LynxEngine is a high-performance, AI-powered semantic search engine built on the MERN stack. Unlike traditional keyword-based search, LynxEngine understands user intent using high-dimensional vector embeddings and features an automated "Hunter" logic that scrapes and indexes Wikipedia in real-time when local data is insufficient.
| Component | Technology |
|---|---|
| Frontend | React.js, React Router, Lucide-React |
| Backend | Node.js, Express.js, Axios, Cheerio |
| Database | MongoDB Atlas (Vector Search Index) |
| AI Models | Ollama (mxbai-embed-large & Gemma 2:2b) |
| Styling | Custom CSS (Windows XP "Luna" Design System) |
LynxEngine operates through a Three-Stage Retrieval Pipeline designed for maximum precision:
The user query is processed by Gemma 2:2b to resolve entities.
- Example: "lui hamilto" →
Lewis_Hamilton. - This ensures the "Hunter" lands on the correct Wikipedia URL immediately.
The normalized query is converted into a 1024-dimension vector using mxbai-embed-large. The system performs a
$vectorSearch against MongoDB to find matches based on Cosine Similarity.
If the highest similarity score falls below a dynamic threshold (e.g., 0.67), the engine:
- Triggers a real-time crawl of Wikipedia.
- Parses and cleans the content using Cheerio.
- Generates new embeddings and indexes the data for future users.
- Vector Stability Check: Uses a custom scoring algorithm to determine if search results are high-quality or if a fresh crawl is required.
- Flag Prioritization: Specialized scraper logic to detect and prioritize national flags and high-res SVG media for country-based queries.
- Resilient Async Flow: Implements
AbortControllerwatchdogs to prevent system hangs during local AI generation. - Nostalgic UX: A complete, custom-built UI inspired by the Windows XP (2005) desktop environment, optimized for low-glare viewing.
- Ollama installed and running.
- MongoDB Atlas account with a Vector Search Index named
vector_index. - Node.js & NPM.
- Clone the Repository:
git clone [https://github.com/swayamvaza/AI-Search-Engine.git](https://github.com/swayamvaza/AI-Search-Engine.git) cd AI-Search-Engine - Pull AI Models:
ollama pull mxbai-embed-large ollama pull gemma2:2b
- Launch Backend Engine:
- Create a directory for backend
cd backend npm install node server.js
- Launch Frontend Engine:
- Create a directory for frontend
cd frontend npm install npm start
├── client/ # React Frontend (XP Design System) ├── server.js # Node.js Server & Scraper Logic ├── Item.js # Mongoose Schema for Vector Documents ├── .env # System Environment Variables └── README.md # System Documentation
| Swayam Kumar |  |
|---|---|
| Project Repository | AI-Search-Engine |
| Tech Stack | MERN + AI (Ollama) |
Developed by Swayam Kumar
LynxEngine v0.1-alpha