Knowledge Graph

The Knowledge Graph automatically extracts entities and relationships from your bookmarks, creating an interconnected web of knowledge. It helps you discover connections between seemingly unrelated content.

How It Works

When you save a bookmark, Arivu’s AI (Gemini 2.5 Flash) analyzes the content and:

1. Extracts entities — people, organizations, concepts, technologies, and topics
2. Maps relationships — discovers connections between entities based on context
3. Generates embeddings — converts content to vector representations for semantic similarity
4. Builds the graph — entities become nodes, relationships become weighted edges

Processing happens in the background. The graph updates automatically as you save more bookmarks.

Entity Types

The AI identifies five entity categories:

Using the Graph

Navigate to Knowledge Graph from the sidebar. The interactive visualization shows:

• Nodes — Entities sized by how frequently they appear across your bookmarks
• Edges — Connections between entities, weighted by co-occurrence strength
• Clusters — Groups of related entities that form natural knowledge areas

Interactions

• Click a node to see all bookmarks containing that entity
• Hover to highlight connected nodes and edges
• Zoom and pan to explore different areas of the graph
• Filter by type to focus on specific entity categories (people, tech, concepts)

Graph Statistics

The graph view displays aggregate stats:

• Total entities — How many unique entities have been extracted
• Total relationships — Number of connections between entities
• Top entities — Most frequently appearing entities across your collection
• Cluster count — Number of distinct knowledge areas detected

Building a Richer Graph

The graph becomes more useful as your collection grows. Tips:

• Save diverse content — Bookmarks across different topics create more interesting cross-domain connections
• Use collections — Group related bookmarks; the graph respects collection context
• Check back regularly — New bookmarks often reveal unexpected connections to older content

API Endpoints