Victor is a web-optimized vector database written in Rust, designed for efficient vector storage and retrieval. It targets developers building applications that require semantic search or similarity matching, offering both Rust and JavaScript APIs for seamless integration. The database boasts a highly efficient vector storage format, reducing storage needs significantly compared to naive JSON encoding.

How It Works

Victor utilizes a custom, space-efficient vector storage format, achieving a 1.5 KB representation for a 1536-dimensional vector, a substantial reduction from JSON's 20.6 KB. It also incorporates PCA for vector compression when storage is constrained. The database can operate using the native filesystem or a transient in-memory filesystem, and exposes a Web API leveraging the Private Origin File System.

Quick Start & Requirements

• JavaScript: npm install victor-db
• Rust: cargo add victor-db
• Prerequisites: Rust toolchain, Node.js/npm for JS examples. The Rust API can automatically generate embeddings using fastembed-rs's default model.
• Examples: See www/ directory for a comprehensive JS example, including fetching embeddings from OpenAI. Rust examples are in the /examples directory.

Highlighted Details

• Web-optimized vector storage format.
• Rust and JavaScript APIs available.
• Supports native filesystem and in-memory storage.
• Includes PCA for vector compression.

Maintenance & Community

Maintained by Sam Hall and Andre Popovitch. Further community or roadmap information is not detailed in the README.

Licensing & Compatibility

The README does not explicitly state the license.

Limitations & Caveats

The README does not detail limitations, known bugs, or deprecation status. The project appears to be actively developed with clear examples for both Rust and JavaScript usage.