OpenAI Agents SDK Memory Integration

pip install openai-agents mcp-client-cli python-dotenv

import os
import asyncio
from dotenv import load_dotenv
from agents import Agent, Runner, function_tool
from mcp_client import StdioMCPClient

load_dotenv()

# Define global MCP client instance
mcp_client = StdioMCPClient(command="npx", args=["-y", "@sovseal/mcp-server"])

# 1. Define memory search tool
@function_tool
async def search_memory(query: str, user_id: str) -> str:
    """Search through past conversations and memories for user preferences."""
    response = await mcp_client.call_tool("recall_memory", {"query": query, "topK": 3})
    results = response.get("content", [])
    if results:
        return "\n".join([f"- {r.get('text', '')}" for r in results])
    return "No relevant memories found."

# 2. Define memory save tool
@function_tool
async def save_memory(content: str, user_id: str) -> str:
    """Save important facts or preferences about the user to persistent memory."""
    await mcp_client.call_tool("store_memory", {"content": content})
    return "Information saved successfully."

# 3. Create specialized agents
travel_agent = Agent(
    name="Travel Planner",
    instructions="""You are a travel planning specialist. Always query search_memory first
    to see if there are travel preferences or past destinations stored for the user.""",
    tools=[search_memory, save_memory],
    model="gpt-4o"
)

health_agent = Agent(
    name="Health Advisor",
    instructions="""You are a health and wellness advisor. Query search_memory first
    to see if there are dietary restrictions or health preferences stored for the user.""",
    tools=[search_memory, save_memory],
    model="gpt-4o"
)

# 4. Create triage agent with handoffs
triage_agent = Agent(
    name="Personal Assistant",
    instructions="""You are a triage assistant. Route the user to the Travel Planner or
    Health Advisor based on their request. Use available memory tools to personalize interactions.""",
    handoffs=[travel_agent, health_agent],
    model="gpt-4o"
)

async def chat_session(user_input: str, user_id: str):
    async with mcp_client:
        # Run the agent execution loop
        result = await Runner.run_async(triage_agent, user_input, user_id=user_id)
        
        # Save the current interaction summary
        await mcp_client.call_tool(
            "store_memory",
            {"content": f"User '{user_id}' requested: '{user_input}'. Agent responded: '{result.final_output}'"}
        )
        
        return result.final_output

if __name__ == "__main__":
    response = asyncio.run(chat_session("Plan a healthy meal for my trip to Rome", "alice_12"))
    print("Assistant Response:", response)