🚀 Semantic Search Platform

A powerful full-stack platform combining intelligent vector search capabilities with a modern web interface. This platform enables semantic search across multiple file types using state-of-the-art ML models, wrapped in a user-friendly interface with modularity and scalability.

🌟 System Overview

The platform consists of two main microservices:

Vector Search Service (Python)

• Advanced semantic search engine
• Multi-modal search capabilities (text & images)
• Multiple file format processing (ZIP, RAR archives, PDF documents, TXT files and images)
• High-performance vector database (Milvus)
• Neural processing with CLIP & Sentence Transformers

Web Application (Next.js)

• Secure authentication and authorization with OAuth
• LLM integration
• High-performance gRPC communication
• Database management
• Modern UI with Ant Design

✨ Key Features

• Intelligent Search

  • 🔍 Semantic search across documents and images
  • 📄 Support for multiple file formats (PDF, TXT, Images, Archives)
  • 🤖 State-of-the-art ML models for understanding content
  • ⚡ High-performance vector similarity search

• Modern Web Interface

  • 🎨 Intuitive, responsive design
  • 🔐 Secure OAuth authentication
  • 🔄 Context-based LLM itnegration

🏗️ Architecture

The platform follows a microservices architecture with clean separation of concerns:

System Components

1. Frontend (Next.js)

   • App Router for routing
   • Ant Design components
   • Server-side rendering
   • SEO optimization

2. Backend API (Next.js)

   • API routes
   • MySQL with Prisma ORM

3. Vector Search Service (Python)

   • File processing pipeline
   • Embedding generation
   • Vector storage and retrieval
   • gRPC/REST APIs

4. Databases

   • MySQL for application data
   • Milvus for vector storage

Communication Flow

1. User signs in → Web App
2. User uploads files → Web App
3. Web App processes request → Vector Service
4. Vector Service generates embeddings → Milvus
5. Search queries flow through similar pipeline
6. Results aggregated, constructed into context prompt for LLM
7. User gets relevant response regarding their documents

---

Built with ❤️ using Next.js, Python, and cutting-edge ML models