FalkorDB: Ultra-Fast Open-Source Graph Database for Knowledge Graphs and GraphRAG

FalkorDB is an ultra-fast, open-source multi-tenant property graph database built specifically for LLM Knowledge Graphs and GraphRAG (Graph-based Retrieval-Augmented Generation). As the direct successor to RedisGraph – which was discontinued by Redis Inc. in 2023 – FalkorDB has been adopted by a growing community of AI practitioners who need graph database performance optimized for the age of large language models.

Under the hood, FalkorDB uses sparse matrix operations via the GraphBLAS standard to represent and query graph adjacency matrices. This approach is fundamentally different from the index-based traversal used by most graph databases, and it is the key to FalkorDB’s millisecond-latency query performance even on graphs with millions of nodes.

graph TD
    A[LLM Application] --> B[Natural Language Query]
    B --> C[GraphRAG Pipeline]
    C --> D[Cypher Query Translation]
    D --> E[FalkorDB Engine]
    E --> F[GraphBLAS Sparse Matrices]
    F --> G[Adjacency Matrix A]
    F --> H[Adjacency Matrix B]
    G --> I[Query Result]
    H --> I
    I --> C
    C --> J[Context-Enriched LLM Response]

What Is FalkorDB?

FalkorDB is a property graph database implemented as a Redis module. It stores data as nodes (entities) and edges (relationships), with properties on both. What distinguishes it from other graph databases like Neo4j is its computational engine: instead of traversing indexes node by node, FalkorDB represents the entire graph as sparse adjacency matrices and performs graph queries using linear algebra operations.

This approach, known as GraphBLAS-based graph computation, enables FalkorDB to achieve sub-millisecond query latency on graphs that would take seconds to traverse with traditional index-based approaches. For AI workloads where query speed directly impacts user experience, this performance difference is critical.

FalkorDB is licensed under the Server Side Public License v1 (SSPLv1), which allows free use, modification, and redistribution as long as the software is not offered as a database-as-a-service product.

How Does FalkorDB Compare to RedisGraph?

RedisGraph was a pioneering graph database module for Redis, but it was discontinued by Redis Inc. in 2023 as part of a shift in the company’s open-source strategy. The FalkorDB team forked the final open-source version and has continued active development, adding significant new capabilities:

FalkorDB maintains API compatibility with the OpenCypher query language that RedisGraph used, so existing RedisGraph users can migrate with minimal changes.

What Is GraphRAG?

GraphRAG stands for Graph-based Retrieval-Augmented Generation. It is an alternative to the standard vector-based RAG approach that has become the dominant pattern for grounding LLM responses in factual data.

Traditional RAG works by:

1. Embedding documents as vectors
2. Storing them in a vector database
3. Finding the most similar vectors to a query
4. Feeding the matched content to the LLM as context

GraphRAG works differently:

1. Entities and relationships are extracted from documents and stored as a graph
2. When a query comes in, the graph is traversed along relationship paths
3. Traversed entities and their connections are returned as context
4. The LLM receives structured, relationship-aware context rather than flat document chunks

The key advantage of GraphRAG is multi-hop reasoning. A vector search can find documents similar to a query, but it cannot answer “how does concept A relate to concept B through intermediary C?” without having been explicitly trained on that relationship. GraphRAG can traverse the path A -> X -> Y -> C -> B and return the full chain of relationships.

FalkorDB supports both pure GraphRAG and hybrid approaches through its integration with frameworks like LangChain, LlamaIndex, and the GraphRAG Toolkit.

What Query Language Does FalkorDB Use?

FalkorDB uses OpenCypher, the open standard property graph query language. A typical FalkorDB query looks like this:

MATCH (p:Person)-[:WORKS_AT]->(c:Company)
WHERE c.name = 'FalkorDB'
RETURN p.name, p.role
ORDER BY p.name

FalkorDB extends OpenCypher with proprietary extensions for graph-specific operations, including graph projection, matrix operations, and multi-tenant database management.

Quick Start and Deployment

Getting started with FalkorDB takes less than a minute:

docker run -p 6379:6379 -p 3000:3000 -it --rm falkordb/falkordb:edge

This single command starts FalkorDB as a Redis module on port 6379 and launches the FalkorDB Browser visualization UI on port 3000. You can connect with any Redis client (redis-cli on port 6379) or use one of the client libraries:

• Python: pip install falkordb (FalkorDB/falkordb-py)
• TypeScript: npm install falkordb (FalkorDB/falkordb-ts)
• Java: (FalkorDB/JFalkorDB)
• Go: (Community: falkordb/falkordb-go)
• Rust: (FalkorDB/redismodule-rs)

FAQ

What is FalkorDB? FalkorDB is an ultra-fast, open-source multi-tenant property graph database designed for LLM Knowledge Graphs and GraphRAG. It is the successor to RedisGraph and uses sparse matrix GraphBLAS operations for high-performance graph querying.

How does FalkorDB compare to RedisGraph? FalkorDB is the direct evolution of RedisGraph. When RedisGraph was discontinued by Redis Inc., the FalkorDB team forked and continued development, adding multi-tenant support, improved GraphRAG features, and continued performance optimizations.

What is GraphRAG? GraphRAG (Graph-based Retrieval-Augmented Generation) is an approach that uses knowledge graphs as the retrieval backend for LLMs. Instead of vector similarity search, GraphRAG traverses graph relationships to find contextually connected information, enabling multi-hop reasoning.

What query language does FalkorDB use? FalkorDB uses OpenCypher with proprietary extensions. OpenCypher is the open standard for property graph query languages, originally developed by Neo4j. FalkorDB extends it with graph database-specific optimizations.

How do I get started with FalkorDB? Run docker run -p 6379:6379 -p 3000:3000 -it –rm falkordb/falkordb:edge. This starts FalkorDB as a Redis module on port 6379 with the FalkorDB Browser UI on port 3000. Client libraries are available for Python, TypeScript, Java, Go, .NET, and Rust.

Further Reading

• FalkorDB GitHub Repository
• FalkorDB Documentation
• GraphRAG Toolkit
• FalkorDB Docker Hub