Kuzu V0 136 Fixed [repack] Jun 2026

If you are currently running v0.135 or earlier, follow these steps to safely upgrade. Due to the concurrency control changes, the on-disk format for transaction logs has changed. While data files remain compatible, pending WAL (Write-Ahead Log) files from v0.135 are not forward-compatible.

With v0.136 , this functionality has been and moved to a dedicated, standalone extension: djc-ext-pydantic .

Native HNSW vector indices for similarity searches, essential for AI and RAG applications. kuzu v0 136 fixed

Accelerated textual search queries embedded within native Cypher execution blocks.

Data ingestion has also seen important refinements. Previous versions occasionally encountered issues with specific CSV formatting or NULL value handling during the bulk loading process. The v0.1.3.6 patch resolves these inconsistencies, ensuring that the COPY FROM command remains robust. This means smoother transitions from flat data files to a structured graph schema without manual data scrubbing. If you are currently running v0

Kùzu is an MIT-licensed, embedded graph database system designed to run locally within application processes, similar to how SQLite operates for relational data or DuckDB operates for analytical workloads. Built in C++, Kùzu implements the structured property graph model and uses Cypher as its query language. It is optimized for multi-core architectures, utilizing a vectorized execution engine and innovative graph join algorithms (such as Worse-Case Optimal Joins) to handle massive datasets with minimal memory overhead. Core Fixes and Enhancements in v0.13.6

[Cypher Query] ──> [Optimizer (Fixed Filter Pushdowns)] ──> [Vectorized Engine (Stabilized Memory)] ──> Fast Results With v0

Below is a blog post summarizing the recent critical fixes and the current status of the project, including the 2025 archiving announcement. Kùzu Update: Recent Fixes and the State of the Graph

The open-source graph database management system (DBMS) community recently welcomed the release of . As an in-memory, embedded property graph database library designed for query speed and scalability, Kùzu has rapidly gained traction among data scientists and developers building advanced knowledge graphs and graph machine learning applications.

: Fixed an OS-specific path parsing error on Windows that previously disrupted local database importing routines. Performance Enhancements & Graph RAG Optimization