Automatic Connections
Lexic automatically discovers relationships between your notes, turning isolated thoughts into connected knowledge. Here's how it works.
Connection Discovery Methods
Lexic uses three complementary approaches to find connections between your notes:
1. Semantic Similarity
Notes that discuss similar concepts are connected, even if they use different words. "Q4 revenue projections" connects to "end-of-year financial forecast" because they mean the same thing.
How it works: Each note is converted to a vector embedding using text-embedding-3-small (1536 dimensions). Notes with similar vectors are semantically related.
2. Shared Entities
Notes that mention the same people, projects, organizations, or concepts are automatically connected. If two notes both mention "Project Atlas," Lexic links them.
3. Explicit References
Links you create manually. When you reference one note from another, that connection is stored with maximum strength. Your explicit links always take priority.
Connection Strength
Not all connections are equal. Lexic assigns a strength score to every connection, helping you focus on the most relevant relationships.
Strong
Highly related notes
Moderate
Related concepts
Weak
Tangential relationship
Minimum threshold: By default, connections below 0.3 are filtered out. This prevents noise from weak associations. You can adjust this threshold in your settings.
Graph Traversal Engine
When you ask Lexic a question, it doesn't just search notesโit explores the knowledge graph to find relevant context. This traversal is designed to be both thorough and cost-efficient.
$ Budget-Constrained Exploration
Every traversal has a word budget. Lexic prioritizes the most promising paths to maximize insight within your cost limits.
^ Exponential Depth Cost
Each level deeper costs more. Direct connections are cheap; second-degree connections cost more. This focuses exploration on the most relevant paths.
* Intelligent Priority Scoring
Connections are ranked by strength and relevance to your query. High-priority paths are explored first.
~ Circular Reference Detection
The engine tracks visited notes to prevent infinite loops when traversing densely connected areas of your graph.
Performance Optimization
Traversal results are cached for 30 minutes. If you ask similar questions in quick succession, Lexic can reuse previous explorations, saving both time and words.
Connection Storage
Understanding how connections are stored helps you understand Lexic's privacy and access model.
Lexicon Boundaries
Connections only exist between notes in the same lexicon. Your personal notes never connect to team notes unless they're in a shared lexicon.
RLS-Enforced Access
Row-Level Security ensures you only see connections to notes you have access to. Even if a connection exists, it's invisible if you can't see the target note.
Privacy note: Connections are bidirectional for discovery but access is always one-way. Seeing that Note A connects to Note B doesn't grant access to Note B.
Quick Reference
| Connection Type | How It's Found | Typical Strength |
|---|---|---|
| Semantic | Vector similarity (embeddings) | 0.3 - 0.8 |
| Entity | Shared people, projects, concepts | 0.5 - 0.9 |
| Explicit | Manual links you create | 1.0 |