fix(inference): honor host-local serving recipe contracts

[prekshivyas]

Summary

Serving presets can now evaluate schema-valid readiness observation comparisons. Host-local vLLM recipes with an empty runtime environment now produce a valid serving command instead of a leading &&.

Related Issue

Refs #8246

Changes

• Evaluate equals, one-of, at-least, and version-at-least observation comparisons against typed readiness values.
• Compare numeric dotted versions segment by segment and reject invalid version values.
• Omit the vLLM environment prefix when runtime.environment is empty.
• Add regression tests for matching and rejected readiness comparisons and for the empty-environment command.

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)

Quality Gates

• Tests added or updated for changed behavior
• Existing tests cover changed behavior — justification:
• Tests not applicable — justification:
• Docs updated for user-facing behavior changes
• Docs not applicable — justification: The change restores existing internal serving-catalog and vLLM command contracts. It adds no command, flag, schema field, default, or public workflow.
• Sensitive paths changed (security, policy, credentials, preflight, onboarding, inference, runner, sandbox, or messaging)
• Sensitive-path review completed or maintainer-approved waiver recorded — reviewer/approval link/justification:
• Non-success, skipped, or missing CI check accepted by maintainer — check name, approval link, and follow-up issue:

Documentation Writer Review

• Documentation writer subagent reviewed the completed changes
• Result: no-docs-needed
• Evidence: The review of commit d6491799c found that the change restores existing internal behavior and does not change a documented user-facing surface. No terminology, structure, voice, or test-title findings were reported.
• Agent: Codex Desktop

DGX Station Hardware Evidence

• Tested on DGX Station
• Tested commit: Not applicable.
• Station profile/scenario: Not applicable.
• Result: Not applicable.
• Supporting evidence: Not applicable.

Verification

• PR description includes a Signed-off-by: line and every commit appears as Verified in GitHub
• Normal pre-commit, commit-msg, and pre-push hooks passed, or npm run validate:pr passed after refreshing origin/main when hooks were skipped or unavailable
• Targeted behavior tests pass for the current change set, or tests are marked not applicable above — npx vitest run --project cli src/lib/inference/serving/resolver.test.ts src/lib/inference/vllm-models.test.ts: 2 files and 73 tests passed.
• Applicable broad gate passed — npm test for broad runtime/test-harness changes; npm run check for repo-wide validation/coverage changes — command/result: Not run; targeted source tests cover both changed behaviors.
• Quality Gates section completed with required justifications or waivers
• No secrets, API keys, or credentials committed
• 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: Prekshi Vyas prekshiv@nvidia.com

Summary by CodeRabbit

• Bug Fixes
  • Readiness requirements now correctly support string, membership, numeric minimum, and version minimum comparisons.
  • Version comparisons safely handle dotted numeric versions and reject insufficient or malformed values.
  • Serving commands no longer include an invalid command separator when no environment overrides are configured.

[github-code-quality]

Code Coverage Overview

Languages: TypeScript

TypeScript / code-coverage/plugin

The overall coverage in commit d649179 in the agent/8246-host-loca... branch remains at 96%, unchanged from commit b148bc1 in the main branch.

TypeScript / code-coverage/cli

The overall coverage in commit d649179 in the agent/8246-host-loca... branch remains at 82%, unchanged from commit b148bc1 in the main branch.

Show a code coverage summary of the most impacted files.

File	main b148bc1	agent/8246-host-loca... d649179	+/-
src/lib/domain/.../connect-env.ts	97%	89%	-8%
src/lib/inferen.../vllm-models.ts	86%	86%	0%
src/lib/sandbox...rce-identity.ts	87%	87%	0%
src/lib/inferen...ing/resolver.ts	89%	90%	+1%

---

_{Updated August 05, 2026 20:48 UTC}

[coderabbitai]

📝 Walkthrough

Walkthrough

The resolver now evaluates equality, membership, numeric minimum, and semantic version minimum readiness comparisons. The vLLM command builder omits empty serving environment exports.

Changes

Readiness comparison evaluation

Layer / File(s)	Summary
Typed readiness comparison flow src/lib/inference/serving/resolver.ts, src/lib/inference/serving/resolver.test.ts	The resolver evaluates comparison requirements against matching observations. Tests cover successful matches and rejection for mismatched, insufficient, and malformed values.

vLLM command generation

Layer / File(s)	Summary
Conditional serving environment prefix src/lib/inference/vllm-models.ts, src/lib/inference/vllm-models.test.ts	buildVllmServeCommand omits the export prefix when serveEnv is empty. A regression test verifies the command starts with the installation and serve commands.

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

Sequence Diagram(s)

sequenceDiagram
  participant ReadinessResolver
  participant ReadinessObservations
  participant ServingPreset
  ReadinessResolver->>ReadinessObservations: Find observation by ID
  ReadinessResolver->>ReadinessResolver: Evaluate comparison
  ReadinessResolver->>ServingPreset: Accept or reject readiness requirements

Loading

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
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.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly describes the inference serving changes that enforce host-local serving recipe contracts.

✨ 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/8246-host-local-spark-fixes

---

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

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@src/lib/inference/serving/resolver.test.ts`:
- Around line 502-596: Extend the resolver-level test “matches typed readiness
observation comparisons (`#8246`)” with a malformed observed
host.gpu.driver_version value such as “580.65.x”. Invoke
resolveManagedInferenceServing through the existing resolverInput and
comparedCatalog boundary, and assert the result is rejected with code
“requirements-not-met”.

🪄 Autofix

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 03122dc9-6014-4d00-a1f7-1e8423f18575

📥 Commits

Reviewing files that changed from the base of the PR and between c6713c9 and bcedc5b.

📒 Files selected for processing (4)

• src/lib/inference/serving/resolver.test.ts
• src/lib/inference/serving/resolver.ts
• src/lib/inference/vllm-models.test.ts
• src/lib/inference/vllm-models.ts

[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 · high confidence · 0 blockers · 0 warnings · 0 suggestions
• Nemotron 3 Ultra (second opinion): Completed · high confidence · 0 blockers · 0 warnings · 0 suggestions
• Model comparison: normalized findings match; normalized terminology decisions differ; normalized E2E selections differ; severity counts match.

3 additional E2E selections from the second opinion

Advisory only. The primary lane did not select these E2E jobs or targets.

• spark-install: The completed second-opinion lane identified E2E coverage that the primary lane omitted.
• vllm-docker-storage: The completed second-opinion lane identified E2E coverage that the primary lane omitted.
• llama-cpp-dgx-spark-qualification: The completed second-opinion lane identified E2E coverage that the primary lane omitted.

Second-opinion terminology and E2E selections are advisory. They do not change the primary assessment or E2E / PR Gate.

1 semantic terminology decision

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

• justified — readiness observation comparison at src/lib/inference/serving/resolver.test.ts:502: Keep “readiness observation comparisons” where the contrast with other readiness requirement types matters.

E2E guidance

Advisory only. E2E / PR Gate selects and runs jobs independently.

Recommended E2E: inference-routing, network-policy

Workflow run details

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

[apurvvkumaria]

Approve — reviewed current head d649179 with focus on readiness comparison semantics, version parsing, fail-closed selection, vLLM command construction, compatibility, and tests. No blocking defect found; required checks are green.