A Graph in R2R is a knowledge graph that is associated with a specific Collection. Each Graph contains:
- Entities: Extracted information nodes from documents (e.g., people, places, concepts)
- Relationships: Connections between entities defining how they relate
- Communities: LLM generated descriptions of groupings of related entities found from Leiden clustering.
- Document Mappings: Tracking which documents have contributed to the graph
Key features of Graphs in R2R:
- Git-like Model
- Each Collection has an associated Graph that can diverge independently
- The pull operation syncs document knowledge into the graph
- Changes can be experimental without affecting the base Collection and underlying documents
- Knowledge Organization
- Automatic entity and relationship extraction from documents
- Community detection for hierarchical knowledge organization
- Support for manual creation and editing of entities, relationships and communities
- Rich metadata and property management
- Access Control
- Graph operations are tied to Collection permissions
- Superuser privileges required for certain operations like community building
- Document-level access checks when pulling content
Core Operations
| Method | Endpoint | Description |
|---|---|---|
| GET | /graphs/ | Get graph overview |
| POST | /graphs/{collection_id}/pull | Sync documents with graph |
| POST | /graphs/{collection_id}/ | Get graph details |
Entity Management
| Method | Endpoint | Description |
|---|---|---|
| GET | /graphs/{collection_id}/entities | List entities ([r2r-docs.sciphi.ai][1]) |
| POST | /graphs/{collection_id}/entities | Create entity |
| GET | /graphs/{collection_id}/entities/{entity_id} | Get entity details |
Similar CRUD endpoints exist by mapping /entities above to for relationships (/relationships) and communities (/communities).