import asyncio
import logging
import os
from dotenv import load_dotenv
from pydantic_ai import Agent
from pydantic_ai.mcp import MCPServerStdio
logger = logging.getLogger(__name__)
load_dotenv()
model_identifier = os.getenv("MODEL_IDENTIFIER", "openai:gpt-5-mini")
logger.info(f"Creating agent with model: {model_identifier}")
try:
server = MCPServerStdio(
"uv",
args=["run", "agentic-investor", "stdio"],
timeout=60,
env=dict(os.environ),
)
agent = Agent(model_identifier, toolsets=[server])
agent.set_mcp_sampling_model() # Allows MCP server to make LLM calls via the MCP client
logger.info("Agent created successfully")
except Exception as e:
logger.error(f"Failed to create Investor agent: {e}")
raise
async def main():
print("Chat with Investor Agent (type 'quit' to exit)")
async with agent:
result = None
while True:
try:
user_input = input("\nš¤: ").strip()
if user_input.lower() in ["quit", "exit", "q"]:
logger.info("Chat session ended by user")
break
if not user_input:
continue
logger.debug(f"Processing user input: {user_input}")
result = await agent.run(
user_input,
message_history=result.new_messages() if result else None,
)
print(f"š¤: {result.output}")
except (KeyboardInterrupt, EOFError):
logger.info("Chat session interrupted by user")
break
except Exception as e:
logger.error(f"Error processing user input '{user_input}': {e}")
print(f"ā {e}")
continue
if __name__ == "__main__":
asyncio.run(main())
