fix(e2e): pass --no-verify to inference set in switch E2E tests

[hunglp6d]

Summary

The hermes-inference-switch-e2e and openclaw-inference-switch-e2e nightly jobs fail because the nemohermes inference set / nemoclaw inference set verification request to the nvidia-prod endpoint for model z-ai/glm-5.1 times out at https://integrate.api.nvidia.com/v1/chat/completions. These E2E tests validate the inference switch mechanism (route, config patching, hashes, live requests via inference.local), not external model endpoint availability. Adding --no-verify skips the remote health check that is orthogonal to the test surface.

Related Issue

Fixes #4145

Changes

• test/e2e/test-hermes-inference-switch.sh: add --no-verify to nemohermes inference set call
• test/e2e/test-openclaw-inference-switch.sh: add --no-verify to nemoclaw inference set call

Validation

The --no-verify flag is a first-class CLI option (see src/lib/actions/inference-set.ts:309, src/lib/cli/public-display-defaults.ts:74). It is already used throughout the onboard flow (src/lib/onboard.ts:5550, 5637, 5714, 5796, 5840) and in sandbox connect (src/lib/actions/sandbox/connect.ts:277). The E2E tests continue to validate the full switch path — route assertion, config.yaml patching, config hash integrity, Hermes gateway PID stability, .env immutability, and live inference through inference.local — the only step removed is the upfront remote endpoint probe.

• Original failing run: 26347272005 on 42ac98dc84b9493cb50dcb3edd0f7c0f8d1831c5
• Targeted jobs: hermes-inference-switch-e2e (#77559129270), openclaw-inference-switch-e2e (#77559129278)

Note: A custom-e2e validation run could not be created because the CI token lacks workflows scope for pushing workflow files. The fix is mechanical (adding a documented flag) and does not alter test assertions.

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

AI Disclosure

• AI-assisted — tool: Claude Code (nemoclaw-diagnosis skill)

---

Signed-off-by: Hung Le hple@nvidia.com

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[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: ebd92171-2e61-42fc-b13e-6bad5e3ad46d

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
• Commit unit tests in branch fix/nightly-e2e-inference-switch-no-verify-42ac98d

---

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

[github-actions]

E2E Advisor Recommendation

Required E2E: None
Optional E2E: hermes-inference-switch-e2e, openclaw-inference-switch-e2e

Dispatch hint: hermes-inference-switch-e2e,openclaw-inference-switch-e2e

Workflow run

Full advisor summary

E2E Recommendation Advisor

Base: origin/main
Head: HEAD
Confidence: high

Required E2E

• None. No merge-blocking E2E is required because this PR changes only existing E2E test scripts, not runtime/user-flow code. Running the two affected jobs is useful to validate the test harness change but is optional rather than required by the runtime-impact policy.

Optional E2E

• hermes-inference-switch-e2e (medium; requires Docker, NVIDIA_API_KEY, full install/onboard, live inference): Optional self-validation for the modified Hermes inference-switch E2E script; confirms the updated nemohermes inference set --no-verify test path still installs, switches routes/config, and performs live requests.
• openclaw-inference-switch-e2e (medium; requires Docker, NVIDIA_API_KEY, full install/onboard, live inference): Optional self-validation for the modified OpenClaw inference-switch E2E script; confirms the updated nemoclaw inference set --no-verify test path still installs, switches routes/config, and performs live requests.

New E2E recommendations

• None.

Dispatch hint

• Workflow: .github/workflows/nightly-e2e.yaml
• jobs input: hermes-inference-switch-e2e,openclaw-inference-switch-e2e

[github-actions]

E2E Scenario Advisor Recommendation

Required scenario E2E: None
Optional scenario E2E: None

Workflow run

Full scenario advisor summary

E2E Scenario Advisor

Base: origin/main
Head: HEAD
Confidence: high

Required scenario E2E

• None. No scenario workflow, scenario metadata, scenario runtime, or validation-suite files changed.

Optional scenario E2E

• None.

Relevant changed files

• None.

[github-actions]

PR Review Advisor

Findings: 0 needs attention, 2 worth checking, 0 nice ideas
Top item: Default inference verification coverage removed from switch E2Es

Review findings

🛠️ Needs attention

• None.

🔎 Worth checking

• Default inference verification path is no longer exercised by these switch E2Es (test/e2e/test-hermes-inference-switch.sh:466): Both changed scripts now pass `--no-verify` to `nemohermes inference set` / `nemoclaw inference set`. That matches the PR goal of avoiding an external endpoint preflight, and the scripts still perform post-switch live requests through `inference.local`, but these E2Es no longer cover the CLI's default upstream verification behavior for the switched provider/model path.
  • Recommendation: Confirm there is separate stable coverage for the default `inference set` verification behavior, or add a targeted test that exercises verification without depending on the flaky external model endpoint. If this suite is intentionally scoped only to routing/config patching, document that boundary in the test or PR notes.
  • Evidence: Diff changes `nemohermes inference set --provider "$SWITCH_PROVIDER" --model "$SWITCH_MODEL"` to include `--no-verify`; `test/e2e/test-openclaw-inference-switch.sh` makes the same change for `nemoclaw inference set`. Existing post-switch checks still call `check_inference_local`, `check_hermes_api_chat`, `check_sandbox_inference`, and `check_openclaw_agent_turn`.
• Active overlapping E2E changes should be reconciled (test/e2e/test-hermes-inference-switch.sh:466): The patch applies to files that still exist, but trusted drift evidence shows active PR overlap on both touched E2E scripts, including an apparently similar Hermes inference-switch verification change. This raises a small risk of duplicate or conflicting test adjustments.
  • Recommendation: Before landing, compare this diff with the overlapping E2E PRs and either consolidate the duplicated Hermes change or ensure the OpenClaw addition here remains compatible with the other pending changes.
  • Evidence: Trusted overlap data lists PR fix(e2e): skip redundant inference verify in hermes-inference-switch #4109 touching `test/e2e/test-hermes-inference-switch.sh` with title `fix(e2e): skip redundant inference verify in hermes-inference-switch`, PR fix(e2e): widen routing retry window in full-e2e security-posture test #4110 also touching the Hermes script, and PR fix(e2e): handle top-level payloads in openclaw agent JSON output #4030 touching `test/e2e/test-openclaw-inference-switch.sh`.

🌱 Nice ideas

• None.

Workflow run details

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

[hunglp6d]

Closed as a duplicate of #4111