旅行聊天机器人

graph TD
    A[User Question] --> B[MCP Client]
    B --> C[Gemini LLM]
    C --> D{Need MCP Tool?}
    D -->|No| E[Direct LLM Response]
    D -->|Yes| F[Parse Intent & Extract Parameters]
    F --> G{Which Tools?}
    G -->|Places| H[Places MCP Server]
    G -->|Food| I[Food MCP Server]
    G -->|Hotels| J[Hotels MCP Server]
    H --> K[Combine Results]
    I --> K
    J --> K
    K --> L[Gemini Synthesizes Answer]
    L --> M[User Response]
    E --> M

travel-mcp-chatbot/
│
├── 📂 client/
│   ├── mcp_client.py              # Original client
│   └── mcp_chatbot_enhanced.py    # ✨ Enhanced chatbot with memory
│
├── 📂 servers/
│   ├── food_server.py             # 🍜 Food recommendations MCP server
│   ├── hotels_server.py           # 🏨 Hotel suggestions MCP server
│   └── places_server.py           # 🏛️ Tourist places MCP server
│
├── 📂 docs/
│   ├── ENHANCEMENTS.md            # Detailed enhancement guide
│   └── examples.md                # Usage examples
│
├── 🔒 .env                        # API keys (not in git)
├── ⚙️ mcp.json                    # MCP server configuration
├── 📦 requirements.txt            # Python dependencies
├── 🧪 test_gemini.py             # Gemini API test
└── 📖 README.md                   # You are here!

git clone https://github.com/your-username/travel-mcp-chatbot.git
cd travel-mcp-chatbot

# Create virtual environment
python -m venv venv

# Activate virtual environment
# Windows:
venv\Scripts\activate
# Mac/Linux:
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

GOOGLE_API_KEY=your_gemini_api_key_here

{
  "client": {
    "model": "gemini-2.5-flash"
  },
  "servers": [
    {
      "name": "tour-places",
      "url": "http://127.0.0.1:8001/mcp",
      "tools": ["top_places_to_visit"]
    },
    {
      "name": "tour-food",
      "url": "http://127.0.0.1:8002/mcp",
      "tools": ["top_food_to_try"]
    },
    {
      "name": "tour-hotels",
      "url": "http://127.0.0.1:8003/mcp",
      "tools": ["top_hotels_to_stay"]
    }
  ]
}

# Terminal 1 - Places Server
fastmcp run servers/places_server.py:mcp --transport http --port 8001

# Terminal 2 - Food Server
fastmcp run servers/food_server.py:mcp --transport http --port 8002

# Terminal 3 - Hotels Server
fastmcp run servers/hotels_server.py:mcp --transport http --port 8003

# In a new terminal
cd client
python mcp_chatbot_enhanced.py

You: What are the top places to visit in Paris?

🔍 Looking up top_places_to_visit for Paris...

🤖 Here are the must-visit places in Paris:

• Eiffel Tower - The iconic iron tower offering stunning city views
• Louvre Museum - World's largest art museum, home to Mona Lisa
• Notre-Dame Cathedral - Gothic masterpiece (currently under restoration)
• Arc de Triomphe - Monument honoring French military victories
• Sacré-Cœur - Beautiful basilica in Montmartre
• Champs-Élysées - Famous avenue for shopping and dining

Would you like recommendations for food or hotels as well?

You: Tell me about Tokyo

🔍 Looking up top_places_to_visit for Tokyo...

🤖 [Tokyo attractions listed...]

You: What should I eat there?

🔍 Looking up top_food_to_try for Tokyo...

🤖 Here are must-try foods in Tokyo:

• Sushi - Fresh from Tsukiji Outer Market
• Ramen - Try Ichiran or local ramen-ya
• Tempura - Light, crispy perfection
• Tonkatsu - Breaded pork cutlet
...

You: Budget hotels please

🔍 Looking up top_hotels_to_stay for Tokyo...

🤖 Great budget options in Tokyo:

Shinjuku Area:
• Capsule Inn - Unique Japanese experience
• Kimi Ryokan - Traditional budget accommodation
...

You: I'm planning a trip to Barcelona - what should I see, eat, and where should I stay on a midrange budget?

🔍 Looking up top_places_to_visit, top_food_to_try, top_hotels_to_stay for Barcelona...

🤖 Here's your complete Barcelona guide:

📍 Top Places to Visit:
• La Sagrada Familia
• Park Güell
• Las Ramblas
...

🍽️ Must-Try Foods:
• Paella
• Tapas
• Crema Catalana
...

🏨 Midrange Hotels:
Gothic Quarter:
• Hotel Barcelona Cathedral
...

You: Top attractions in Rome

🤖 [Rome attractions...]

You: Actually, let's talk about Venice instead

🔍 Looking up top_places_to_visit for Venice...

🤖 [Venice attractions...]

You: And food?

🔍 Looking up top_food_to_try for Venice...

🤖 [Venice food - still remembers Venice context]

python test_gemini.py

✅ Gemini API is working!
Response: [Gemini's response to test query]

# Test Places Server
curl http://127.0.0.1:8001/mcp

# Test Food Server
curl http://127.0.0.1:8002/mcp

# Test Hotels Server
curl http://127.0.0.1:8003/mcp

{
  "client": {
    "model": "gemini-2.5-flash"  // or "gemini-2.5-pro"
  }
}

recent = conversation_history[-3:]  # Last 3 exchanges (adjust number)

🔍 Looking up top_places_to_visit for Paris...
🤖 [Detailed place list with descriptions]

You: Thanks!
🤖 You're welcome! Have a wonderful trip! ✈️

# servers/activities_server.py
from fastmcp import FastMCP
import google.generativeai as genai

mcp = FastMCP("tour-activities")

@mcp.tool
def top_activities(city: str) -> str:
    """Get top activities and experiences in a city."""
    # Your implementation
    return response

if __name__ == "__main__":
    mcp.run()

{
  "servers": [
    {
      "name": "tour-activities",
      "url": "http://127.0.0.1:8004/mcp",
      "tools": ["top_activities"]
    }
  ]
}

Tool descriptions:
- top_places_to_visit: Returns tourist attractions and landmarks
- top_food_to_try: Returns local cuisine and food recommendations
- top_hotels_to_stay: Returns hotel and accommodation recommendations
- top_activities: Returns activities and experiences  # ← Add this

fastmcp run servers/activities_server.py:mcp --transport http --port 8004

# Install dev dependencies
pip install -r requirements-dev.txt

# Run tests
pytest tests/

# Format code
black .