fix(ci): prefer completions for hosted inference

.github/workflows/e2e-script.yaml

@@ -231,6 +231,7 @@ jobs:
             printf 'NEMOCLAW_ENDPOINT_URL=https://inference-api.nvidia.com/v1\n'
             printf 'NEMOCLAW_MODEL=nvidia/nvidia/nemotron-3-super-v3\n'
             printf 'NEMOCLAW_COMPAT_MODEL=nvidia/nvidia/nemotron-3-super-v3\n'
+            printf 'NEMOCLAW_PREFERRED_API=openai-completions\n'
             printf 'COMPATIBLE_API_KEY=%s\n' "${NVIDIA_INFERENCE_API_KEY}"
           } >> "$GITHUB_ENV"
 

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

@@ -992,6 +992,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
         run: |
           set -euo pipefail
           npx vitest run --project e2e-scenarios-live \

.github/workflows/nightly-e2e.yaml

@@ -470,6 +470,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -545,6 +546,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_ISSUE_4434_LIVE: "1"
           NEMOCLAW_NON_INTERACTIVE: "1"
@@ -973,6 +975,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -1270,6 +1273,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -1576,6 +1580,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           E2E_ARTIFACT_DIR: ${{ github.workspace }}/e2e-artifacts/vitest/credential-migration
           NEMOCLAW_RUN_E2E_SCENARIOS: "1"
@@ -1801,6 +1806,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -1813,6 +1819,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -1851,6 +1858,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -1863,6 +1871,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -1901,6 +1910,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -1913,6 +1923,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -1952,6 +1963,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -1964,6 +1976,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -2003,6 +2016,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -2016,6 +2030,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -2057,6 +2072,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -2070,6 +2086,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
@@ -2149,6 +2166,7 @@ jobs:
           NEMOCLAW_ENDPOINT_URL: https://inference-api.nvidia.com/v1
           NEMOCLAW_MODEL: nvidia/nvidia/nemotron-3-super-v3
           NEMOCLAW_COMPAT_MODEL: nvidia/nvidia/nemotron-3-super-v3
+          NEMOCLAW_PREFERRED_API: openai-completions
           COMPATIBLE_API_KEY: ${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
           NEMOCLAW_NON_INTERACTIVE: "1"
           NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"

src/lib/onboard/providers.test.ts

@@ -59,6 +59,7 @@ function withProviderEnv(next: Record<string, string | undefined>, testBody: ()
     "NEMOCLAW_ENDPOINT_URL",
     "NEMOCLAW_MODEL",
     "NEMOCLAW_COMPAT_MODEL",
+    "NEMOCLAW_PREFERRED_API",
     "NEMOCLAW_CLOUD_EXPERIMENTAL_MODEL",
     "NEMOCLAW_E2E_USE_HOSTED_INFERENCE",
     "COMPATIBLE_API_KEY",
@@ -303,11 +304,26 @@ describe("onboard provider helpers", () => {
         expect(process.env.NEMOCLAW_ENDPOINT_URL).toBe(HOSTED_INFERENCE_ENDPOINT_URL);
         expect(process.env.NEMOCLAW_MODEL).toBe(HOSTED_INFERENCE_MODEL);
         expect(process.env.NEMOCLAW_COMPAT_MODEL).toBe(HOSTED_INFERENCE_MODEL);
+        expect(process.env.NEMOCLAW_PREFERRED_API).toBe("openai-completions");
         expect(process.env.COMPATIBLE_API_KEY).toBe("repo-hosted-key");
       },
     );
   });
 
+  it("does not override an explicit hosted inference API preference", () => {
+    withProviderEnv(
+      {
+        NVIDIA_INFERENCE_API_KEY: "repo-hosted-key",
+        NEMOCLAW_E2E_USE_HOSTED_INFERENCE: "1",
+        NEMOCLAW_PREFERRED_API: "openai-responses",
+      },
+      () => {
+        expect(stageHostedInferenceSourceSecretEnv()).toBe(true);
+        expect(process.env.NEMOCLAW_PREFERRED_API).toBe("openai-responses");
+      },
+    );
+  });
+
   it("keeps explicit cloud provider selection on the Build provider path", () => {
     withProviderEnv(
       {

src/lib/onboard/providers.ts

@@ -254,6 +254,8 @@ function stageHostedInferenceSourceSecretEnv() {
     HOSTED_INFERENCE_MODEL;
   process.env.NEMOCLAW_MODEL = model;
   process.env.NEMOCLAW_COMPAT_MODEL = (process.env.NEMOCLAW_COMPAT_MODEL || "").trim() || model;
+  process.env.NEMOCLAW_PREFERRED_API =
+    (process.env.NEMOCLAW_PREFERRED_API || "").trim() || "openai-completions";
   process.env[HOSTED_INFERENCE_CREDENTIAL_ENV] = sourceKey;
   return true;
 }

test/e2e-advisor-dispatch.test.ts

@@ -19,6 +19,23 @@ jobs:
   cloud-e2e:
     if: github.event_name != 'workflow_dispatch' || inputs.jobs == '' || contains(format(',{0},', inputs.jobs), ',cloud-e2e,')
     steps: []
+  cloud-onboard-e2e:
+    if: github.event_name != 'workflow_dispatch' || inputs.jobs == '' || contains(format(',{0},', inputs.jobs), ',cloud-onboard-e2e,')
+    uses: ./.github/workflows/e2e-script.yaml
+    with:
+      ref: \${{ inputs.target_ref || github.ref }}
+      nvidia_api_key: true
+    secrets: *nightly-e2e-default-secrets
+  launchable-smoke-e2e:
+    if: github.event_name != 'workflow_dispatch' || inputs.jobs == '' || contains(format(',{0},', inputs.jobs), ',launchable-smoke-e2e,')
+    runs-on: ubuntu-latest
+    steps:
+      - name: Run launchable install-flow smoke test
+        env:
+          NVIDIA_INFERENCE_API_KEY: \${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
+          NEMOCLAW_E2E_USE_HOSTED_INFERENCE: "1"
+          COMPATIBLE_API_KEY: \${{ (github.event_name != 'workflow_dispatch' || inputs.target_ref == '') && secrets.NVIDIA_INFERENCE_API_KEY || '' }}
+        run: bash test/e2e/test-launchable-smoke.sh
   report-to-pr:
     steps: []
   notify-on-failure:
@@ -68,6 +85,8 @@ describe("E2E advisor auto-dispatch planning", () => {
 
     expect(jobs).toContain("network-policy-e2e");
     expect(jobs).toContain("cloud-e2e");
+    expect(jobs).toContain("cloud-onboard-e2e");
+    expect(jobs).toContain("launchable-smoke-e2e");
     expect(jobs).not.toContain("report-to-pr");
     expect(jobs).not.toContain("notify-on-failure");
     expect(jobs).not.toContain("scorecard");
@@ -122,6 +141,56 @@ describe("E2E advisor auto-dispatch planning", () => {
     expect(plan.inputs?.jobs).toBe("network-policy-e2e,cloud-e2e");
   });
 
+  it("filters hosted-inference jobs that cannot receive secrets on target-ref dispatches", () => {
+    const workflowText = nightlyWorkflowText();
+    const plan = planAutoDispatch({
+      result: {
+        confidence: "high",
+        requiredTests: [
+          { id: "network-policy-e2e", workflow: "nightly-e2e.yaml" },
+          { id: "cloud-onboard-e2e", workflow: "nightly-e2e.yaml" },
+          { id: "launchable-smoke-e2e", workflow: "nightly-e2e.yaml" },
+        ],
+      },
+      workflowText,
+      event: pullRequest("MEMBER"),
+      env: {
+        GITHUB_EVENT_NAME: "pull_request",
+        GITHUB_REPOSITORY: "NVIDIA/NemoClaw",
+      },
+    });
+
+    expect(plan.status).toBe("ready");
+    expect(plan.jobs).toEqual(["network-policy-e2e"]);
+    expect(plan.inputs?.jobs).toBe("network-policy-e2e");
+    expect(plan.ignoredJobs).toEqual(["cloud-onboard-e2e", "launchable-smoke-e2e"]);
+    expect(plan.targetRefSecretBlockedJobs).toEqual(["cloud-onboard-e2e", "launchable-smoke-e2e"]);
+  });
+
+  it("skips cleanly when every recommended target-ref job requires withheld secrets", () => {
+    const workflowText = nightlyWorkflowText();
+    const plan = planAutoDispatch({
+      result: {
+        confidence: "high",
+        requiredTests: [
+          { id: "cloud-onboard-e2e", workflow: "nightly-e2e.yaml" },
+          { id: "launchable-smoke-e2e", workflow: "nightly-e2e.yaml" },
+        ],
+      },
+      workflowText,
+      event: pullRequest("MEMBER"),
+      env: {
+        GITHUB_EVENT_NAME: "pull_request",
+        GITHUB_REPOSITORY: "NVIDIA/NemoClaw",
+      },
+    });
+
+    expect(plan.status).toBe("skipped");
+    expect(plan.reason).toMatch(/hosted inference secrets are withheld/);
+    expect(plan.ignoredJobs).toEqual(["cloud-onboard-e2e", "launchable-smoke-e2e"]);
+    expect(plan.targetRefSecretBlockedJobs).toEqual(["cloud-onboard-e2e", "launchable-smoke-e2e"]);
+  });
+
   it("plans dispatch for allowlisted authors whose private org membership appears as contributor", () => {
     const workflowText = nightlyWorkflowText();
     const plan = planAutoDispatch({

test/e2e-script-workflow.test.ts

@@ -547,6 +547,7 @@ describe("E2E reusable workflow contract", () => {
     expect(runStep?.env?.NEMOCLAW_ENDPOINT_URL).toBe("https://inference-api.nvidia.com/v1");
     expect(runStep?.env?.NEMOCLAW_MODEL).toBe("nvidia/nvidia/nemotron-3-super-v3");
     expect(runStep?.env?.NEMOCLAW_COMPAT_MODEL).toBe("nvidia/nvidia/nemotron-3-super-v3");
+    expect(runStep?.env?.NEMOCLAW_PREFERRED_API).toBe("openai-completions");
     expect(runStep?.env?.COMPATIBLE_API_KEY).toBe(GUARDED_HOSTED_INFERENCE_SECRET);
     expect(runStep?.env?.GITHUB_TOKEN).toBeUndefined();
     expect(runStep?.env?.NEMOCLAW_RUN_E2E_SCENARIOS).toBe("1");
@@ -905,6 +906,7 @@ describe("E2E reusable workflow contract", () => {
     expect(exportStep?.run).toContain("NEMOCLAW_ENDPOINT_URL=https://inference-api.nvidia.com/v1");
     expect(exportStep?.run).toContain("NEMOCLAW_MODEL=nvidia/nvidia/nemotron-3-super-v3");
     expect(exportStep?.run).toContain("NEMOCLAW_COMPAT_MODEL=nvidia/nvidia/nemotron-3-super-v3");
+    expect(exportStep?.run).toContain("NEMOCLAW_PREFERRED_API=openai-completions");
     expect(exportStep?.run).toContain("COMPATIBLE_API_KEY=%s");
 
     expect(hostedJobs.length).toBeGreaterThan(20);
@@ -962,6 +964,7 @@ describe("E2E reusable workflow contract", () => {
       expect(step.env?.NEMOCLAW_ENDPOINT_URL, jobName).toBe("https://inference-api.nvidia.com/v1");
       expect(step.env?.NEMOCLAW_MODEL, jobName).toBe("nvidia/nvidia/nemotron-3-super-v3");
       expect(step.env?.NEMOCLAW_COMPAT_MODEL, jobName).toBe("nvidia/nvidia/nemotron-3-super-v3");
+      expect(step.env?.NEMOCLAW_PREFERRED_API, jobName).toBe("openai-completions");
       expect(step.env?.COMPATIBLE_API_KEY, jobName).toBe(GUARDED_HOSTED_INFERENCE_SECRET);
     }
 

test/e2e/lib/ci-compatible-inference.sh

@@ -37,6 +37,7 @@ nemoclaw_e2e_configure_compatible_inference() {
   export NEMOCLAW_ENDPOINT_URL="${NEMOCLAW_ENDPOINT_URL:-https://inference-api.nvidia.com/v1}"
   export NEMOCLAW_MODEL="${NEMOCLAW_MODEL:-${NEMOCLAW_CLOUD_EXPERIMENTAL_MODEL:-$NEMOCLAW_E2E_COMPATIBLE_INFERENCE_MODEL_DEFAULT}}"
   export NEMOCLAW_COMPAT_MODEL="${NEMOCLAW_COMPAT_MODEL:-$NEMOCLAW_MODEL}"
+  export NEMOCLAW_PREFERRED_API="${NEMOCLAW_PREFERRED_API:-openai-completions}"
   export COMPATIBLE_API_KEY="$NVIDIA_INFERENCE_API_KEY"
 }