## Release range
- Previous release: `v0.0.137` at `f00fb1dafe2113940b89bcd82f0561adf2696a7b`
- Candidate: `7312e7f49e3e96125b8237968e2552a48e2cb1b6`
- Candidate selection: current-main
- Commits: 12
- Risky files detected: 62
## QA context
### Risky areas
- Workflow / enforcement
- Credentials / inference
- Sandbox / policy / SSRF
- Onboarding / host glue
### Suggested test focus
- Onboarding wizard and sandbox creation
- Policy enforcement, network egress, and SSRF protections
- CI checks, pre-commit hooks, and DCO declarations
- Credential storage and inference provider routing
## Canonical release entry
- Path: `docs/changelog/2026-10-22.mdx`
- Entry:
## v0.0.138
NemoClaw v0.0.138 tightens installation and image-build security and fixes onboarding, local inference, and timeout recovery.
The inference guides also clarify when hosted custom endpoints use native providers or the shared route.
- Linux system-mode Ollama installation verifies a pinned official installer before running it and stops if the downloaded content cannot be verified.
The repository `.dockerignore` now excludes credential, key, and certificate names at every directory depth, including nested files in Docker build contexts.
Related change: [PR #13614](https://github.com/NVIDIA/NemoClaw/pull/13614).
- New hosted custom endpoint selections outside the Personal tier can use sandbox-attached providers when supported.
Fresh Personal-tier hosted selections stay on `inference.local` without a native attachment, and retained shared selections remain on their recorded route.
The provider guides now describe these paths for OpenClaw, Hermes, Deep Agents Code, and Pi.
Related changes: [PR #13528](https://github.com/NVIDIA/NemoClaw/pull/13528), [PR #13609](https://github.com/NVIDIA/NemoClaw/pull/13609).
- New local-inference profiles for `host.openshell.internal` and `host.docker.internal` admit IPv6 unique-local addresses as well as private IPv4 addresses.
Existing sandboxes keep their recorded permissions until the local provider is reselected.
Public, link-local, and loopback IPv6 addresses remain excluded.
Related change: [PR #13582](https://github.com/NVIDIA/NemoClaw/pull/13582).
- `NEMOCLAW_STATUS_PROBE_TIMEOUT_MS` and `NEMOCLAW_SANDBOX_EXEC_TIMEOUT_MS` values above `2147483647` milliseconds now fall back to their defaults instead of timing out immediately.
Related change: [PR #13550](https://github.com/NVIDIA/NemoClaw/pull/13550).
- Onboarding records the requested compatible-endpoint reasoning effort with the new sandbox so resume and rebuild use that selection.
Related change: [PR #12814](https://github.com/NVIDIA/NemoClaw/pull/12814).
- Interactive onboarding keeps `Ctrl-C` working when prompt output is captured.
Closing stdin during a prompt or between questions cancels with a nonzero exit and preserves the session for `nemoclaw onboard --resume`.
Related changes: [PR #12210](https://github.com/NVIDIA/NemoClaw/pull/12210), [PR #12216](https://github.com/NVIDIA/NemoClaw/pull/12216).
- The Hermes managed Dockerfile pins a base image that satisfies the current security inventory.
Related change: [PR #13581](https://github.com/NVIDIA/NemoClaw/pull/13581).
## Documentation coverage
- Latest included cumulative docs PR: [#13586](https://github.com/NVIDIA/NemoClaw/pull/13586), `docs: prepare v0.0.138 documentation`.
- Final PR commit: `6a7a0105b963d0ea19d735363550afcac10d9584`; squash merge commit: `7312e7f49e3e96125b8237968e2552a48e2cb1b6`.
- Final automated refresh coverage commit: `95d0ed24726403b717e62bc29a303ed9e8350ec4`.
- First-parent commits and merged PRs after that coverage point through the candidate:
- `f88e1358a0d912b574206d0518c079213734008b`: [PR #13583](https://github.com/NVIDIA/NemoClaw/pull/13583), E2E public NVIDIA model selection.
- `fb4a61d1ffaf6dc63da3b079fd3a639a79ba650e`: [PR #12216](https://github.com/NVIDIA/NemoClaw/pull/12216), stdin EOF cancellation.
- `1e2b7d311ae585959cb9bfb21a90b1a603d4cffd`: [PR #12814](https://github.com/NVIDIA/NemoClaw/pull/12814), requested reasoning effort.
- `841b01d30e6cda5ae2e0ceb93c0604ebf9443364`: [PR #13581](https://github.com/NVIDIA/NemoClaw/pull/13581), Hermes base-image pin.
- `33cc11b635072f2f6fa8c049343c105592898903`: [PR #12210](https://github.com/NVIDIA/NemoClaw/pull/12210), Ctrl-C during captured prompts.
- `6669f7047fc05e3a9897f87411a1fe9e2810d45f`: [PR #13582](https://github.com/NVIDIA/NemoClaw/pull/13582), private IPv6 host bridges.
- `db9dffbda4d721d7b8954cb79ad780a0a2ce2756`: [PR #13550](https://github.com/NVIDIA/NemoClaw/pull/13550), timer limit handling.
- `86d3c838c023dcfebf4d4538ec3b9981a22322a2`: [PR #13609](https://github.com/NVIDIA/NemoClaw/pull/13609), hosted custom routing guidance.
- `169c038ca088a73a10ad5e76310be455147b261b`: [PR #13614](https://github.com/NVIDIA/NemoClaw/pull/13614), installer and build-context security.
- `efda41bde34a60900ca5bc3c7bb136e4c86cfca6`: [PR #13597](https://github.com/NVIDIA/NemoClaw/pull/13597), E2E registry process boundary.
- `7312e7f49e3e96125b8237968e2552a48e2cb1b6`: [PR #13586](https://github.com/NVIDIA/NemoClaw/pull/13586), cumulative release documentation.
- Changed paths: all four under `docs/`: `docs/changelog/2026-10-22.mdx`, `docs/inference/set-up-ollama.mdx`, `docs/manage-sandboxes/recover-rebuild-sandboxes.mdx`, and `docs/reference/commands.mdx`.
- Review and checks: eligible approval on final PR head; 20 successful checks (label-release-target, changes, Analyze (javascript-typescript), Docs / Validate and publish PR preview, Docs / Verify CLI and installer reference parity, codebase-growth-guardrails, commit-lint, require-maintainer-edits, dco-check, candidate-upgrade-admission, CodeQL (javascript-typescript), check-hash, Analyze (python), CodeQL (python), docs-only-checks, candidate-sdk-package, ShellCheck SARIF, checks, CodeQL, ShellCheck); 12 skipped checks (request, hugging-face-models, openshell-sdk-package, reviewed-npm-audit, real-openclaw-dist-harness, static-checks, compile-artifacts, installer-integration, plugin-tests, build-typecheck, cli-test-shards, cli-tests); no failed or pending checks.
- Automated review: [PR Review Advisor run 39810875932](https://github.com/NVIDIA/NemoClaw/actions/runs/39810875932) passed its blocker gate on the final PR head; nine specialists recorded zero findings and zero E2E recommendations.
- Open managed docs PRs: none.
- Maintainer decision: Proceed with the candidate as shown.
## Base and managed image evidence
- Base-image candidate: `7312e7f49e3e96125b8237968e2552a48e2cb1b6`
- Evidence: [E2E push run 39820777714, attempt 1](https://github.com/NVIDIA/NemoClaw/actions/runs/39820777714) on `main`, with successful [base-image-publication job 119972269845](https://github.com/NVIDIA/NemoClaw/actions/runs/39820777714/job/119972269845), completed 2026-10-22 17:17:48 UTC. The candidate-bound aggregate verified applicable base and managed-image publication.
## General E2E decision
- Newest identifiable full manual main run: [39803230756, attempt 2](https://github.com/NVIDIA/NemoClaw/actions/runs/39803230756), testing `efda41bde34a60900ca5bc3c7bb136e4c86cfca6`, which differs from the planned candidate. It began independently after the first decision. Created 2026-10-22 17:37:56 UTC, started 17:37:55 UTC, and last updated 17:45:31 UTC. At inspection 17:48:39 UTC, age was 10 minutes 43 seconds. The attempt was in progress: 50 jobs succeeded, 3 failed, 29 were in progress, and 15 were skipped. Release qualification had not started.
- Attempt 2 [base-image-publication](https://github.com/NVIDIA/NemoClaw/actions/runs/39803230756/job/119983726898) succeeded at 17:39:38 UTC. [Exact staging Brev Launchable](https://github.com/NVIDIA/NemoClaw/actions/runs/39803230756/job/119983726865) was still in progress.
- Attempt 2 failures at inspection: [Hermes GPU startup fallback](https://github.com/NVIDIA/NemoClaw/actions/runs/39803230756/job/119984576243) failed after its sandbox container exited before stable Ready; the retained artifact shows cleanup succeeded, while observed network denials do not establish the cause. [Hermes native NVIDIA inference switch](https://github.com/NVIDIA/NemoClaw/actions/runs/39803230756/job/119984584159) failed when a CLI adapter form timed out after 150 seconds. [Pi managed runtime qualification](https://github.com/NVIDIA/NemoClaw/actions/runs/39803230756/job/119984583929) failed; its cause was not classified at this inspection.
- Earlier [attempt 1](https://github.com/NVIDIA/NemoClaw/actions/runs/39803230756), created and started 15:16:15 UTC and last updated 16:09:13 UTC, completed with failure. Its [Release qualification](https://github.com/NVIDIA/NemoClaw/actions/runs/39803230756/job/119938175059) started 16:08:20 UTC and failed at 16:08:29 UTC. [Base-image-publication](https://github.com/NVIDIA/NemoClaw/actions/runs/39803230756/job/119910239130) failed on a workflow-history pagination race. [Exact staging Brev Launchable](https://github.com/NVIDIA/NemoClaw/actions/runs/39803230756/job/119910239886) failed after restart because the in-sandbox `inference.local` health probe did not return a trusted result. Dependent full-suite jobs were skipped, including catalogue, managed-image, Hermes, messaging, and cloud-onboard paths.
- Exact-candidate automatic push run: [39820777714, attempt 1](https://github.com/NVIDIA/NemoClaw/actions/runs/39820777714), tested `7312e7f49e3e96125b8237968e2552a48e2cb1b6`, completed successfully at 2026-10-22 17:23:05 UTC. [Relevant E2E](https://github.com/NVIDIA/NemoClaw/actions/runs/39820777714/job/119975573648) passed. Its Release qualification was skipped because it was a push run, not a full manual run.
- Maintainer E2E choice: Proceed with the active status shown. This choice preceded the three attempt-2 failures in the inspection snapshot; final tag confirmation follows review of this updated brief.
- Requested focused or full runs by this release decision: none. Attempt 2 was started independently.
Exceptions: The newest full manual E2E run tested earlier commit efda41bde34a60900ca5bc3c7bb136e4c86cfca6 and is unresolved, with three failed jobs and other jobs still running at the recorded inspection. Attempt 1 failed image evidence collection and staging inference health. Proceeding because the exact candidate passed automatic Relevant E2E and image publication.