[prekshivyas]

Summary

• remove an orphaned nemoclaw-vllm container during full uninstall when Docker verifies its NemoClaw managed label
• retain fail-closed behavior for a missing or mismatched label, malformed inspection, or failed removal
• add focused uninstall regression coverage for the labeled and unlabeled cases

Root cause

The host-local vLLM cleanup only ran when its local key state existed. A legacy/orphaned managed container with the reserved name and managed label but no local state was skipped, then caused the residual-container guard to abort the rest of uninstall.

Validation

• npx vitest run src/lib/actions/uninstall/run-plan-local-model-profile.test.ts
• npm run typecheck:cli
• npx @biomejs/biome check src/lib/actions/uninstall/run-plan.ts src/lib/actions/uninstall/run-plan-local-model-profile.test.ts
• live DGX Spark full uninstall: completed successfully after removing the reproduced nemoclaw-vllm orphan and all NemoClaw resources

Closes #8981

Summary by CodeRabbit

• Bug Fixes

  • Improved full uninstall cleanup for orphaned host-local model containers.
  • Managed containers are removed when no active runtime remains.
  • Unmanaged, unlabeled, or malformed containers are preserved.
  • Cleanup safely handles inspection failures and stops if removal fails.

• Documentation

  • Documented ownership validation for the reserved model container during uninstall.

• Tests

  • Added coverage for managed, unmanaged, malformed, and failed-removal scenarios.

Documentation Writer Review

• Documentation writer subagent reviewed the completed changes
• Result: docs-updated
• Evidence: docs/reference/commands.mdx
• Agent: Codex Desktop

---

Signed-off-by: Prekshi Vyas prekshiv@nvidia.com

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[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: b938079f-1731-4a3e-930e-34071af20f96

📥 Commits

Reviewing files that changed from the base of the PR and between 5632a33 and bd8943d.

📒 Files selected for processing (1)

• src/lib/actions/uninstall/run-plan-local-model-profile.test.ts

🚧 Files skipped from review as they are similar to previous changes (1)

• src/lib/actions/uninstall/run-plan-local-model-profile.test.ts

---

📝 Walkthrough

Walkthrough

The uninstall flow now inspects the expected host-local vLLM container and removes it only when ownership is confirmed. Tests cover managed, unlabeled, malformed, and removal-failure scenarios.

Changes

Host-local vLLM orphan cleanup

Layer / File(s)	Summary
Managed container cleanup src/lib/actions/uninstall/run-plan.ts, src/lib/inference/local-model-profile/cleanup.ts, docs/reference/commands.mdx	The uninstall flow validates the container ID and managed label before removal. Inspection or removal failures block later cleanup. The ownership constants are re-exported and the fallback behavior is documented.
Ownership cleanup validation src/lib/actions/uninstall/run-plan-local-model-profile.test.ts	Tests verify managed-container removal, preservation of unlabeled or malformed containers, removal failure handling, and reserved-name inspection.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Mergeability Score: ⚪ Minimal · up to bd894

This PR makes a localized uninstall cleanup change for orphaned managed containers, with focused regression coverage and validation; no actionable merge-blocking risk remains.

Sequence Diagram(s)

sequenceDiagram
  participant UninstallFlow
  participant Docker
  participant nemoclaw-vllm
  UninstallFlow->>Docker: Inspect nemoclaw-vllm
  Docker-->>UninstallFlow: Return container ID and managed label
  UninstallFlow->>Docker: Force-remove verified container
  Docker->>nemoclaw-vllm: Remove container
  Docker-->>UninstallFlow: Return removal result

Loading

Possibly related PRs

• NVIDIA/NemoClaw#8919: Both changes modify host-local model runtime cleanup and uninstall behavior.
• NVIDIA/NemoClaw#8976: Both changes modify Docker cleanup in run-plan.ts.
• NVIDIA/NemoClaw#9031: This change extends the same managed vLLM uninstall cleanup and tests.

Suggested labels: area: sandbox, bug-fix

Suggested reviewers: ericksoa

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	The changes address issue #8981 by removing only validated managed vLLM containers and preserving fail-closed behavior for unsafe cases.
Out of Scope Changes check	✅ Passed	The implementation, tests, exports, and documentation directly support the managed vLLM uninstall fix and contain no unrelated changes.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: removing an orphaned managed vLLM container during uninstall.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch agent/fix-orphaned-vllm-uninstall-dco

---

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

[cv]

The replacement commit is signed and verified. The PR description still needs the contributor DCO declaration required by this repository. Please add your own line in this form:

Signed-off-by: Prekshi Vyas prekshiv@nvidia.com

I am carrying the reviewed ownership-validation and documentation fixes onto this replacement branch. I will recheck the description and current CI after that push.

[cv]

The maintainer fixes are now on commit 5632a33b02: strict Docker ID validation, malformed-inspection and removal-failure coverage, the canonical uninstall documentation update, and the matching documentation review receipt. The focused tests, docs build, pre-commit hooks, and pre-push TypeScript checks passed. The contributor DCO line in the PR description remains the only description-level action before approval.

[github-actions]

🌿 Preview your docs: https://nvidia-preview-pr-9032.docs.buildwithfern.com/nemoclaw

[github-code-quality]

Code Coverage Overview

Languages: TypeScript

TypeScript / code-coverage/plugin

The overall coverage in commit 79f4737 in the agent/fix-orphaned-v... branch remains at 96%, unchanged from commit 626b75d in the main branch.

---

_{Updated August 13, 2026 20:58 UTC}

[cv]

The new merge commits are signed, but the repository DCO check also requires the contributor declaration in the PR description. Please edit the description, not the branch, and add:

Signed-off-by: Prekshi Vyas prekshiv@nvidia.com

No further branch commits are needed.

[prekshivyas]

Reviewed the current head. No blocking issues found in the orphaned managed-vLLM cleanup; CI remains the merge gate. This is a comment rather than an approval because the authenticated reviewer is also the PR author.

[cv]

Reviewed the accepted uninstall scope, ownership proof, inspected-ID removal, fail-closed error paths, regression coverage, documentation update, resolved feedback, DCO declaration, and current CI. The implementation is ready to merge.

[github-actions]

PR Review Advisor — No blocking findings reported

Advisor assessment: No blocking advisor findings reported
Next action: No advisor follow-up needed.
Findings: 0 blockers · 0 warnings · 0 suggestions

Model lanes

• GPT-5.6 Terra (primary): Completed · medium confidence · 0 blockers · 0 warnings · 0 suggestions
• Nemotron 3 Ultra (second opinion): Completed · high confidence · 1 blocker · 0 warnings · 0 suggestions
• Model comparison: normalized findings differ; normalized terminology decisions differ; normalized E2E selections differ; Nemotron reported 1 more blocker, the same number of warnings, the same number of suggestions.

Second-opinion terminology and E2E selections are advisory. Live E2E does not run automatically for pull requests.

3 semantic terminology decisions

Terminology decisions are advisory. They affect the assessment only when a separate finding identifies concrete semantic impact.

• justified — orphaned host-local vLLM container at src/lib/actions/uninstall/run-plan-local-model-profile.test.ts:96: Retain this term for the tested cleanup case.
• established — ownership state at docs/reference/commands.mdx:4085: Retain the established repository term.
• define — reserved `nemoclaw-vllm` container at docs/reference/commands.mdx:4085: Retain the term because the same sentence defines its Docker label and container-ID conditions.

E2E guidance

Advisory only. A maintainer can dispatch the default E2E suite for the commit under review.

Recommended E2E: inference-routing

Manual-only E2E: network-policy
The manual PR workflow does not run these selectors for the commit under review. Run them from reviewed code on main.

1 optional E2E recommendation

• vllm-docker-storage

Workflow run details

This automated review informs maintainers. Warnings and suggestions do not require a response. A maintainer decides whether to merge.