GraphRAG SDK 0.4.0

It is a graph database that combines graph traversal with vector search to support GraphRAG and retrieval-augmented generation, reducing LLM hallucinations and improving accuracy by grounding responses in organizational knowledge graphs. It sells to developers and data teams in businesses building GenAI applications such as chatbots, agentic AI, fraud detection, and security analytics. It is positioned as an open-source, Redis-based database using sparse matrices and linear algebra with AVX acceleration for low-latency queries, offered as SaaS and infrastructure API with multi-tenant support.

Key features

• GraphRAG with ontology auto-detection
• Built-in agent orchestration
• Graph traversal with vector search
• Persistent conversation history and memory
• Entity extraction and fact linking
• Real-time fraud ring detection
• Security graph for vulnerabilities and assets
• Multi-tenant multi-graph support
• Linear scalability
• Ultra-low latency queries
• Redis-based architecture
• Sparse matrix adjacency representation
• AVX acceleration for queries