Quick Start

Welcome to Azcore! This guide will get you building your first multi-agent system in under 5 minutes. We'll create a simple agent team that can answer questions and use tools.

🔑 2. Set Up Your API Key

Create a .env file in your project directory:

# .env
OPENAI_API_KEY=your-openai-api-key-here

⚡ 3. Create Your First Agent Team

Create a file called quick_start.py:

from azcore import Config, TeamBuilder, GraphOrchestrator
from langchain_core.tools import tool

# Load configuration
config = Config.from_env()
llm = config.get_llm()

# Define a simple tool
@tool
def search_web(query: str) -> str:
    """Search the web for information."""
    return f"Found information about: {query}"

@tool
def calculate(math_expression: str) -> str:
    """Calculate a mathematical expression."""
    try:
        result = eval(math_expression)
        return f"Result: {result}"
    except:
        return "Invalid expression"

# Create an agent team
agent_team = (TeamBuilder("assistant_team")
    .with_llm(llm)
    .with_tools([search_web, calculate])
    .with_prompt("You are a helpful assistant that can search the web and do calculations.")
    .build())

# Set up a simple orchestrator
orchestrator = GraphOrchestrator()
orchestrator.add_team(agent_team)
orchestrator.set_entry_point("assistant_team")
orchestrator.add_edge("assistant_team", "END")

# Compile the workflow
graph = orchestrator.compile()

# Test it!
result = graph.invoke({
    "messages": [{"role": "user", "content": "What is 15 + 27? Also search for information about Python programming."}]
})

print("Agent Response:")
print(result["messages"][-1].content)

▶️ 4. Run Your Agent

Execute the script:

python quick_start.py

You should see output like:

Agent Response:
The sum of 15 + 27 is 42. Regarding Python programming, Python is a high-level programming language known for its simplicity and readability...

🎯 5. Next Steps

Congratulations! You've just created your first multi-agent system. Here's what you can do next:

Add More Tools

Extend your agent with more capabilities:

@tool
def get_weather(city: str) -> str:
    """Get weather information for a city."""
    return f"Weather in {city}: Sunny, 72°F"

# Add to your team
agent_team = (TeamBuilder("assistant_team")
    .with_llm(llm)
    .with_tools([search_web, calculate, get_weather])  # Add new tool
    .with_prompt("You are a helpful assistant with web search, calculation, and weather capabilities.")
    .build())

Create Multiple Teams

Build a more complex system with specialized teams:

from azcore import Supervisor

# Create specialized teams
research_team = (TeamBuilder("research_team")
    .with_llm(llm)
    .with_tools([search_web])
    .with_prompt("You are a research specialist.")
    .build())

math_team = (TeamBuilder("math_team")
    .with_llm(llm)
    .with_tools([calculate])
    .with_prompt("You are a mathematics specialist.")
    .build())

# Create supervisor to coordinate teams
supervisor = Supervisor(
    llm=llm,
    members=["research_team", "math_team"]
)

# Build hierarchical orchestrator
orchestrator = GraphOrchestrator()
orchestrator.set_supervisor(supervisor)
orchestrator.add_team(research_team)
orchestrator.add_team(math_team)
orchestrator.set_entry_point("supervisor")
orchestrator.add_edge("supervisor", "END")

graph = orchestrator.compile()

Enable Reinforcement Learning

Make your agents learn optimal tool usage:

from azcore.rl import RLManager, HeuristicRewardCalculator

# Set up RL
rl_manager = RLManager(
    tool_names=["search_web", "calculate"],
    q_table_path="rl_data/agent_q_table.pkl"
)
reward_calc = HeuristicRewardCalculator()

# Create RL-enabled team
smart_team = (TeamBuilder("smart_team")
    .with_llm(llm)
    .with_tools([search_web, calculate])
    .with_prompt("You are an intelligent assistant.")
    .with_rl(rl_manager, reward_calc)  # Enable learning!
    .build())

🛠️ Troubleshooting

Common Issues

"ModuleNotFoundError: No module named 'azcore'"

pip install azcore

"AuthenticationError: OpenAI API key required"

• Check your .env file has the correct API key
• Make sure the .env file is in your current directory

"Tool execution failed"

• Verify your tools are properly defined with the @tool decorator
• Check that tool functions return strings

Need Help?

• Check the Installation Guide for detailed setup
• Visit GitHub Issues for community support

🎉 You're All Set!

You've successfully created your first Azcore multi-agent system! The framework provides:

• Hierarchical Architecture - Coordinator-planner-supervisor-team design
• Flexible Workflows - Sequential, concurrent, swarm, and graph-based patterns
• Reinforcement Learning - Agents that learn optimal tool usage
• Production Ready - Caching, error handling, and monitoring built-in