Wiki-Plus: Custom Wikipedia

React-based Wikipedia browser with AI-powered semantic search using Transformers.js and PGlite.

Features

• AI-Powered Semantic Search: Uses Transformers.js to run local AI embeddings for intelligent article search
• PGlite Database: Lightweight PostgreSQL-compatible database running entirely in the browser
• Real-time Search: Toggle between keyword search and AI semantic search
• Modern UI: Clean, responsive interface with gradient design
• Vector Embeddings: Articles are automatically indexed with vector embeddings for semantic similarity
• No Backend Required: Everything runs in your browser

How It Works

Semantic Search

1. Initialization: When the app loads, it:

   • Initializes the Transformers.js embedding model (Xenova/all-MiniLM-L6-v2)
   • Creates a PGlite database
   • Loads sample articles
   • Generates embeddings for each article
   • Stores articles with their embeddings in the database

2. Search Process:

   • Keyword Search: Simple text matching in titles and content
   • AI Semantic Search:
     • Generates an embedding for your query
     • Compares it with all article embeddings using cosine similarity
     • Returns the most semantically similar articles
     • Shows relevance scores