🍜 Restaurant Dish Finder - Local RAG System

A simple RAG (Retrieval-Augmented Generation) system for finding dishes in Frankfurt Asian restaurants.

✨ Key Features

• No LLM API required! - Runs completely locally
• No API keys needed - Free to use
• Two versions available:
  • simple_dish_finder.py - Zero dependencies (pure Python)
  • dish_finder.py - Enhanced with semantic search (optional ML libraries)

🚀 Quick Start

Option 1: Simple Version (No Dependencies)

# Just run it!
python simple_dish_finder.py

Option 2: Enhanced Version (With Semantic Search)

# Install dependencies
pip install -r requirements.txt

# Run
python dish_finder.py

📁 Files

File	Description
simple_dish_finder.py	Zero-dependency version (pure Python)
dish_finder.py	Enhanced version with semantic search
requirements.txt	Dependencies for enhanced version
asian-restaurants-frankfurt-guide.md	Restaurant data (optional)

🔍 How It Works

Search Methods

1. Exact Match - Direct dish name match
2. Partial Match - Substring matching
3. Fuzzy Match - Handles typos (using difflib or rapidfuzz)
4. Keyword Match - Word-based matching
5. Semantic Search - Meaning-based search (enhanced version only)

Architecture

┌─────────────────────────────────────────────────────┐
│                    User Query                       │
│                    "pad thai"                       │
└─────────────────┬───────────────────────────────────┘
                  │
                  ▼
┌─────────────────────────────────────────────────────┐
│              Search Pipeline                        │
│  ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌──────────┐   │
│  │ Exact   │→│ Partial │→│ Fuzzy   │→│ Keyword  │   │
│  │ Match   │ │ Match   │ │ Match   │ │ Match    │   │
│  └─────────┘ └─────────┘ └─────────┘ └──────────┘   │
│                      │                              │
│                      ▼ (if ML libraries installed)  │
│              ┌──────────────┐                       │
│              │   Semantic   │                       │
│              │   Search     │                       │
│              └──────────────┘                       │
└─────────────────┬───────────────────────────────────┘
                  │
                  ▼
┌─────────────────────────────────────────────────────┐
│              Ranked Results                         │
│  #1 Phad Thai @ Thong Thai - 7€ (exact, 100%)       │
│  #2 Pad Thai @ Zenzakan - 18€ (partial, 90%)        │
└─────────────────────────────────────────────────────┘

💻 Usage Examples

Interactive Mode

🔍 Search: pho

🍜 Found 3 result(s) for 'pho':
=======================================================

#1 Pho Bo (Beef noodle soup)
   💰 Price: 14€
   🏠 Restaurant: Góc Phố - Vietnamese Street Food
   🍴 Cuisine: Vietnamese
   📂 Category: Nudelsuppen (Pho & Soups)
   📍 Address: Schärfengäßchen 6, 60311 Frankfurt
   🎯 Match: exact (100%)

#2 Pho Ga (Chicken noodle soup)
   💰 Price: 13€
   🏠 Restaurant: Góc Phố - Vietnamese Street Food
   🎯 Match: partial (90%)

Programmatic Usage

from simple_dish_finder import SimpleRAG

# Initialize
rag = SimpleRAG()
rag.load_from_markdown("restaurants.md")

# Search
results = rag.search("curry", top_k=5)

for item, score, match_type in results:
    print(f"{item.dish_name} @ {item.restaurant} - {item.price}")

🤔 Why No LLM API?

For this specific use case, an LLM is overkill because:

Task	LLM Needed?	Our Solution
Find dish in document	❌	Keyword/semantic search
Handle typos	❌	Fuzzy string matching
Return restaurant info	❌	Structured data lookup
Natural language understanding	❌	Simple keyword extraction

When you WOULD need an LLM:

• Answering complex questions about the food
• Generating recommendations based on preferences
• Having a conversation about restaurants
• Summarizing reviews

📊 Search Quality Comparison

Method	Speed	Typo Tolerance	Semantic Understanding
Exact Match	⚡⚡⚡	❌	❌
Fuzzy Match	⚡⚡	✅	❌
Keyword Match	⚡⚡⚡	⚠️	❌
Semantic Search	⚡	✅	✅

🛠️ Customization

Add Your Own Restaurant Data

Create a markdown file with this structure:

## 1. Restaurant Name ⭐⭐⭐⭐

**Cuisine:** Thai
**Price Range:** €€
**Address:** Street 123, City

### Menu

**Category:**
- Dish Name - Price€
- Another Dish - Price€

Adjust Search Sensitivity

In simple_dish_finder.py:

# Fuzzy match threshold (0.0 - 1.0)
if ratio > 0.5:  # Lower = more results, higher = stricter

# Keyword match threshold
if score > 0.3:  # Lower = more results

📝 License

MIT License - Feel free to use and modify!

🙏 Credits

Restaurant data compiled from various sources including:

• TripAdvisor
• Yelp
• Restaurant websites
• Wolt/Lieferando

---

No API keys. No costs. Just search! 🍜