Feature

Installation

K2HVEC provides official packages for major Linux distributions(Debian, Ubuntu, Rocky Linux, Fedora, and Alpine Linux).
You can easily install it from packagecloud.io without building from source code.

For other operating systems and environments, you can build from source.

Key-Value Flexibility

K2HVEC stores and retains pairs of vector embeddings(used as keys) and their associated arbitrary user data(values).

• No Length Restrictions:
  There are no limits on the array length(dimensions) of vector embeddings or the size of user data.
• Flexible Storage Engine:
  Depending on your use case, you can choose between in-memory storage(high-speed, volatile) or file-backed storage(large-capacity, persistent).

Fast Vector Search

Performs high-speed and accurate vector similarity searches using specified input vector embeddings.

Full Inheritance of K2HASH Features

Inherits all native capabilities of the underlying K2HASH engine.
You can directly apply features such as data encryption, expiration limits (TTL), and custom attributes to the stored user data.

Bundled Tools

K2HVEC is provided as a C/C++ library, with its features accessible via APIs. In addition to the core library, the following tools are included:

• CLI Tool (Interactive / Command Execution):
  A command-line tool for intuitively operating, testing, and performing vector searches on the database from an external interface.
• Projection Matrix Management Tool:
  A dedicated tool for generating and loading the Projection Matrix used internally during data storage and similarity search processing.