GraphRAG Explorer

Every article published on graphwiz.ai is distilled into a knowledge graph: entities, the relations between them, and confidence scores for each connection. The explorer below lets you query that graph directly — in plain language, no query syntax required.

Type a question and the explorer retrieves the matching subgraph from the Neo4j-backed store, then generates a written analysis over it — retrieval-augmented generation where the retrieval layer is a graph instead of a flat text index. That matters for questions about how concepts relate: “how does GraphRAG differ from classic RAG?” or “what connects Kubernetes ingress controllers to service meshes?” return structured relationships rather than a bag of loosely related paragraphs.

The tier selector controls how deep the traversal goes: demo stays shallow and fast (depth 1, up to 25 entities), while production walks up to three hops and 200 entities for the fullest picture. Clicking any node in the result takes you to the matching topic page with the articles behind it. To understand how this pipeline is built — entity extraction, graph construction, and retrieval — read the GraphRAG articles.