docs(devnotes): add Nemotron-Personas dev note

[3mei]

📋 Summary

Adds the Inside Nemotron-Personas dev note covering how the multi-locale Nemotron-Personas HF collection is built (4-stage compound-AI pipeline) and how it's used as a seeding primitive across Nemotron training (long-context, tool-use, formal logic, safety refusals, instruction-following). Ships alongside a runnable Tutorial 7 demonstrating reproduction + customization, plus a Colab variant

🔗 Related Issue

N/A

🔄 Changes

✨ Added

• docs/devnotes/posts/nemotron-personas.md — new dev note
• docs/devnotes/posts/assets/nemotron-personas/ — four images: three pipeline-stage diagrams from the partner repo plus a black-background Nemotron-Personas world-map hero
• docs/notebook_source/7-nemotron-personas.py — jupytext source for the Reproducing & Customizing Nemotron-Personas tutorial;
• docs/colab_notebooks/7-nemotron-personas.ipynb — committed Colab variant; i

🔧 Changed

• docs/scripts/generate_colab_notebooks.py — adds an ADDITIONAL_SETUP_CELLS map paralleling ADDITIONAL_DEPENDENCIES; injects NGC CLI install + NGC_API_KEY cells. Future devnote-paired tutorials needing extra Colab bootstrap can register one-line entries in the same map.
• mkdocs.yml — adds Reproducing & Customizing Nemotron-Personas under the Tutorials nav

🧪 Testing

• make test passes
• Notebook runs end-to-end via jupytext --to ipynb --execute
• make generate-colab-notebooks regenerates the Colab .ipynb cleanly with the NGC setup cells in the expected position
• Unit tests added/updated (N/A — this PR is docs + tutorial assets; no engine code changed)
• E2E tests added/updated (N/A — Tutorial 7 is opt-in via make convert-execute-notebooks and gated on NVIDIA_API_KEY + on-disk NGC dataset, matching how Tutorials 5/6 are gated on OPENROUTER_API_KEY)

✅ Checklist

• Follows commit message conventions
• Commits are signed off (DCO)
• Architecture docs updated (N/A — no architectural changes)

[github-actions]

Docs preview: https://06744256.dd-docs-preview.pages.dev

Notebook tutorials are placeholder-only in previews.

[danecor]

Looks good! Some possible issues / suggestions attached.

[danecor]

This shows up in the notebook as a completely independent block from COLAB_API_KEY_CELL, with duplicated imports. Can we just append the try-except to that block conditionally?

import getpass
import os

from google.colab import userdata

try:
    os.environ["NVIDIA_API_KEY"] = userdata.get("NVIDIA_API_KEY")
except userdata.SecretNotFoundError:
    os.environ["NVIDIA_API_KEY"] = getpass.getpass("Enter your NVIDIA API key: ")

...

import getpass
import os

from google.colab import userdata

try:
    os.environ["NGC_API_KEY"] = userdata.get("NGC_API_KEY")
except userdata.SecretNotFoundError:
    os.environ["NGC_API_KEY"] = getpass.getpass("Enter your NGC API key: ")

[danecor]

This fails for me even with a populated NGC_API_KEY:

---------------------------------------------------------------------------
CalledProcessError                        Traceback (most recent call last)
Cell In[5], line 12
      8     for p in existing:
      9         print(f"  - {p.name}")
     10 else:
     11     print(f"Nemotron-Personas-{personas_locale} not found. Downloading via the Data Designer CLI...")
---> 12     subprocess.run(
     13         shlex.split(f"data-designer download personas --locale {personas_locale}"),
     14         check=True,
     15     )

File ~/.local/share/uv/python/cpython-3.11.9-macos-aarch64-none/lib/python3.11/subprocess.py:571, in run(input, capture_output, timeout, check, *popenargs, **kwargs)
    569     retcode = process.poll()
    570     if check and retcode:
--> 571         raise CalledProcessError(retcode, process.args,
    572                                  output=stdout, stderr=stderr)
    573 return CompletedProcess(process.args, retcode, stdout, stderr)

CalledProcessError: Command '['data-designer', 'download', 'personas', '--locale', 'en_US']' returned non-zero exit status 1.

A similar issue was raised by codex in review:

Avoid interactive NGC downloads in executed notebooks — The notebook cell depends on data-designer being runnable from the notebook kernel, ngc being on that kernel’s PATH, NGC_API_KEY being in that kernel’s environment, and the interactive “Proceed with download?” prompt being answerable.

My problem was that I didn't have ngc installed locally yet. This might all be fine in colab, if that's the expected or enforced route, but a bit confusing locally given that the error message is uninformative.

[danecor]

So colab suggests to me that we need to cast this to age in order to avoid breaking the age > 6 limit checks below, i.e.

config_builder.add_column(dd.ExpressionColumnConfig(name="age", expr="{{ person.age }}", dtype="int"))

But the checks work for me, even though the underlying value appears to be a string. Not sure why, something in jinja? So I guess this cast is optional

[danecor]

This image link points to the branch - will it update automatically to main when the branch is merged and deleted?

[danecor]

This is fine, but does read a bit LLM-y (em-dashes, "that's not a x; it's a why").

[danecor]

Seems like an unnecessary sentence.

[danecor]

I would just say "They're a seeding primitive used across many post-training stages." without the side-quest bit.

[danecor]

Nitpick - each place we say "and from there into Super v3" (here and once below) I would just say "and Super v3". Current phrasing sounds a bit like Super v3 inherited from the Nano v3 model, rather than just using the same training dataset.

[danecor]

\mu and \sigma don't render as latex for me, might want to insert the actual unicode characters.