline.
"Only untagged files" toggle
On by default. With this on, Sortio skips files that already have any Finder tag, so it never overwrites your existing work. Turn it off if you want suggestions for every file in the Space, including ones that already have tags.
A reasonable starting pattern: leave both toggles at their defaults (AI on, untagged only). If you're working through a very large folder and want to keep credits low, run the heuristic first to handle the obvious cases, then run AI on whatever remains.
Works with TagFiler / Hazel-Style Workflows
Auto-tag was designed to slot in next to existing tag-driven workflows on macOS. If you already use Brett Terpstra's TagFiler or rules in Hazel to route files based on Finder tags, Auto-tag handles the part those tools don't: getting tags onto files in the first place.
A typical chain
1. You drop a batch of new documents into a Space's folder.
2. Sortio's Auto-tag panel suggests tags for the untagged files based on what you've already tagged.
3. You approve the chips that look right and click Apply selected tags.
4. TagFiler, Hazel, or Sortio's own Finder-tag Automations pick up the newly tagged files and route them.
The corpus model is intentional here: rather than asking you to maintain a tag taxonomy, Sortio extracts one from the folder you point it at. That keeps the friction of bringing new files into an established workflow close to zero.
Tips & Caveats
Start with a folder that's well-tagged
The corpus is only as good as the folder you learn from. If you're starting fresh, tag 20-50 representative files by hand first, then build the corpus from that folder.
Confidence scores aren't accuracy guarantees
Treat the per-chip confidence as a sort order, not a verdict. When in doubt, click into the file in Finder and verify before approving the tag.
Suggestions are non-destructive
Nothing is written to disk until you click Apply selected tags. You can rebuild the corpus and re-run suggestions as many times as you want without affecting any files.
Combine with Finder-tag Automations
Once tags are on the files, build deterministic rules in macOS Finder Tag Automations to route, color, or rename based on those tags. Auto-tag and Automations were designed to work together.
Credit usage
When Use AI is on, Sortio spends 1 credit per AI suggestion call. The deterministic heuristic is free. Both modes only run when you click Suggest, so you're never charged unintentionally.