Repository navigation
docs: restore Hermes variant page structure - #11926
Conversation
Signed-off-by: Julie Yaunches <jyaunches@nvidia.com>
|
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. |
|
Caution Review failedThe pull request is closed. ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Enterprise Run ID: 📒 Files selected for processing (2)
📝 WalkthroughWalkthroughChangesThe PR adds a Hermes-specific NemoClaw quickstart and an Ollama setup page. The documentation covers installation, onboarding, sandbox launch, integrations, observability, scripted workflows, and local inference prerequisites. NemoClaw documentation
Estimated code review effort: 2 (Simple) | ~12 minutes Change: Bug fix Suggested reviewers: ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
|
🌿 Preview your docs: https://nvidia-preview-pr-11926.docs.buildwithfern.com/nemoclaw |
## Outcome Initial cleanup in Pi qualification, agent-turn latency, and sessions/agents CLI E2E checks the named OpenShell gateway before deleting a sandbox. A verified absent registration lets onboarding proceed. Failed or ambiguous observations and terminal cleanup failures remain errors. ## Reason Main E2E run 35087621077 failed these jobs before onboarding because a fresh runner had no `nemoclaw` gateway. The fixtures handled an absent sandbox but not an absent gateway. ## Changes - Add a shared gateway observation to `SandboxClient`. It accepts only complete, ordered missing-registration guidance from the pinned gateway-info command. - Use that observation in the three initial cleanup paths. Preserve strict terminal cleanup and forward cleanup. - Resolve the latency cleanup gateway once, so inspection and registration removal use the same name when `OPENSHELL_GATEWAY` overrides it. - Cover malformed, incomplete, repeated, reordered, failed, timed-out, and signaled observations; gateway overrides; and cleanup ordering. - Map Pi qualification to the shared cleanup-client regression suite. ## Verification Candidate: `fcef6f7b17abdf8b18c5fe178975accd7f4ab43e`. Integrated trusted main: `8657616c8171de9371f80c6c1cdaf93f54dffb09`. - 196 focused tests passed across `e2e-clients`, `agent-turn-latency-progress`, and `agent-variant-docs`. Command: `node_modules/.bin/vitest run --project e2e-support --project integration test/e2e/support/e2e-clients.test.ts test/e2e/support/agent-turn-latency-progress.test.ts test/generation/agent-variant-docs.test.ts --testTimeout 120000`. The explicit timeout accommodates cold imports. - Normal commit hooks passed, including formatting, lint, repository checks, E2E phases, source-shape and growth budgets, secret scanning, and commitlint. - The former shard 4 and 12 documentation failures were inherited. Merged repairs #11908 and #11926 are integrated; the retained documentation test passes. - The diff contains no secrets, API keys, or credentials. Normal pre-push publication validation and JavaScript/CLI type-checks passed with `NODE_OPTIONS=--max-old-space-size=8192`. GitHub verified every PR commit. Latest-commit CI and selected trusted E2E results are being collected; those gates remain pending. ## Review notes Alternative review completed for the candidate above. Both CodeRabbit findings and the matching maintainer change request were repaired with protecting tests. The full seven-file diff was reviewed for correctness, gateway identity, absence classification, error propagation, cleanup ordering, and coverage. The local Docker Advisor used trusted main, pinned OpenShell 0.0.116, Landlock, and an isolated controller. It failed before its first specialist: the pinned Pi image pull reported `no space left on device` in the shared VM. No specialist completed. The task sandbox, gateway, controller, and failed cache were cleaned up without stopping other tasks. This is self-review supplemented by the existing CodeRabbit and human findings, not Advisor clearance or independent review. Selected E2E: `cloud-onboard`, `cloud-inference`, `security-posture`, `pi-agent-qualification`, `agent-turn-latency`, and `sessions-agents-cli`. --- Signed-off-by: Julie Yaunches <jyaunches@nvidia.com> <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit - **Bug Fixes** - Improved sandbox cleanup reliability by verifying the configured gateway before attempting deletion. - Missing gateways are handled safely, while malformed, unexpected, timed-out, or failed checks are surfaced. - Sandbox destruction continues even when gateway cleanup cannot proceed. - Standardized pre-onboarding cleanup across end-to-end workflows. - Cleanup now respects configured gateway names and verifies cleanup ordering. - **Tests** - Expanded coverage for gateway detection, custom gateway names, cleanup ordering, timeout behavior, failure handling, and deletion errors. <!-- end of auto-generated comment: release notes by coderabbit.ai --> --------- Signed-off-by: Julie Yaunches <jyaunches@nvidia.com>
Outcome
Restore the Hermes quickstart and Ollama page structure that commit
755f28daccidentally removed, while retaining its Portable Ollama guidance.Why
Main's
CI / Main Branchrun fails two CLI shards because the affected MDX pages no longer start with YAML frontmatter and the Hermes quickstart lost its managed-image onboarding contract.Changes
Verification
npm exec -- vitest run --project integration test/generation/agent-variant-docs.test.tsnpm exec -- vitest run --project integration test/onboarding/onboard-managed-image-buildless-e2e.test.tsnpm run docsgit diff --checkSigned-off-by: Julie Yaunches jyaunches@nvidia.com
Summary by CodeRabbit