VectorSearch - AI-Powered RAG Search for WordPress

Transform your WordPress site into an intelligent knowledge base with AI-powered semantic search and a beautiful chat interface powered by Google Gemini.

Features

• 🤖 AI-Powered Semantic Search - Understands meaning, not just keywords
• 💬 Interactive Chat Widget - Beautiful Vue.js chat interface
• 🔍 Hybrid Search - Combines semantic + keyword search with RRF ranking
• ⚡ Google Gemini Integration - Latest AI models for embeddings and responses
• 📱 Mobile Responsive - Works perfectly on all devices
• 🔒 Session-Based Conversations - Maintains context during chat
• 🎯 Source Attribution - Always shows where answers come from
• ⚙️ Easy Setup - Just add your Google Gemini API key

Screenshots

How It Works

User Question → RAG Search → Top 3 Relevant Posts → Google Gemini → AI Answer
                    ↓
        (Semantic Search + Keyword Search)
                    ↓
            Reciprocal Rank Fusion

Installation

Prerequisites

• WordPress 5.8 or higher
• PHP 7.4 or higher
• Google Gemini API Key (Get one free)

Configuration

Admin Dashboard

The VectorSearch admin panel (/wp-admin → VectorSearch) provides:

1. API Key Configuration - Enter and save your Gemini API key
2. Indexing Tool - Index all posts and pages with one click
3. Search Tester - Test your RAG system before going live

Modify Prompts

Edit the prompt in index.php (line ~77):

$prompt = "Your custom instructions here...";

Technical Architecture

Tech Stack

• Backend: PHP, WordPress APIs
• Frontend: Vue.js 3 (CDN)
• AI: Google Gemini (embedding-001 + gemini-2.5-flash)
• Search: Hybrid (Semantic + Keyword) with RRF

File Structure

vectorsearch/
├── assets/
│   ├── chat-widget.js      # Vue.js chat component
│   └── chat-widget.css     # Chat widget styles
├── index.php               # Main plugin file
├── readme.txt             # WordPress.org readme
├── README.md              # This file
└── LICENSE.txt            # GPL v2 license

Database Schema

Vector embeddings are stored as WordPress post meta:

wp_postmeta
├── meta_key: 'vs_embedding'
└── meta_value: JSON array of 768 floats

API Calls

Indexing (per post):

• POST to Gemini embedding-001 API

Search (per query):

• Embedding API call (query → vector)
• Cosine similarity calculation (in PHP)
• Gemini generative API call (context → answer)