ztagsuggest

name: ztagsuggest description: AI-assisted tag suggestions for untagged notes using semantic similarity to find common tags from related notes

/ztagsuggest - Tag Suggestions

Suggest tags for untagged or under-tagged notes by analyzing semantically similar notes.

Usage

/ztagsuggest                  # Interactive suggestion workflow
/ztagsuggest <note-id>        # Suggest tags for specific note

How It Works

1. Find untagged notes - Scan permanent/ folder for notes with empty or minimal tags
2. Query similar notes - Use /zrelated to find semantically similar tagged notes
3. Extract tag patterns - Collect tags from top 5 similar notes
4. Score suggestions - Rank tags by frequency across similar notes
5. Present for review - Human approves before applying

Implementation

Step 1: Find Untagged Notes

List permanent notes without tags or with <2 tags:

# Scan for untagged notes
export OBSIDIAN_VAULT="${OBSIDIAN_VAULT:-$HOME/Documents/Obsidian Vault}"
find "$OBSIDIAN_VAULT/knowledge-base/permanent" -name "*.md" -exec grep -l "tags: \[\]" {} \; 2>/dev/null | head -20

Or notes with minimal tags:

export OBSIDIAN_VAULT="${OBSIDIAN_VAULT:-$HOME/Documents/Obsidian Vault}"
grep -l "tags:" "$OBSIDIAN_VAULT/knowledge-base/permanent"/*.md | while read f; do
  count=$(grep "tags:" "$f" | grep -oE '\[.*\]' | tr ',' '\n' | wc -l)
  if [ "$count" -lt 2 ]; then echo "$f: $count tags"; fi
done

Step 2: Get Related Notes

For each untagged note, find similar notes using semantic search:

isengardcli run --account 806230523044 -- bash -c '
export ZETTELKASTEN_BUCKET=zettelkasten-cajias
export ZETTELKASTEN_INDEX=knowledge-index
export OBSIDIAN_VAULT="${OBSIDIAN_VAULT:-$HOME/Documents/Obsidian Vault}"
zk-related "<note title or content excerpt>" --top 5
'

Step 3: Extract Tag Patterns

From the related notes output, collect tags and count frequencies:

Step 4: Review and Apply

Present suggestions to user:

Note: subagent-delegation-for-context-management.md
Current tags: []
Hub: AI Engineering

Suggested tags (from 5 similar notes):
  ★ context-management (4/5 matches, 80%)
  ★ ai-agents (3/5 matches, 60%)
  ○ workflows (2/5 matches, 40%)

Apply suggestions? [y/n/edit]

Only apply tags with ≥40% confidence unless user explicitly approves lower.

Step 5: Update Note

If approved, manually add tags to the note's frontmatter:

tags: [context-management, ai-agents]

Then run /zsync to update the vector store.

Scoring Algorithm

Tag confidence = (# similar notes with tag) / (# similar notes checked) × 100

Batch Processing

For bulk suggestions:

1. Run audit to identify all untagged notes
2. For each note:
   • Query related notes
   • Collect tag suggestions
   • Batch review with user
3. Apply approved tags
4. Re-sync vectors

Human Review Gates

IMPORTANT: Always get human approval before applying tags:

• AI suggests based on patterns
• Human validates relevance
• Human can override or edit suggestions
• Never auto-apply without review

Success Metrics

Tag Taxonomy Compliance

Suggestions should comply with [[hubs/tag-taxonomy|Tag Taxonomy]]:

• Only suggest approved tags
• Warn if suggesting deprecated tags
• Respect maximum 6 tags guideline

Related

• /ztagaudit - Generate tag health report
• /ztagfix - Tag remediation workflow
• /zrelated - Find semantically related notes
• [[hubs/tag-taxonomy|Tag Taxonomy]] - Tag governance