Skip to content

Latest commit

 

History

History
68 lines (47 loc) · 2.95 KB

File metadata and controls

68 lines (47 loc) · 2.95 KB
description Explore entity relationships and enhance search with knowledge graph features.

Knowledge Graph

The Knowledge Graph automatically extracts entities and relationships from your documents, creating a connected web of knowledge that goes beyond keyword search.

What the Knowledge Graph does

When you upload documents, the system doesn't just chunk and embed them — it also extracts:

  • Entities — people, concepts, technologies, processes mentioned in your documents
  • Relationships — how entities connect (references, depends on, implements, etc.)
  • Context — the surrounding information that gives relationships meaning

This creates a graph structure that agents can traverse to find connections that vector search alone would miss.

Browsing the graph

Go to Knowledge Bases → Knowledge Graph to:

  • Search entities — find specific concepts with full-text search
  • View entity details — see all relationships and source documents for an entity
  • Explore connections — navigate from entity to entity, following relationships
  • Visualise — see a graph visualisation of entities and their connections
  • Find paths — discover how two concepts are connected through intermediate entities

Graph-enhanced RAG

When an agent searches your knowledge base, the Knowledge Graph enriches results:

  1. Vector search finds the most relevant document chunks
  2. Graph traversal discovers related entities and their connected documents
  3. Combined results give the agent both directly relevant and contextually related information

This means agents can answer questions that require connecting information across multiple documents — something pure vector search struggles with.

Entity types

The system automatically categorises extracted entities:

  • Technologies and tools
  • People and organisations
  • Processes and workflows
  • Concepts and terms
  • APIs and endpoints
  • Data structures

Filter by type when browsing to focus on specific categories.

Managing the graph

  • Rebuild — trigger a full re-extraction from all documents
  • Import — import entity data from external sources
  • Statistics — view entity counts, relationship counts, and graph density
  • Snapshots — the system takes daily snapshots and computes diffs so you can track how your knowledge graph evolves

Team access control

Entity visibility respects team access settings. If a document is restricted to certain team members, its extracted entities are only visible to those members.

{% hint style="info" %} The Knowledge Graph builds automatically as you upload documents. No manual configuration needed — but you can trigger a rebuild if you want to re-extract entities after updating the extraction model. {% endhint %}

Related

  • Documents — uploading and managing source documents
  • Database — structured data querying
  • CodeGraph — code-specific knowledge graph