docs: add tracing notebook

[tgasser-nv]

Description

Related Issue(s)

Checklist

• I've read the CONTRIBUTING guidelines.
• I've updated the documentation if applicable.
• I've added tests if applicable.
• @mentions of the person or team responsible for reviewing proposed changes.

[github-actions]

Documentation preview

https://nvidia.github.io/NeMo-Guardrails/review/pr-1337

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 71.68%. Comparing base (3fdd65d) to head (1182843).
⚠️ Report is 3 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #1337      +/-   ##
===========================================
+ Coverage    71.59%   71.68%   +0.09%     
===========================================
  Files          168      168              
  Lines        16862    16862              
===========================================
+ Hits         12072    12088      +16     
+ Misses        4790     4774      -16     

Flag	Coverage Δ
python	71.68% <ø> (+0.09%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.
see 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[tgasser-nv]

Attaching a PDF printout of the notebook to make reviewing easy (cc @cparisien , @Pouyanpi , @miyoungc )
1_tracing_pdf_dump.pdf

[Pouyanpi]

Thank you @tgasser-nv. This is a valuable addition 👍🏻

Merge plan: Let's ensure this is merged once all pending PRs for 0.16.0 are merged to develop and do a rebase on top of this so that it reflects the latest changes.

Docs build: @miyoungc can you confirm that build_notebook_docs.py is the right path for rendering this notebooks to
https://docs.nvidia.com/nemo/guardrails/latest/getting-started ?

Tutorial usage: right now users can’t follow the docs due to missing configs. Two options, happy with either here:

• Inline the needed configs in the notebook and instruct users to create a config/ folder (like other tutorials), or
• Use from_content to avoid file creation entirely.

Nice to haves (follow-ups, separate PR):

• Create NemoGuard NIMs configs to examples/configs so cloned users can reference them directly.
• Use the most recent span_format: opentelemetry. (I’ll switch the default in #1331; this PR can keep current unless trivial.) The current default format is considered legacy and we might deprecate it.
• Show how to toggle parallel/sequential dynamically so we don’t need two folders:

config.rails.input.parallel = True
config.rails.output.parallel = False

• point users to the “Parallel Execution of Input and Output Rails” section.
  docs: https://docs.nvidia.com/nemo/guardrails/latest/user-guides/configuration-guide/guardrails-configuration.html#parallel-rails

• quiet installs to keep notebook output clean:

!pip install pandas plotly langchain_nvidia_ai_endpoints aiofiles -q

If we pick one of the two config approaches above and rebase after 0.16.0 lands it'd be great.

[cparisien]

This is looking really good! Needs a little bit of cleanup but it's in pretty good shape already.

[tgasser-nv]

Thank you @tgasser-nv. This is a valuable addition 👍🏻

Merge plan: Let's ensure this is merged once all pending PRs for 0.16.0 are merged to develop and do a rebase on top of this so that it reflects the latest changes.

Docs build: @miyoungc can you confirm that build_notebook_docs.py is the right path for rendering this notebooks to https://docs.nvidia.com/nemo/guardrails/latest/getting-started ?

Tutorial usage: right now users can’t follow the docs due to missing configs. Two options, happy with either here:

• Inline the needed configs in the notebook and instruct users to create a config/ folder (like other tutorials), or
• Use from_content to avoid file creation entirely.

Nice to haves (follow-ups, separate PR):

• Create NemoGuard NIMs configs to examples/configs so cloned users can reference them directly.
• Use the most recent span_format: opentelemetry. (I’ll switch the default in feat: enhance tracing system with OpenTelemetry semantic conventions #1331; this PR can keep current unless trivial.) The current default format is considered legacy and we might deprecate it.
• Show how to toggle parallel/sequential dynamically so we don’t need two folders:

config.rails.input.parallel = True
config.rails.output.parallel = False

• point users to the “Parallel Execution of Input and Output Rails” section.
  docs: https://docs.nvidia.com/nemo/guardrails/latest/user-guides/configuration-guide/guardrails-configuration.html#parallel-rails
• quiet installs to keep notebook output clean:

!pip install pandas plotly langchain_nvidia_ai_endpoints aiofiles -q

If we pick one of the two config approaches above and rebase after 0.16.0 lands it'd be great.

Quick update:

• All the PRs for v0.16 were merged today, I rebased on top of those and did a clean run of the notebook. I can see all the spurious ERROR messages are gone, and the names of rails follow the new guardrails.request format
• I rewrote all the configs and created them as Python dicts before converting to RailsConfig objects using Pydantic's model_validate() function. The prompts are now loaded from elsewhere in the repo instead of copy-pasting them into new files. The only file in the PR is the notebook.
• I'm picking up the most recent span format, I updated the clean_trace_dataframe() method to match.
• Added -q to the pip install

I met with @miyoungc who's looking at the best way to link the notebook from the rest of the docs. Am holding off merging until Monday to give her time to work on this.

[Pouyanpi]

Thank you Tim! It looks good to me. Just a very minor issue that I commented and 2 questions. Please merge it when you are ready 👍🏻