Embeddings And Semantic Search
2 min read
Find notes by meaning rather than exact keywords with SystemSculpt's embeddings feature. This Pro feature enables intelligent note discovery.
Setting Up Embeddings
-
Enable in Settings
- Go to Embeddings tab
- Toggle "Enable Embeddings" (requires Pro license)
- Processing starts automatically
-
Processing
- Runs in background
- Progress shown in settings
- New/modified files processed automatically
-
Provider
- SystemSculpt uses OpenAI embeddings for privacy-first semantic search
- Privacy-first by default: Data sent via the API is not used to train OpenAI models
- Model: text-embedding-3-small
- Requires SystemSculpt credits
Using Similar Notes
Once embeddings are enabled:
-
Open Similar Notes Panel
- Use Command Palette: "Open Similar Notes Panel"
- Shows notes similar to current file
- Updates automatically as you navigate
-
How It Works
- Calculates semantic similarity between notes
- Shows most related content
- Based on meaning, not just keywords
Embeddings Settings
Exclusions
- Exclude chat history (default: enabled)
- Respect Obsidian exclusions (default: enabled)
- Add custom folder exclusions
- Add specific file exclusions
Management
- View processing progress
- Clear embeddings data
- Check database statistics (debug command)
Performance Notes
- Large files (>5MB) are automatically excluded
- Binary files are skipped
- Processing is adaptive to device performance
- Uses 1 credit per embedding request
Best Practices
- Let initial processing complete
- Exclude large/irrelevant folders
- Use Similar Notes panel for discovery
- Monitor credit usage
Troubleshooting
Embeddings not processing?
- Verify Pro license active
- Check credits available
- Look for errors in settings
- Try clearing and reprocessing
Similar Notes not updating?
- Ensure embeddings enabled
- Check current file is processed
- Verify not in excluded folder
- Try reopening panel
Next Steps
- Set up Agent Mode for automation
- Process Documents with AI
- Transcribe Audio to searchable notes