Fix Asynchronous Event Loop Blocking in Mesa-LLM Parallel Stepping

[Nandhan-Golla]

Summary

Fixes asynchronous event loop blocking when executing synchronous agent.step() functions and synchronous tool calls during parallel agent execution.

Bug / Issue

When utilizing parallel_stepping.py's asynchronous concurrency features in Mesa-LLM, standard Mesa Agents invoking the synchronous agent.step() method were doing so directly inside the asyncio event loop. Similarly, non-async tool functions parsed inside tools/tool_manager.py were invoked synchronously. Because Python asyncio runs on a single thread, any long-running or computationally expensive synchronous method call fully blocked the entire event loop, stalling all agents from operating concurrently and adding significant latency.

Implementation

• mesa_llm/parallel_stepping.py: Updated _sync_step function to use await asyncio.to_thread(agent.step) instead of blindly calling agent.step().
• mesa_llm/tools/tool_manager.py: Updated _process_tool_call to check if a function is synchronous. If so, it invokes the tool using await asyncio.to_thread(function_to_call, **filtered_args).

Code Changes:

diff --git a/mesa_llm/parallel_stepping.py b/mesa_llm/parallel_stepping.py
--- a/mesa_llm/parallel_stepping.py
+++ b/mesa_llm/parallel_stepping.py
@@ -33,7 +33,7 @@ async def step_agents_parallel(agents: list[Agent | LLMAgent]) -> None:
 
 async def _sync_step(agent: Agent) -> None:
     """Run synchronous step in async context."""
-    agent.step()
+    await asyncio.to_thread(agent.step)
 
 
 def step_agents_multithreaded(agents: list[Agent | LLMAgent]) -> None:
diff --git a/mesa_llm/tools/tool_manager.py b/mesa_llm/tools/tool_manager.py
--- a/mesa_llm/tools/tool_manager.py
+++ b/mesa_llm/tools/tool_manager.py
@@ -137,7 +137,7 @@ class ToolManager:
             if inspect.iscoroutinefunction(function_to_call):
                 function_response = await function_to_call(**filtered_args)
             else:
-                function_response = function_to_call(**filtered_args)
+                function_response = await asyncio.to_thread(function_to_call, **filtered_args)
 
             # Only treat None as empty
             if function_response is None:

Testing

• Verified by running the entire test_parallel_stepping.py and test_tools test directories using pytest.
• 66 comprehensive tests passed successfully locally with zero regressions.

Pytest Output:

$ uv run pytest tests/test_parallel_stepping.py tests/test_tools/
========================= test session starts ==========================
platform linux -- Python 3.12.3, pytest-9.0.2, pluggy-1.6.0
rootdir: /home/hari/gsoc/mesa-llm
configfile: pyproject.toml
plugins: anyio-4.8.0, html-4.1.1, cov-7.0.0, asyncio-1.3.0, metadata-3.1.1, mock-3.15.1
asyncio: mode=Mode.STRICT, openai_api_base=None, openai_api_key=None
collected 66 items                                                     

tests/test_parallel_stepping.py ....                             [  6%]
tests/test_tools/test_tool_decorator.py ..........               [ 21%]
tests/test_tools/test_tool_manager.py .......................... [ 60%]
..........................                                       [100%]

========================= warnings summary =========================
tests/test_parallel_stepping.py::test_step_agents_parallel
tests/test_parallel_stepping.py::test_step_agents_multithreaded
tests/test_parallel_stepping.py::test_automatic_parallel_shuffle_do
tests/test_parallel_stepping.py::test_step_agents_parallel_sync_in_running_loop
  /home/hari/gsoc/mesa-llm/.venv/lib/python3.12/site-packages/mesa/mesa_logging.py:112: FutureWarning: The use of the `seed` keyword argument is deprecated, use `rng` instead. No functional changes.
    res = func(*args, **kwargs)

-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
================== 66 passed, 4 warnings in 0.42s ==================

Additional Notes

This correctly delegates blocking sync I/O calls to background threads, yielding control back to the asyncio event loop and providing major latency and concurrency improvements in large Mesa agent simulations.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: c7ce6580-9629-4219-9541-51bd1e733998

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}