Knowledge Graph Service Guide
Capture structured facts and relationships to enrich planning and ensure the assistant never forgets important context.
Fact storage
Use the graph service to register entities, relationships, and attributes that orchestrator plans can query deterministically.
- `kg.register_item` ingests triples with provenance metadata.
- Graph namespaces organize data by workspace or project.
- Retention policies align with compliance or privacy rules.
Querying
`kg.query` supports pattern matching and traversal so workflows can answer who-owns-what or dependency questions instantly.
- Supports filtering by relationship weight, freshness, and source.
- Integrates with vector memory for hybrid reasoning pipelines.
- Results include evidence references back to the original documents.
Governance
All graph mutations are logged. Optional approvals can require review before new facts are accepted into high-trust namespaces.
- `kg.delete` provides reversible tombstone operations.
- Namespace permissions tie into future role-based access controls.
- Upcoming sync adapters will mirror facts into external systems like Neo4j.