Fix LiteLLM logging worker reset on proxy restart

[ultmaster]

Summary

• reset LiteLLM's global logging worker when the proxy boots so it binds to the fresh event loop
• add a regression test ensuring a restart replaces the logging worker singleton

Testing

• pytest tests/test_llm_proxy_restart.py

---

https://chatgpt.com/codex/tasks/task_e_68f4dab56bc4832e9e5ef8a9f053433f

[Copilot]

Pull Request Overview

This PR fixes a bug where LiteLLM's global logging worker would raise a RuntimeError when the LLM proxy is restarted, because the logging worker's asyncio queue remains bound to the old event loop after Uvicorn creates a new one.

• Adds a new function _reset_litellm_logging_worker() to recreate LiteLLM's global logging worker on proxy restart
• Calls the reset function in LLMProxy.start() to ensure the logging worker uses the fresh event loop
• Includes a regression test to verify the logging worker is properly replaced between restarts

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
agentlightning/llm_proxy.py	Adds logging worker reset functionality and calls it during proxy startup
tests/test_llm_proxy_restart.py	New test file verifying the logging worker is refreshed on restart

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

The bare except Exception is too broad and could mask unexpected errors. Consider catching more specific exceptions like ImportError or AttributeError that are expected when the module doesn't exist or the attribute is missing.

Suggested change

	except Exception: # pragma: no cover - best-effort hygiene
	except (ImportError, AttributeError): # pragma: no cover - best-effort hygiene

[ultmaster]

Exception:

the following exception happens sometimes when I try to restart llmproxy. Figure out what might be wrong here:

(APIServer pid=126879) INFO:     127.0.0.1:39246 - "POST /v1/chat/completions HTTP/1.1" 200 OK
Task exception was never retrieved
future: <Task finished name='Task-7390' coro=<LoggingWorker._worker_loop() done, defined at /mnt/vss/_work/agent-lightning/agent-lightning/.venv/lib/python3.13/site-packages/litellm/litellm_core_utils/logging_worker.py:57> exception=RuntimeError('<Queue at 0x783cbf633f20 maxsize=50000> is bound to a different event loop')>
Traceback (most recent call last):
  File "/mnt/vss/_work/agent-lightning/agent-lightning/.venv/lib/python3.13/site-packages/litellm/litellm_core_utils/logging_worker.py", line 65, in _worker_loop
    task = await self._queue.get()
           ^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/cloudtest/.local/share/uv/python/cpython-3.13.9-linux-x86_64-gnu/lib/python3.13/asyncio/queues.py", line 183, in get
    getter = self._get_loop().create_future()
             ~~~~~~~~~~~~~~^^
  File "/home/cloudtest/.local/share/uv/python/cpython-3.13.9-linux-x86_64-gnu/lib/python3.13/asyncio/mixins.py", line 20, in _get_loop
    raise RuntimeError(f'{self!r} is bound to a different event loop')
RuntimeError: <Queue at 0x783cbf633f20 maxsize=50000> is bound to a different event loop
INFO:     10.1.72.12:59254 - "POST /chat/completions HTTP/1.1" 200 OK
2025-10-19 12:24:09,412 [INFO] (Process-124055 agentlightning.store.client_server)   127.0.0.1:50034 - "GET /ge