Show HN: PardusDB – SQLite-like vector database in Rust (github.com)

Pardus AI has launched PardusDB, a lightweight, SQLite-like embedded vector database built in Rust, designed to enhance local AI application development. This open-source project facilitates developers in creating retrieval-augmented generation (RAG) pipelines, semantic search, and recommendation systems, offering a privacy-focused solution with persistent vector storage and zero external dependencies. Key features include single-file storage, SQL-like syntax for table management, fast graph-based approximate nearest neighbor search, and optional GPU acceleration for improved performance. PardusDB stands out in the AI/ML community due to its impressive performance metrics, significantly outpacing competitors like Neo4j and HelixDB in both speed and accuracy. It boasts single inserts taking around 160 microseconds and batch inserts for 10,000 vectors completed in just 6 milliseconds, demonstrating a remarkable 1983x speed advantage over Neo4j for inserts. The platform emphasizes ease of deployment with minimal setup requirements, making it an attractive choice for developers seeking a powerful yet simple database for AI applications without the overhead of complex server management.