<!-- markdownlint-disable MD041 -->
## Summary

Expand the existing "Verify the Active Model" section in
`docs/inference/switch-inference-providers.md` to fully document
`nemoclaw inference get`. The command landed in #3424 with a one-line
entry in `docs/reference/commands.md`, and the inference how-to page
picked it up at a minimum level (command + `--json` flag only). This
change adds the output shape, the JSON shape, and the unconfigured-route
error so users have a documented inverse to `nemoclaw inference set` on
the same page.

## Problem

After running `nemoclaw inference set --provider X --model Y`, the
inference how-to page tells users to run `nemoclaw inference get` and
`nemoclaw inference get --json` but does not show what either prints.
Users either have to run the command themselves to find out what to
expect, or rely on `nemoclaw status` diffs to confirm the route changed.
The unconfigured-route case (gateway has no registered inference route)
was not mentioned at all, so users hitting that error had no docs hint
at what state caused it or how to recover.

## Changes

- `docs/inference/switch-inference-providers.md`: rework the existing
"Verify the Active Model" H2 to add (1) human-readable output
(`Provider: <name>` / `Model: <id>`), (2) JSON output body, and (3) the
non-zero exit + `OpenShell inference route is not configured.` error
string with a pointer to `nemoclaw onboard`. Section title, position,
and the trailing `nemoclaw <name> status` cross-reference are preserved.

The new content uses the verified output format from
[`src/lib/actions/inference-get.ts`](src/lib/actions/inference-get.ts)
and the verified error message.

## Test plan

- [x] `npx prek run --files
docs/inference/switch-inference-providers.md` passes (gitleaks,
markdownlint, docs-to-skills dry-run, skills YAML)
- [x] `commitlint` passes
- [ ] `make docs` (no Sphinx in container; rendered preview will appear
via NemoClaw's docs build job)

## 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)

Signed-off-by: latenighthackathon
<latenighthackathon@users.noreply.github.com>


<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **Documentation**
* Updated instructions for viewing the active inference provider and
model using the `nemoclaw inference get` command, including a `--json`
machine-readable example.
* Clarified behavior when no route is configured (command exits non-zero
with "OpenShell inference route is not configured.") and added guidance
to run `nemoclaw onboard` and verify state before/after switching
routes.

<!-- review_stack_entry_start -->

[![Review Change
Stack](https://storage.googleapis.com/coderabbit_public_assets/review-stack-in-coderabbit-ui.svg)](https://app.coderabbit.ai/change-stack/NVIDIA/NemoClaw/pull/3431)

<!-- review_stack_entry_end -->
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

Signed-off-by: latenighthackathon <latenighthackathon@users.noreply.github.com>
Co-authored-by: latenighthackathon <latenighthackathon@users.noreply.github.com>
Co-authored-by: Carlos Villela <cvillela@nvidia.com>