Kuzu V0 136 Fixed <4K>

The , cementing its status as the premier in-process property graph database before its repository was officially archived. By addressing deep-seated execution bugs, this specific version became a foundational, highly stable release widely adopted by developers looking for serverless, vectorized graph computing. The Core Bug Fixes in Kùzu v0.13.6

Reduces CPU overhead during complex, multi-core query parallelism execution.

Numbers do not lie. The Kuzu team released a public benchmark comparing v0.135 (buggy) vs. on a standard dataset (TPC-H-like workload with 10 million records). kuzu v0 136 fixed

Kuzu v0.1.36 represented a maturation step for the query engine, moving beyond simple graph traversal to include stricter relational constraints. The headline "fix" for this version was the implementation of , a feature highly requested by users attempting to model complex schemas. Additionally, the release included significant upgrades to the LOAD FROM capabilities, closing the gap between Kuzu’s Cypher implementation and standard SQL-like data ingestion.

The release expanded the system's HNSW vector index capabilities to natively support precision floating-point columns. This change allowed developers to load highly precise embeddings from top-tier LLM models directly into the graph without losing mathematical fidelity during cosine similarity calculations. Memory Compression The , cementing its status as the premier

The v0.13.6 update mitigates a rare race condition during structural modifications (like adding a new property column to an existing node table while concurrent read queries are actively scanning the database catalog).

For developers running Kùzu in production or utilizing it inside local data science pipelines, upgrading to v0.13.6 provides several immediate operational advantages: Numbers do not lie

Kuzu Graph Database: Stability and Evolution in v0.11.x and Beyond

Before upgrading your application library dependencies, launch your current Kùzu environment and export your database contents to directory-backed formats like Parquet or CSV:

Beyond stability, the v0.13.6 release introduces iterative performance upgrades that maximize hardware utilization: