feat(installer): add Spark express vLLM profile option by ericksoa · Pull Request #8512 · NVIDIA/NemoClaw

ericksoa · 2026-08-07T02:58:32Z

Summary

Completes the accepted DGX Spark Express option 2 capability in #8379:

Adds the fixed catalog-backed vLLM profile choice and physical qualification target.
Publishes authenticated host-local vLLM on exactly 127.0.0.1:8000 and the current private openshell-docker bridge gateway on port 8000.
Keeps the sandbox route at https://inference.local while host validation remains on loopback.
Pins every authenticated managed-vLLM Docker operation to the physical default daemon.
Revalidates exact sandbox and container identities before live-target cleanup.

Type of Change

Code change with doc updates

Quality Gates

Tests added or updated for changed behavior
Docs updated for user-facing behavior changes
Sensitive paths changed (onboarding, inference, Docker publication, and E2E cleanup)
Sensitive-path review completed; private-publication and Docker-daemon findings are resolved

Documentation Writer Review

Documentation writer subagent reviewed the completed changes
Result: docs-updated
Evidence:
- docs/get-started/prerequisites.mdx
- docs/inference/choose-local-inference-server.mdx
- docs/inference/set-up-vllm-on-two-dgx-sparks.mdx
- docs/inference/set-up-vllm.mdx
- docs/reference/platform-support.mdx
- docs/resources/prompt-assets/dgx-spark.md
- docs/resources/starter-prompt.md
- test/e2e/README.md
Agent: Codex Desktop

Verification

DGX Spark evidence

Host: DGX Spark with NVIDIA GB10 and 124,610 MiB RAM.
Final hardware-tested source commit: 28cbe0e85088a22b4031298f39b812fbc0627b9d; final rebased PR head 7f1d958562bbc2915704d551f97b1132991c62f2 has the identical Git tree 544ea15a168b5091c9b21437fb7036d4f600dfa8.
OpenShell 0.0.101 final-tree target: 1/1 passed in 5m46s after vLLM readiness in 4m30s.
Managed vLLM published exactly 127.0.0.1:8000 and 172.18.0.1:8000; no wildcard or additional binding was present.
Authenticated sandbox chat, unrelated-egress denial, and identity-guarded cleanup passed.
The earlier OpenShell 0.0.99 target and manual source-checkout onboarding also passed; four dashboard probes and two clean recovery cycles returned HTTP 200.
Full evidence and limitations: feat(installer): add Spark express vLLM profile option #8512 (comment)

Host-specific SSH finding

This machine has a pre-existing custom wildcard Host * SSH stanza that applies Compression yes, ControlMaster auto, a socket ControlPath, and ControlPersist 600 to the OpenShell sandbox host. These are machine-specific settings, not DGX Spark defaults. With them active, both OpenShell 0.0.99 and 0.0.101 failed the primary dashboard forward. A temporary first-match Host sandbox override set compression, multiplexing, and persistence off; the final OpenShell 0.0.101 target then passed onboarding, runtime checks, inference, denial, and cleanup. The original SSH file was restored byte-for-byte.

One earlier failed diagnostic relay attempt left stale OpenShell PID state until the exact record was moved aside. Track that behavior separately with OpenShell. It does not establish a DGX Spark requirement or block this PR's source-checkout qualification.

Signed-off-by: Aaron Erickson aerickson@nvidia.com
Signed-off-by: Carlos Villela cvillela@nvidia.com

Summary by CodeRabbit

New Features
- DGX Spark Express now offers managed vLLM with automatic profile selection or a fixed Qwen3.6 35B-A3B NVFP4 profile.
- Fixed profiles include guided validation, protected settings, and offline model-cache support.
Bug Fixes
- Improved host-local vLLM connectivity, recovery, authentication, and Docker networking.
- Restricted published inference ports to validated loopback and private bridge addresses.
Documentation
- Updated setup, prerequisites, platform support, and troubleshooting guidance for Spark inference profiles.
- Added DGX Spark Express qualification and safety-test documentation.

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

…lm-e2e

copy-pr-bot · 2026-08-07T02:58:35Z

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.

coderabbitai · 2026-08-07T02:58:38Z

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 0c0484e1-a826-4415-8135-44d43d429931

📥 Commits

Reviewing files that changed from the base of the PR and between ea39cc4 and 7f1d958.

📒 Files selected for processing (45)

ci/platform-matrix.json
docs/get-started/prerequisites.mdx
docs/inference/choose-local-inference-server.mdx
docs/inference/set-up-vllm-on-two-dgx-sparks.mdx
docs/inference/set-up-vllm.mdx
docs/reference/platform-support.mdx
docs/resources/prompt-assets/dgx-spark.md
docs/resources/starter-prompt.md
scripts/checks/run-llama-cpp-dgx-spark-qualification.mts
scripts/install.sh
src/lib/agent/onboard.test.ts
src/lib/agent/onboard.ts
src/lib/inference/local-vllm-auth.test.ts
src/lib/inference/local.test.ts
src/lib/inference/local.ts
src/lib/inference/serving/host-local-vllm-selection.test.ts
src/lib/inference/serving/host-local-vllm-selection.ts
src/lib/inference/serving/vllm-host-local-lifecycle.test.ts
src/lib/inference/serving/vllm-host-local-lifecycle.ts
src/lib/inference/serving/vllm-host-local-network.test.ts
src/lib/inference/serving/vllm-host-local-network.ts
src/lib/inference/serving/vllm-managed-support.ts
src/lib/inference/vllm.test.ts
src/lib/inference/vllm.ts
src/lib/onboard.ts
src/lib/onboard/command.test.ts
src/lib/onboard/command.ts
src/lib/onboard/config-sync.test.ts
src/lib/onboard/config-sync.ts
src/lib/onboard/inference-providers/types.ts
src/lib/onboard/openclaw-setup.test.ts
src/lib/onboard/openclaw-setup.ts
src/lib/onboard/setup-inference.ts
src/lib/onboard/setup-nim-vllm.test.ts
src/lib/onboard/setup-nim-vllm.ts
test/e2e/README.md
test/e2e/live/inference-routing-helpers.ts
test/e2e/live/spark-express-vllm.test.ts
test/e2e/mock-parity.json
test/e2e/support/spark-express-vllm-safety.test.ts
test/e2e/support/spark-express-vllm-safety.ts
test/helpers/installer-express-prompt-pty-harness.ts
test/install-express-prompt.test.ts
test/llama-cpp-dgx-spark-qualification-runner.test.ts
test/starter-prompt-docs.test.ts

📝 Walkthrough

Walkthrough

DGX Spark Express now supports automatic managed-vLLM selection and a fixed catalog-backed vLLM profile. Managed host-local vLLM uses validated bridge networking and separate inference and validation endpoints. Onboarding, sandbox sync, documentation, and physical E2E qualification were updated.

Changes

DGX Spark vLLM onboarding and qualification

Layer / File(s)	Summary
Express profile selection and onboarding `scripts/install.sh`, `src/lib/onboard/command.ts`, `docs/...`, `test/install-express-prompt.test.ts`	Express installation now selects managed or fixed vLLM profiles. Fixed-profile overrides are rejected, and profile provenance is preserved during resume.
Managed vLLM bridge networking and validation `src/lib/inference/...`, `docs/inference/...`, `src/lib/onboard/setup-nim-vllm.ts`	Managed vLLM resolves a private Docker bridge gateway, publishes loopback and bridge bindings, uses a dedicated Docker environment, and separates inference from validation URLs.
Noninteractive sandbox configuration sync `src/lib/onboard/config-sync.ts`, `src/lib/agent/onboard.ts`, `src/lib/onboard/openclaw-setup.ts`	Sandbox configuration scripts use shared noninteractive arguments with `--no-tty` and `bash -s`.
Physical Spark qualification and safety validation `test/e2e/live/spark-express-vllm.test.ts`, `test/e2e/support/spark-express-vllm-safety.ts`, `scripts/checks/run-llama-cpp-dgx-spark-qualification.mts`	Physical qualification validates local Docker use, exact resource identity, catalog-derived container configuration, inference, egress denial, diagnostics, and cleanup.

Estimated code review effort: 5 (Critical) | ~90 minutes

Possibly related issues

Restore loopback port publishing in the llama.cpp qualification runner #8667 — The qualification runner restores loopback-only port publication described by this issue.

Possibly related PRs

NVIDIA/NemoClaw#8521 — Adjusts E2E expectations for managed vLLM Docker recovery inspection.
NVIDIA/NemoClaw#8532 — Modifies qualification container port publication in the same runner.
NVIDIA/NemoClaw#8588 — Updates DGX Spark fixed vLLM profile documentation and onboarding behavior.

Suggested labels: feature, area: inference, provider: vllm, platform: dgx-spark, area: onboarding

Suggested reviewers: apurvvkumaria, cv, jyaunches

✨ Finishing Touches 💡 1

🛠️ Fix failing CI checks 💡

Create stacked PR
Commit on current branch

📝 Generate docstrings

Create stacked PR
Commit on current branch

🧪 Generate unit tests (beta)

Create PR with unit tests
Commit unit tests in branch feat/spark-express-vllm-e2e

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

github-actions · 2026-08-07T02:59:22Z

🌿 Preview your docs: https://nvidia-preview-pr-8512.docs.buildwithfern.com/nemoclaw

github-code-quality · 2026-08-07T02:59:22Z

Code Coverage Overview

Languages: TypeScript

TypeScript / code-coverage/plugin

The overall coverage in commit 7f1d958 in the feat/spark-express-v... branch remains at 96%, unchanged from commit ea39cc4 in the main branch.

_{Updated August 10, 2026 02:08 UTC}

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

github-actions · 2026-08-07T03:04:37Z

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 · low confidence · 0 blockers · 0 warnings · 0 suggestions
Model comparison: normalized findings match; normalized terminology decisions differ; normalized E2E selections differ; severity counts match.

5 terminology differences from the second opinion

Advisory only. These are normalized differences from the primary terminology receipt.

fixed catalog-backed vLLM profile at docs/resources/prompt-assets/dgx-spark.md:14: primary classified it as justified; the second opinion classified it as define.
local-model profile at src/lib/onboard/command.ts:39: selected only by the second-opinion lane as established.
host-local vLLM at src/lib/inference/serving/vllm-host-local-network.ts:15: selected only by the second-opinion lane as established.
bounded host publication at docs/inference/choose-local-inference-server.mdx:133: selected only by the second-opinion lane as define.
exact private IPv4 gateway at docs/inference/choose-local-inference-server.mdx:116: selected only by the second-opinion lane as define.

1 additional E2E selection 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.

Second-opinion terminology and E2E selections are advisory. Live E2E does not run automatically for pull requests.

3 semantic terminology decisions

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

justified — automatic serving-profile selection at docs/resources/prompt-assets/dgx-spark.md:13: Keep this term. The option text and documentation state the option-1 and option-2 distinction.
justified — fixed catalog-backed vLLM profile at docs/resources/prompt-assets/dgx-spark.md:14: Keep this term. The modifiers identify catalog ownership and fixed behavior that affect configuration.
define — private bridge gateway at docs/inference/choose-local-inference-server.mdx:145: Keep the term with the surrounding definition of one private IPv4 Gateway value and exact binding requirement.

E2E guidance

Advisory only. A maintainer can dispatch the default E2E suite against this exact revision.

Recommended E2E: cloud-inference, cloud-onboard, managed-image-multiarch-startup, managed-image-protected-runtime, security-posture, full-e2e, hermes-e2e, inference-routing, network-policy, onboard-repair, onboard-resume

Workflow run details

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

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

…lm-e2e

cv

Product scope and security review are incomplete. This draft adds a second DGX Spark Express option, fixed vLLM profile, installer behavior, public documentation, and a physical qualification target without a linked accepted issue or design decision. Record the decision that defines ownership, lifecycle, compatibility, security, and hardware validation for this supported surface, and complete the required sensitive-path review and exact DGX Spark evidence before marking the PR ready. Refresh onto current main and rerun all required checks afterward.

copy-pr-bot · 2026-08-08T06:18:47Z

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.

Signed-off-by: Carlos Villela <cvillela@nvidia.com>

cv · 2026-08-10T01:10:33Z

DGX Spark hardware evidence — final source tree

Tested on a physical DGX Spark with NVIDIA GB10 and 124,610 MiB RAM. The final hardware run used source commit 28cbe0e85088a22b4031298f39b812fbc0627b9d. The rebased PR head 7f1d958562bbc2915704d551f97b1132991c62f2 has the identical Git tree 544ea15a168b5091c9b21437fb7036d4f600dfa8.

OpenShell 0.0.101 passing scope

The protected spark-express-vllm target passed 1/1 in 5m46s.
Express option 2 selected local-model-profile.vllm.spark.v1 and recipe vllm.qwen3-6-35b-a3b-nvfp4.spark-single.v1.
vLLM became ready after 4m30s.
The runtime used model revision 491c2f1ea524c639598bf8fa787a93fed5a6fbce and image nvcr.io/nvidia/vllm@sha256:9204569b17ee4c0eff75194b8e6e458479c8aee18953b5ab9cf359fcdac659e2.
Managed vLLM published only 127.0.0.1:8000 and 172.18.0.1:8000.
Authenticated sandbox chat passed. Unrelated HTTPS CONNECT was denied with curl exit 56 and proxy HTTP 403.
Cleanup rechecked the full sandbox and managed-container identities before removal. No test sandbox, vLLM container, dashboard listener, inference listener, or OpenShell forward remained.
The original SSH configuration was restored byte-for-byte. The test-preparation gateway service was returned to its prior stopped state. OpenShell 0.0.101 remains installed.

The earlier OpenShell 0.0.99 target passed 1/1 in 5m58s. A separate source-checkout onboarding on that version passed all eight phases, returned exactly SPARK_FINAL_OK, produced four dashboard HTTP 200 responses, and passed two clean recovery cycles.

Host-specific SSH finding

This machine has a pre-existing custom wildcard Host * SSH stanza. For the OpenShell sandbox host, its effective settings are Compression yes, ControlMaster auto, a socket ControlPath, and ControlPersist 600. These are machine-specific settings, not DGX Spark defaults.

With those settings active, the OpenShell 0.0.101 run completed vLLM readiness and sandbox creation, but the primary dashboard forward was refused. Onboarding exited 1, the target stopped before inference assertions, and its exact-identity cleanup passed.

A temporary first-match Host sandbox override set Compression no, ControlMaster no, and ControlPersist no. The otherwise identical OpenShell 0.0.101 target then passed onboarding, exact runtime checks, inference, denial, and cleanup. This shows that 0.0.101 alone does not remove the host-specific interference; disabling connection sharing and compression for the OpenShell sandbox host circumvents it.

One earlier failed diagnostic relay attempt left a stale OpenShell PID record until that exact record was moved aside. Clean stops removed their records. Track that stale-record behavior separately with OpenShell; it does not establish a DGX Spark requirement or block this PR.

Signed-off-by: Carlos Villela <cvillela@nvidia.com>

## Summary Make CLI tests independent of scheduler timing and host OpenShell installation state. Increase CLI coverage from 8 to 12 shards; the measured run model lowers the slowest shard from 586.3 seconds to 344.3 seconds while issue #8669 tracks the underlying slow tests. ## Related Issue Relates #8669. Follow-up to #8512. ## Changes - Inject a deterministic monotonic clock into the stale deadline-generation lifecycle-lock test. - Keep the stale-rebuild fixture on supported OpenShell 0.0.101 binaries and directly verify that rebuild does not invoke the installer. - Give the legacy Hermes compatibility fixture setup a 30-second hook timeout so an overloaded coverage shard does not fail Vitest’s default 10-second timeout. - Run PR and main CLI coverage with 12 shards, merge 12 reports, and report failures from every shard. - Exercise shard 12 as valid input, reject shard 13, and verify failed-shard links for shard 12. ## Type of Change - [x] 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 - [x] Tests added or updated for changed behavior - [ ] Existing tests cover changed behavior — justification: - [ ] Tests not applicable — justification: - [ ] Docs updated for user-facing behavior changes - [x] Docs not applicable — justification: This changes internal CI parallelism and deterministic test fixtures only; no supported product surface changes. - [ ] 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 - [x] Documentation writer subagent reviewed the completed changes - Result: `no-docs-needed` - Evidence: - The workflow changes only increase internal CLI coverage parallelism and update failed-shard validation. - The test changes make existing CI and lifecycle fixtures deterministic, including direct verification that rebuild does not invoke the OpenShell installer. - The legacy Hermes compatibility change only extends the heavy test-fixture setup timeout; it does not change production behavior. - No production command, configuration, user workflow, output contract, or supported product surface changes. - Agent: Codex Desktop   ## DGX Station Hardware Evidence - [ ] Tested on DGX Station - Tested commit: Not applicable - Station profile/scenario: Not applicable; `scripts/prepare-dgx-station-host.sh` is unchanged. - Result: Not applicable - Supporting evidence: Not applicable ## Verification - [x] PR description includes a `Signed-off-by:` line and every commit appears as `Verified` in GitHub - [x] 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 - [x] Targeted behavior tests pass for the current change set, or tests are marked not applicable above — the two formerly flaky files passed five runs, 225 test executions; the updated installer probe passed five additional runs; the workflow contract passed 5/5; the source-shape budget passed. The legacy Hermes compatibility suite passed 18/18 tests in each of five normal runs and 18/18 once with V8 coverage; that coverage command exited only on unrelated whole-repository coverage thresholds. - [ ] Applicable broad gate passed — `npm run check` passed every pre-commit check. Its CLI coverage lane completed 23,741 passing tests and found 27 unrelated DGX Spark host-state failures: an installed user service, a group-writable Codex workspace parent, and ARM64 Python executed in an amd64 container. An isolated HOME cleared all 24 service-related failures. - [x] Quality Gates section completed with required justifications or waivers - [x] No secrets, API keys, or credentials committed - [ ] `npm run docs` builds without warnings (doc changes only) - [ ] Doc pages follow the [style guide](https://github.com/NVIDIA/NemoClaw/blob/main/docs/CONTRIBUTING.md) (doc changes only) - [ ] New doc pages include SPDX header and frontmatter (new pages only) --- Signed-off-by: Carlos Villela <cvillela@nvidia.com>  ## Summary by CodeRabbit * **Tests** * Expanded CLI coverage testing from 8 to 12 parallel shards. * Updated coverage validation and reporting to include all 12 shards. * Improved test reliability with deterministic timing and extended setup timeouts. * Verified rebuilds avoid unnecessary installation attempts and related messaging.  --------- Signed-off-by: Carlos Villela <cvillela@nvidia.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

ericksoa added 7 commits August 6, 2026 19:13

feat(installer): route Spark express through local profile

68f74a2

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

docs: repin Spark express prompt asset

abbd226

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

chore: merge current main

ac51ca2

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

feat(installer): add Spark express profile choice

e7073ef

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

docs: clarify Spark express profile choices

4394858

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

docs: repin Spark express prompt asset

bcb9a8c

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

Merge remote-tracking branch 'origin/main' into feat/spark-express-vl…

98615cc

…lm-e2e

ericksoa self-assigned this Aug 7, 2026

test(e2e): keep Spark profile target linear

ec2def0

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

ericksoa added 4 commits August 6, 2026 20:08

test(installer): cover Spark profile override conflicts

f52f8c0

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

test(e2e): map Spark profile mock parity

8641844

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

Merge remote-tracking branch 'origin/main' into feat/spark-express-vl…

26e901a

…lm-e2e

Merge branch 'main' into feat/spark-express-vllm-e2e

d1e76fe

cv requested changes Aug 7, 2026

View reviewed changes

merge: resolve conflicts with main

7c17f31

ericksoa and others added 9 commits August 8, 2026 17:27

Merge branch 'main' into feat/spark-express-vllm-e2e

a987bb7

Merge branch 'main' into feat/spark-express-vllm-e2e

f209185

Merge branch 'main' into feat/spark-express-vllm-e2e

94a9b29

Merge branch 'main' into feat/spark-express-vllm-e2e

d695c10

merge: refresh from main

ab58d9b

fix(onboard): recover Spark fixed-profile installs

eb7dde6

fix(onboard): keep installer profile out of generic selector

a1e2d32

fix(inference): route Spark vLLM through private bridge

34dd5e4

test(inference): type managed bridge resolver

ad4fc15

cv added 7 commits August 9, 2026 14:43

chore(onboard): meet entrypoint growth budget

ff22a4b

test(e2e): keep Spark qualification linear

9175c7b

test(onboard): cover non-default config sync

05363b4

Signed-off-by: Carlos Villela <cvillela@nvidia.com>

Merge branch 'main' into feat/spark-express-vllm-e2e

533cf69

Merge branch 'main' into feat/spark-express-vllm-e2e

9f2a423

fix(inference): align vllm diagnostics with bridge

e8b2faf

test(e2e): allow reasoning chat probe to finish

64cf09f

cv and others added 3 commits August 9, 2026 18:14

test(inference): keep diagnostic mock linear

2f5e9fc

Signed-off-by: Carlos Villela <cvillela@nvidia.com>

Merge branch 'main' into feat/spark-express-vllm-e2e

87f322a

fix(inference): publish llama.cpp qualification port

7f1d958

Signed-off-by: Carlos Villela <cvillela@nvidia.com>

cv marked this pull request as ready for review August 10, 2026 02:16

cv approved these changes Aug 10, 2026

View reviewed changes

cv merged commit 05d607c into main Aug 10, 2026
60 of 67 checks passed

cv deleted the feat/spark-express-vllm-e2e branch August 10, 2026 02:21

github-actions Bot added the v0.0.106 Release target label Aug 10, 2026

cv mentioned this pull request Aug 10, 2026

ci: reduce slow CLI test hotspots #8669

Closed

coderabbitai Bot mentioned this pull request Aug 10, 2026

fix(inference): restore qualification loopback publish #8672

Merged

23 tasks

cv mentioned this pull request Aug 10, 2026

ci: stabilize CLI coverage shards #8675

Merged

23 tasks

coderabbitai Bot mentioned this pull request Aug 10, 2026

fix(inference): restore loopback port publishing in the llama.cpp qualification runner #8676

Closed

22 tasks

This was referenced Aug 10, 2026

fix(inference): stop the vLLM install when the serving port is taken #8699

Open

fix(inference): diagnose blocked llama.cpp bridge #8716

Merged

Navigation Menu

Uh oh!

feat(installer): add Spark express vLLM profile option - #8512

feat(installer): add Spark express vLLM profile option#8512
cv merged 32 commits into
mainfrom
feat/spark-express-vllm-e2e

ericksoa commented Aug 7, 2026 •

edited by coderabbitai Bot

Loading

Uh oh!

copy-pr-bot Bot commented Aug 7, 2026

Uh oh!

coderabbitai Bot commented Aug 7, 2026 •

edited

Loading

Review failed

Walkthrough

Changes

Uh oh!

github-actions Bot commented Aug 7, 2026

Uh oh!

github-code-quality Bot commented Aug 7, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Aug 7, 2026 •

edited

Loading

Uh oh!

cv left a comment

Uh oh!

copy-pr-bot Bot commented Aug 8, 2026

Uh oh!

cv commented Aug 10, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Conversation

ericksoa commented Aug 7, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Type of Change

Quality Gates

Documentation Writer Review

Verification

DGX Spark evidence

Host-specific SSH finding

Summary by CodeRabbit

Uh oh!

copy-pr-bot Bot commented Aug 7, 2026

Uh oh!

coderabbitai Bot commented Aug 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review failed

Walkthrough

Changes

Uh oh!

github-actions Bot commented Aug 7, 2026

Uh oh!

github-code-quality Bot commented Aug 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Code Coverage Overview

TypeScript / code-coverage/plugin

Uh oh!

github-actions Bot commented Aug 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR Review Advisor — No blocking findings reported

Model lanes

E2E guidance

Uh oh!

cv left a comment

Choose a reason for hiding this comment

Uh oh!

copy-pr-bot Bot commented Aug 8, 2026

Uh oh!

cv commented Aug 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

DGX Spark hardware evidence — final source tree

OpenShell 0.0.101 passing scope

Host-specific SSH finding

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

ericksoa commented Aug 7, 2026 •

edited by coderabbitai Bot

Loading

coderabbitai Bot commented Aug 7, 2026 •

edited

Loading

github-code-quality Bot commented Aug 7, 2026 •

edited

Loading

github-actions Bot commented Aug 7, 2026 •

edited

Loading

cv commented Aug 10, 2026 •

edited

Loading