Building an AI Travel Planner with AutoGen: A Step-by-Step Guide

This guide will help you create an AI-powered travel planner using Microsoft's AutoGen framework. The application will utilize multiple AI agents to collaborate and plan a personalized travel itinerary based on user preferences. We'll use Python and the AgentChat API of AutoGen to build this system.

---

Table of Contents

1. Introduction
2. Prerequisites
3. Project Setup
4. Installing Dependencies
5. Creating the Agents
   • 1. UserAgent
   • 2. FlightAgent
   • 3. HotelAgent
   • 4. ActivityAgent
6. Implementing the Main Program
7. Running the Application
8. Conclusion
9. Additional Notes

---

Introduction

AutoGen is an open-source framework for building AI agent systems. It simplifies the creation of event-driven, distributed, scalable, and resilient agentic applications. In this guide, we'll build an AI Travel Planner where different AI agents collaborate to plan a travel itinerary based on user input.

Use Case: An AI Travel Planner that interacts with the user to gather preferences and coordinates multiple specialized agents (FlightAgent, HotelAgent, ActivityAgent) to plan flights, accommodations, and activities.

---

Prerequisites

• Python 3.8+ installed on your machine.
• OpenAI API Key: Obtain one from OpenAI.
• Terminal Access: Ability to run commands in your operating system's terminal.
• Git (optional): For version control.
• Basic Knowledge of Python: Understanding of Python programming and asynchronous programming with asyncio.

---

Project Setup

1. Create a Project Directory

Open your terminal and create a new directory for the project:

bashCopy
1mkdir ai_travel_planner
2cd ai_travel_planner

2. Initialize a Git Repository (Optional)

bashCopy
1git init

3. Create a Virtual Environment

bashCopy
1python3 -m venv venv

4. Activate the Virtual Environment

• On Linux/macOS:

  bashCopy
  1source venv/bin/activate

• On Windows:

  bashCopy
  1venv\Scripts\activate

---

Installing Dependencies

1. Upgrade pip

bashCopy
1pip install --upgrade pip

2. Install AutoGen Packages

Install the required AutoGen packages and the OpenAI extension:

bashCopy
1pip install 'autogen-agentchat==0.4.0.dev8' 'autogen-ext[openai]==0.4.0.dev8'

3. Install python-dotenv for Environment Variables

bashCopy
1pip install python-dotenv

---

Creating the Agents

We'll create four agents:

1. UserAgent: Interacts with the user to gather preferences.
2. FlightAgent: Handles flight booking queries.
3. HotelAgent: Handles accommodation booking.
4. ActivityAgent: Suggests activities based on destination.

---

1. UserAgent

This agent will initiate the conversation with the user, gather preferences, and coordinate with other agents.

Code: user_agent.py

pythonCopy
1# user_agent.py
2
3from autogen_agentchat.agents import UserProxyAgent
4from autogen_agentchat.message import AssistantMessage
5
6class UserAgent(UserProxyAgent):
7    pass  # Inherits functionality from UserProxyAgent

---

2. FlightAgent

Handles flight-related queries and bookings.

Code: flight_agent.py

pythonCopy
1# flight_agent.py
2
3import asyncio
4from autogen_agentchat.agents import AssistantAgent
5from autogen_ext.models import OpenAIChatCompletionClient
6
7async def search_flights(departure_city: str, destination_city: str, departure_date: str, return_date: str):
8    # Mock implementation of flight search
9    await asyncio.sleep(1)  # Simulate network delay
10    return f"Found flights from {departure_city} to {destination_city} departing on {departure_date} and returning on {return_date}."
11
12flight_agent = AssistantAgent(
13    name="FlightAgent",
14    model_client=OpenAIChatCompletionClient(
15        model="gpt-4",
16        # api_key will be loaded from environment variable
17    ),
18    instructions="""
19You are an AI agent specialized in booking flights. Assist in finding flights based on user preferences.
20""",
21    tools=[search_flights],
22)

---

3. HotelAgent

Handles accommodation queries and bookings.

Code: hotel_agent.py

pythonCopy
1# hotel_agent.py
2
3import asyncio
4from autogen_agentchat.agents import AssistantAgent
5from autogen_ext.models import OpenAIChatCompletionClient
6
7async def search_hotels(destination_city: str, check_in_date: str, check_out_date: str):
8    # Mock implementation of hotel search
9    await asyncio.sleep(1)  # Simulate network delay
10    return f"Found hotels in {destination_city} from {check_in_date} to {check_out_date}."
11
12hotel_agent = AssistantAgent(
13    name="HotelAgent",
14    model_client=OpenAIChatCompletionClient(
15        model="gpt-4",
16    ),
17    instructions="""
18You are an AI agent specialized in booking accommodations. Assist in finding hotels based on user preferences.
19""",
20    tools=[search_hotels],
21)

---

4. ActivityAgent

Suggests activities at the destination.

Code: activity_agent.py

pythonCopy
1# activity_agent.py
2
3import asyncio
4from autogen_agentchat.agents import AssistantAgent
5from autogen_ext.models import OpenAIChatCompletionClient
6
7async def suggest_activities(destination_city: str, interests: str):
8    # Mock implementation of activity suggestions
9    await asyncio.sleep(1)  # Simulate processing time
10    return f"Suggested activities in {destination_city} based on your interests ({interests}): Visit the museum, explore downtown, enjoy local cuisine."
11
12activity_agent = AssistantAgent(
13    name="ActivityAgent",
14    model_client=OpenAIChatCompletionClient(
15        model="gpt-4",
16    ),
17    instructions="""
18You are an AI agent specialized in suggesting activities and attractions. Provide recommendations based on user interests.
19""",
20    tools=[suggest_activities],
21)

---

Implementing the Main Program

We'll now create the main script that ties everything together.

Code: main.py

pythonCopy
1# main.py
2
3import asyncio
4import os
5from dotenv import load_dotenv
6from autogen_agentchat.agents import UserProxyAgent
7from autogen_agentchat.teams import SequentialTeam
8from autogen_agentchat.task import Console
9from autogen_ext.models import OpenAIChatCompletionClient
10
11# Import agents
12from flight_agent import flight_agent
13from hotel_agent import hotel_agent
14from activity_agent import activity_agent
15
16# Load environment variables
17load_dotenv()
18openai_api_key = os.getenv("OPENAI_API_KEY")
19
20# Ensure API key is set
21if not openai_api_key:
22    raise ValueError("OPENAI_API_KEY is not set in the environment variables.")
23
24# Set the API key for model clients
25flight_agent.model_client.api_key = openai_api_key
26hotel_agent.model_client.api_key = openai_api_key
27activity_agent.model_client.api_key = openai_api_key
28
29async def main():
30    # Create the user agent
31    user_agent = UserProxyAgent(
32        name="UserAgent",
33    )
34
35    # Define the travel planning team
36    travel_team = SequentialTeam(
37        agents=[
38            flight_agent,
39            hotel_agent,
40            activity_agent,
41        ],
42        user_agent=user_agent,
43    )
44
45    # Initial user message
46    user_message = input("You: ")
47
48    # Run the team
49    stream = travel_team.run_stream(task=user_message)
50    await Console(stream)
51
52if __name__ == "__main__":
53    asyncio.run(main())

---

Running the Application

1. Set Up Environment Variables

Create a .env file in your project directory:

bashCopy
1touch .env

Add your OpenAI API key to the .env file:

iniCopy
1# .env
2OPENAI_API_KEY=your_openai_api_key_here

Note: Replace your_openai_api_key_here with your actual API key.

2. Run the Application

bashCopy
1python main.py

3. Interact with the Travel Planner

Example Interaction:

textCopy
1You: I want to plan a trip to Paris from New York next month.
2
3FlightAgent: Found flights from New York to Paris departing on 2024-12-01 and returning on 2024-12-10.
4
5HotelAgent: Found hotels in Paris from 2024-12-01 to 2024-12-10.
6
7ActivityAgent: Suggested activities in Paris based on your interests (art, history): Visit the Louvre Museum, explore the Eiffel Tower, enjoy local French cuisine.

---

Conclusion

You've successfully built an AI Travel Planner using AutoGen! This application demonstrates how multiple AI agents can collaborate to perform complex tasks. Each agent specializes in a particular domain and communicates to provide a cohesive service to the user.

---

Additional Notes

• Asynchronous Programming: The use of asyncio allows agents to perform tasks concurrently.
• Mock Implementations: The functions search_flights, search_hotels, and suggest_activities are mock implementations. In a real-world application, you'd integrate with actual APIs.
• Error Handling: For production use, add proper error handling and input validation.
• Extensibility: You can extend this application by adding more agents, such as a CarRentalAgent or RestaurantAgent.

---

Happy Coding!