NeuroSearch

An open-source AI-powered search engine with real-time streaming responses

Live Demo • Documentation • Report Bug

---

📖 About

NeuroSearch is an open-source AI search engine that combines the power of large language models with real-time web search to provide accurate, cited answers. Built with Next.js and powered by Groq's ultra-fast inference, it delivers streaming responses with source attribution.

💡 Want to learn how to build this? Check out the official tutorial!

✨ Features

• 🔍 Real-time Web Search - Leverages Exa.ai for high-quality, relevant search results
• ⚡ Ultra-fast Inference - Powered by Groq's LPU technology for near-instant responses
• 📝 Streaming Responses - Answers stream in real-time for a smooth user experience
• 📚 Source Attribution - Displays top search results alongside AI-generated answers
• 🔄 Related Questions - Automatically suggests follow-up questions based on your query
• 📊 Observability - Integrated with Helicone for monitoring and analytics
• 🎨 Modern UI - Clean, responsive interface built with Tailwind CSS

🛠️ Tech Stack

Category	Technology
Framework	Next.js 16 (App Router, Turbopack)
UI Library	React 19
Styling	Tailwind CSS 3.4
Language	TypeScript 5
LLM Inference	Groq AI
LLM Models	OpenAI gpt-oss-120b & Qwen qwen3-32b
Search API	Exa.ai
AI SDK	Vercel AI SDK 5
Observability	Helicone
Analytics	Plausible

🔄 How It Works

graph LR
    A[User Question] --> B[Exa Search API]
    B --> C[Top 9 Results]
    C --> D[Process Top 5 Sources]
    D --> E[LLM Context Window]
    A --> E
    E --> F[gpt-oss-120b]
    F --> G[Streaming Answer]
    D --> H[qwen3-32b]
    H --> I[3 Related Questions]

Loading

1. Search - User's question is sent to Exa.ai to retrieve the top 9 relevant results
2. Process - Text is extracted from the top 5 sources (optimized for token limits)
3. Generate - Combined context and question are sent to gpt-oss-120b for answer generation
4. Stream - Response is streamed back to the user in real-time
5. Suggest - Qwen qwen3-32b generates 3 related questions using the top 2 sources

🚀 Getting Started

Prerequisites

• Node.js 18+
• npm or yarn
• API accounts for the services below

API Keys Required

Service	Purpose	Sign Up
Groq	LLM inference	Get API Key
Exa	Search API	Get API Key
Helicone	Observability	Get API Key

Installation

1. Clone the repository

   git clone https://github.com/yourusername/neurosearch.git
   cd neurosearch

2. Install dependencies

   npm install
   # or
   yarn install

3. Configure environment variables

   Create a .env.local file in the root directory:

   GROQ_API_KEY=your_groq_api_key
   EXA_API_KEY=your_exa_api_key
   HELICONE_API_KEY=your_helicone_api_key

   Note: Helicone integration is automatic. All Groq requests are routed through Helicone for observability—no additional configuration needed.

4. Start the development server

   npm run dev

5. Open in browser

   Navigate to http://localhost:3000

📁 Project Structure

neurosearch/
├── app/
│   ├── api/              # API routes
│   │   ├── getAnswer/    # Answer generation endpoint
│   │   ├── getSimilarQuestions/  # Related questions endpoint
│   │   └── getSources/   # Source fetching endpoint
│   ├── globals.css       # Global styles
│   ├── layout.tsx        # Root layout
│   └── page.tsx          # Main page
├── components/           # React components
│   ├── Answer.tsx        # Answer display
│   ├── Hero.tsx          # Landing section
│   ├── InputArea.tsx     # Search input
│   ├── Sources.tsx       # Source cards
│   └── ...
├── public/               # Static assets
├── utils/                # Utilities & types
└── ...

🗺️ Roadmap

• Smart Tokenization - Implement tokenizer to optimize source content within token limits
• Regenerate Feature - Allow users to regenerate answers
• Enhanced Citations - Numbered in-text citations with source linking
• Share Answers - Generate shareable links for search results
• Auto-scroll - Smooth scrolling during streaming (especially mobile)
• SPA Navigation - Migrate to page-based routing to fix hard refresh issues
• Caching & Rate Limiting - Integrate Upstash Redis
• Advanced RAG - Implement keyword search and question rephrasing
• Authentication - Add Clerk auth with PostgreSQL/Prisma for user sessions

🙏 Inspiration

This project was inspired by:

• Perplexity - AI-powered answer engine
• You.com - AI search with citations
• Lepton Search - Open-source AI search

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

👨‍💻 Author

MontaCoder

• GitHub: @montacoder

---

⭐ Star this repo if you found it helpful!

Made with ❤️ and Groq