Langfuse Integration

[LStromann]

=Hey guys, has anyone successfuly implemented Langfuse tracing in their project? I am following the quick startup guide on the Cognee docs for the integration but I am getting the highlighted error Object of type "None" cannot be called for the first line in the code ```@observe(name="example_run", as_type="example")
async def simple_test():
await cognee.prune.prune_data()
await cognee.prune.prune_system(metadata=True)

await cognee.add("Generalized linear models can be used when the response variable is not normally distributed.")
await cognee.cognify()
results = await cognee.search(
    query_text="What type of models can be used when the response variable is not normally distributed?",
)
for result in results:
    print(result)```When I run it my terminal outputs ```ModuleNotFoundError: No module named 'langfuse'```, even though the Cognee docs specify that Cognee comes with a lightweight Langfuse wrapper. If this is broken I am willing to switch to a different tracing/eval service, I am not particularly attached to Langfuse. I appreciate any help. Thank you so much!

This discussion was automatically pulled from Discord.

[LStromann]

Also, I am getting a weird bug where deleting all the langfuse stuff in my code breaks the code that used to work: running the pipeline again gives me the same langfuse error, though I have not mentioned it once in my codebase.

[LStromann]

hey @russkiyximik, did you install langfuse?

[LStromann]

you can install the monitoring dependencies - pip install "cognee[monitoring]" let us know if this works thanks!

[LStromann]

Okay, so I got rid of the error by installing langfuse==2.60.10. It seems that they changed their imports in newer versions

[LStromann]

However, I am still getting object of type "none" cannot be called

[LStromann]

can you share the script

[LStromann]

import cognee
from cognee import visualize_graph

from cognee.modules.observability.get_observe import get_observe
observe = get_observe()

@observe(name="example_run", as_type="example")
async def simple_test():
    await cognee.prune.prune_data()
    await cognee.prune.prune_system(metadata=True)
    
    await cognee.add("Generalized linear models can be used when the response variable is not normally distributed.")
    await cognee.cognify()
    results = await cognee.search(
        query_text="What type of models can be used when the response variable is not normally distributed?",
    )
    for result in results:
        print(result)

if __name__ == '__main__':
    asyncio.run(simple_test())```

[LStromann]

It's running but as I said I am getting the highlighted calling error and the trace isn't showing up in langfuse cloud

[LStromann]

got it did you set these env in root

LANGFUSE_PUBLIC_KEY=pk-lf-...
LANGFUSE_SECRET_KEY=sk-lf-...
LANGFUSE_HOST=https://cloud.langfuse.com

[LStromann]

Yes I have

[LStromann]

can you try remove these lines from script and run it -

from cognee.modules.observability.get_observe import get_observe
observe = get_observe()
@observe(name="example_run", as_type="example")

[LStromann]

Yes the script works after removing those

[LStromann]

I get the correct answer from cognee.search

[LStromann]

what about the trace is it showing up in langfuse cloud

[LStromann]

Nope

[LStromann]

That import does not work

[LStromann]

I'm using langfuse==2.60.10

[LStromann]

There is ``` from langfuse.client import LangfuseClient

[LStromann]

https://python-sdk-v2.docs-snapshot.langfuse.com/docs/observability/sdk/python/decorators/

try this import -
from langfuse.decorators import langfuse_context
then -
langfuse_context.flush()

[LStromann]

Ah unfortunately it has not sent the traces

[LStromann]

can you please try this

[LStromann]

This code ```from langfuse.decorators import observe, langfuse_context

@observe()
def main():
print("Hello, Langfuse!")

main()

langfuse_context.flush()```

[LStromann]

Yields this python .\langfuse-test.py Langfuse client is disabled since no public_key was provided as a parameter or environment variable 'LANGFUSE_PUBLIC_KEY'. See our docs: https://langfuse.com/docs/sdk/python/low-level-sdk#initialize-client Hello, Langfuse!

[LStromann]

ill load dotenv

[LStromann]

Okay so now it runs with connection to the client but there is no trace in the lcoud

[LStromann]

cloud

[LStromann]

can you pleae update it in your script import observe, and @observe() at top of the function like this

[LStromann]

import cognee
from cognee import visualize_graph

from langfuse.decorators import observe, langfuse_context

@observe()
async def simple_test():
    from cognee.base_config import get_base_config
    config = get_base_config()
    print(f"Monitoring tool: {config.monitoring_tool}") # Prints Observer.LANGFUSE
    print(f"Langfuse keys set: {bool(config.langfuse_public_key and config.langfuse_secret_key)}") # Prints True if keys are set

    await cognee.prune.prune_data()
    await cognee.prune.prune_system(metadata=True)
    
    await cognee.add("Generalized linear models can be used when the response variable is not normally distributed.")
    await cognee.cognify()
    results = await cognee.search(
        query_text="What type of models can be used when the response variable is not normally distributed?",
    )
    for result in results:
        print(result)


if __name__ == '__main__':
    asyncio.run(simple_test())

    #flush traces to langfuse cloud
    langfuse_context.flush()
    print("traces sent to Langfuse cloud!")```

[LStromann]

Is this what you mean?

[LStromann]

If so I am still not seeing the trace in the cloud