[Bug] Error: cannot pickle '_thread.RLock' object when using OpenAIChat with Memori + Agno (Streaming Mode)

[makramab]

First Check

• I added a very descriptive title to this issue.
• I searched existing issues and documentation.

Memori Version

3.0.6

OS / Python Version

macOS

LLM Provider

OpenAI

LLM Model & Version

gpt-5-mini

Database

SQLite

Description

When using Memori (v3.0.6) with Agno (v2.3.8) and running an agent with OpenAIChat in streaming mode (Agent.arun(..., streaming=True)), the process consistently throws:

Error from OpenAI API: cannot pickle '_thread.RLock' object
Model provider error after 1 attempts: cannot pickle '_thread.RLock' object

This issue does not occur when switching the model provider to Gemini, suggesting the bug is specific to the OpenAIChat integration.

---

Environment

• Memori: 3.0.6
• Agno: 2.3.8
• Python: (same environment using uv run)
• Model Provider: OpenAIChat → ❌ error
• Alternative provider (Gemini) → ✅ works correctly
• Running example based on Agno documentation:
  https://docs.agno.com/integrations/memory/memori#memori

---

Steps to Reproduce

1. Install latest Memori (3.0.6) and Agno (2.3.8).

2. Use the code example from Agno’s documentation and modify the call to:

      async for chunk in agent.arun(
        "Hi, I'd like to order a large pepperoni pizza with extra cheese", stream=True
    ):
        if chunk.event == RunEvent.run_content:
            print(chunk.content, end="", flush=True)
    print("\n")

3. Run the file:

   uv run python main.py
   

4. Observe the error after the first user message.

---

Expected Behavior

• Agent should stream responses normally when using OpenAIChat.
• Behavior should match Gemini provider (which works as expected).

---

Actual Behavior

Execution fails with:

Error from OpenAI API: cannot pickle '_thread.RLock' object
Model provider error after 1 attempts: cannot pickle '_thread.RLock' object
Traceback (most recent call last):
   ...

The error appears immediately after the agent receives the first instruction.

---

Additional Notes

Here's my full code that I use

import asyncio
import os

from agno.agent import Agent, RunOutput
from agno.models.google import Gemini
from agno.models.openai import OpenAIChat
from agno.run.agent import RunEvent
from dotenv import load_dotenv
from sqlalchemy import create_engine
from sqlalchemy.orm import sessionmaker

from memori import Memori

load_dotenv()

db_path = os.getenv("DATABASE_PATH", "memori_agno.db")
engine = create_engine(f"sqlite:///{db_path}")
Session = sessionmaker(bind=engine)

model = Gemini(
    id="gemini-2.5-flash",
)
model = OpenAIChat(
    id="gpt-5-mini",
)

# mem = Memori(conn=Session).agno.register(gemini=model)
mem = Memori(conn=Session).agno.register(openai_chat=model)
mem.attribution(entity_id="customer-456", process_id="support-agent")
mem.config.storage.build()

agent = Agent(
    model=model,
    instructions=[
        "You are a helpful customer support agent.",
        "Remember customer preferences and history from previous conversations.",
    ],
    markdown=True,
)


async def main():
    print("Customer: Hi, I'd like to order a large pepperoni pizza with extra cheese")
    print("Agent: ", end="")
    async for chunk in agent.arun(
        "Hi, I'd like to order a large pepperoni pizza with extra cheese", stream=True
    ):
        if chunk.event == RunEvent.run_content:
            print(chunk.content, end="", flush=True)
    print("\n")

    print("Customer: Actually, can you remind me what I just ordered?")
    print("Agent: ", end="")
    async for chunk in agent.arun(
        "Actually, can you remind me what I just ordered?", stream=True
    ):
        if chunk.event == RunEvent.run_content:
            print(chunk.content, end="", flush=True)
    print("\n")

    print("Customer: Perfect! And what size was that again?")
    print("Agent: ", end="")
    async for chunk in agent.arun(
        "Perfect! And what size was that again?", stream=True
    ):
        if chunk.event == RunEvent.run_content:
            print(chunk.content, end="", flush=True)
    print()


if __name__ == "__main__":
    asyncio.run(main())