DotDB - Vector Database from Scratch

A fully in-browser vector database implementation with semantic search capabilities. Here's the vector DB in 300 lines of code

You can try it here.

Features

• Pure Browser Implementation - No backend required, runs entirely in your browser
• Persistent Storage - Uses IndexedDB to store vectors across sessions
• Embeddings - Powered by HuggingFace Transformers.js (bge-small-en-v1.5))
• Cosine Similarity - Fast vector similarity search with configurable top-K results
• Auto-chunking - Automatically splits documents by paragraphs for better search granularity

Tech Stack

• TypeScript
• IndexedDB - Browser-native persistent storage of vectors & metadata
• React & Tailwind - For UI
• Transformers.js - Client-side embeddings generation
• Vite - Fast development & build

Getting Started

# Install dependencies
pnpm install

# Run development server
pnpm dev

# Build for production
pnpm build

How It Works

1. Add Documents - Paste text, it's automatically split into paragraphs and embedded
2. Search - Query using natural language
3. Get Results - Ranked by cosine similarity with configurable top-K

All embeddings are generated client-side and stored in IndexedDB for instant recall.

Live Demo

dotdb.summerlabs.xyz