Add in Datadog LLM Observability Documentation

[sabrenner]

This PR adds in Datadog LLM Observability documentation for instrumenting stacks using the Google ADK. Datadog has released auto-instrumentation support for this in the ddtrace python package within the past couple of months, and are looking to add our presence here as part of the suite of existing listed observability vendors.

Please let me know what extra work is needs to be done on this as applicable and I am happy to iterate and push up changes! Thanks to reviewers in advance!

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[koverholt]

Hi @sabrenner, thanks for the PR and contribution. This functionality looks great, and the mention of ADK on the LLM observability page is really nice!

I'd like to write up a sample agent on this page to help users get started, and update the ddtrace-run command to actually run the agent and generate traces. I can help with that. However, when trying things out, I ran into the following error:

https://gist.github.com/koverholt/8e754e2c26e3c16635c051f524ab049d

What's the best way for us to troubleshoot that together? Should I open an issue in https://github.com/DataDog/dd-trace-py?

[koverholt]

Actually it looks like this issue captures exactly what I ran into, which is preventing traces from being sent:

DataDog/dd-trace-py#15174

So we would want to get that resolved before publishing these docs.

[sabrenner]

@koverholt Thank you for flagging that issue, yes definitely agree we should get that issue resolved first. I'll get it fixed and released, and then follow up back here once it is! ETA is hopefully within the next couple of weeks (just due to release cycle timings, although taking a look at the issue hopefully the fix won't be too complicated) 😄

[sabrenner]

Hi @koverholt! Just wanted to leave an update here that the fix for this has been merged, and will be released next week, at which point I think we'll be unblocked here. I'll ping you here again when it is released but just wanted to keep you in the loop!

[sabrenner]

Hi @koverholt, this fix was released in ddtrace==3.19.1 and will be in the upcoming 4.1.0 release as well. Let me know if the 3.19.1 version works or if there are any more issues, but otherwise I think if it's all good we're good to move forward here!

[koverholt]

Thanks for the update and the fix! ❤️ I will test this again with the latest version.

In the meantime, the Contributor License Agreement (CLA) test is failing for two GitHub accounts in this PR, can you check the build output for more information on signing CLAs? https://github.com/google/adk-docs/pull/998/checks?check_run_id=56814475201

[koverholt]

Suggest updating this command to reflect ADK usage, as in:

ddtrace-run adk web

or

ddtrace-run adk run my_agent

[koverholt]

Suggest adding a link or code snippet for a sample agent, something along the lines of:

Follow the steps in the ADK Getting Started guide to create a sample ADK agent.

[koverholt]

@sabrenner, the good news is that the error that I ran into before is resolved in ddtrace==3.19.1. However, ddtrace==4.0.0 is the latest package version that gets installed when pip install ddtrace is run, and the error still occurs since (as you mentioned) the fix is not available in that branch until 4.1.0 is released.

Pinning to version 3 of ddtrace in these docs seems suboptimal, so let's wait until 4.1.0 is available before we publish to minimize the likelihood of users running into the error with 4.0.0.