[Bug]: No module named 'opentelemetry.exporter.otlp.proto.grpc'

[asmaier]

What happened?

Because of the bug #13081 I'm running litellm<1.75 (1.74.15.post2).
I'm setting up langfuse with litellm like

langfuse = get_langfuse_configuration()
if langfuse["enabled"]:
    import os
    import litellm

    os.environ["LANGFUSE_HOST"] = langfuse["host"]
    os.environ["LANGFUSE_PUBLIC_KEY"] = langfuse["public_key"]
    os.environ["LANGFUSE_SECRET_KEY"] = fetch_langfuse_secret_key()

    # see https://github.com/BerriAI/litellm/issues/11500
    litellm.success_callback = ["langfuse_otel"]
    litellm.failure_callback = ["langfuse_otel"]

Since the newest version of langfuse 3.2.5 today I see the error No module named 'opentelemetry.exporter.otlp.proto.grpc'

I can avoid that error by setting "langfuse<3.2.5". So maybe this is caused by my specific combination of litellm and langfuse versions.

Relevant log output

15:18:38 - LiteLLM:ERROR: litellm_logging.py:3554 - [Non-Blocking Error] Error initializing custom logger: No module named 'opentelemetry.exporter.otlp.proto.grpc'
Traceback (most recent call last):
  File "/app/.venv/lib/python3.11/site-packages/litellm/litellm_core_utils/litellm_logging.py", line 3478, in _init_custom_logger_compatible_class
    _otel_logger = OpenTelemetry(
                   ^^^^^^^^^^^^^^
  File "/app/.venv/lib/python3.11/site-packages/litellm/integrations/opentelemetry.py", line 136, in __init__
    provider.add_span_processor(self._get_span_processor())
                                ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/app/.venv/lib/python3.11/site-packages/litellm/integrations/opentelemetry.py", line 972, in _get_span_processor
    from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import (
ModuleNotFoundError: No module named 'opentelemetry.exporter.otlp.proto.grpc'

Are you a ML Ops Team?

Yes

What LiteLLM version are you on ?

1.74.15.post2

Twitter / LinkedIn details

No response

[aelawson]

Running into this as well on the latest versions of langfuse (3.7.0) and litellm (1.78.2). Would be great if it can be addressed!