Core Concepts

Understand the fundamental concepts and architecture that make TrustGraph a powerful AI agent intelligence platform.

What is TrustGraph?

TrustGraph is an Open Source Agent Intelligence Platform that transforms AI agents from simple task executors into intelligent, contextually-aware systems. Unlike traditional AI approaches that work with isolated data points, TrustGraph creates interconnected knowledge structures that enable agents to understand relationships and context.

Core Concepts

Knowledge Graphs

Knowledge Graphs are the foundation of TrustGraph’s intelligence. They represent information as interconnected networks of entities and relationships, rather than isolated documents or data points.

  • Entities: People, places, concepts, or objects in your data
  • Relationships: How entities connect and relate to each other
  • Context: The meaning that emerges from understanding these connections

GraphRAG (Graph Retrieval-Augmented Generation)

GraphRAG is TrustGraph’s advanced approach to information retrieval that goes beyond traditional RAG systems:

Traditional RAG:

  • Retrieves similar documents based on vector similarity
  • Works with isolated pieces of information
  • Limited contextual understanding

GraphRAG:

  • Understands relationships between different pieces of information
  • Retrieves contextually relevant knowledge based on graph structure
  • Provides more accurate, nuanced responses
  • Significantly reduces AI hallucinations

Knowledge Packages

Knowledge Packages combine the best of both worlds:

  • Knowledge Graphs: For structured relationships and context
  • Vector Embeddings: For semantic similarity search
  • Unified Access: Single interface for complex knowledge retrieval

This hybrid approach enables both precise relationship-based queries and flexible semantic search.

AI Agent Intelligence

TrustGraph enables AI agents to:

  • Reason about relationships: Understand how different facts connect
  • Provide contextual responses: Draw insights from interconnected knowledge
  • Reduce hallucinations: Ground responses in structured knowledge
  • Learn continuously: Build and refine knowledge over time

Architecture Overview

Knowledge Graph Builder

Extracts entities and relationships from your enterprise data:

  • Document Processing: Analyzes text, PDFs, and other formats
  • Entity Extraction: Identifies key concepts and objects
  • Relationship Mapping: Discovers how entities connect
  • Graph Construction: Builds interconnected knowledge structures

Vector Embedding Engine

Creates semantic representations of knowledge elements:

  • Semantic Encoding: Converts text into mathematical representations
  • Similarity Mapping: Enables finding related concepts
  • Hybrid Search: Combines with graph structure for powerful queries

GraphRAG Processor

Combines graph and vector search for contextual retrieval:

  • Relationship-Aware Retrieval: Finds information based on connections
  • Context Assembly: Builds comprehensive context for AI responses
  • Multi-Hop Reasoning: Follows relationship chains for deeper insights

AI Agent Runtime

Executes intelligent agents with access to knowledge graphs:

  • Contextual Understanding: Agents know how information relates
  • Grounded Responses: Answers based on structured knowledge
  • Transparent Reasoning: Clear path from question to answer

Integration Layer

Connects with existing enterprise infrastructure:

  • LLM Integration: Works with multiple AI models
  • Data Connectors: Integrates with databases, documents, APIs
  • API Gateway: Provides unified access to all capabilities

How TrustGraph Works

1. Knowledge Ingestion 

Documents → Entity Extraction → Relationship Discovery → Knowledge Graph

2. Query Processing 

User Question → GraphRAG → Contextual Retrieval → AI Response

3. Continuous Learning 

New Data → Graph Updates → Enhanced Knowledge → Better Responses

Key Benefits

Reduced Hallucinations

By grounding AI responses in structured knowledge graphs, TrustGraph significantly reduces the likelihood of AI generating false or misleading information.

Contextual Intelligence

Agents understand not just what information exists, but how different pieces of information relate to each other.

Enterprise Integration

Unifies fragmented organizational knowledge into coherent, queryable knowledge systems.

Transparency

Full visibility into how data is processed and how AI agents arrive at their responses.

Flexibility

Open-source architecture prevents vendor lock-in and enables customization.

From Your First Steps

When you followed the First Steps guide, you experienced these concepts in action:

  • Document Loading: Your PDFs became entities and relationships in a knowledge graph
  • Graph Visualization: You saw how TrustGraph represents knowledge as interconnected data
  • Vector Search: You found relevant information using semantic similarity
  • Graph RAG: You asked questions and received contextually-aware answers

Essential Terminology

Knowledge Graph: Network of interconnected entities and relationships GraphRAG: Graph-enhanced retrieval and generation for AI responses Knowledge Package: Combined graph and vector representation of knowledge Entity: A person, place, concept, or object in your data Relationship: A connection between two entities Vector Embedding: Mathematical representation of text for similarity search Agent Intelligence: AI that understands context and relationships N-Triples: Standard format for representing graph data as subject-predicate-object statements

Next Steps

Now that you understand TrustGraph’s core concepts: