knowledge-graph

You can build a knowledge graph by submitting KnowledgeUpdateRequest triples to populate entities and relationships, consolidating diverse geopolitical, regulatory, and supply-chain data into a single queryable graph to reduce data silos.

GraphRAG-enabled retrieval combines graph traversal with retrieval-augmented generation. The system uses NetworkX-based traversal across an in-memory adjacency graph to query relationships, enabling intelligent retrieval across geopolitical and supply-chain data sources.

Entity deduplication is handled through a built-in dedup window during batch updates. The system processes KnowledgeUpdateRequests using a write-buffered batch workflow, ensuring safe and scalable updates without creating duplicate graph nodes.

Yes, the knowledge graph supports optional SQLite persistence alongside its in-memory adjacency structure. You can flush the write buffer using the KG_FLUSH task API to persist entities, relationships, and intelligence data permanently.

You can query the graph by submitting KG_QUERY or KG_STATS task API requests. These endpoints handle graph traversal and retrieval across the consolidated geopolitical, regulatory, and supply-chain intelligence data sources stored in the graph.

Use the KG_STATS task API to retrieve graph statistics and the KG_CANARY endpoint to verify health. These APIs monitor the in-memory graph after processing KnowledgeUpdateRequest batches, ensuring your geopolitical and supply-chain intelligence layer remains stable.