"""AG2 (formerly AutoGen) multi-agent research team wrapped with Bindu.
Three specialized agents collaborate under GroupChat with LLM-driven
speaker selection to produce structured research reports. The team is
exposed as a single Bindu agent, callable via A2A protocol.
Architecture:
User (via Bindu A2A) → handler → AG2 GroupChat
├── researcher — gathers information
├── analyst — evaluates findings
└── writer — produces the final report
"""
import os
from autogen import ConversableAgent, LLMConfig
from autogen.agentchat import initiate_group_chat
from autogen.agentchat.group.patterns import AutoPattern
from bindu.penguin import bindufy
from dotenv import load_dotenv
load_dotenv()
llm_config = LLMConfig(
{
"model": os.getenv("LLM_MODEL", "openai/gpt-4o-mini"),
"api_key": os.environ.get("OPENROUTER_API_KEY", ""),
"base_url": "https://openrouter.ai/api/v1",
}
)
config = {
"author": "ag2-community",
"name": "ag2-research-team",
"description": (
"A multi-agent research team powered by AG2 (formerly "
"AutoGen). Three specialists — researcher, analyst, and "
"writer — collaborate to produce structured reports on "
"any topic."
),
"deployment": {
"url": "http://localhost:3773",
"expose": True,
"cors_origins": ["http://localhost:5173"],
},
"skills": ["skills/question-answering"],
}
def handler(messages: list[dict[str, str]]):
"""Run the AG2 research team and return the final report."""
if not messages:
return [{"role": "assistant", "content": "No input provided."}]
user_input = messages[-1].get("content", "")
if not user_input:
return [{"role": "assistant", "content": "Empty message."}]
# Fresh agents per request to avoid state leakage
researcher = ConversableAgent(
name="researcher",
system_message=(
"You are a research specialist. Investigate the "
"given topic thoroughly. Present key facts, data "
"points, and sources in a structured format."
),
llm_config=llm_config,
)
analyst = ConversableAgent(
name="analyst",
system_message=(
"You are an analyst. Review the researcher's "
"findings. Identify trends, implications, risks, "
"and opportunities. Be quantitative where possible."
),
llm_config=llm_config,
)
writer = ConversableAgent(
name="writer",
system_message=(
"You are a report writer. Once the researcher and "
"analyst have contributed, synthesize their work "
"into a structured report with: Executive Summary, "
"Key Findings, Analysis, and Recommendations. "
"Keep it under 500 words. End with TERMINATE."
),
llm_config=llm_config,
)
user = ConversableAgent(
name="user", human_input_mode="NEVER"
)
pattern = AutoPattern(
initial_agent=researcher,
agents=[researcher, analyst, writer],
user_agent=user,
group_manager_args={"llm_config": llm_config},
)
result, ctx, last = initiate_group_chat(
pattern=pattern,
messages=f"Research and produce a report on: {user_input}",
max_rounds=10,
)
# Extract the last substantive message
for msg in reversed(result.chat_history):
content = msg.get("content", "")
if content and "TERMINATE" not in content:
return [{"role": "assistant", "content": content}]
return [{"role": "assistant", "content": "Research complete."}]
if __name__ == "__main__":
bindufy(config, handler)
