fix(onboard): restore Qwen3.6 27B FP8 as DGX Station vLLM default

[zyang-dev]

Summary

Restores Qwen 3.6 27B FP8 as the default managed-vLLM model for DGX Station because the DeepSeek V4 Flash recipe needs more accuracy validation.

Changes

• Switched the DGX Station vLLM profile default back to Qwen/Qwen3.6-27B-FP8.
• Kept deepseek-v4-flash registered as a supported managed-vLLM override.
• Updated profile tests and docs to describe Qwen 27B FP8 as the DGX Station default.

Type of Change

• Code change (feature, bug fix, or refactor)
• Code change with doc updates
• Doc only (prose changes, no code sample modifications)
• Doc only (includes code sample changes)

Verification

• npx prek run --all-files passes
• npm test passes
• Tests added or updated for new or changed behavior
• No secrets, API keys, or credentials committed
• Docs updated for user-facing behavior changes
• npm run docs builds without warnings (doc changes only)
• Doc pages follow the style guide (doc changes only)
• New doc pages include SPDX header and frontmatter (new pages only)

---

Signed-off-by: Your Name your-email@example.com

Summary by CodeRabbit

• Documentation

  • Updated the default model for DGX Station's managed vLLM profile from DeepSeek V4 Flash to Qwen3.6-27B-FP8.
  • Reordered documentation tables listing available model overrides.
  • Updated environment variable documentation for model configuration.

• Tests

  • Updated test cases to verify the new default model behavior for DGX Station vLLM profiles.

[coderabbitai]

Too many files changed? Review this PR in Change Stack to see how the pieces fit before you dive in.

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: cbe899ec-e049-4a30-a220-a879eb7e45ad

📥 Commits

Reviewing files that changed from the base of the PR and between 1f0b4c5 and a11e7f4.

📒 Files selected for processing (8)

• docs/inference/inference-options.mdx
• docs/reference/commands-nemohermes.mdx
• docs/reference/commands.mdx
• src/lib/inference/vllm-models.test.ts
• src/lib/inference/vllm-models.ts
• src/lib/inference/vllm.test.ts
• src/lib/inference/vllm.ts
• test/detect-vllm-profile.test.ts

---

📝 Walkthrough

Walkthrough

This PR updates the default vLLM model for DGX Station from DeepSeek V4 Flash to Qwen3.6-27B-FP8. The change spans platform profile configuration, test assertions, documentation tables, and environment variable references to ensure consistent behavior across the codebase.

Changes

DGX Station vLLM Default Model Switch

Layer / File(s)	Summary
Core model selection logic src/lib/inference/vllm.ts	Introduces qwen27bFP8Model() helper and updates STATION_PROFILE.defaultModel to use the new model. The installVllm comment documenting per-platform defaults is adjusted to reflect Station: Qwen3.6-27B and Spark: Qwen3.6-35B-A3B NVFP4.
Model selector documentation src/lib/inference/vllm-models.ts	selectVllmModelFromEnv documentation is updated to specify per-platform defaults for unset NEMOCLAW_VLLM_MODEL (Station: Qwen3.6-27B; Spark: Qwen3.6-35B-A3B NVFP4).
Test expectations for new model default src/lib/inference/vllm.test.ts, src/lib/inference/vllm-models.test.ts, test/detect-vllm-profile.test.ts	vLLM profile detection tests assert Qwen/Qwen3.6-27B-FP8 with env value qwen3.6-27b for DGX Station. Test descriptions are clarified to distinguish DeepSeek as a managed-vLLM override rather than a Station-specific default.
User-facing documentation docs/inference/inference-options.mdx, docs/reference/commands-nemohermes.mdx, docs/reference/commands.mdx	The inference options table promotes qwen3.6-27b to "default on the DGX Station profile" and demotes deepseek-v4-flash to "supported override." Environment variable reference docs reorder recognized slugs to reflect the new ordering.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Possibly related PRs

• NVIDIA/NemoClaw#4867: Prior vLLM model default changes for DGX Station and related platform profile wiring updates.

Suggested labels

Platform: Station, Provider: vLLM, area: inference, bug-fix

Suggested reviewers

• cv

Poem

🐰 A Station that once held a DeepSeek so bright,
Now shines with a Qwen's efficient light,
Tests and docs aligned in perfect accord,
Models swapped swift by reviewer's word! ✨

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 25.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The PR title accurately describes the main change: reverting DGX Station's default vLLM model from DeepSeek V4 Flash back to Qwen3.6 27B FP8, which is reflected consistently across documentation and test updates.
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.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ 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 fix/revert-dgx-station-vllm-to-qwen27b

---

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

[github-actions]

E2E Advisor Recommendation

Required E2E: None
Optional E2E: None

Workflow run

Full advisor summary

E2E Recommendation Advisor

Failed: Could not parse JSON from advisor output; see /home/runner/work/NemoClaw/NemoClaw/artifacts/e2e-advisor/e2e-advisor-raw-output.txt

[github-actions]

E2E Scenario Advisor Recommendation

Required scenario E2E: None
Optional scenario E2E: None

Workflow run

Full scenario advisor summary

E2E Scenario Advisor

Failed: Could not parse JSON from advisor output; see /home/runner/work/NemoClaw/NemoClaw/artifacts/e2e-advisor/e2e-scenario-advisor-raw-output.txt

[github-actions]

🌿 Preview your docs: https://nvidia-preview-pr-4888.docs.buildwithfern.com/nemoclaw

[github-actions]

PR Review Advisor

Findings: 0 needs attention, 0 worth checking, 0 nice ideas
Top item: No actionable findings

Consider writing more tests for

• **Runtime validation** — DGX Station managed-vLLM install with `NEMOCLAW_VLLM_MODEL` unset pre-downloads/serves `Qwen/Qwen3.6-27B-FP8` and reports that model in the install summary.. Unit tests cover the static profile/default selection and DeepSeek override registry behavior. Because the touched path starts containers and serves a model in managed-vLLM onboarding, runtime validation would improve confidence without being required to understand the code-level change.
• **Runtime validation** — DGX Station managed-vLLM install with `NEMOCLAW_VLLM_MODEL=deepseek-v4-flash` still selects DeepSeek as an override and keeps the DeepSeek-specific serve flags.. Unit tests cover the static profile/default selection and DeepSeek override registry behavior. Because the touched path starts containers and serves a model in managed-vLLM onboarding, runtime validation would improve confidence without being required to understand the code-level change.

Workflow run details

This is an automated advisory review. A human maintainer must make the final merge decision.