Correct usage of langfuse with Spring AI

Dear langfuse team,

I have been trying to figure out the best way to setup langfuse with Spring AI for some time now and wanted to share where I am at.

For this, I have created a fork of the current sample application here.

Would you mind taking a look to see if:

• This is the correct way to add the session and user id?
• This is the correct way to set the input and output to show up in the UI on the root level?

To me, it looks like, the spans are not correctly nested:

This is also the case if I don't filter out any spans that are not AI related:

However, if I add tool calls, this seems to nest the spans correctly again, but won't show the parameters of the tool calls.

Some other notes

• Maybe the files in the config folder help others to see how to reduce the noise coming from opentelemetry spring boot starter. If you don't filter out the spans, you also get all other traces into langfuse, e.g.

I am happy to make this example a bit nicer and more elaborate (with a tool call etc)and submit a pull request to improve the documentation a bit.

Thanks for your help

Start with the linked davidbellem/langfuse-examples fork and inspect its config folder and Spring AI setup. Compare the session and user ID, root input/output, span nesting, tool-call parameters, and OpenTelemetry filtering with the current examples. Done means the correct setup is verified and documented in a reproducible example.