RuntimeError: no running event loop when using cl.make_async alongside cl.instrument_openai()

[dominpm]

Description

When cl.instrument_openai() is enabled, any OpenAI request goes through Chainlit’s instrumentation. However, if that request is made within a function wrapped by cl.make_async(...), the instrumentation eventually calls asyncio.create_task() from a background thread (which lacks a running event loop). As a result, the following error occurs:

Error: RuntimeError: no running event loop

2025-02-06 11:06:06 - no running event loop
Traceback (most recent call last):
  File "C:\path\to\env\Lib\site-packages\chainlit\utils.py", line 44, in wrapper
    return await user_function(**params_values)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\path\to\env\Lib\site-packages\chainlit\callbacks.py", line 118, in with_parent_id
    await func(message)
  File "C:\path\to\project\app.py", line 1052, in main
    await for_custom_starter_translate_doc()
  File "C:\path\to\project\app.py", line 601, in for_custom_starter_translate_doc
    translated_file = await cl.make_async(translate_file)(
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\path\to\env\Lib\site-packages\asyncer_main.py", line 365, in wrapper
    return await anyio.to_thread.run_sync(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\path\to\env\Lib\site-packages\anyio\to_thread.py", line 56, in run_sync
    return await get_async_backend().run_sync_in_worker_thread(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\path\to\env\Lib\site-packages\anyio_backends_asyncio.py", line 2461, in run_sync_in_worker_thread
    return await future
           ^^^^^^^^^^^^
  File "C:\path\to\env\Lib\site-packages\anyio_backends_asyncio.py", line 962, in run
    result = context.run(func, *args)
             ^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\path\to\project\custom_starters\translate.py", line 143, in translate_file
    translated_text = translate_text(text, model_name=model_name, target_language=target_language)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\path\to\project\custom_starters\translate.py", line 50, in translate_text
    response = translation_chain.invoke({"text_to_translate": text})
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\path\to\env\Lib\site-packages\langchain_core\runnables\base.py", line 3016, in invoke
    input = context.run(step.invoke, input, config)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\path\to\env\Lib\site-packages\langchain_core\language_models\chat_models.py", line 284, in invoke
    self.generate_prompt(
  File "C:\path\to\env\Lib\site-packages\langchain_core\language_models\chat_models.py", line 860, in generate_prompt
    return self.generate(prompt_messages, stop=stop, callbacks=callbacks, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\path\to\env\Lib\site-packages\langchain_core\language_models\chat_models.py", line 690, in generate
    self._generate_with_cache(
  File "C:\path\to\env\Lib\site-packages\langchain_core\language_models\chat_models.py", line 925, in _generate_with_cache
    result = self._generate(
             ^^^^^^^^^^^^^^^
  File "C:\path\to\env\Lib\site-packages\langchain_openai\chat_models\base.py", line 781, in generate
    response = self.client.create(**payload)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\path\to\env\Lib\site-packages\literalai\wrappers.py", line 66, in wrapped
    result = after_func(result, context, *args, **kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\path\to\env\Lib\site-packages\literalai\instrumentation\openai.py", line 362, in after
    on_new_generation(
  File "C:\path\to\env\Lib\site-packages\chainlit\openai_init.py", line 51, in on_new_generation
    asyncio.create_task(step.send())
  File "C:\path\to\env\Lib\asyncio\tasks.py", line 381, in create_task
    loop = events.get_running_loop()
           ^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: no running event loop

Steps to Reproduce

1. Enable instrumentation with cl.instrument_openai().
2. Create a function that makes an OpenAI API call (e.g., via a LangChain Chain).
3. Wrap that function with cl.make_async(...), causing it to run in a worker thread.
4. Call the wrapped function within an async Chainlit handler.
5. Observe that the call fails with RuntimeError: no running event loop.

---

Expected Behavior

Calls to an OpenAI endpoint (instrumented by Chainlit) within a background thread should either succeed or bypass streaming instrumentation gracefully without causing a runtime error.

---

Actual Behavior

Chainlit’s instrumentation tries to schedule an async task (asyncio.create_task(...)) in a worker thread, where no event loop is running. This raises RuntimeError: no running event loop.

---

Environment

• Chainlit version: 1.3.2

I believe the root cause is that anyio.to_thread.run_sync does not share the main event loop with the worker thread. Meanwhile, Chainlit’s instrumentation tries to stream or log messages by calling asyncio.create_task within that thread. This triggers the no running event loop error.