group-chat

You are creating an AG2 multi-agent group chat workflow. Follow these patterns exactly.

Instructions

1. Ask the user for:

   • What task the group needs to solve
   • How many agents and their specializations
   • Speaker selection pattern: auto (LLM picks), round_robin, or manual handoff
   • Maximum conversation rounds (default: 10)
   • Termination condition

2. Create the group chat following this pattern:

Group Chat Pattern

from autogen import ConversableAgent, GroupChat, GroupChatManager

# --- Define Specialist Agents ---

researcher = ConversableAgent(
    name="Researcher",
    description="Finds and synthesizes information",
    system_message="""You are a research specialist.
- Search for relevant information
- Summarize findings clearly
- Cite sources when possible
When your research is complete, say TERMINATE.""",
    llm_config={"model": "gpt-4o-mini"},
)

analyst = ConversableAgent(
    name="Analyst",
    description="Analyzes data and draws conclusions",
    system_message="""You are an analytical specialist.
- Analyze information provided by other agents
- Identify patterns and insights
- Provide evidence-based conclusions""",
    llm_config={"model": "gpt-4o-mini"},
)

writer = ConversableAgent(
    name="Writer",
    description="Drafts clear, structured output",
    system_message="""You are a writing specialist.
- Take research and analysis from other agents
- Produce clear, well-structured output
- When the final output is ready, say TERMINATE.""",
    llm_config={"model": "gpt-4o-mini"},
)

# --- Configure Group Chat ---

group_chat = GroupChat(
    agents=[researcher, analyst, writer],
    messages=[],
    max_round=10,
    speaker_selection_method="auto",  # "auto", "round_robin", "random", or callable
)

manager = GroupChatManager(
    groupchat=group_chat,
    llm_config={"model": "gpt-4o-mini"},
)

# --- Run the Group Chat ---

result = researcher.initiate_chat(
    manager,
    message="Your task description here",
)