¶ Knowledge Graph
The Knowledge Graph in Fuse enables structured, semantic representation of relationships between knowledge artifacts, concepts, systems, and facts.
It forms the foundation for explainable reasoning, context traversal, cross-entity linking, and real-time enrichment of AI and workflow experiences.
¶ Overview
A knowledge graph consists of KnowledgeNodes (entities, concepts, resources) and KnowledgeLinks (relationships). These are derived from pages, resources, workflows, and external providers.
Graphs can be static, dynamic, or composite — built through extraction, ingestion, or manual curation — and support multi-domain traversal during prompt composition, validation, or orchestration.
¶ Core Entities
| Entity | Description |
|---|---|
KnowledgeNode |
Represents a concept, entity, term, document, or structured artifact |
KnowledgeLink |
Represents a semantic relationship between two nodes |
KnowledgeGraph (composite) |
Represents a scoped or contextual subset of nodes and links |
Nodes and links may be typed, versioned, scored, scoped, and filtered dynamically.
¶ Key Capabilities
| Capability | Description |
|---|---|
 Semantic Relationships |
Encode meaning between entities, facts, systems, and concepts |
 Context-Aware Traversal |
Navigate relationships based on role, intent, or scope |
 Ontology Alignment |
Enforce valid node/link types through a defined schema |
 Reasonable Explainability |
Enable traceability and logical paths to knowledge used by AI |
 Graph Composition |
Build scoped graphs dynamically per query, assistant, or workflow |
¶ Graph Sources
Graphs may be composed from:
- Structured document metadata
- Extracted facts from Resource content
- Workflow outputs or API responses
- Manual relationships defined by admins
- External knowledge providers with structured APIs
Graphs can be layered (e.g., global + KB + session) to support multi-context enrichment.
¶ Usage Across Domains
- Intelligence Domain: Traverse knowledge for AI prompt injection
- Workflow Domain: Trigger decisions or pathing based on graph states
- Security Domain: Filter traversal results based on identity or access policies
- UI Domain: Visualize linked facts, concept clusters, or knowledge trails
- Automation Domain: Drive rule engines or recommendations using relationship chains
¶ Governance & Trust
- Nodes and links inherit trust scores from sources
- Graph traversal respects access control and session context
- Versioning ensures traceable evolution of relationships
- Graphs may be scoped by tenant, domain, or purpose (e.g., policy graph, assistant graph)
¶ Future Extensions
- Visual graph navigation tools
- Time-based or temporal relationships
- Graph-based validation and dependency checks
- Hybrid graph + embedding retrieval orchestration
Next: Ontology & Type Definitions — Learn how node types and link semantics are defined, enforced, and extended.