Show DEV: PardusDB – The 'SQLite of Vector DBs' written in Rust

Source: Dev.to

Overview

Hey everyone! I’m excited to share a project I’ve been working on: PardusDB.

If you’ve ever built a RAG (Retrieval‑Augmented Generation) app or a local AI agent, you’ve probably realized that setting up a full‑blown vector cluster like Milvus or Pinecone is often overkill for smaller, local‑first projects.

I wanted something that felt like SQLite—simple, embedded, and zero‑config—but built specifically for vector search and document storage. So, I built it in Rust.

What is PardusDB?

PardusDB is a lightweight, high‑performance, document‑oriented NoSQL database designed to handle high‑dimensional embeddings with a tiny footprint.

Key Features

• Embedded & Local‑First – No server overhead; it lives directly in your application.
• Schema‑less – Store your metadata and vectors together without rigid table structures.
• Rust‑Powered – Fast, memory‑safe, and optimized for low‑latency similarity search.
• Open Source – The code is public for anyone to use, break, and improve.

Why I Built It

Most vector databases are designed for the cloud. PardusDB is for the edge, the desktop, and the developer who wants to get a vector‑enabled app running in minutes, not hours.

Check out the repository here: https://github.com/JasonHonKL/PardusDB

I’d love to hear your thoughts, answer any questions, or see how you might use this in your next AI project!