memory_agents/memo-openai_agents.py

# --------------------------------------------------------------
#  chat_with_memory_redis.py   (100% FIXED – Sync, Simple, Works)
# --------------------------------------------------------------
import os
from agents import Agent, Runner, function_tool
from mem0 import Memory
from dotenv import load_dotenv

# ---- Load .env ----
load_dotenv()
os.environ["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY")

# ---- Redis Mem0 Config (EXACT from your working comment) ----
REDIS_URL = os.getenv(
    "REDIS_URL",
    "redis://default:FO6JF6mPJ0bgPj1Rpof84OtYyZUeIxUm@redis-17714.c81.us-east-1-2.ec2.redns.redis-cloud.com:17714"
)

mem0_config = {
    "vector_store": {
        "provider": "redis",
        "config": {
            "collection_name": "mem0_chat",
            "embedding_model_dims": 1536,
            "redis_url": REDIS_URL,
        }
    },
    "version": "v1.1"
}

# ---- FIXED: Use from_config (matches your commented code) ----
memory = Memory.from_config(mem0_config)
print("✅ Mem0 + Redis ready")

# --------------------------------------------------------------
#  SYNC Memory Tools (no async – simple & fast)
# --------------------------------------------------------------
@function_tool
def search_memory(query: str, user_id: str) -> str:
    """Search user's memories in Redis."""
    results = memory.search(query, user_id=user_id, limit=5)
    if results and results.get("results"):
        return "\n".join(f"- {r['memory']}" for r in results["results"])
    return "No memories found."

@function_tool
def save_memory(content: str, user_id: str) -> str:
    """Save fact to Redis memory."""
    memory.add(content, user_id=user_id)
    return "Saved to memory."

# --------------------------------------------------------------
#  Personal Assistant Agent
# --------------------------------------------------------------
def personal_assistant(user_id: str) -> Agent:
    return Agent(
        name="Assistant",
        instructions=f"""You are a helpful assistant with long-term memory.

**Always follow this order:**
1. Call `search_memory` with the query and user_id="{user_id}".
2. Personalize your answer using any memories found.
3. Answer the question clearly.
4. If user shares a new fact/preference, call `save_memory` with a short summary and same user_id.

Only use tool results – no guessing.
""",
        model="gpt-4o-mini",
        tools=[search_memory, save_memory],
    )

# --------------------------------------------------------------
#  MAIN FUNCTION: chat_with_memory (SYNC – call directly!)
# --------------------------------------------------------------
def chat_with_memory(query: str, user_id: str) -> str:
    """
    Chat with Redis-memory agent. Returns final response as string.

    >>> chat_with_memory("I love pizza", "alice")
    >>> chat_with_memory("What food do I like?", "alice")  # Recalls pizza!
    """
    agent = personal_assistant(user_id)
    
    try:
        result = Runner.run_sync(agent, query)  # FIXED: run_sync (from your original)
        return result.final_output.strip()
    except Exception as e:
        return f"Error: {e}"

# --------------------------------------------------------------
#  Test it!
# --------------------------------------------------------------
if __name__ == "__main__":
    # # Save fact
    # print(chat_with_memory("I love pizza and live in New York", "alice"))
    
    # # Recall
    # print(chat_with_memory("Where do I live?", "alice"))
    # print(chat_with_memory("What food do I like?", "alice"))
    while True:
        query = input("You: ")
        response = chat_with_memory(query, "alice")
        print("Assistant:", response)