fix(profiler): explain inaccessible model paths during trust checks

[tedzhouhk]

Summary

When a local model path cannot be inspected (for example, a protected symlink on a model-cache PVC), DGD materialization currently leaks an unhandled filesystem error. Report an actionable error with the original cause, including directory/symlink ownership checks and the option to use the actual snapshot path in modelCache.pvcModelPath.

Preserve config-inspection errors in resolve_model_path() before it can fall back to a Hub model ID. Missing configs and non-directory paths retain the existing fallback.

Use one explicit stat() so filesystem errors remain distinguishable from missing/non-directory references. Preserve explicit worker trust flags and the existing remote-model trust policy. Correct the auto-map warning/docstring so inconclusive detection does not claim that --trust-remote-code was injected.

Validation

• Profiler protocol, DGD materialization, and model-path resolution suites: 131 passed, including resolution through materialization and persistence of inaccessible-PVC errors in profiler_status.yaml. The new inaccessible-config regressions fail before the resolver fix and pass afterward.
• Actual run_profile with a real chmod(000) parent directory: preserves the PermissionError cause and complete actionable status, without mocked filesystem calls.
• pre-commit run --all-files and git diff --check: passed.
• QA cluster reproduction with stock 1.5.0-rc.7: protected alias fails; the direct snapshot path completes profiling/spec generation.
• Repeated both RAPID DGDR cases with the two updated functions overlaid onto rc.7 through a ConfigMap: alias reports the new actionable error; snapshot still completes with profiler exit 0 and DGDR Ready / SpecGenerated. This was a source-overlay check, not a rebuilt image or worker inference test.

The inaccessible alias intentionally remains an error; fix its ownership or use the snapshot path. No operator change is required.

Summary by CodeRabbit

• Bug Fixes
  • Improved validation for local model paths when determining whether implicit remote-code trust is allowed.
  • Invalid, missing, or non-directory model paths are now handled consistently, with clearer runtime errors and diagnostic logging.
  • Explicit --trust-remote-code settings continue to work without being overridden or modified.
  • Added coverage for symlinked paths and other inaccessible or invalid model-path scenarios.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 187cab32-1c0c-4aa4-a616-c16c8b93b9ae

📥 Commits

Reviewing files that changed from the base of the PR and between 5292fc9 and 1d93100.

📒 Files selected for processing (2)

• components/src/dynamo/profiler/tests/unit/test_profiler_protocol.py
• components/src/dynamo/profiler/utils/model_info.py

Included review availability: Your plan provides up to 12 included reviews per hour; 11 remain after this review.

---

Walkthrough

The change updates model path inspection for implicit remote-code trust. It distinguishes valid directories from inaccessible paths and adds tests for symlinks, invalid paths, wrapped errors, configuration preservation, and diagnostic logging.

Changes

Model path trust validation

Layer / File(s)	Summary
Model path inspection and trust-policy messaging components/src/dynamo/profiler/utils/model_info.py	model_ref_allows_implicit_trust_remote_code uses stat() and S_ISDIR. Missing and non-directory paths return False; other OSError failures raise diagnostic RuntimeError exceptions. Auto-map documentation and logging describe inconclusive trust-policy results.
Trust-policy behavior tests components/src/dynamo/profiler/tests/unit/test_profiler_protocol.py	Tests cover eligible directories and symlinks, invalid paths, inaccessible paths, explicit trust configuration, preserved input configuration, original error causes, and diagnostic logging.

Priority: ⬇️ Low

Estimated code review effort: 3 (Moderate) | ~25 minutes

Merge Risk: ⚪ Minimal · up to 1d931

No actionable merge-blocking risk remains in the reviewed change.

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 57.14% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 7 functions across 2 files.	Write docstrings for the functions missing them to satisfy the coverage threshold.
Description check	⚠️ Warning	The description explains the change and validation results, but it does not follow the required template. It omits the required Related Issues section and the Where should the reviewer start? section.	Add the required Related Issues section and select either the linked-issue option with a real issue number or confirm that no related issue exists. Add a Where should the reviewer start? section with the key files for review. Align the exis…

✅ Passed checks (3 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Title check	✅ Passed	The title clearly and concisely describes the main change: improving diagnostics for inaccessible model paths during profiler trust checks.

Full details: Description check

Resolution

Add the required Related Issues section and select either the linked-issue option with a real issue number or confirm that no related issue exists. Add a Where should the reviewer start? section with the key files for review. Align the existing Summary and Validation content with the template's Overview and Details sections.

• Fix all pre-merge checks with AI

✨ Finishing Touches 💡 1

🛠️ Fix failing CI checks 💡

• Create stacked PR
• Commit on current branch

---

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

[glamr-agent]

Automated evidence record — validation complete

factory: This is the machine-generated evidence record for the automated review of this
request. It is advisory: it is not an approval, and it does not replace human review.

Reviewed commit: 5e1dbbc80e3a616c5b149ccd84184815a76520ea — the head of this
request when the checks below ran. The change was reproduced on a local branch at
that head and examined there; nothing was pushed to this request's branch.

Validation status: complete

Evidence summary: [3/3 validated]

AI review assessment: sound. This assessment is produced by an automated reviewer
and is advisory only.

Validation result: complete — pass. Every check listed below ran to completion and
exited 0.

Evidence audit: complete [3/3 validated] — the command report below comes from
recorded runs.

Two things the checks below do not cover, stated so the record is not read as more
than it is: the Kubernetes deploy-and-infer retest with the corrected PVC alias needs a
cluster and was not run here, and this request's own full pipeline was neither triggered
nor waited on by this run.

Commands and results [3/3 validated]

Generated from the commands recorded during this run.

Check 1

Builds the changed Dynamo source and confirms that Python can import its compiled extension.

Result: Passed (exit 0)

Command:

Not shown because the exact command contained private run data.

Check 2

Checks the changed files with the repository's fast lint and formatting commands.

Result: Passed (exit 0)

Command:

Not shown because the exact command contained private run data.

Check 3

Runs the relevant Python unit tests without requiring a GPU.

Result: Passed (exit 0)

Command:

Not shown because the exact command contained private run data.