"""AG2 (formerly AutoGen) agent wrapped with Bindu.
A simple Q&A agent using AG2's ConversableAgent, exposed as a
Bindu-compatible networked service with identity, communication,
and payment capabilities.
"""
import os
from autogen import ConversableAgent, LLMConfig
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-assistant",
"description": (
"A research assistant powered by AG2 (formerly AutoGen). "
"Ask any question and get a clear, concise answer."
),
"deployment": {
"url": "http://localhost:3773",
"expose": True,
"cors_origins": ["http://localhost:5173"],
},
"skills": ["skills/question-answering"],
}
def handler(messages: list[dict[str, str]]):
"""Forward messages to the AG2 agent and return its response."""
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
agent = ConversableAgent(
name="assistant",
system_message=(
"You are a helpful research assistant. Provide "
"clear, concise answers to questions. When you "
"don't know something, say so."
),
llm_config=llm_config,
)
user_proxy = ConversableAgent(
name="user", human_input_mode="NEVER",
)
result = user_proxy.initiate_chat(
agent, message=user_input, max_turns=1
)
if not result.chat_history:
return [{"role": "assistant", "content": "No response."}]
reply = result.chat_history[-1].get("content", "")
return [{"role": "assistant", "content": reply}]
if __name__ == "__main__":
bindufy(config, handler)
