Skip to main content

Knowledge Graph

The knowledge graph is a highlight feature of SparkNoteAI. It uses LLM-powered intelligent extraction of concepts and relationships, stored in Neo4j graph database with visual display.

Core Capabilities​

LLM-Powered Extraction​

  • Calls the configured LLM model (OpenAI / Anthropic / Azure / Alibaba Cloud) to automatically analyze note content
  • Extracts key concepts and relationships between concepts
  • Concept normalization: merges similar concepts to improve reusability
  • Supports both full rebuild and incremental update modes

Graph Storage​

  • Uses Neo4j graph database to store nodes and edges
  • GraphNode model stores concept information
  • GraphEdge model stores relationships between concepts
  • Nodes can be linked to their corresponding notes

Visual Display​

  • 2D force-directed graph visualization (react-force-graph)
  • Nodes rendered with lucide icons
  • Supports node dragging and zooming
  • Real-time polling displays graph build progress

Knowledge Graph

Scene Configuration​

Which LLM model the knowledge graph uses is managed through FeatureSetting and can be modified in the settings page. Supports:

  • Selecting different LLM providers
  • Configuring model parameters
  • Build status management: maintains LLM configuration state during rebuilds and marks build-in-progress

Usage Flow​

  1. Configure LLM: Set up the LLM for knowledge graph in the settings page
  2. Build Graph: Trigger the knowledge graph build task
  3. View Status: Check build progress via status polling
  4. Explore Graph: Drag and explore concept relationships in the force-directed graph

API Overview​

For detailed API documentation, see API Reference - Knowledge Graph.