Skip to main content
Auto Tagger provides several settings to customize how and when tags are suggested. All settings are available in Settings → Community Plugins → Auto Tagger.

​
Model Statistics

At the top of the settings page, you’ll see statistics about the current model (from src/main.ts:439-445): 
42 tags learned from 156 tagged notes (320 total). 1,847 unique words indexed.
This tells you:
  • Unique tags: Total number of distinct tags the plugin has learned
  • Tagged notes: Number of notes that contained at least one tag
  • Total documents: All markdown files in your vault
  • Unique words: Size of the vocabulary after tokenization and filtering
If this section doesn’t appear, the vault hasn’t been scanned yet. Wait a moment after enabling the plugin, or use the “Rescan vault” command.

​
Suggestion Behavior

​
Auto-suggest While Editing

Controls whether the plugin automatically suggests tags as you type.Default: true (enabled)Interface location: src/main.ts:447-457

​
Check Delay

What it controls: Debounce time in milliseconds after you stop typing before the plugin checks for suggestions. Default: 2000 (2 seconds) Range: Minimum 500ms Interface location: src/main.ts:459-477
When you type in a note, the plugin waits for this duration of inactivity before analyzing the content. This prevents excessive computation while you’re actively writing.
  • Lower values (500-1000ms): More responsive, but may impact performance
  • Higher values (3000-5000ms): Less intrusive, suggestions appear after longer pauses
Recommended values:
  • Fast typists: 2500-3000ms
  • Prefer immediate feedback: 1000-1500ms
  • Large vaults (1000+ notes): 3000ms+

​
Tag Placement

What it controls: Where accepted tags are inserted into your note. Default: first-line Options:
Value: first-lineInserts tags as inline tags at the beginning of the note content (after frontmatter if present).Behavior (from src/main.ts:298-326):
  • Skips frontmatter block if present
  • If first content line already has tags, appends to that line
  • Otherwise, creates a new tag line before first content
Example:
---
title: My Note
---

#python #data-science

Note content starts here...
Best for:
  • Visual tag organization
  • Quick scanning in preview mode
  • Consistency with manual inline tagging
Interface location: src/main.ts:479-499

​
Suggestion Filtering

​
Max Suggestions

What it controls: Maximum number of tags to suggest at once. Default: 5 Range: 1-10 (slider) Interface location: src/main.ts:501-513
When using the “Show all” modal (from the suggestion notice), up to double this number will be displayed (see src/main.ts:265).
How to choose:
  • 1-3 tags: Minimalist approach, only strongest matches
  • 5 tags: Balanced (default recommendation)
  • 7-10 tags: Exploratory, discover more tag relationships

​
Minimum Confidence

What it controls: Threshold for cosine similarity score. Tags must score at least this value to be suggested. Default: 0.01 Range: 0.005 to 0.1 (increments of 0.005) Interface location: src/main.ts:515-529
The confidence score represents the cosine similarity between your note’s content and the tag’s characteristic word profile (after co-occurrence boosting).
  • Lower threshold (0.005-0.02): More suggestions, including weaker matches
  • Higher threshold (0.05-0.1): Fewer suggestions, only strong matches
The “right” value depends on:
  • How consistently you tag notes
  • How specific vs. general your tags are
  • Vault size and tagging density
Recommended starting points:
  • Small vault (under 100 notes): 0.02-0.03
  • Medium vault (100-500 notes): 0.01 (default)
  • Large vault (500+ notes): 0.005-0.01
  • Highly consistent tagging: 0.02-0.05

​
Vault Management

​
Rescan Vault

What it does: Rebuilds the tag model by re-scanning all markdown files in your vault. Interface location: src/main.ts:531-547 When to use:
  • After bulk-tagging many notes
  • After importing notes with tags
  • After restructuring your tag system
  • When suggestions seem stale or inaccurate
Rescanning will clear the suggestion cache. You may see repeated suggestions for currently open notes.
What happens during rescan:
  1. All existing model data is cleared
  2. Every markdown file is read and processed
  3. New TF-IDF vectors are computed
  4. Co-occurrence matrix is rebuilt
  5. Suggestion cache is cleared
  6. A notice displays the new statistics
The process is the same as the initial scan at startup (batched, with UI thread yielding).

​
Commands

Auto Tagger registers two commands in the command palette:

​
Suggest Tags for Current Note

Command ID: suggest-tags Interface location: src/main.ts:96-104 Behavior:
  • Clears the suggestion cache for the current file
  • Immediately shows all suggestions in a modal
  • Works even if “Auto-suggest while editing” is disabled

​
Rescan Vault for Tag Patterns

Command ID: rescan-vault Interface location: src/main.ts:106-118 Behavior:
  • Same as clicking “Rescan now” button in settings
  • Shows progress notice during scan
  • Displays completion notice with statistics

​
Advanced: Settings Interface

Settings are stored in the AutoTaggerSettings interface (from src/main.ts:16-22): 
interface AutoTaggerSettings {
  debounceMs: number;           // Check delay
  tagLocation: "first-line" | "frontmatter" | "inline-end";
  minScore: number;             // Minimum confidence
  maxSuggestions: number;       // Max suggestions
  autoSuggest: boolean;         // Auto-suggest while editing
}
Default values are defined in src/main.ts:24-30: 
const DEFAULT_SETTINGS: AutoTaggerSettings = {
  debounceMs: 2000,
  tagLocation: "first-line",
  minScore: 0.01,
  maxSuggestions: 5,
  autoSuggest: true,
};
Settings are persisted to .obsidian/plugins/auto-tagger/data.json in your vault.

Build docs developers (and LLMs) love

Get started for freeTalk to us