fix(onboard): reuse a running Ollama daemon on WSL mirrored networking

[hunglp6d]

Summary

WSL express install aborted when Ollama was already running on the Windows host and WSL used mirrored networking. The daemon answers on the distro's own 127.0.0.1 there, so isWindowsHostOllama read false, the provider menu kept the install entry, and onboarding tried to reinstall through PowerShell interop. Onboarding now reuses the running daemon on that route, and the model pull no longer shells out to an ollama binary that does not exist in WSL.

Related Issue

Fixes #7472

Changes

• resolveRequestedProviderSelection collapses an install-windows-ollama request to the running ollama entry when a daemon already answers, before the requested key is matched. A Windows-host daemon on an unsupported container runtime still falls through to the existing unsupported-windows-host-ollama rejection.
• pullOllamaModel takes the HTTP pull path whenever no local ollama binary exists, not only when the resolved host is host.docker.internal. On mirrored networking the CLI branch ran and failed with ollama: command not found against a daemon that answered /api/tags.
• setup-nim-flow passes the existing ollamaRunning probe result into provider selection.
• install-ollama is deliberately untouched: resolveOllamaInstallMenuEntry keeps that entry for a running-but-stale daemon, and collapsing it would skip the Ollama upgrade path. provider-selection.test.ts locks that scope.

The new ollamaRunning input on ResolveRequestedProviderSelectionInput is optional rather than required. Current requirement: test/onboard-selection.test.ts sits exactly at its 4769-line legacy budget in ci/test-file-size-budget.json, and adding the field to its helper pushed it to 4770 and failed the test-file-size-budget hook. A required field would grow a file the repository is ratcheting down. The single production consumer, setup-nim-flow.ts, passes it explicitly. preferManagedVllmDefault on the same interface is already optional. The absent-value path is exercised by test/onboard-selection.test.ts and test/onboard-selection-windows-provider-rejection.test.ts, whose helpers omit the field, and the false path by still installs on the Windows host when no daemon responds (#7472).

Type of Change

• 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

• Tests added or updated for changed behavior
• Existing tests cover changed behavior — justification:
• Tests not applicable — justification:
• Docs updated for user-facing behavior changes
• Docs not applicable — justification: The documentation already describes the intended behavior. docs/get-started/windows-preparation.mdx:170 states that NemoClaw can use an already-running Windows-host daemon; this fix restores that promise on mirrored networking. No command, flag, environment variable, configuration key, or error string is added or renamed.
• 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

• Documentation writer subagent reviewed the completed changes
• Result: no-docs-needed
• Evidence: The reviewer searched docs/ for WSL onboarding, Windows-host Ollama, mirrored networking, express install, and OLLAMA_HOST=0.0.0.0 and examined docs/get-started/windows-preparation.mdx:169-180, docs/inference/set-up-ollama.mdx:39-40,101-127, docs/reference/troubleshooting.mdx:2650-2662, and docs/resources/prompt-assets/windows-wsl.md:20. No passage becomes inaccurate: the pages describe the documented reuse behavior this fix restores, and the host.docker.internal references remain valid for bridged networking. The reviewer also checked the new code comments and test titles against WRITING.md and .agents/skills/_shared/controlled-words.md; all findings were applied before this commit. No docs/changelog/YYYY-MM-DD.mdx entry is required, because that file belongs to the pre-tag release-notes workflow rather than a contributor PR.
• Agent: Claude Code documentation writer subagent

DGX Station Hardware Evidence

• Tested on DGX Station
• Tested commit:
• Station profile/scenario:
• Result:
• Supporting evidence:

Verification

• PR description includes a Signed-off-by: line and every commit appears as Verified in GitHub
• 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
• Targeted behavior tests pass for the current change set, or tests are marked not applicable above — command/result or justification: npx vitest run --project cli src/lib/onboard/provider-selection.test.ts src/lib/onboard/setup-nim-flow.test.ts src/lib/inference/ollama/proxy.test.ts passed 53/53. npx vitest run --project integration test/onboard-selection.test.ts test/onboard-selection-windows-provider-rejection.test.ts test/ollama-pull-timeout.test.ts passed 73/73. npm run typecheck:cli and npm run checks:repository passed. Reverting only the three source files turns exactly the three new #7472 tests red and leaves the other 50 green.
• Applicable broad gate passed — npm test for broad runtime/test-harness changes; npm run check for repo-wide validation/coverage changes — command/result:
• Quality Gates section completed with required justifications or waivers
• No secrets, API keys, or credentials committed
• npm run docs builds without warnings (doc changes only)
• Doc pages follow the style guide (doc changes only)
• New doc pages include SPDX header and frontmatter (new pages only)

---

Signed-off-by: Hung Le hple@nvidia.com

Summary by CodeRabbit

• New Features

  • Improved Ollama setup across Windows, WSL, and local environments.
  • Reuses an already running Ollama service when available.
  • Automatically selects CLI- or HTTP-based model downloads based on the available environment.
  • Supports running Ollama through WSL mirrored networking.

• Bug Fixes

  • Added fallback handling when the local Ollama command-line tool is unavailable.
  • Improved provider selection for Windows-hosted and WSL-based Ollama instances.

[copy-pr-bot]

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]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 8090337d-2a2f-41ef-8608-6f5315cc4202

📥 Commits

Reviewing files that changed from the base of the PR and between be686a2 and bf003cd.

📒 Files selected for processing (3)

• src/lib/inference/ollama/proxy.test.ts
• src/lib/onboard/setup-nim-flow.test.ts
• src/lib/onboard/setup-nim-flow.ts

🚧 Files skipped from review as they are similar to previous changes (3)

• src/lib/onboard/setup-nim-flow.ts
• src/lib/onboard/setup-nim-flow.test.ts
• src/lib/inference/ollama/proxy.test.ts

---

📝 Walkthrough

Walkthrough

Ollama onboarding now reuses detected Windows-host daemons in supported WSL environments. Model pulling now selects HTTP when the local CLI is unavailable or the host is Docker-internal, and otherwise uses the CLI.

Changes

Ollama runtime selection

Layer / File(s)	Summary
Daemon-aware provider resolution src/lib/onboard/provider-selection.ts, src/lib/onboard/provider-selection.test.ts	Provider selection accepts ollamaRunning and maps supported Windows-host installation requests to the running ollama option. Tests cover running, unavailable, and WSL-local daemon cases.
Setup flow runtime wiring src/lib/onboard/setup-nim-flow.ts, src/lib/onboard/setup-nim-flow.test.ts	Setup passes Ollama running state into provider resolution. Tests cover WSL mirrored networking and loopback access.
Model pull transport dispatch src/lib/inference/ollama/proxy.ts, src/lib/inference/ollama/proxy.test.ts	Model pulls use HTTP for Docker-internal hosts or missing local CLIs. CLI pulls remain the default when the executable is available. Dispatch tests mock CLI detection and streamed HTTP output.

Estimated code review effort: 3 (Moderate) | ~25 minutes

Possibly related PRs

• NVIDIA/NemoClaw#5432: Related Ollama onboarding changes cover Windows-host daemon state and setup flow handling.

Suggested labels: bug-fix, area: onboarding, area: inference, platform: wsl

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 40.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: reusing a running Ollama daemon on WSL mirrored networking.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches 💡 1

📝 Generate docstrings 💡

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/mirrored-ollama-reuse

---

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

[github-code-quality]

Code Coverage Overview

Languages: TypeScript

TypeScript / code-coverage/plugin

The overall coverage in commit bf003cd in the fix/mirrored-ollama-... branch remains at 96%, unchanged from commit ad26164 in the main branch.

TypeScript / code-coverage/cli

The overall coverage in commit bf003cd in the fix/mirrored-ollama-... branch remains at 81%, unchanged from commit 643a4ab in the main branch.

Show a code coverage summary of the most impacted files.

File	main 643a4ab	fix/mirrored-ollama-... bf003cd	+/-
src/lib/actions...oy-execution.ts	95%	89%	-6%
src/lib/platform.ts	89%	84%	-5%
src/lib/shields/index.ts	70%	69%	-1%
src/lib/shields...nsition-lock.ts	88%	87%	-1%
src/lib/policy/index.ts	59%	59%	0%
src/lib/actions...licy-channel.ts	80%	81%	+1%
src/lib/actions...ess-recovery.ts	82%	84%	+2%
src/lib/inferen...ollama/proxy.ts	26%	31%	+5%
src/lib/onboard...ization-deps.ts	58%	83%	+25%
src/lib/actions...ce-preflight.ts	50%	78%	+28%

---

_{Updated August 04, 2026 07:31 UTC}

[coderabbitai]

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 `@src/lib/inference/ollama/proxy.test.ts`:
- Around line 432-440: Update the test case around loadProxyForDispatch and
“pulls over HTTP when the daemon resolves on the Windows host” to set
hasLocalCli: true, isolating the Docker-host predicate from the
missing-local-CLI condition while preserving the existing HTTP-versus-bash
assertions.

🪄 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: 8da61e8d-8a3b-402c-b18d-d621b763a4b9

📥 Commits

Reviewing files that changed from the base of the PR and between 95a9d6f and be686a2.

📒 Files selected for processing (6)

• src/lib/inference/ollama/proxy.test.ts
• src/lib/inference/ollama/proxy.ts
• src/lib/onboard/provider-selection.test.ts
• src/lib/onboard/provider-selection.ts
• src/lib/onboard/setup-nim-flow.test.ts
• src/lib/onboard/setup-nim-flow.ts

[github-actions]

PR Review Advisor — Informational

Advisor assessment: Informational / low confidence
Next action: No advisor follow-up needed.
Findings: 0 blockers · 0 warnings · 0 suggestions
Status: Same-session synthesis validation failed; the advisor result is incomplete.

Model lanes

• GPT-5.6 Terra (primary): Completed · low confidence · 0 blockers · 0 warnings · 0 suggestions
• Nemotron 3 Ultra (second opinion): Failed

Second-opinion terminology and E2E selections are advisory. They do not change the primary assessment or E2E / PR Gate.

3 semantic terminology decisions

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

• established — Windows-host daemon at src/lib/inference/ollama/proxy.ts:868: Keep the established term.
• established — mirrored networking at src/lib/onboard/provider-selection.ts:112: Keep the established term.
• justified — local `ollama` binary at src/lib/inference/ollama/proxy.ts:866: Keep the modifier because CLI-versus-HTTP pull dispatch depends on executable availability, not daemon reachability.

E2E guidance

Advisory only. E2E / PR Gate selects and runs jobs independently.

Recommended E2E: inference-routing, network-policy, onboard-repair, onboard-resume, cloud-onboard

Workflow run details

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