Intelligent document search powered by agentic AI

Agentic AI Search

An intelligent document search system powered by agentic AI search. Uses autonomous multi-step reasoning to search, analyze, and synthesize information from document collections using Cloudflare Workers, AI Search, and Google Gemini.

Features

• 🤖 Agentic AI Search - Autonomous multi-step reasoning with Google Gemini
• 🔍 Semantic Search - Powered by Cloudflare AI Search
• ✏️ Contextual Query Rewriting - Automatically consolidates follow-up questions into context-aware queries
• 🧠 Knowledge Synthesis - Iteratively gathers and evaluates information
• 💬 Modern Chat Interface - Built with vanilla HTML, CSS (Tailwind), and jQuery
• ⚡ Edge Deployment - Runs on Cloudflare Workers
• 💾 Persistent Sessions - Room-based conversation history with automatic loading on reconnect
• 🔄 Multi-RAG Support - Switch between multiple RAG instances via dropdown (creates new room automatically)

Setup

1. Install Dependencies

npm install

2. Configure Google API Key

Get your Google API key from Google AI Studio

For local development: Create a .dev.vars file in the root directory:

cp .dev.vars.example .dev.vars

Then edit .dev.vars and add your API key:

GOOGLE_GENERATIVE_AI_API_KEY=your_actual_api_key_here

For production deployment: Set the secret using Wrangler:

npx wrangler secret put GOOGLE_GENERATIVE_AI_API_KEY

Or set it in the Cloudflare dashboard under Workers > Your Worker > Settings > Variables

3. Configure AI Search Instance

The application uses AI Search instances for document retrieval. Make sure:

• You have one or more AI Search instances in your Cloudflare account
• Your documents are indexed in the AI S