Knowledge Graph
VisualizationVisualize how your saved knowledge connects through semantic similarity and shared topics.
Overview
The Knowledge Graph is an interactive force-directed visualization of your entire bookmark archive. Each saved link becomes a node, and edges connect links that share semantic similarity (via embedding cosine distance) or common tags. Navigate to it from the sidebar graph icon or visit /dashboard/connections.
Smart Clustering
K-Means++ Clustering
Relink automatically groups your bookmarks into semantic clusters using k-means++ on embedding vectors. The number of clusters is auto-chosen as ceil(√(n/2)), capped at 8. Each cluster gets a label derived from its most frequent tags.
Visual Cluster Hulls
When in "Clusters" color mode, soft convex-hull backgrounds are drawn behind each cluster group on the canvas. These semi-transparent regions with dashed borders visually group related nodes.
Cluster vs Domain Coloring
Toggle between Clusters (semantic grouping) and Domains(source website) coloring using the toggle in the top-right. Each mode affects node colors and the filter pills shown.
Timeline Scrubber
The timeline at the bottom of the graph has a dual-handle range slider. Drag the handles to filter nodes by date range and watch how your knowledge clusters formed and evolved over time. Colored dots on the track represent individual bookmark creation dates.
Interactions
How Connections Are Made
Two links are connected with an edge if they meet either criterion:
- Cosine similarity > 0.5 between their embedding vectors
- Shared tags — at least one tag in common
Edge thickness is proportional to the similarity weight. Node size scales with connection count.
Tip: Save at least 5–10 links on related topics to see meaningful clusters form. The graph becomes more insightful as your knowledge base grows.