%pip install langchain langchain-openai python-dotenv --quiet

import os
from dotenv import load_dotenv
from langchain.agents import create_agent
from langchain_openai import ChatOpenAI
from langchain.tools import tool
from langchain_core.messages import HumanMessage

# Load environment variables from .env file
load_dotenv()

# Configure OpenRouter model
model = ChatOpenAI(
    api_key=os.getenv("OPENROUTER_API_KEY"),
    base_url="https://openrouter.ai/api/v1",
    model="gpt-4o-mini",
    temperature=0.7,
    max_tokens=1000,
)

# Create specialized logistics planning agent
logistics_agent = create_agent(
    model=model,
    system_prompt="""You are a travel logistics expert. You handle practical travel planning:
    - Calculate distances between locations and travel times
    - Estimate costs for transportation, accommodation, and activities
    - Optimize routes and suggest efficient itineraries
    - Consider time zones, weather, and practical constraints
    Always provide short, clear, practical logistics information."""
)

# Create specialized recommendations agent
recommendations_agent = create_agent(
    model=model,
    system_prompt="""You are a travel recommendations specialist. You suggest experiences and activities:
    - Recommend top attractions, landmarks, and must-see places
    - Suggest restaurants, local cuisine, and dining experiences
    - Recommend cultural activities, events, and local experiences
    - Provide insights about local customs, best times to visit, and hidden gems
    Always provide brief, engaging, personalized recommendations."""
)

@tool
def plan_logistics_agent(trip_request: str) -> str:
    """
    Plan travel logistics including distances, times, costs, and routes.
    Use this to calculate practical travel information and optimize itineraries.
    
    Args:
        trip_request: Trip details (e.g., "3 days in Paris, budget $1500, from London")
    
    Returns:
        Logistics information: distances, travel times, costs, and route suggestions
    """
    response = logistics_agent.invoke({"messages": [HumanMessage(f"Plan logistics for this trip: {trip_request}")]})
    return response["messages"][-1].content

@tool
def get_recommendations_agent(trip_details: str) -> str:
    """
    Get travel recommendations for attractions, restaurants, and activities.
    Use this to suggest what to see, do, and eat at the destination.
    
    Args:
        trip_details: Destination and trip information (e.g., "3 days in Paris, interested in art and food")
    
    Returns:
        Recommendations for attractions, restaurants, activities, and cultural insights
    """
    response = recommendations_agent.invoke({"messages": [HumanMessage(f"Provide recommendations for: {trip_details}")]})
    return response["messages"][-1].content

orchestrator = create_agent(
    model=model,
    system_prompt="""You are a travel planning coordinator. 
    When planning trips, use both specialists:
    1. Use plan_logistics_agent to calculate practical details: distances, times, costs, and routes
    2. Use get_recommendations_agent to suggest attractions, restaurants, and activities
    Always combine both the practical logistics and exciting recommendations in your final response.""",
    tools=[plan_logistics_agent, get_recommendations_agent]
)

# Test 1: City trip planning - uses both agents
response = orchestrator.invoke({"messages": [HumanMessage("Plan a 3-day trip to Rome. I'm coming from London with a budget of $2000. Calculate travel costs and time, and suggest must-see attractions and restaurants.")]})
print(response["messages"][-1].content)

# Test 2: Multi-city itinerary - both agents work together
response = orchestrator.invoke({"messages": [HumanMessage("I want to visit Paris, Amsterdam, and Berlin in 7 days starting from New York. Plan the logistics (flights, trains, costs) and recommend top 3 things to do in each city.")]})
print(response["messages"][-1].content)

# Test 3: Weekend getaway - comprehensive planning
response = orchestrator.invoke({"messages": [HumanMessage("Plan a weekend trip to Barcelona from Madrid. Budget is $500. Calculate travel time and costs, and suggest the best places to visit, eat, and experience local culture.")]})
print(response["messages"][-1].content)

Multi-Agent Systems: Agents as Tools¶

What Are Multi-Agent Systems?¶

The "Agents as Tools" Pattern¶

What You'll Build¶

Prerequisites¶

Quick Setup Checklist¶

Setup Environment¶

Installation¶

Setup¶

Scenario: Travel Planning System¶

System Architecture¶

Why This Architecture?¶

Create Specialized Agents¶

Convert Agents to Tools¶

How It Works¶

Design Pattern¶

Create Orchestrator Agent¶

Orchestrator Design¶

System Prompt Strategy¶

The Power of This Pattern¶

Test the Multi-Agent System¶

Understanding Multi-Agent Orchestration¶

Test Scenarios¶

What You've Learned¶

Key Benefits¶

How Multi-Agent Systems Work¶

Design Patterns You've Seen¶

Best Practices for Multi-Agent Systems¶

Extending This System¶

Next Steps¶

Experiment¶