Multi-Agent Systems — Agents Collaborating

import autogen
from autogen import AssistantAgent, UserProxyAgent, GroupChat, GroupChatManager

# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
# MULTI-AGENT SYSTEM WITH AUTOGEN
# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

config_list = [{"model": "gpt-4o-mini", "api_key": "your-key"}]
llm_config = {"config_list": config_list, "temperature": 0.1}

# Specialist agents
planner = AssistantAgent(
    name="Planner",
    llm_config=llm_config,
    system_message='''You are a project planner. Break down tasks into steps.
Coordinate with Engineer, Critic, and Writer.
Synthesize the final output.''',
)

engineer = AssistantAgent(
    name="Engineer",
    llm_config=llm_config,
    system_message='''You are a Python software engineer.
Write clean, well-tested, production-ready code.
When writing code, check for edge cases and error handling.''',
)

critic = AssistantAgent(
    name="Critic",
    llm_config=llm_config,
    system_message='''You are a code reviewer and quality assurance agent.
Review code for bugs, security issues, performance, and correctness.
Flag issues clearly with severity (HIGH/MEDIUM/LOW).''',
)

writer = AssistantAgent(
    name="Writer",
    llm_config=llm_config,
    system_message='''You are a technical writer.
Write clear documentation, docstrings, README files.
Make complex technical concepts understandable.''',
)

# User proxy executes code automatically
user_proxy = UserProxyAgent(
    name="UserProxy",
    human_input_mode="NEVER",  # fully autonomous
    max_consecutive_auto_reply=5,
    code_execution_config={
        "work_dir": "./agent_workspace",
        "use_docker": False,
    },
    is_termination_msg=lambda msg: "TASK_COMPLETE" in msg.get("content", ""),
)

# Group chat: all agents can speak and hear each other
group_chat = GroupChat(
    agents=[planner, engineer, critic, writer, user_proxy],
    messages=[],
    max_round=15,
    speaker_selection_method="auto",  # GPT-4 selects who speaks next
)

manager = GroupChatManager(groupchat=group_chat, llm_config=llm_config)

# Start collaborative task
user_proxy.initiate_chat(
    manager,
    message='''Build a production-ready Python API endpoint that:
1. Accepts a CSV file upload with columns: date, price, volume
2. Returns: trend (bullish/bearish/neutral), simple moving average (7-day), volatility score
3. Include error handling, input validation, type hints
4. Write comprehensive docstrings and usage examples
When done, output TASK_COMPLETE'''
)

# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
# CREWAI ALTERNATIVE -- simpler, role-based
# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
from crewai import Agent, Task, Crew, Process

researcher = Agent(
    role="Research Analyst",
    goal="Find comprehensive, accurate information about the given topic",
    backstory="Expert researcher with 10 years experience in data analysis",
    verbose=True, allow_delegation=False,
)

writer_crew = Agent(
    role="Content Writer",
    goal="Write engaging, accurate content based on research",
    backstory="Experienced technical writer who makes complex topics clear",
    verbose=True, allow_delegation=False,
)

research_task = Task(description="Research the latest advances in quantum computing", agent=researcher, expected_output="Comprehensive research report with key facts and citations")
writing_task  = Task(description="Write a blog post explaining quantum computing to beginners", agent=writer_crew, expected_output="800-word engaging blog post")

crew = Crew(
    agents=[researcher, writer_crew],
    tasks=[research_task, writing_task],
    process=Process.sequential,   # or Process.hierarchical for manager-based
    verbose=True,
)
result = crew.kickoff()