# Trip Planner MCP PoC

This is a Proof of Concept for a Trip Planner application using the Model Context Protocol (MCP), Postgres, and Neo4j. It demonstrates how an Orchestrator Agent can coordinate with specialized agents (Team, Food, Travel) to answer complex queries.

## Prerequisites

1.  **Docker Desktop**: Ensure Docker is installed and running.
2.  **Python 3.10+**: Ensure Python is installed.

## Setup

1.  **Start the Databases**:
    Open a terminal in this folder and run:
    ```bash
    docker-compose up -d
    ```
    This will start Postgres (port 5432) and Neo4j (ports 7474, 7687).

2.  **Create Virtual Environment**:
    ```bash
    python -m venv venv
    # Windows
    venv\Scripts\activate
    # Mac/Linux
    # source venv/bin/activate
    ```

3.  **Install Dependencies**:
    ```bash
    pip install -r requirements.txt
    ```

4.  **Configure Environment Variables**:
    Copy `.env.example` to `.env` and add your API keys:
    ```bash
    # Windows
    copy .env.example .env
    # Mac/Linux
    # cp .env.example .env
    ```
    Edit `.env` and add:
    - `GOOGLE_API_KEY`: Get from https://makersuite.google.com/app/apikey
    - `OPENWEATHER_API_KEY`: Get from https://openweathermap.org/api (free tier)

5.  **Seed Neo4j Data**:
    The Postgres data is seeded automatically. For Neo4j, run this script once:
    ```bash
    python src/db_utils.py
    ```

## Running the Application

### Option 1: Test Trip Recommendations Tool
Test the new trip recommendations tool with real-time weather data:
```bash
python test_trip_recommendations.py
```

### Option 2: Orchestrator (Simulated)
Run the hardcoded orchestrator to see the agent workflow without an API key:
```bash
python src/orchestrator.py
```

### Option 3: Chatbot (Single Agent)
Run the interactive chatbot (requires Google API Key):

1.  Create a `.env` file in the root folder:
    ```
    GOOGLE_API_KEY=your-gemini-api-key
    OPENWEATHER_API_KEY=your-openweather-api-key
    ```
    Get free OpenWeather API key from: https://openweathermap.org/api (1000 calls/day free)
    
2.  Run the chatbot:
    ```bash
    python src/chatbot.py
    ```

### Option 4: A2A Chatbot (Multi-Agent System) ✨ **RECOMMENDED**
Run the advanced multi-agent system where an Orchestrator coordinates with specialized Team, Food, and Travel agents:

1.  Ensure `.env` has your `GOOGLE_API_KEY` and `OPENWEATHER_API_KEY`.
2.  Run the A2A chatbot:
    ```bash
    python src/a2a_chatbot.py
    ```
    
**Features:**
- 🤖 **Intelligent Agent Orchestration**: Automatically routes queries to specialized agents (Team, Food, Travel)
- 💬 **Conversational Context**: Maintains chat history across sessions (`chat_history.json`)
- 🔄 **Parallel Processing**: Multiple agents work together to provide comprehensive responses
- 🎯 **Smart Query Understanding**: Uses Gemini AI to analyze and decompose complex travel planning requests
- 🌐 **Full MCP Integration**: Seamless access to Neo4j, Postgres, Weather API, and Country Info API

The A2A chatbot is **fully functional** and provides the most advanced trip planning experience with multi-agent collaboration.

## How it Works

### Core Components

1.  **`src/mcp_server.py`**: The MCP Server exposing multiple tools:
    *   `run_cypher_query`: Execute raw Cypher queries against Neo4j (employee org graph).
    *   `run_sql_query`: Execute raw SQL queries against Postgres (food preferences, trip data).
    *   `get_trip_recommendations`: Get comprehensive travel recommendations with real-time weather data, customizable preferences, dates, group size, and budget.
    *   `get_country_info`: Get country information (currency, languages) from RestCountries API.

2.  **`src/orchestrator.py`**: Basic orchestrator with simulated agent workflow (no API key needed).

3.  **`src/a2a_chatbot.py`**: **Advanced Multi-Agent System** ✨
    *   **Orchestrator Agent**: Analyzes user queries and coordinates specialized agents
    *   **Team Agent**: Handles employee data queries using Neo4j graph database
    *   **Food Agent**: Manages food preferences and dietary requirements from Postgres
    *   **Travel Agent**: Provides comprehensive trip planning with weather and location data
    *   **Intelligent Routing**: Uses Gemini AI to determine which agents are needed for each query
    *   **Context Aware**: Maintains conversation history and builds on previous interactions
    *   **Parallel Execution**: Multiple agents can work simultaneously for complex queries

4.  **`src/chatbot.py`**: Single-agent chatbot for basic interactions.

## Project Structure

*   `docker-compose.yml`: Database configuration.
*   `data/`: Initialization scripts for databases.
*   `src/`: Source code.
    *   `mcp_server.py`: The MCP tool provider with database query tools.
    *   `orchestrator.py`: Basic orchestrator with simulated workflow.
    *   `db_utils.py`: Database connection helpers.
    *   **`a2a_chatbot.py`**: **Advanced multi-agent chatbot system** ✨ (Fully functional)
    *   `chatbot.py`: Single-agent chatbot for basic use.
*   `test_trip_recommendations.py`: Test script for trip recommendations tool.
*   `TRIP_RECOMMENDATIONS_GUIDE.md`: Comprehensive guide for using trip recommendations.
*   `ARCHITECTURE.md`: Detailed system architecture documentation.
*   `AGENT_UPDATES.md`: Agent implementation updates and changes.
*   `.env.example`: Template for environment variables.

## New Features

### 🌦️ Real-Time Weather Integration
The `get_trip_recommendations` tool now fetches live weather data from OpenWeatherMap API:
- Current conditions and 5-day forecast
- Temperature, humidity, wind speed
- Weather-based packing tips
- Automatic location geocoding

### 🎯 Smart Trip Planning
- **Customizable Preferences**: adventure, food, culture, relaxation, shopping, nature
- **Group Travel Support**: Tips for traveling with multiple people
- **Budget-Aware**: Recommendations based on budget constraints
- **Date Flexibility**: Auto-defaults to upcoming weekend if dates not specified
- **External Resources**: Links to TripAdvisor, Google Maps, Booking.com, WikiTravel

### 📊 Example Queries You Can Try (A2A Chatbot)

1. **Basic Trip**: "Plan a weekend trip to Paris"
2. **Detailed Planning**: "Plan a 5-day adventure trip to Udaipur from 2025-12-20 to 2025-12-25 for 4 people with preferences: adventure, nature, food. Budget is 5000 INR per person"
3. **Group Trip**: "We are 8 friends planning a relaxation and shopping trip to Bali"
4. **Team Outing**: "Get my team members and plan a 3-day team outing to Goa considering everyone's food preferences"
5. **Multi-Agent Query**: "Who are the employees under Bob Chen and what are their food preferences? Plan a team dinner in San Francisco"
6. **Complex Planning**: "I want to organize a corporate retreat to Jaipur for 15 people next month with cultural activities"

### 🔧 Advanced Database Queries
- **`run_cypher_query`**: Execute raw Cypher queries for complex Neo4j graph traversals
- **`run_sql_query`**: Execute raw SQL queries for flexible Postgres data retrieval
- Full schema documentation in tool descriptions

For detailed usage instructions, see `TRIP_RECOMMENDATIONS_GUIDE.md`.