docs: split starter prompt platform instructions - #7061
Conversation
Signed-off-by: Miyoung Choi <miyoungc@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. |
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
📝 WalkthroughWalkthroughThe starter prompt now routes platform-specific Express instructions through dedicated DGX Spark, DGX Station, and Windows WSL Markdown assets. Platform detection, provider and policy handling, contribution guidance, and tests were updated to validate asset loading and installer defaults. ChangesPlatform Express asset routing
Estimated code review effort: 3 (Moderate) | ~20 minutes Sequence Diagram(s)sequenceDiagram
participant Environment
participant StarterPrompt
participant PlatformAsset
participant Installer
Environment->>StarterPrompt: detect OS, WSL, and hardware
StarterPrompt->>PlatformAsset: load one matching asset
PlatformAsset->>Installer: apply accepted platform settings
PlatformAsset-->>StarterPrompt: continue with fallback providers if declined
Suggested labels: 🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
|
🌿 Preview your docs: https://nvidia-preview-pr-7061.docs.buildwithfern.com/nemoclaw |
Code Coverage OverviewLanguages: TypeScript TypeScript / code-coverage/pluginThe overall coverage remains at 96%, unchanged from the TypeScript / code-coverage/cliThe overall coverage in the Show a code coverage summary of the most impacted files.
Updated |
PR Review Advisor — InformationalAdvisor assessment: Informational / high confidence Model lanes
Nemotron output stays in workflow artifacts and does not change the assessment above. E2E guidanceAdvisory only. E2E / PR Gate selects and runs jobs independently. Recommended E2E: None 1 optional E2E recommendation
This automated review informs maintainers. Warnings and suggestions do not require a response. A maintainer decides whether to merge. |
There was a problem hiding this comment.
Actionable comments posted: 4
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
docs/resources/starter-prompt.md (1)
218-223: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winDo not re-request final approval for Express.
The loaded platform assets treat the Express confirmation as approval and explicitly skip the later final-permission prompt, but Line 222 unconditionally asks for final permission. This can produce duplicate consent prompts and break the non-interactive Express contract. Scope the summary and final-permission rule to non-Express flows, or explicitly exempt Express here.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@docs/resources/starter-prompt.md` around lines 218 - 223, Update the installation approval flow around the final-permission step to skip that prompt for Express selections, since Express confirmation already constitutes approval. Keep the summary and final-permission request for non-Express installations, and preserve setting NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE and NEMOCLAW_YES only after the applicable approval.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@docs/CONTRIBUTING.md`:
- Around line 134-135: Document in docs/CONTRIBUTING.md that all platform asset
links use one shared immutable commit SHA or release tag, and specify who
updates the revision and how related assets are updated together. In
docs/resources/starter-prompt.md, replace the mutable main references with the
same pinned revision used by the starter prompt.
In `@docs/resources/prompt-assets/dgx-station.md`:
- Around line 10-14: Add the established third-party-software disclosure used by
the other platform assets before the Express confirmation question in the DGX
Station prompt. Preserve the existing Express defaults and choices, and add a
regression assertion verifying the disclosure appears before acceptance and
before setting NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE.
In `@test/starter-prompt-docs.test.ts`:
- Around line 676-689: Extend the assertions in the expressAssets loop to
require NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE=1 in every asset source. Also add
behavioral coverage that verifies Express proceeds with third-party software
approval rather than pausing or failing at that prompt.
- Around line 24-37: Update the promptAssets validation and its corresponding
assertions in the starter prompt test so each configured URL is explicitly
associated with the asset loaded from its matching path. Verify the observable
routing result for every platform entry, rather than independently checking that
paths and URLs appear in the prompt, preventing swapped platform mappings from
passing.
---
Outside diff comments:
In `@docs/resources/starter-prompt.md`:
- Around line 218-223: Update the installation approval flow around the
final-permission step to skip that prompt for Express selections, since Express
confirmation already constitutes approval. Keep the summary and final-permission
request for non-Express installations, and preserve setting
NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE and NEMOCLAW_YES only after the applicable
approval.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 4484b9a8-2c64-427a-bb46-1c2aff29becb
📒 Files selected for processing (6)
docs/CONTRIBUTING.mddocs/resources/prompt-assets/dgx-spark.mddocs/resources/prompt-assets/dgx-station.mddocs/resources/prompt-assets/windows-wsl.mddocs/resources/starter-prompt.mdtest/starter-prompt-docs.test.ts
Signed-off-by: San Dang <sdang@nvidia.com>
Signed-off-by: San Dang <sdang@nvidia.com>
apurvvkumaria
left a comment
There was a problem hiding this comment.
Requesting changes at exact head 7953cac5cbeeb8c2ac5f882e72085caad292b03c.
-
The DGX Station asset currently labels a provider-preseeded DeepSeek path as Express, but that path skips Station Express host preparation.
docs/resources/prompt-assets/dgx-station.mdsetsNEMOCLAW_PROVIDER=install-vllmbefore invoking the installer. Inscripts/install.sh, a preselected provider bypassesmaybe_offer_express_install, soactivate_express_installnever records the Station platform andensure_station_express_hostdoes not run the pinned driver, Docker, NVIDIA Container Toolkit, CDI, or reboot/resume preparation. A fresh Station can therefore follow these instructions, bypass required preparation, and still be told it selected the maintained Express flow. Please either route unprepared hosts through the official--station-deepseekExpress entry point, or explicitly scope this provider-preseeded path to an already-prepared host and fail closed when the documented readiness prerequisites are absent. It should not be described as the general Station Express path while host automation is disabled. -
The final confirmation omits security-relevant host effects. This asset makes its first confirmation final and suppresses the later summary, but the prompt currently discloses only the model/container, storage, and Deferred status. Before
Choices:, disclose every applicable host mutation, possible reboot/resume behavior, and that Docker-group membership grants root-equivalent control. If the path is changed to prepared-host-only, disclose that host automation will not run and require the readiness checks before accepting. Add regression assertions that these disclosures precede both the choice and the non-interactive acceptance variables.
Also tighten the immutable-asset contract: document the atomic shared-revision update procedure and make CI prove each local platform asset still matches the content at its pinned revision. The current tests validate URL shape and local prose independently, so a future local edit can silently drift from what the starter prompt actually downloads.
The automated checks are green, but they did not cross-check this prompt against the installer's control flow; these are release-blocking documentation and safety issues.
Co-authored-by: Miyoung Choi <miyoungc@nvidia.com> Signed-off-by: Apurv Kumaria <akumaria@nvidia.com>
Co-authored-by: Miyoung Choi <miyoungc@nvidia.com> Signed-off-by: Apurv Kumaria <akumaria@nvidia.com>
Co-authored-by: Miyoung Choi <miyoungc@nvidia.com> Signed-off-by: Apurv Kumaria <akumaria@nvidia.com>
Co-authored-by: Miyoung Choi <miyoungc@nvidia.com> Signed-off-by: Apurv Kumaria <akumaria@nvidia.com>
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@docs/resources/prompt-assets/dgx-station.md`:
- Around line 20-22: Update the documentation around the --check and --verify
modes to call them “non-repairing” rather than “read-only.” Explicitly disclose
that --verify starts GPU test containers through CDI and Docker, which may
create Docker state, consume GPU and storage resources, and produce logs, before
requesting permission; preserve the existing pinned-image and no-pull behavior.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 0978104f-ab22-4a4d-ac50-86fafde4d5a1
📒 Files selected for processing (4)
docs/CONTRIBUTING.mddocs/resources/prompt-assets/dgx-station.mddocs/resources/starter-prompt.mdtest/starter-prompt-docs.test.ts
🚧 Files skipped from review as they are similar to previous changes (1)
- docs/resources/starter-prompt.md
Co-authored-by: Miyoung Choi <miyoungc@nvidia.com> Signed-off-by: Apurv Kumaria <akumaria@nvidia.com>
Co-authored-by: Miyoung Choi <miyoungc@nvidia.com> Signed-off-by: Apurv Kumaria <akumaria@nvidia.com>
Co-authored-by: Miyoung Choi <miyoungc@nvidia.com> Signed-off-by: Apurv Kumaria <akumaria@nvidia.com>
Signed-off-by: Aaron Erickson <aerickson@nvidia.com>
<!-- markdownlint-disable MD041 --> ## Summary Align the DGX Station coding-agent instructions with the versioned installer. The ordinary Station Express path now remains authoritative for its Nemotron Ultra default, while DeepSeek remains available only through the installer’s explicit `--station-deepseek` override. ## Changes - [#7061](#7061) -> `docs/resources/prompt-assets/dgx-station.md`: Replace the DeepSeek-only provider-preseeded path with the installer’s Ultra-default and DeepSeek-override contract. - Delegate Station qualification, host preparation, third-party notice, Express summary, and reboot or login resume to the selected maintained release’s installer. - Stop the coding-agent path from pre-setting provider, model, non-interactive, approval, or Express-skip variables, and stop it from running the Station preparation helper separately. - Preserve Ultra’s approximately 352 GB disclosure, shared Station host-safety boundaries, the temporary metadata override guard, and Deferred support status. - Pin every starter-prompt platform asset URL to the immutable asset commit and update its SHA-256 contract. - Update focused tests to require both installer-supported Station model outcomes and reject the former DeepSeek-only path. ## Type of Change - [ ] Code change (feature, bug fix, or refactor) - [ ] Code change with doc updates - [x] 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: - [x] Docs updated for user-facing behavior changes - [ ] Docs not applicable — justification: - [ ] 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: ## DGX Station Hardware Evidence - [ ] Tested on DGX Station - Tested commit: Not applicable; this PR does not change `scripts/prepare-dgx-station-host.sh` or runtime behavior. - Station profile/scenario: Not applicable. - Result: Not applicable. - Supporting evidence: Installer behavior is covered by the focused automated Express tests below. ## 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 check:diff` passed when hooks were skipped or unavailable - [x] Targeted behavior tests pass for the current change set, or tests are marked not applicable above — `npx vitest run test/starter-prompt-docs.test.ts test/install-express-prompt.test.ts` (88 passed, 1 skipped) - [ ] Applicable broad gate passed — `npm test` for broad runtime/test-harness changes; `npm run check` for repo-wide validation/coverage changes — command/result: Not run; this is a focused documentation contract change. - [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) — passed with 0 errors and 1 Fern warning - [x] 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: Miyoung Choi <miyoungc@nvidia.com> <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **Documentation** * Updated DGX Station guidance to use the official Station Express installer after hardware detection. * Added support for Nemotron Ultra by default and DeepSeek V4 Flash as an explicit option. * Clarified installer consent, interactive prompts, reboot/resume behavior, and fallback provider selection. * Refreshed platform instruction links and clarified installer-specific prompt handling. * **Tests** * Updated validation for revised Station Express instructions, consent messaging, and asset revisions. <!-- end of auto-generated comment: release notes by coderabbit.ai --> --------- Signed-off-by: Miyoung Choi <miyoungc@nvidia.com>
Summary
Move platform-specific Express install guidance out of the general starter prompt so non-DGX users do not receive irrelevant DGX instructions. The starter prompt now loads focused raw Markdown assets only after detecting DGX Spark, DGX Station, or Windows WSL.
Changes
Type of Change
Quality Gates
Verification
Signed-off-by:line and every commit appears asVerifiedin GitHubpre-commit,commit-msg, andpre-pushhooks passed, ornpm run check:diffpassed when hooks were skipped or unavailablenpx vitest run test/starter-prompt-docs.test.ts(15 passed)npm testfor broad runtime/test-harness changes;npm run checkfor repo-wide validation/coverage changes — not applicable to this focused documentation changenpm run docsbuilds without warnings (doc changes only) — build passed; Fern reported two existing warningsSigned-off-by: Miyoung Choi miyoungc@nvidia.com
Summary by CodeRabbit