For production analytical workloads or academic research requiring graph traversal, choose full . For embedded edge devices or simple key-value-like graph storage, Lite suffices.
The development roadmap indicates that after , the team plans to release v0.14x with a focus on distributed queries. However, v0.136 will likely remain a Long-Term Support (LTS) candidate due to its stability. Given the release cadence, v0.136 is expected to receive security backports until at least late 2025. kuzu v0 136 full
Kuzu utilizes a factorized query processing engine. In graph traversals (joins), intermediate results can grow exponentially. Factorization allows Kuzu to avoid materializing full cross-products during joins, passing compressed representations of lists between operators. In v0.1.36, optimizations to the factorization logic have reduced memory consumption during deep recursive queries (such as Variable-Length Path traversals). However, v0
Full-text search (BM25), HNSW Vector Index, Graph Algorithms Serializable ACID compliance License MIT (Open Source) In graph traversals (joins), intermediate results can grow
v0.1.36 maintained parity across its language bindings. The Python API ( kuzu package) is particularly notable for its tight integration with the PyData ecosystem, allowing query results to be returned directly as Pandas DataFrames or Arrow tables.