[Bug]: acompletion logs duplicates to langfuse

[N13T]

What happened?

When using acompletion and logging to langfuse we create duplicate traces. Here's the code im running

import os
import asyncio

os.environ["OPENAI_API_KEY"] = "openai_api_key"
os.environ["ANTHROPIC_API_KEY"] = "anthropic_api_key"

os.environ['LITELLM_LOG'] = 'DEBUG'
os.environ["LANGFUSE_DEBUG"] = "True"

os.environ["LANGFUSE_PUBLIC_KEY"] = "pk"
os.environ["LANGFUSE_SECRET_KEY"] = "sk"

import litellm
litellm.set_verbose = True
litellm.success_callback = ["langfuse"]
litellm.failure_callback = ["langfuse"] # logs errors to langfuse

async def main():
    models = ["gpt-4o-mini", "claude-3-5-haiku-20241022"]

    messages = [
    {
        "role": "user", 
        "content": "Hello, how are you?"},
    ]

    resp = await litellm.acompletion(
        model=models[0],
        messages=messages,
        temperature=0.0,
        fallbacks=models[1:],
        metadata = {
        "generation_name": "test-gen",
        "project": "litellm-test" 
    }
    )
    return resp

if __name__ == "__main__":
    asyncio.run(main())

And here are the traces in langfuse, you can see the completion one is a single trace while the acompletion ones are duplicates.

Relevant log output

No response

Are you a ML Ops Team?

Yes

What LiteLLM version are you on ?

v1.56.4

Twitter / LinkedIn details

No response

[krrishdholakia]

able to repro with script. investigating

[krrishdholakia]

seems to be caused by the fallbacks: param

investigating

[krrishdholakia]

Fixed - 94fa390

Will be live in today's release.

---

@N13T Can we do a 10min call sometime this/next week?

Want to learn how you're using litellm sdk, so we can improve for your team.

Attaching my calendly, for your convenience - https://calendly.com/d/4mp-gd3-k5k/litellm-1-1-onboarding-chat