Restaurant Recommendation RAG System 🍽️

A sophisticated Retrieval-Augmented Generation (RAG) system that provides intelligent restaurant recommendations using vector search and conversational AI.

🚀 Features

• Vector-Powered Search: Semantic similarity search using Google Gemini embeddings
• Intelligent Recommendations: Context-aware restaurant suggestions with explanations
• Multi-Modal Queries: Support for restaurant discovery, detailed reviews, and Q&A
• Real-time Processing: Fast vector search with Elasticsearch backend
• Flexible Architecture: Both direct Elasticsearch and LlamaIndex implementations

📊 Data Setup

The system uses a restaurant reviews dataset with the following structure:

• business_name: Restaurant name
• text: Customer review text
• rating: Numerical rating (1-5)
• rating_category: Review category (taste, menu, atmosphere, etc.)

🔍 System Architecture

Core Components

1. Vector Store: Elasticsearch with cosine similarity
2. Embedding Pipeline: Text → Gemini embeddings → 768-dimensional vectors
3. Retrieval: KNN search for semantically similar reviews
4. Generation: Gemini LLM for contextual responses

Data Flow

User Query → Embedding → Vector Search → Context Retrieval → LLM Generation → Response

🔧 API Functions

get_suggestions(query)

Returns restaurant recommendations based on semantic search

• Input: Natural language query
• Output: JSON with restaurant names, reviews, and notes

restaurant_qna(name, query)

Answers specific questions about a restaurant

• Input: Restaurant name and question
• Output: JSON with restaurant-specific answer

get_summary(name)

Provides a comprehensive restaurant summary

• Input: Restaurant name
• Output: JSON with highlights, dishes, and overall rating

general_qna(query)

Handles general food-related questions

• Input: General query
• Output: JSON with informative answer

🚀 Performance

• Vector Dimensions: 768 (Gemini embedding size)
• Search Latency: Sub-second response times
• Index Size: 1,100 restaurant reviews
• Similarity Method: Cosine similarity for semantic matching

🔮 Future Enhancements

• Multi-language support
• Image-based restaurant recommendations
• Real-time review ingestion
• Advanced filtering (cuisine type, price range, location)
• User preference learning
• Integration with restaurant APIs

🙏 Acknowledgments

• Google AI for Gemini API
• Elasticsearch for vector search capabilities
• LlamaIndex for RAG framework
• Open source community for tools and libraries