from agno.agent import Agent
from agno.models.openai import OpenAIChat
from agno.os import AgentOS
from dotenv import load_dotenv
from agno.tools.mcp import MCPTools
from agno.db.sqlite import SqliteDb
db = SqliteDb(
db_file="tmp/ibmi_agent.db",
memory_table="ibmi_agent_memory",
session_table="ibmi_agent_sessions",
metrics_table="ibmi_agent_metrics",
eval_table="ibmi_agent_evals",
knowledge_table="ibmi_agent_knowledge"
)
mcp_tools = MCPTools(
transport="streamable-http",
url = "http://127.0.0.1:3010/mcp"
)
load_dotenv() # Load environment variables from .env file
assistant = Agent(
name="IBM i Agent",
model=OpenAIChat(id="gpt-4o"),
instructions=["You are a helpful assistant that helps users with IBM i related tasks."],
db=db,
tools=[mcp_tools],
markdown=True,
enable_agentic_memory=True,
enable_user_memories=True,
search_knowledge=True,
add_history_to_context=True,
read_chat_history=True,
debug_mode=True
)
agent_os = AgentOS(
os_id="ibmi-agentos",
description="IBM i AgentOS",
agents=[assistant]
)
app = agent_os.get_app()
if __name__ == "__main__":
# Default port is 7777; change with port=...
agent_os.serve(app="agentos:app", reload=True)
