fix(e2e): use hosted validation models in Vitest lanes

.github/workflows/e2e-vitest-scenarios.yaml

@@ -1719,7 +1719,6 @@ jobs:
       NEMOCLAW_NON_INTERACTIVE: "1"
       NEMOCLAW_RECREATE_SANDBOX: "1"
       NEMOCLAW_SANDBOX_NAME: e2e-hermes
-      NEMOCLAW_MODEL: minimaxai/minimax-m2.7
       NEMOCLAW_ONBOARD_VALIDATION_TIMEOUT_SECONDS: "60"
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6.0.3

test/e2e-scenario/live/cron-preflight-inference-local.test.ts

@@ -18,14 +18,17 @@ import { resultText } from "../fixtures/clients/index.ts";
 import { type SandboxClient, validateSandboxName } from "../fixtures/clients/sandbox.ts";
 import { expect, test } from "../fixtures/e2e-test.ts";
 import { shouldRunLiveE2EScenarios } from "../fixtures/live-project-gate.ts";
-import { requireHostedInferenceConfig } from "../fixtures/hosted-inference.ts";
+import {
+  DEFAULT_HOSTED_INFERENCE_MODEL,
+  requireHostedInferenceConfig,
+} from "../fixtures/hosted-inference.ts";
 import type { ShellProbeResult } from "../fixtures/shell-probe.ts";
 import { isTransientProviderValidationFailure } from "./network-policy-transient-provider.ts";
 
 const REPO_ROOT = path.resolve(import.meta.dirname, "../../..");
 const SANDBOX_NAME = process.env.NEMOCLAW_SANDBOX_NAME ?? "e2e-cron-preflight";
 validateSandboxName(SANDBOX_NAME);
-const MODEL = process.env.NEMOCLAW_CRON_PREFLIGHT_MODEL ?? "nvidia/nemotron-3-super-120b-a12b";
+const MODEL = process.env.NEMOCLAW_CRON_PREFLIGHT_MODEL ?? DEFAULT_HOSTED_INFERENCE_MODEL;
 const INSTALL_ATTEMPTS = process.env.CI === "true" || process.env.GITHUB_ACTIONS === "true" ? 3 : 1;
 const LIVE_TIMEOUT_MS = 35 * 60_000;
 const PROBE_SOURCE = String.raw`

test/e2e-scenario/live/diagnostics.test.ts

@@ -18,13 +18,17 @@ import path from "node:path";
 import { buildAvailabilityProbeEnv } from "../fixtures/availability-env.ts";
 import { validateSandboxName } from "../fixtures/clients/sandbox.ts";
 import { expect, test } from "../fixtures/e2e-test.ts";
+import {
+  DEFAULT_HOSTED_INFERENCE_MODEL,
+  requireHostedInferenceConfig,
+} from "../fixtures/hosted-inference.ts";
 import { shouldRunLiveE2EScenarios } from "../fixtures/live-project-gate.ts";
 import type { ShellProbeResult } from "../fixtures/shell-probe.ts";
 
 const REPO_ROOT = path.resolve(import.meta.dirname, "../../..");
 const CLI_ENTRYPOINT = path.join(REPO_ROOT, "bin", "nemoclaw.js");
 const SANDBOX_NAME = process.env.NEMOCLAW_SANDBOX_NAME ?? `e2e-diag-${process.pid}`;
-const DIAGNOSTICS_MODEL = process.env.NEMOCLAW_MODEL ?? "openai/gpt-oss-120b";
+const DIAGNOSTICS_MODEL = process.env.NEMOCLAW_MODEL ?? DEFAULT_HOSTED_INFERENCE_MODEL;
 const DEBUG_QUICK_TIMEOUT_MS = 30_000;
 const INSTALL_TIMEOUT_MS = 35 * 60_000;
 const TEST_TIMEOUT_MS = 55 * 60_000;
@@ -138,7 +142,8 @@ runDiagnosticsTest(
       "run `npm run build:cli` before live repo CLI scenarios",
     ).toBe(true);
 
-    const apiKey = secrets.required("NVIDIA_INFERENCE_API_KEY");
+    const hosted = requireHostedInferenceConfig(secrets, process.env, { model: DIAGNOSTICS_MODEL });
+    const apiKey = hosted.apiKey;
     await artifacts.writeJson("scenario.json", {
       id: "diagnostics",
       runner: "vitest",
@@ -195,7 +200,7 @@ runDiagnosticsTest(
       fs.rmSync(home, { recursive: true, force: true });
     });
 
-    const env = testEnv(home, { NVIDIA_INFERENCE_API_KEY: apiKey });
+    const env = testEnv(home, hosted.env);
     await bestEffort(() =>
       host.command("node", [CLI_ENTRYPOINT, SANDBOX_NAME, "destroy", "--yes"], {
         artifactName: "pre-cleanup-nemoclaw-destroy-diagnostics",

test/e2e-scenario/live/hermes-e2e.test.ts

@@ -11,7 +11,10 @@ import { trustedProviderEndpoint } from "../fixtures/clients/provider.ts";
 import { trustedSandboxShellScript, validateSandboxName } from "../fixtures/clients/sandbox.ts";
 import { expect, test } from "../fixtures/e2e-test.ts";
 import { shouldRunLiveE2EScenarios } from "../fixtures/live-project-gate.ts";
-import { requireHostedInferenceConfig } from "../fixtures/hosted-inference.ts";
+import {
+  DEFAULT_HOSTED_INFERENCE_MODEL,
+  requireHostedInferenceConfig,
+} from "../fixtures/hosted-inference.ts";
 import type { ShellProbeResult } from "../fixtures/shell-probe.ts";
 
 // Migrated from test/e2e/test-hermes-e2e.sh.
@@ -33,7 +36,7 @@ const HERMES_DASHBOARD_INTERNAL_PORT =
 const SESSION_FILE = path.join(os.homedir(), ".nemoclaw", "onboard-session.json");
 const REGISTRY_FILE = path.join(os.homedir(), ".nemoclaw", "sandboxes.json");
 const LIVE_TIMEOUT_MS = 70 * 60_000;
-const CHAT_MODEL = process.env.NEMOCLAW_MODEL ?? "nvidia/nemotron-3-super-120b-a12b";
+const CHAT_MODEL = process.env.NEMOCLAW_MODEL ?? DEFAULT_HOSTED_INFERENCE_MODEL;
 const ONBOARD_VALIDATION_TIMEOUT_SECONDS =
   process.env.NEMOCLAW_ONBOARD_VALIDATION_TIMEOUT_SECONDS ?? "60";
 
@@ -65,18 +68,18 @@ function hermesDashboardE2eEnabled(): boolean {
   );
 }
 
-function commandEnv(apiKey?: string): NodeJS.ProcessEnv {
+function commandEnv(hostedEnv: NodeJS.ProcessEnv = {}): NodeJS.ProcessEnv {
   const env: NodeJS.ProcessEnv = {
     ...buildAvailabilityProbeEnv(),
+    ...hostedEnv,
     NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1",
     NEMOCLAW_AGENT: "hermes",
     NEMOCLAW_NON_INTERACTIVE: "1",
     NEMOCLAW_RECREATE_SANDBOX: "1",
-    NEMOCLAW_MODEL: CHAT_MODEL,
+    NEMOCLAW_MODEL: hostedEnv.NEMOCLAW_MODEL ?? CHAT_MODEL,
     NEMOCLAW_ONBOARD_VALIDATION_TIMEOUT_SECONDS: ONBOARD_VALIDATION_TIMEOUT_SECONDS,
     NEMOCLAW_SANDBOX_NAME: SANDBOX_NAME,
   };
-  if (apiKey) env.NVIDIA_INFERENCE_API_KEY = apiKey;
   if (process.env.NEMOCLAW_E2E_HERMES_DASHBOARD) {
     env.NEMOCLAW_E2E_HERMES_DASHBOARD = process.env.NEMOCLAW_E2E_HERMES_DASHBOARD;
   }
@@ -232,7 +235,7 @@ test.skipIf(!shouldRunLiveE2EScenarios())(
       dashboardEnabled: hermesDashboardE2eEnabled(),
     });
 
-    const env = commandEnv(apiKey);
+    const env = commandEnv(hosted.env);
     const redactionValues = [apiKey];
 
     const cleanupHermes = async (label: string) => {

test/e2e-scenario/live/upgrade-stale-sandbox-helpers.ts

@@ -43,18 +43,17 @@ function assertSafeSandboxName(): void {
   }
 }
 
-export function commandEnv(apiKey?: string): NodeJS.ProcessEnv {
-  const env: NodeJS.ProcessEnv = {
+export function commandEnv(extra: NodeJS.ProcessEnv = {}): NodeJS.ProcessEnv {
+  return {
     ...buildAvailabilityProbeEnv(),
+    ...extra,
     NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1",
     NEMOCLAW_NON_INTERACTIVE: "1",
     NEMOCLAW_REBUILD_VERBOSE: "1",
     NEMOCLAW_RECREATE_SANDBOX: "1",
     NEMOCLAW_SANDBOX_NAME: SANDBOX_NAME,
     OPENSHELL_GATEWAY: process.env.OPENSHELL_GATEWAY ?? "nemoclaw",
   };
-  apiKey && Object.assign(env, { NVIDIA_INFERENCE_API_KEY: apiKey });
-  return env;
 }
 
 export async function bestEffort(run: () => Promise<unknown>): Promise<void> {
@@ -201,7 +200,7 @@ export async function cleanupOldImage(host: HostCliClient): Promise<void> {
 
 export async function installCurrentNemoclaw(
   host: HostCliClient,
-  apiKey: string,
+  hosted: { apiKey: string; env: NodeJS.ProcessEnv },
 ): Promise<ShellProbeResult> {
   let install: ShellProbeResult | undefined;
   for (let attempt = 1; attempt <= INSTALL_ATTEMPTS; attempt += 1) {
@@ -211,8 +210,8 @@ export async function installCurrentNemoclaw(
           ? "phase-1-install-current-nemoclaw"
           : `phase-1-install-current-nemoclaw-attempt-${attempt}`,
       cwd: REPO_ROOT,
-      env: commandEnv(apiKey),
-      redactionValues: [apiKey],
+      env: commandEnv(hosted.env),
+      redactionValues: [hosted.apiKey],
       timeoutMs: 20 * 60_000,
     });
     const retry =

test/e2e-scenario/live/upgrade-stale-sandbox.test.ts

@@ -13,6 +13,7 @@
 import { buildAvailabilityProbeEnv } from "../fixtures/availability-env.ts";
 import { resultText } from "../fixtures/clients/index.ts";
 import { expect, test } from "../fixtures/e2e-test.ts";
+import { requireHostedInferenceConfig } from "../fixtures/hosted-inference.ts";
 import { shouldRunLiveE2EScenarios } from "../fixtures/live-project-gate.ts";
 import {
   assertDeleteInstalledSandboxAllowed,
@@ -37,7 +38,7 @@ test.skipIf(!shouldRunLiveE2EScenarios())(
   "upgrade-sandboxes detects and rebuilds stale OpenClaw sandboxes (#1904)",
   { timeout: LIVE_TIMEOUT_MS },
   async ({ artifacts, cleanup, host, sandbox, secrets, skip }) => {
-    const apiKey = secrets.required("NVIDIA_INFERENCE_API_KEY");
+    const hosted = requireHostedInferenceConfig(secrets);
 
     await artifacts.writeJson("scenario.json", {
       id: "upgrade-stale-sandbox",
@@ -67,7 +68,7 @@ test.skipIf(!shouldRunLiveE2EScenarios())(
     cleanup.add("remove stale OpenClaw test image", () => cleanupOldImage(host));
     await cleanupStaleSandbox(host, sandbox);
 
-    const install = await installCurrentNemoclaw(host, apiKey);
+    const install = await installCurrentNemoclaw(host, hosted);
     expect(install.exitCode, resultText(install)).toBe(0);
 
     const deleteInstalledSandbox = await sandbox.openshell(["sandbox", "delete", SANDBOX_NAME], {
@@ -119,8 +120,8 @@ test.skipIf(!shouldRunLiveE2EScenarios())(
 
     const staleCheck = await host.nemoclaw(["upgrade-sandboxes", "--check"], {
       artifactName: "phase-5-upgrade-sandboxes-check-stale",
-      env: commandEnv(apiKey),
-      redactionValues: [apiKey],
+      env: commandEnv(hosted.env),
+      redactionValues: [hosted.apiKey],
       timeoutMs: 120_000,
     });
     expect(staleCheck.exitCode, resultText(staleCheck)).toBe(0);
@@ -129,8 +130,8 @@ test.skipIf(!shouldRunLiveE2EScenarios())(
 
     const rebuild = await host.nemoclaw([SANDBOX_NAME, "rebuild", "--yes"], {
       artifactName: "phase-6-rebuild-stale-sandbox",
-      env: commandEnv(apiKey),
-      redactionValues: [apiKey],
+      env: commandEnv(hosted.env),
+      redactionValues: [hosted.apiKey],
       timeoutMs: 25 * 60_000,
     });
     expect(rebuild.exitCode, resultText(rebuild)).toBe(0);
@@ -148,8 +149,8 @@ test.skipIf(!shouldRunLiveE2EScenarios())(
 
     const cleanCheck = await host.nemoclaw(["upgrade-sandboxes", "--check"], {
       artifactName: "phase-7-upgrade-sandboxes-check-clean",
-      env: commandEnv(apiKey),
-      redactionValues: [apiKey],
+      env: commandEnv(hosted.env),
+      redactionValues: [hosted.apiKey],
       timeoutMs: 120_000,
     });
     expect(cleanCheck.exitCode, resultText(cleanCheck)).toBe(0);