pip install agentcp agently python-dotenv
{
"OPENAI_API_KEY": "your_api_key",
"BASE_URL": "https://api.example.com/v1",
"MODEL": "qwen-72b-chat"
}
import agentcp
import Agently
import os
from dotenv import load_dotenv
load_dotenv()
if __name__ == "__main__":
agent_id = 'your_agent_id'
acp = agentcp.AgentCP('.', seed_password='')
aid = acp.load_aid(agent_id)
openai_api_key = os.getenv("OPENAI_API_KEY")
model_url = os.getenv("BASE_URL")
model_name = os.getenv("MODEL")
agent = (
Agently.create_agent()
.set_settings("current_model", "OAIClient")
.set_settings("model.OAIClient.auth", {'api_key': openai_api_key})
.set_settings("model.OAIClient.url", model_url)
.set_settings("model.OAIClient.options", {'model': model_name})
)
@aid.message_handler() #消息处理函数
async def sync_message_handler(msg):
receiver = aid.get_receiver_from_message(msg)
if aid.id not in receiver:
return
session_id = aid.get_session_id_from_message(msg)
sender = aid.get_sender_from_message(msg)
sender_content = aid.get_content_from_message(msg)
result = (
agent
.general("输出规定", "必须使用中文进行输出")
.role({
"姓名": "ACP小助手",
"任务": "使用自己的知识为用户解答常见问题",
})
.input(sender_content)
.instruct(["你需要根据用户的问题提供相关的回答", "你可以适当的有点幽默"])
.start()
)
aid.send_message_content(to_aid_list=[sender], session_id=session_id, llm_content=result)
return True
aid.online()
acp.serve_forever()
