PDF → Knowledge Graph Builder

An interactive web application that transforms documents into visual knowledge graphs. Extract concepts and relationships from text using AI (Ollama), store them in SQLite, and visualize them with an interactive force-directed graph.

Quick Start

Prerequisites

• Node.js (v18+)
• Ollama (for AI topic generation - optional)

Installation

1. Clone the repository and install dependencies:

   npm install

2. Set up the database:

   npm run setup-db

3. (Optional) Seed with sample data:

   npm run seed-data

4. Start the server:

   npm start

5. Open your browser:

   http://localhost:3001
   

Using Ollama (Optional)

To enable AI-powered topic generation:

1. Install Ollama: ollama.ai

2. Start Ollama server:

   ollama serve

3. Install a model:

   ollama pull llama2:7b

4. Configure model (optional): Create a .env file:

   OLLAMA_URL=http://localhost:11434
   OLLAMA_MODEL=llama2:7b
   

Database Schema

Nodes Table

• id (INTEGER, PRIMARY KEY) - Unique identifier
• name (TEXT, NOT NULL) - Concept/entity name
• type (TEXT) - Node type (concept, algorithm, architecture, field, etc.)
• description (TEXT) - Detailed description of the concept

Edges Table

• id (INTEGER, PRIMARY KEY) - Unique identifier
• src_id (INTEGER, NOT NULL) - Source node ID (foreign key)
• relation (TEXT, NOT NULL) - Relationship type (e.g., "uses", "includes", "depends_on")
• dst_id (INTEGER, NOT NULL) - Destination node ID (foreign key)
• evidence (TEXT) - Optional evidence/context supporting the relationship

Usage

Visualizing the Graph

1. Open http://localhost:3001 in your browser
2. The graph automatically loads from the database
3. Interact with nodes:
   • Click any node to see its description and connections
   • Drag nodes to reposition them
   • Zoom with mouse wheel
   • Pan by dragging the background

Generating New Topics

1. Type a topic or description in the input box (top-left)
2. Click "Generate Edges & Descriptions" or press Ctrl+Enter (Mac: Cmd+Enter)
3. Ollama will generate related concepts and relationships
4. Review the generated edges and concepts in the panel

API Endpoints

GET /api/graph

Returns the full graph data for visualization.

Response:

{
  "nodes": [
    {
      "id": "1",
      "name": "Machine Learning",
      "type": "concept",
      "description": "A method of data analysis..."
    }
  ],
  "links": [
    {
      "source": "1",
      "target": "2",
      "relation": "includes",
      "evidence": "Neural networks are a subset..."
    }
  ]
}

GET /api/nodes

Returns all nodes in the database.

GET /api/edges

Returns all edges in the database.

POST /api/generate-topic

Generates nodes and edges for a topic using Ollama.

Request:

{
  "description": "Machine Learning"
}

Response:

{
  "success": true,
  "topic": "Machine Learning",
  "description": "A brief description...",
  "edges": [...],
  "concepts": [...]
}

GET /api/health

Health check endpoint.

Project Structure

CMU-Claude-Builder/
├── data/
│   └── knowledge_graph.db    # SQLite database
├── public/
│   └── index.html            # Frontend visualization
├── scripts/
│   ├── setup-db.js          # Database initialization
│   └── seed-data.js          # Sample data seeding
├── db.js                     # Database helper class
├── server.js                 # Express server & API
├── package.json
└── README.md

Development

Available Scripts

• npm start - Start the server
• npm run setup-db - Initialize/create database tables
• npm run seed-data - Clear and seed database with sample data

Database Operations

The db.js file provides a Database class with methods:

const Database = require('./db');

const db = new Database();
await db.connect();

// Create a node with description
const node = await db.createNode('Machine Learning', 'concept', 'Description here');

// Create an edge
const edge = await db.createEdge(1, 'relates_to', 2, 'Evidence text');

// Get full graph
const graph = await db.getFullGraph();

// Clear all data
await db.clearAll();

await db.close();

Features Explained

Force-Directed Graph

• Nodes are positioned using a physics simulation
• More connected nodes appear larger and more central
• Clusters form naturally based on relationships

Node Types & Colors

• concept - Blue (#60a5fa)
• algorithm - Green (#34d399)
• architecture - Pink (#f472b6)
• field - Yellow (#fbbf24)

Centrality Visualization

Node size is calculated based on the number of connections:

• More connections = larger node
• Helps identify key concepts in the knowledge graph

Configuration

Environment Variables

Create a .env file (optional):

PORT=3001
OLLAMA_URL=http://localhost:11434
OLLAMA_MODEL=llama2:7b

License

MIT