Data scientists building RAG systems (Retrieval-Augmented Generation) need to store entity relationships. The new LIST of STRUCT type allows you to attach vector embeddings directly to nodes as a list of floats, eliminating the need for a separate vector database.
Kùzu v0.13.6 introduces a more aggressive memory reclaimer within its buffer manager. When running intensive graph algorithms—such as PageRank or Weakly Connected Components—over billions of edges, the engine dynamically shrinks its internal hash tables to prevent Out-Of-Memory (OOM) errors on memory-constrained environments. 2. Zero-Copy Ingestion Speedups
To upgrade or install the latest version, you can use standard package managers like pip for Python: pip install kuzu==0.1.36 Use code with caution.
Your (e.g., fraud detection, recommendation engine, or knowledge graphs). kuzu v0 136
Whether you are building the next-generation fraud detection system or a personal knowledge graph, Kuzu v0.136 provides the tooling you need—without the complexity.
Regardless of the version, Kùzu is built on a "wisdom" (Sumerian meaning) of high-performance architecture: Cypher Support : Uses the property graph model and Cypher query language , making it intuitive for those familiar with Neo4j. Blazing Speed : Utilizes columnar disk-based storage and vectorized query processing
: Works seamlessly with tools like LangChain , PyTorch Geometric , and Pandas . Getting Started with v0.1.36 Your (e
Version 0.3.6 brings optimizations to the Cypher query engine. The implementation of smarter join orderings and improved predicate pushdowns ensures that complex multi-hop queries execute with minimal overhead. The engine is specifically tuned for Large Language Model (LLM) applications where graph retrieval-augmented generation (GraphRAG) requires low-latency lookups. Expanded Integration Ecosystem
Enhanced "Copy From" capabilities allow users to ingest data directly from DuckDB tables or Parquet files with higher throughput.
Compresses intermediate query results to drastically speed up multi-hop graph joins. Core Improvements and Updates in Kùzu v0.13.6 define a schema
Kùzu’s claim to fame is its ability to handle complex, multi-hop joins without the exponential memory explosion common in traditional graph databases. Version 0.13.6 introduces further updates to its factorized execution engine. By optimizing how intermediate sub-graphs are compressed in memory, v0.13.6 reduces the peak memory footprint of dense graph queries. This allows data scientists to execute deep, 4+ hop relationship scans on local machines without encountering Out-Of-Memory (OOM) errors. 2. Expanded Cypher Language Support
represents a significant milestone in the evolution of KuzuDB, an embeddable graph database built for query speed and scalability . This release focuses on enhancing the database's performance and expanding its features for analytical workloads. Core Features of Kuzu v0.1.36
Financial institutions use graph databases to flag circular transactions or sudden connection to known bad actors. With , the improved recursive joins allow you to run variable-length pattern matching on the fly. For example:
: This is the foundational paper describing Kùzu's architecture, including its factorized query processor and use of columnar storage.
The following script demonstrates how to initialize an on-disk database, define a schema, and execute basic Cypher queries.