Python: [Bug]: deepcopy fails on ChatAgent.run() when agent has tools - TypeError: cannot pickle '_thread.RLock' object

[droideronline]

Description

When using ChatAgent with tools and calling agent.run() or agent.run_stream(), the method fails with a TypeError because _prepare_thread_and_messages calls deepcopy(self.default_options) which contains AIFunction objects that hold references to OpenTelemetry Histogram objects containing threading.Lock.

Code Sample

from agent_framework import ChatAgent, ai_function

@ai_function
def my_tool(param: str) -> str:
    """A simple tool."""
    return f"Result: {param}"

# Create agent with tools
agent = ChatAgent(
    chat_client=some_chat_client,
    name="test-agent",
    tools=[my_tool],  # Any tool causes the issue
)

# This will fail
response = await agent.run("Hello")

Error Messages / Stack Traces

Traceback (most recent call last):
  File ".../agent_framework/_agents.py", line 806, in run
    thread, run_chat_options, thread_messages = await self._prepare_thread_and_messages(
  File ".../agent_framework/_agents.py", line 1229, in _prepare_thread_and_messages
    chat_options = deepcopy(self.default_options) if self.default_options else {}
  File "/usr/lib/python3.11/copy.py", line 172, in deepcopy
    y = _reconstruct(x, memo, *rv)
  ...
  File "/usr/lib/python3.11/copy.py", line 271, in _reconstruct
    state = deepcopy(state, memo)
  File "/usr/lib/python3.11/copy.py", line 146, in deepcopy
    y = copier(x, memo)
  File "/usr/lib/python3.11/copy.py", line 231, in _deepcopy_dict
    y[deepcopy(key, memo)] = deepcopy(value, memo)
  ...
TypeError: cannot pickle '_thread.RLock' object

Package Versions

agent-framework v1.0.0b260116

Python Version

No response

Additional Context

No response

[droideronline]

This only happens when OTel is enabled

[droideronline]

CC - @moonbox3

[MarShaikh]

Hi @droideronline quick question, does your my_tool use some object/method that might contain some HTTP connection pools, sessions or thread locks? I was facing similar issues with the same version and realised this to be the case. Might be worth looking into. To be transparent though, I was facing this issue with AzureOpenAIResponsesClient and creating agents using as_agent method.