HyperStreamDB

HyperStreamDB is a high-performance, indexed streaming lakehouse built in Rust. It is designed to provide ultra-low latency queries on streaming data by maintaining real-time indexes (HNSW for vectors, Inverted for scalars) directly on the storage format.

Key Features

• Streaming-First: Designed for continuous ingestion and real-time query availability.

• Indexed Storage: Native support for Vector (HNSW) and Scalar (Inverted) indexes, enabling sub-second latency on massive datasets.

• Multiglot: Core in Rust, with high-performance bindings for Python, and connectors for Trino and Spark via JNI.

• Zero-Copy: Leverages Arrow C Data Interface for zero-copy data sharing between the storage engine and compute engines.

• Storage Agnostic: Supports local filesystem, S3, GCS, and Azure Blob Storage.

Getting Started

Check out the Architecture guide to understand how HyperStream works, or jump into the Python Bindings to start using it today.