Eidetic

Knowledge Graph

AI
Agentic
Visualization
New

An interactive, AI-agent-built knowledge graph explorer. Renders nodes and edges as an SVG force-directed graph with draggable nodes, zoom controls, search/filter, minimap, detail panel, and real-time agent activity logging.

Preview

Interactive AI system knowledge graph with 18 nodes and 20 edges. Click nodes, drag to reposition, scroll to zoom.

Knowledge Graph
100%
Node Types
18 nodes / 20 edges
Agent Activity

Graph construction complete: 18 nodes, 20 edges

7:41:25 AM

Added evaluation and feedback loops

7:41:15 AM

Connected planning module to agent orchestration

7:41:05 AM

Added AI Agent and tool execution capabilities

7:40:50 AM

Identified hallucination risk and linked mitigations

7:40:35 AM

Added vector database and embedding infrastructure

7:40:20 AM

Connected RAG Pipeline -> LLM (augments)

7:40:10 AM

Added "RAG Pipeline" as concept node

7:39:55 AM

Added "Large Language Model" as model node

7:39:45 AM

Started knowledge graph construction

7:39:30 AM

Agent Building Animation

Watch the AI agent build a graph node by node in real time. Click the button to start.

2 nodes, 1 edges
Knowledge Graph
100%
Node Types
2 nodes / 1 edges
Agent Activity

Agent ready. Click "Start Building" to begin.

7:41:30 AM

Hierarchical Layout

Nodes arranged in layers by type for a structured, top-down view.

Knowledge Graph
100%
Node Types
18 nodes / 20 edges

Radial Layout

Nodes arranged in a circle, useful for seeing all connections at a glance.

Knowledge Graph
100%
Node Types
18 nodes / 20 edges

Props

KnowledgeGraph component API reference

PropTypeDefaultDescription
nodesGraphNode[][]Array of graph nodes to display
edgesGraphEdge[][]Array of edges connecting nodes
isAgentBuildingbooleanfalseWhether to show the agent building animation and pulse effects
agentLogAgentLogEntry[][]Activity log entries from the AI agent
layout'force-directed' | 'hierarchical' | 'radial''force-directed'Layout algorithm for node positioning
onNodeSelect(node: GraphNode | null) => voidCallback when a node is selected or deselected
onLayoutChange(layout: LayoutMode) => voidCallback when layout mode changes via toolbar
showMinimapbooleantrueShow the minimap overview in the bottom-right corner
showLegendbooleantrueShow the node type legend in the bottom-left corner
showAgentLogbooleantrueShow the agent activity log side panel
heightnumber | string600Height of the graph visualization area
classNamestringAdditional CSS classes for the root container

GraphNode Type

Structure for individual graph nodes

interface GraphNode {
id: string // Unique identifier
label: string // Display label
type: NodeType // 'concept' | 'entity' | 'action' | 'data' | 'model' | 'risk'
description?: string // Shown in detail panel
metadata?: Record<string, string> // Key-value pairs
group?: string // Optional grouping
}

GraphEdge Type

Structure for graph edges

interface GraphEdge {
id: string // Unique identifier
source: string // Source node ID
target: string // Target node ID
label?: string // Relationship label (e.g., "depends on")
weight?: number // Affects layout attraction strength
}

AgentLogEntry Type

Structure for agent activity log entries

interface AgentLogEntry {
timestamp: Date
message: string
type: 'add-node' | 'add-edge' | 'remove' | 'update' | 'info'
}

Usage

Import and implementation example

import { KnowledgeGraph, type GraphNode, type GraphEdge } from '@/blocks/agentic/knowledge-graph'
const nodes: GraphNode[] = [
{ id: 'a', label: 'Machine Learning', type: 'concept', description: 'Core ML concepts' },
{ id: 'b', label: 'Training Data', type: 'data' },
{ id: 'c', label: 'Neural Network', type: 'model' },
{ id: 'd', label: 'Bias Risk', type: 'risk' },
]
const edges: GraphEdge[] = [
{ id: 'e1', source: 'b', target: 'a', label: 'feeds' },
{ id: 'e2', source: 'a', target: 'c', label: 'trains' },
{ id: 'e3', source: 'd', target: 'b', label: 'originates from' },
]
export default function Example() {
return (
<KnowledgeGraph
nodes={nodes}
edges={edges}
isAgentBuilding={false}
layout="force-directed"
onNodeSelect={(node) => console.log('Selected:', node)}
height={500}
/>
)
}

Built With

3 components

This block uses the following UI components from the design system:

Features

Built-in functionality

  • Force-directed layout: Physics-based simulation with node repulsion and edge attraction for organic graph layouts
  • Three layout modes: Force-directed, hierarchical (layered by type), and radial (circular) layouts
  • Draggable nodes: Click and drag any node to reposition it; simulation pauses during drag
  • Node detail panel: Click a node to view its description, connections, and metadata in a side panel
  • Agent building animation: Pulse effects, floating particles, and activity log when the AI is constructing the graph
  • Zoom and pan: Scroll to zoom, drag background to pan, plus zoom controls and fit-to-view button
  • Minimap: Overview minimap in the corner showing the full graph and current viewport
  • Search and filter: Search nodes by name, filter by type, or click legend items to isolate categories
  • Typed nodes: Six node types (concept, entity, action, data, model, risk) each with distinct colors and icons
  • Labeled edges: Edges display relationship labels and curve for clarity
  • Agent activity log: Chronological log of all agent actions: nodes added, edges created, updates
  • Dark mode support: Full dark mode styling for all graph elements and UI controls

Accessibility

Accessibility considerations

ARIA Attributes

Interactive elements are keyboard-focusableNode selection state conveyed visually through selection ring and dimmingAgent building status announced through Badge text

Keyboard Navigation

KeyAction
TabNavigate between toolbar controls
Enter / SpaceActivate buttons, toggle layout, select filters
EscapeClear search or deselect node (when focused)

Notes

  • SVG graph nodes use pointer events for interaction; consider adding keyboard navigation for full accessibility
  • Minimap is decorative and does not require keyboard access
  • Filter and search controls provide text-based alternatives to visual graph exploration
  • Consider adding aria-live regions for agent activity log updates