Our objective is to create an outstanding Knowledge Graph specifically for Large Language Models (LLM) that boasts exceptionally low latency, ensuring swift delivery of information through our Graph Database, known as GraphRAG.
FalkorDB is the first queryable Property Graph database to use sparse matrices to represent the adjacency matrix in graphs and linear algebra to query the graph.
Primary features:
- Adopting the Property Graph Model
- Nodes (vertices) and Relationships (edges) that may have attributes
- Nodes can have multiple labels
- Relationships have a relationship type
- Graphs represented as sparse adjacency matrices
- OpenCypher with proprietary extensions as a query language
- Queries are translated into linear algebra expressions
- GraphRAG — Combine knowledge graphs with retrieval-augmented generation to give LLMs accurate, low-latency context. Get started with the GraphRAG-SDK.
- Code-Graph — Explore, analyze and visualize your codebase as a knowledge graph to power code understanding and AI assistants.
- Agentic Memory — Use FalkorDB as the graph-backed memory layer for AI agents through popular frameworks:
- QueryWeaver — An open-source Text2SQL tool that turns natural language into SQL using graph-powered schema understanding, letting you ask your database questions in plain English.