## Release range

- Previous release: `v0.0.135` at `7a099aeb4f6678d9b9ce6236535e7b41b03c81a2`
- Candidate: `29acfc54caed0a70d77632868ee8f4dc310c8a05`
- Candidate selection: current-main
- Commits: 32
- Risky files detected: 238

## QA context

### Risky areas

- Workflow / enforcement
- Credentials / inference
- Sandbox / policy / SSRF
- Installer / bootstrap
- Onboarding / host glue

### Suggested test focus

- Fresh install and upgrade paths
- 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/2026-10-19.mdx`
- Entry:

## v0.0.136

NemoClaw v0.0.136 expands sandbox-attached inference and adds hosted image input and an experimental local profile for Super VL.
It repairs onboarding and DGX Station Express recovery and strengthens provider ownership checks.
Ordinary operation telemetry now uses the fixed TEST receiver.

- Fresh custom hosted endpoint selections bind the validated endpoint, API, and credential source to a provider attached to the selected sandbox.
  Inference status, health checks, switching, and rebuilds use that recorded route.
  Existing shared-route selections remain on their recorded path.
  Related change: [PR #13201](https://github.com/NVIDIA/NemoClaw/pull/13201).
- New Deep Agents Code, OpenClaw, and Pi local selections use sandbox-attached providers for qualified local endpoints.
  Hermes native-local startup, switching, and rebuild refresh supervisor-issued credential references while preserving the selected route and unrelated settings.
  Related changes: [PR #13203](https://github.com/NVIDIA/NemoClaw/pull/13203), [PR #13293](https://github.com/NVIDIA/NemoClaw/pull/13293), [PR #13007](https://github.com/NVIDIA/NemoClaw/pull/13007), and [PR #13292](https://github.com/NVIDIA/NemoClaw/pull/13292).
- For OpenClaw native-local routes, `status`, `connect`, and `start` now send health requests to the provider recorded for the sandbox.
  They no longer probe an unconfigured shared route for those selections.
  Related change: [PR #13447](https://github.com/NVIDIA/NemoClaw/pull/13447).
- An explicit Experimental Nemotron 3.5 Super VL profile serves one 128 GB DGX Spark with native five-token MTP.
  It rejects 64 GB or unknown-memory systems before installation and retains the Spark Qwen default.
  Physical serving and managed-installer smoke tests passed; complete agent onboarding and sustained-load qualification remain pending.
  Related change: [PR #13357](https://github.com/NVIDIA/NemoClaw/pull/13357).
- The NVIDIA Endpoints model picker includes hosted `nvidia/nemotron-3.5-super-vl-120b-a12b` for OpenClaw.
  Interactive and unattended onboarding enable text and image input for that selection unless an explicit valid input override is set.
  Related change: [PR #13371](https://github.com/NVIDIA/NemoClaw/pull/13371).
- Fresh Model Router onboarding waits for the selected OpenClaw gateway to accept connections before requesting its native restart.
  If startup does not settle within the existing bound or the sandbox identity changes, setup stops before restart.
  Related change: [PR #13481](https://github.com/NVIDIA/NemoClaw/pull/13481).
- DGX Station Express keeps automatic model selection under the serving catalog.
  Installation and resume now retain the selected GPU and model, accept the catalog's automatic receipt, verify an owned vLLM server before reuse, and preserve an unchanged sandbox route.
  Related changes: [PR #13305](https://github.com/NVIDIA/NemoClaw/pull/13305) and [PR #13480](https://github.com/NVIDIA/NemoClaw/pull/13480).
- Credential registration inspects existing operator profiles before changing provider state.
  Sandbox creation verifies NemoClaw-owned extra profiles, and messaging setup rejects a fallback to a provider profile it does not own.
  Related changes: [PR #13353](https://github.com/NVIDIA/NemoClaw/pull/13353), [PR #13356](https://github.com/NVIDIA/NemoClaw/pull/13356), and [PR #13351](https://github.com/NVIDIA/NemoClaw/pull/13351).
- OpenClaw rebuild keeps the recorded dashboard origin when it restores a sandbox.
  The `nemoclaw sandbox mcp` usage message now lists the existing `update` action.
  Related changes: [PR #13198](https://github.com/NVIDIA/NemoClaw/pull/13198) and [PR #12384](https://github.com/NVIDIA/NemoClaw/pull/12384).
- The installer and supported CLI operations now attempt anonymous completion telemetry by default through NVIDIA's fixed TEST receiver.
  QA runs can still add a valid test label; `NEMOCLAW_DISABLE_TELEMETRY=1` suppresses collection and delivery, and Production delivery remains disabled.
  Related change: [PR #13494](https://github.com/NVIDIA/NemoClaw/pull/13494).

## Documentation coverage

- Latest included cumulative docs PR: [#13308](https://github.com/NVIDIA/NemoClaw/pull/13308), merged 2026-10-20 00:46:00 UTC.
- Final PR commit: `372e7aaccf863736b8bb6ba6cd17fdc235c7c261`; merge commit: `29acfc54caed0a70d77632868ee8f4dc310c8a05`.
- Final automated refresh coverage commit: `7db58ebb9abfb187ba0c072f56189eff9dcc589f` (parent of verified bot commit `e3043763b4636543d88ceb9042834f3f489497d5`).
- Later first-parent commits and associated merged PRs (28, including the docs merge):
  - `29acfc54caed0a70d77632868ee8f4dc310c8a05`: [docs: prepare v0.0.136 documentation (#13308)](https://github.com/NVIDIA/NemoClaw/pull/13308).
  - `6d76a7e9aa5aca641df360670aa00394411c302a`: [fix(telemetry): enable ordinary operation capture through TEST (#13494)](https://github.com/NVIDIA/NemoClaw/pull/13494).
  - `c2bfcac9dcedd0fb47186261c11da6d563ecd995`: [test(onboard): honor configured gateway in dashboard fixture (#13500)](https://github.com/NVIDIA/NemoClaw/pull/13500).
  - `29a4a58a7a33ad03e10d388ca4427f3e3ea23db7`: [feat(inference): enable hosted Super VL image inputs (#13371)](https://github.com/NVIDIA/NemoClaw/pull/13371).
  - `d087dbcad90ee639016ba0d259556f024340acb5`: [fix(onboard): wait for routed gateway before native restart (#13481)](https://github.com/NVIDIA/NemoClaw/pull/13481).
  - `88900837e9705c105470c6c3e33835e4b2a9a523`: [test(dcode): set repo root for copied headless check (#13498)](https://github.com/NVIDIA/NemoClaw/pull/13498).
  - `e7edcc49629bd025039b04925de55c9332f499b3`: [fix(e2e): admit approved OpenShell templates in GPU controller (#13486)](https://github.com/NVIDIA/NemoClaw/pull/13486).
  - `ae0dcf273d0ec2d6806d4e9c5ef42281d89ea220`: [test(dcode): validate native inference for issue 13164 (#13363)](https://github.com/NVIDIA/NemoClaw/pull/13363).
  - `af1b779c7c53b015335aa838fd6b71a6994a4956`: [fix(onboard): repair DGX Station Express install and resume (#13480)](https://github.com/NVIDIA/NemoClaw/pull/13480).
  - `b01bb7ffc1bc0e1e7c8b015c7103039998aa9a6e`: [test(e2e): retain portable registry readiness evidence (#13352)](https://github.com/NVIDIA/NemoClaw/pull/13352).
  - `4501ee981d74e2ec88fcd30667625d106575ac82`: [fix(inference): send native local health probes to the recorded provider (#13447)](https://github.com/NVIDIA/NemoClaw/pull/13447).
  - `3198676cf3e7ff99d4ad697f84824cefb0adbdd4`: [fix(e2e): retain PR image contracts for delayed validation (#13417)](https://github.com/NVIDIA/NemoClaw/pull/13417).
  - `762aef4c82dec358e307a771d2b66da352a904e9`: [test(e2e): wait for approved cron completion (#13355)](https://github.com/NVIDIA/NemoClaw/pull/13355).
  - `8da11a7952539e8a990f576c7b353ad8c86cacd0`: [refactor(hermes): share tool gateway profile contract (#13377)](https://github.com/NVIDIA/NemoClaw/pull/13377).
  - `1e82499a048d5a4d85b0876d8c07bd93c8a9a5de`: [fix(hermes): preserve issued native-local credential references (#13292)](https://github.com/NVIDIA/NemoClaw/pull/13292).
  - `4ce30c292c0a518ee713f2d10460d473bb904637`: [fix(providers): verify owned extra profiles before sandbox creation (#13356)](https://github.com/NVIDIA/NemoClaw/pull/13356).
  - `aeada1191114931b5f32a792ceac5497952725db`: [fix(rebuild): keep the recorded OpenClaw dashboard origin (#13198)](https://github.com/NVIDIA/NemoClaw/pull/13198).
  - `74ebf7e9b8c7226da05c53b706ba989bbc7646bb`: [fix(messaging): reject unowned provider profile fallback (#13351)](https://github.com/NVIDIA/NemoClaw/pull/13351).
  - `b5a4030c49e30001aed3646160c7b7ef799423d9`: [fix(credentials): inspect operator profiles before registration (#13353)](https://github.com/NVIDIA/NemoClaw/pull/13353).
  - `48738baa32df99af2ae74248b40201fef1e30220`: [test(pi): require plain-text qualification replies (#13367)](https://github.com/NVIDIA/NemoClaw/pull/13367).
  - `f1dcab784d273a56f1adc6292c7336b81fb41767`: [feat(inference): add Super VL MTP recipe for DGX Spark (#13357)](https://github.com/NVIDIA/NemoClaw/pull/13357).
  - `64ca932954f68f2a3019b404d76435b2039dcd3f`: [feat(pi): support native local inference (#13007)](https://github.com/NVIDIA/NemoClaw/pull/13007).
  - `ae41f9aeb63484b812b0372087bd86b2d0f12cc0`: [fix(cli): list update in mcp usage (#12384)](https://github.com/NVIDIA/NemoClaw/pull/12384).
  - `9edf19440e6b6108ae97dfa9d591a66e293a870b`: [fix(e2e): bind credentialless DCode fixture to loopback (#13339)](https://github.com/NVIDIA/NemoClaw/pull/13339).
  - `321ac1e8d80e518ef6fcbec353ff5c7ef25250cd`: [fix(e2e): select reviewed OpenShell runtime pins (#13335)](https://github.com/NVIDIA/NemoClaw/pull/13335).
  - `f860e2eccb93ccba546cf45c71998e88e60a13d0`: [feat(inference): attach OpenClaw local providers (#13293)](https://github.com/NVIDIA/NemoClaw/pull/13293).
  - `486ff49e9fe58cedd959c1454cb25c875291480c`: [test(e2e): replace credentialed hosted smokes with contracts (#13284)](https://github.com/NVIDIA/NemoClaw/pull/13284).
  - `bdf2172e47538adf94ce5e8b07aeea6010a58dfd`: [feat(inference): bind custom hosted endpoints to native providers (#13201)](https://github.com/NVIDIA/NemoClaw/pull/13201).
- Changed paths: 20 files; all under `docs/**`, with no disallowed paths. Manual maintainer updates cover the work merged after the bot refresh through `6d76a7e9aa5aca641df360670aa00394411c302a`.
- Review and checks: APPROVED on exact final PR head by `rsliter`; merge state CLEAN. All required contexts passed. The complete rollup has 32 successes, 25 skips, and one cancelled older `commit-lint` run; the later `commit-lint` passed. Successes include docs link, preview, CLI/installer parity, docs-only checks, CodeQL, CodeRabbit, DCO, hash, and repository checks. No failed or pending checks. [PR check details](https://github.com/NVIDIA/NemoClaw/pull/13308/checks).
- Open managed docs PRs: None at inspection.
- Maintainer decision: Proceed with the candidate as shown.

## Base and managed image evidence

- Base-image candidate: `29acfc54caed0a70d77632868ee8f4dc310c8a05`
- Evidence: Successful `base-image-publication` on candidate `29acfc54caed0a70d77632868ee8f4dc310c8a05`, completed 2026-10-20 00:49:51 UTC. [E2E workflow run 39349583464, attempt 1](https://github.com/NVIDIA/NemoClaw/actions/runs/39349583464), [verifier job 118391008419](https://github.com/NVIDIA/NemoClaw/actions/runs/39349583464/job/118391008419). The job is bound to `.github/workflows/e2e.yaml`, `event=push`, `head_branch=main`, and the planned SHA; its checked-in verifier covers the applicable GHCR base and managed-image publications.

## General E2E decision

- Candidate SHA: `29acfc54caed0a70d77632868ee8f4dc310c8a05`.
- Newest full main run: [workflow 39344562306, attempt 1](https://github.com/NVIDIA/NemoClaw/actions/runs/39344562306), SHA `6d76a7e9aa5aca641df360670aa00394411c302a` (preceding commit; candidate adds documentation only). Created and started 2026-10-19 23:55:47 UTC; last updated 2026-10-20 01:21:23 UTC; age at 2026-10-20 01:21:57 UTC inspection: 1 hour 26 minutes 10 seconds. Completed with failure.
- Release qualification: failure; started 2026-10-20T01:20:47Z; completed 2026-10-20T01:20:51Z; [job](https://github.com/NVIDIA/NemoClaw/actions/runs/39344562306/job/118400999797).
- Other failed jobs:
  - ubuntu-repo-cloud-langchain-deepagents-code: Repository install onboarding and hosted inference succeed [langchain-deepagents-code; Ubuntu managed-runtime host; NVIDIA hosted inference] [docker]: failure, completed 2026-10-20T00:28:27Z; [job](https://github.com/NVIDIA/NemoClaw/actions/runs/39344562306/job/118379896662).
  - Networking: Personal public fetch without Brave Search or Tavily Search API keys (docker) / NVIDIA inference API key: failure, completed 2026-10-20T00:16:59Z; [job](https://github.com/NVIDIA/NemoClaw/actions/runs/39344562306/job/118379900735).
  - Performance: bounds hosted inference turns for OpenClaw and Hermes (docker) / NVIDIA inference API key: failure, completed 2026-10-20T00:18:29Z; [job](https://github.com/NVIDIA/NemoClaw/actions/runs/39344562306/job/118379901173).
  - Install: bootstraps NemoClaw and completes hosted inference (docker) / NVIDIA inference API key: failure, completed 2026-10-20T00:18:38Z; [job](https://github.com/NVIDIA/NemoClaw/actions/runs/39344562306/job/118379901371).
- Skipped jobs (16): openshell-dev-artifact; Authenticate and compile the DGX Spark qualification plan; external-gateway-health; Build pinned Portable Podman 5.7 toolchain; native-runtime-qualification-producer-plan; Exact staging Brev Launchable identity; Portable Hermes finalization (Podman 5.7, NVIDIA GPU); Build pinned native Podman toolchain / ${{ matrix.architecture }}; Protected llama.cpp DGX Spark qualification; matrix.jobName; Aggregate native runtime qualification evidence; jetson-nvmap-gpu; ${{ matrix.display_name }} (${{ matrix.runtime_provider }}); dgx-station-express; MCP bridge dev (${{ matrix.agent }}, ${{ matrix.runtime_provider }}); Relevant E2E. No cancelled, queued, or still-running jobs at inspection.
- Candidate automatic push E2E: [workflow 39349583464, attempt 1](https://github.com/NVIDIA/NemoClaw/actions/runs/39349583464), SHA `29acfc54caed0a70d77632868ee8f4dc310c8a05`, completed successfully 2026-10-20 00:53:44 UTC; [Relevant E2E](https://github.com/NVIDIA/NemoClaw/actions/runs/39349583464/job/118392584396) passed. This was a push selection, not a full run.
- Maintainer E2E choice: Proceed with the status as shown. No additional E2E run requested.

Exceptions: Proceeding to advance stability work despite unresolved failures in the preceding commit's full E2E (four test jobs and Release qualification); the exact candidate's Relevant E2E and required image-publication verifier passed.