NornicDB

Why we included this project

NornicDB bundles graph traversal and vector similarity search into a single engine, so you don't have to run Neo4j and Qdrant side by side and stitch their results together yourself. It speaks Bolt and Cypher, and accepts Qdrant's gRPC interface as well, which means an application already wired to either driver can point at this server without rewriting its query layer. The AI-oriented extras are where it goes beyond a plain graph store: managed embeddings, LLM-backed reranking, optional GPU acceleration, and policy-based memory decay, the last of which matters for agent-memory and Graph-RAG systems that need to forget or reprioritize older context. The MVCC engine also gives you point-in-time reads, so you can look back and see exactly what the model could have seen at a given moment. Graph, vector, and historical state in one server is a practical way for RAG teams to consolidate their stack.