Concepts
Concepts are named topics or ideas that the AI extracts from your notes. Rather than relying solely on tags or wikilinks, Concepts are discovered automatically and deduplicated across your knowledge base — giving you a shared vocabulary that grows with your thinking.
How Concepts are Extracted
When the AI pipeline processes a note (during ingestion, transcription, or OCR), it identifies the key concepts present in the text. These are merged with existing concepts using a semantic similarity check, so "machine learning" and "ML" resolve to the same concept.
You can also tag a concept manually by adding it as a wikilink and approving the suggestion in the Process Queue.
Browsing Concepts
Open Concepts from the Intelligence section of the sidebar. Each concept card shows its label, description, and how many notes reference it. Click a concept to see all associated notes and related concepts.
Concept Detail View
- Associated notes — all notes that reference this concept, ordered by relevance
- Related concepts — other concepts that frequently appear alongside this one
- Description — an AI-generated summary of the concept based on your notes
Knowledge Graph Integration
When enabled in Concepts settings, concepts appear as distinct nodes in the Knowledge Graph. This lets you visualise how topics cluster and where notes connect through shared concepts.
Settings
Configure concept behaviour in Settings → Concepts:
- Auto-extract — enable or disable automatic extraction when notes are processed.
- Minimum note count — hide concepts referenced by fewer than this many notes, reducing noise.
- Similarity threshold — controls how similar two concept labels must be to be merged. Higher values create more granular, separate concepts.
- Show in graph — toggle concept nodes in the Knowledge Graph.
Tips
- Use a low similarity threshold (0.6–0.7) if your notes cover many related sub-topics you want kept separate.
- Raise the minimum note count to 3–5 to reduce single-mention concepts cluttering the list.
- Concepts work best once you have 20+ processed notes — the deduplication becomes more meaningful with more data.