fix(e2e): load DCode model selector through tsx (#9688)

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

The LangChain Deep Agents Code fresh onboarding E2E check stopped before
model selection because its `.mts` command used named imports from
CommonJS project modules. This change loads those modules through the
boundary that both standalone `tsx` and Vitest expose.

The existing source test did not detect the defect because Vitest
transformed the imports. A subprocess regression test now runs the same
`tsx` command that the live E2E check uses.

## Changes

- Load the URL, onboarding-probe, and provider-model exports through
namespace imports and their CommonJS `default` export when required.
- Add a local module-unwrapping helper because standalone `tsx` and
Vitest expose the same project modules differently. The fresh-onboarding
selector is the current consumer, and the subprocess regression test
protects both execution paths.
- Run the selector without a credential in the regression test and
require the expected credential error after module initialization.

## 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:
- [x] Sensitive paths changed (security, policy, credentials, preflight,
onboarding, inference, runner, sandbox, or messaging)
- [x] Sensitive-path review completed or maintainer-approved waiver
recorded — reviewer/approval link/justification: The change does not
alter credential custody, endpoint validation, network requests, or
sandbox state. The regression test supplies no credential and confirms
that validation rejects the missing value after module initialization.
- [ ] 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:
- Station profile/scenario:
- Result:
- Supporting evidence:

## 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 — `npx vitest run --project e2e-support
test/e2e/support/authorized-chat-model-selection.test.ts` passed 6
tests; `npm run checks:repository` passed.
- [ ] Applicable broad gate passed — `npm test` for broad
runtime/test-harness changes; `npm run check` for repo-wide
validation/coverage changes — command/result:
- [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 docs only)

---
Signed-off-by: Prekshi Vyas <prekshiv@nvidia.com>


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

* **Bug Fixes**
* Improved chat-model selection compatibility across supported module
formats.
* Chat-model selection now reports unsafe non-secure connections with a
clear error and exits with an appropriate failure status.

* **Tests**
* Added end-to-end coverage validating behavior when a credential is
provided but the service uses an unsupported non-loopback HTTP endpoint.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Signed-off-by: Prekshi Vyas <prekshiv@nvidia.com>

Author: prekshivyas

test/e2e/lib/select-authorized-chat-model.mts

@@ -1,11 +1,19 @@
 // SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 // SPDX-License-Identifier: Apache-2.0
 
-import { probeOpenAiLikeEndpointOptimized } from "../../../src/lib/inference/onboard-probes.ts";
-import { fetchOpenAiLikeModels } from "../../../src/lib/inference/provider-models.ts";
-import { isLoopbackHostname } from "../../../src/lib/core/url-utils.ts";
+import * as urlUtilsModule from "../../../src/lib/core/url-utils.ts";
+import * as onboardProbesModule from "../../../src/lib/inference/onboard-probes.ts";
+import * as providerModelsModule from "../../../src/lib/inference/provider-models.ts";
 import type { ModelCatalogFetchResult } from "../../../src/lib/onboard/types.ts";
 
+function unwrapCommonJsModule<T extends object>(module: T): T {
+  return (module as T & { default?: T }).default ?? module;
+}
+
+const { isLoopbackHostname } = unwrapCommonJsModule(urlUtilsModule);
+const { probeOpenAiLikeEndpointOptimized } = unwrapCommonJsModule(onboardProbesModule);
+const { fetchOpenAiLikeModels } = unwrapCommonJsModule(providerModelsModule);
+
 const CHAT_MODEL_HINT = /(?:claude|deepseek|gemma|gpt|kimi|llama|mistral|nemotron|phi|qwen)/iu;
 const NON_CHAT_MODEL_HINT =
   /(?:audio|clip|embed|guard|image|moderation|ocr|rerank|retrieval|reward|safety|speech|video)/iu;

test/e2e/support/authorized-chat-model-selection.test.ts

@@ -1,14 +1,44 @@
 // SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 // SPDX-License-Identifier: Apache-2.0
 
+import { spawnSync } from "node:child_process";
+import path from "node:path";
+
 import { describe, expect, it, vi } from "vitest";
 
 import { selectAuthorizedChatModel } from "../lib/select-authorized-chat-model.mts";
 
 const endpoint = "https://inference.example.test/v1";
 const currentModel = "nvidia/nvidia/nemotron-3-ultra";
+const selectorPath = path.resolve("test/e2e/lib/select-authorized-chat-model.mts");
+const tsxPath = path.resolve("node_modules/.bin/tsx");
 
 describe("authorized alternate chat model selection", () => {
+  it("rejects unsafe credential transport when tsx executes the selector", () => {
+    const result = spawnSync(
+      tsxPath,
+      [
+        selectorPath,
+        "--endpoint",
+        "http://inference.example.test/v1",
+        "--current-model",
+        currentModel,
+      ],
+      {
+        encoding: "utf8",
+        env: { ...process.env, COMPATIBLE_API_KEY: "placeholder-key" },
+        killSignal: "SIGKILL",
+        timeout: 10_000,
+      },
+    );
+
+    expect(result.error).toBeUndefined();
+    expect(result.status).toBe(1);
+    expect(result.stderr).toContain(
+      "authorized model selection failed: the endpoint must use HTTPS unless it targets loopback",
+    );
+  });
+
   it.each([endpoint, "http://127.0.0.1:8000/v1"])(
     "selects the highest-priority catalog model at %s",
     async (permittedEndpoint) => {