To cut down on boilerplate orchestration code, version 0.12.0 expands its ecosystem with a native . This module allows Kuzu to communicate directly with external provider APIs (such as OpenAI, Anthropic, or Hugging Face) to generate text embeddings directly during data ingestion. This limits the reliance on complex, multi-tool data transformation pipelines like LangChain or LlamaIndex for basic embedding generation.
We are thrilled to announce the release of !
Reliability, stability, and maturity
: Managing snapshots and point-in-time recoveries across cloud object stores becomes trivial when dealing with a unified file format.
The search term "kuzu v0 120" refers to the release tag v0.1.2 . However, the "120" is a semantic shorthand—often used by developers migrating from v0.0.x versions. This release is significant because it bridges the gap between a prototype and a production-ready engine. kuzu v0 120 better
At its core, Kùzu v0.12.0 doubles down on its original design goals—combining the lightweight, serverless convenience of an embedded database with the sophisticated query optimization of a world-class OLAP engine. Here is how version 0.12.0 separates itself from its predecessors and legacy graph platforms.
-- Old style (pre‑0.12) MATCH (a:Person)-[:FRIEND_OF]->(b:Person) WHERE a.age > 30 RETURN a.name, COUNT(b) AS friends; To cut down on boilerplate orchestration code, version 0
Managing graph topology efficiently requires map structures that scale linearly. Kùzu utilizes indexing to handle the connections (edges) between entities (nodes). Version 0.12.0 refines these columnar forward and backward star indices. The result is faster structural lookups, ensuring that traversing a graph of hundreds of millions of nodes requires just fractions of a millisecond per hop. 3. Streamlined Native Extensions
To understand why Kuzu v0.12.0 offers a better developer experience and superior performance, let's look at how it stacks up against traditional, server-based graph infrastructure (e.g., Neo4j, TigerGraph): Architectural Feature Traditional Graph Databases (Server-Based) Kuzu v0.12.0 (In-Process) External server/cluster required Serverless, runs in-process Data Transfer Network protocols (TCP/Bolt), Serialization Zero-copy, shared memory space Query Engine Type Row-oriented or bulk-synchronous Vectorized and factorized Storage Layout Slotted-page or pointer-swizzling rows Columnar compressed disk storage Join Efficiency Standard hash joins, nested loops Novel multi-way worst-case optimal joins AI Integration Requires external vector plugins/sync pipelines Native embedded HNSW vector indexing Infrastructure Cost High (dedicated clusters, RAM-heavy servers) Extremely low (scales linearly on a laptop/single node) Key Feature Enhancements in the v0.12.0 Ecosystem We are thrilled to announce the release of
Kùzu's v0.12.0 update focused on expanding the versatility of an already fast engine, particularly strengthening its position in the AI/LLM ecosystem and making it more user-friendly for data management. 1. Superior Vector Search and Hybrid Retrieval
You can copy, move, or archive your graph data by transferring a single file.