Generate a list of NIM models in pre-commit

[dagardner-nv]

Description

• Generate a list of models in CI located at ci/.nim_models_used.json
• model list file is maintained by a new pre-commit hook
• Allows for simple tracking of the NIM models we use.
• Update ci/scripts/model_health_check.py to use logging rather than print statements
• Ignore the locally hosted nim yaml file since the model names are different than their build.nvidia.com counter-parts.

By Submitting this PR I confirm:

• I am familiar with the Contributing Guidelines.
• We require that all contributors "sign-off" on their commits. This certifies that the contribution is your original work, or you have rights to submit it under the same license, or a compatible license.
  • Any contribution which contains commits that are not Signed-Off will not be accepted.
• When the PR is ready for review, new or existing tests cover these changes.
• When the PR is ready for review, the documentation is up to date with these changes.

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: cb355cdb-ac21-4ae6-8567-db36d53f343e

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[yczhang-nv]

LGTM

[mnajafian-nv]

This direction makes sense to me. A checked-in generated manifest is a good way to give build.nvidia.com a stable source of truth and help NAT get notified before model removals break examples or docs.

A few things I think we should tighten before this becomes the consumed contract:

• I would keep ci/.nim_models_used.json if build.nvidia.com has already wired to that path. At this point, path stability matters more than renaming.
• I like keeping llms and embedders split. The replacement path and blast radius are different enough that the distinction is useful.
• I would align the count field with the downstream contract. The Slack example and Smita’s note mention num_configs, but the PR currently generates usage_count. Since the value is counting config files, I think this should be num_configs unless build.nvidia.com explicitly prefers usage_count.
• I would add a top-level schema_version now. Once another team is parsing this file, we need a simple way to evolve it without breaking their cron/parser.
• I would state the inventory scope explicitly. From the implementation, this appears to cover model references discovered from repo YAML configs, not arbitrary mentions in docs/READMEs. That scope is reasonable, but it should be clear so nobody assumes full documentation coverage.
• The pre-commit hook seems appropriate because it runs in dry-run mode and generates from repo-local files. It should stay deterministic and offline. No build.nvidia.com/NIM service calls in pre-commit.
• I would not require config paths in v1 if build.nvidia.com only needs model/count, but paths would be the next useful field for NAT-side remediation.

Once the file contract is aligned with build.nvidia.com, I’m happy to do a code-level pass.