feat(decisions): backport Microsoft-Decision-1 and the decision models Admin UI to rc/1.106.0 (#45639, #45673, #45654, #45717)

litellm/__init__.py

@@ -1684,6 +1684,9 @@ def set_global_gitlab_config(config: Mapping[str, object]) -> None:
     from .llms.openai.decisions.transformation import (
         OpenAIDecisionsConfig as OpenAIDecisionsConfig,
     )
+    from .llms.azure_ai.decisions.transformation import (
+        AzureAIDecisionsConfig as AzureAIDecisionsConfig,
+    )
     from .llms.nvidia_nim.rerank.transformation import (
         NvidiaNimRerankConfig as NvidiaNimRerankConfig,
     )

litellm/_lazy_imports_registry.py

@@ -163,6 +163,7 @@
     "CloudflareDecisionsConfig",
     "StrandsDeciderDecisionsConfig",
     "OpenAIDecisionsConfig",
+    "AzureAIDecisionsConfig",
     "NvidiaNimRerankConfig",
     "NvidiaNimRankingConfig",
     "VertexAIRerankConfig",
@@ -724,6 +725,7 @@
         "StrandsDeciderDecisionsConfig",
     ),
     "OpenAIDecisionsConfig": (".llms.openai.decisions.transformation", "OpenAIDecisionsConfig"),
+    "AzureAIDecisionsConfig": (".llms.azure_ai.decisions.transformation", "AzureAIDecisionsConfig"),
     "NvidiaNimRerankConfig": (
         ".llms.nvidia_nim.rerank.transformation",
         "NvidiaNimRerankConfig",

litellm/llms/azure_ai/decisions/transformation.py

@@ -0,0 +1,24 @@
+import re
+from typing import Final
+
+import httpx
+
+from litellm.llms.base_llm.decisions.transformation import BaseDecisionsConfig
+
+_FOUNDRY_ROUTE_SUFFIX: Final = re.compile(r"(/api/projects/[^/]+)?(/openai/v1|/openai|/models|/v1)?$")
+
+
+class AzureAIDecisionsConfig(BaseDecisionsConfig):
+    """Microsoft Foundry decision models (Microsoft-Decision-1), where model is the deployment name"""
+
+    path = "/providers/microsoft/v1/systemone"
+    api_key_env = ("AZURE_AI_API_KEY",)
+    api_base_env = ("AZURE_AI_API_BASE",)
+
+    def missing_api_base_message(self, custom_llm_provider: str) -> str:
+        return "Missing AZURE_AI_API_BASE - set AZURE_AI_API_BASE or pass api_base, e.g. https://<resource>.services.ai.azure.com"
+
+    def get_complete_url(self, api_base: str, model: str) -> str:
+        url: Final = httpx.URL(api_base)
+        resource_path: Final = _FOUNDRY_ROUTE_SUFFIX.sub("", url.path.rstrip("/"))
+        return str(url.copy_with(path=f"{resource_path}{self.path}", query=None, fragment=None))

litellm/model_prices_and_context_window_backup.json

@@ -81933,5 +81933,12 @@
         "output_cost_per_token": 7.5e-06,
         "source": "https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/index.json",
         "supports_video_input": true
+    },
+    "azure_ai/Microsoft-Decision-1": {
+        "input_cost_per_token": 4.2e-08,
+        "litellm_provider": "azure_ai",
+        "mode": "evaluation",
+        "output_cost_per_token": 4.2e-08,
+        "source": "https://prices.azure.com/api/retail/prices?$filter=productName eq 'Microsoft Decision Models'"
     }
 }

litellm/utils.py

@@ -8991,6 +8991,8 @@ def get_provider_decisions_config(model: str, provider: LlmProviders) -> BaseDec
             return litellm.StrandsDeciderDecisionsConfig()
         if provider == LlmProviders.OPENAI:
             return litellm.OpenAIDecisionsConfig()
+        if provider == LlmProviders.AZURE_AI:
+            return litellm.AzureAIDecisionsConfig()
         return None
 
     @staticmethod

model_prices_and_context_window.json

@@ -81933,5 +81933,12 @@
         "output_cost_per_token": 7.5e-06,
         "source": "https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/index.json",
         "supports_video_input": true
+    },
+    "azure_ai/Microsoft-Decision-1": {
+        "input_cost_per_token": 4.2e-08,
+        "litellm_provider": "azure_ai",
+        "mode": "evaluation",
+        "output_cost_per_token": 4.2e-08,
+        "source": "https://prices.azure.com/api/retail/prices?$filter=productName eq 'Microsoft Decision Models'"
     }
 }

tests/integration/providers/test_decisions_wire.py

@@ -103,6 +103,9 @@ class _Provider:
         True,
         "cloudflare/@cf/cloudflare/clef",
     ),
+    _Provider(
+        "azure_ai", "azure_ai/decision-1", "/providers/microsoft/v1/systemone", "decision-1", _API_KEY, False, None
+    ),
 )
 _PERPLEXITY: Final = _PROVIDERS[0]
 _OPENROUTER: Final = _PROVIDERS[2]

tests/integration/proxy_config.yaml

@@ -346,6 +346,13 @@ model_list:
     litellm_params:
       model: strands_decider/strands-decider-2B-hobson-v19
       api_base: http://127.0.0.1:8191
+  - model_name: azure_ai/decision-1
+    litellm_params:
+      model: azure_ai/decision-1
+      api_base: http://127.0.0.1:8191
+      api_key: synthetic-azure-ai-key
+    model_info:
+      base_model: azure_ai/Microsoft-Decision-1
 general_settings:
   master_key: os.environ/LITELLM_MASTER_KEY
   database_url: os.environ/DATABASE_URL

tests/integration/translation/decisions/bases/azure_ai.py

@@ -0,0 +1,79 @@
+from typing import Final
+
+from integration.translation.case import TranslationTestCase
+
+"""Provider request and reply shape from the Microsoft Foundry Decision playground (POST /providers/microsoft/v1/systemone, model is the deployment name). Mock reply captured live from a Microsoft-Decision-1 deployment on 2026-10-09.
+"""
+MICROSOFT_DECISION_1_TEST_CASE: Final = TranslationTestCase(
+    scenario="basic",
+    litellm_endpoint="/v1/systemone",
+    litellm_request={
+        "model": "azure_ai/decision-1",
+        "state": "Ticket (billing): The export job hangs at 99% and never finishes",
+        "questions": {
+            "defect": {"type": "noul", "instructions": "Is this a defect?"},
+            "severity": {
+                "type": "choice",
+                "instructions": "How severe is it?",
+                "criteria": {"low": "cosmetic", "high": "blocks users"},
+            },
+            "confidence": {"type": "score", "instructions": "How sure are you?", "criteria": ["unsure", "sure"]},
+        },
+        "cache": {"no-cache": True},
+    },
+    expected_provider_endpoint="/providers/microsoft/v1/systemone",
+    expected_provider_headers={"authorization": "Bearer synthetic-azure-ai-key", "content-type": "application/json"},
+    expected_provider_request={
+        "model": "decision-1",
+        "state": "Ticket (billing): The export job hangs at 99% and never finishes",
+        "questions": {
+            "defect": {"type": "noul", "instructions": "Is this a defect?"},
+            "severity": {
+                "type": "choice",
+                "instructions": "How severe is it?",
+                "criteria": {"low": "cosmetic", "high": "blocks users"},
+            },
+            "confidence": {"type": "score", "instructions": "How sure are you?", "criteria": ["unsure", "sure"]},
+        },
+    },
+    mock_provider_response={
+        "model": "microsoft-decision-1",
+        "answers": {
+            "confidence": {
+                "confidence": 0.6351489346076444,
+                "legend": {"0": "unsure", "1": "sure"},
+                "probabilities": {"0": 0.1824255326961778, "1": 0.8175744673038222},
+                "score": 0.8175744673038222,
+                "type": "score",
+            },
+            "defect": {"noul": 0.982013764002278, "type": "noul"},
+            "severity": {
+                "choice": "high",
+                "confidence": 0.9866142978108987,
+                "probabilities": {"high": 0.9933071489054494, "low": 0.006692851094550702},
+                "type": "choice",
+            },
+        },
+        "usage": {"input_tokens": 87, "output_tokens": 3},
+    },
+    expected_litellm_response={
+        "model": "microsoft-decision-1",
+        "answers": {
+            "defect": {"type": "noul", "noul": 0.982013764002278},
+            "severity": {
+                "type": "choice",
+                "choice": "high",
+                "confidence": 0.9866142978108987,
+                "probabilities": {"high": 0.9933071489054494, "low": 0.006692851094550702},
+            },
+            "confidence": {
+                "type": "score",
+                "score": 0.8175744673038222,
+                "confidence": 0.6351489346076444,
+                "legend": {"0": "unsure", "1": "sure"},
+                "probabilities": {"0": 0.1824255326961778, "1": 0.8175744673038222},
+            },
+        },
+        "usage": {"input_tokens": 87, "output_tokens": 3},
+    },
+)

tests/integration/translation/decisions/basic/test_decisions_basic_azure_ai.py

@@ -0,0 +1,11 @@
+import pytest
+from integration._support.client import Gateway
+from integration._support.provider import SharedProvider
+from integration.translation.case import TranslationTestCase
+from integration.translation.decisions.bases.azure_ai import MICROSOFT_DECISION_1_TEST_CASE
+from integration.translation.runner import assert_translation
+
+
+@pytest.mark.parametrize("case", [MICROSOFT_DECISION_1_TEST_CASE], ids=lambda case: case.id)
+def test_decisions_basic_azure_ai(case: TranslationTestCase, gateway: Gateway, provider: SharedProvider) -> None:
+    assert_translation(case, gateway, provider)

tests/unit/decisions/test_main.py

@@ -922,3 +922,88 @@ async def test_requests_a_provider_cannot_serve_are_rejected_before_http(
         await litellm.adecisions(model=model, api_key="caller-key", **request_kwargs)
 
     assert len(respx_mock.calls) == 0
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize(
+    "api_base",
+    (
+        "https://res.services.ai.azure.com",
+        "https://res.services.ai.azure.com/",
+        "https://res.services.ai.azure.com/models",
+        "https://res.services.ai.azure.com/openai/v1",
+        "https://res.services.ai.azure.com/api/projects/proj",
+        "https://res.services.ai.azure.com/api/projects/proj/openai/v1",
+        "https://res.services.ai.azure.com/api/projects/proj/models",
+        "https://res.services.ai.azure.com/models?api-version=2024-05-01-preview",
+    ),
+)
+async def test_azure_ai_decision_posts_deployment_to_foundry_systemone_route(
+    monkeypatch: pytest.MonkeyPatch,
+    respx_mock: respx.MockRouter,
+    api_base: str,
+) -> None:
+    monkeypatch.delenv("AZURE_AI_API_BASE", raising=False)
+    monkeypatch.delenv("AZURE_AI_API_KEY", raising=False)
+    route: Final = respx_mock.post("https://res.services.ai.azure.com/providers/microsoft/v1/systemone").respond(
+        json={**_RESPONSE, "model": "microsoft-decision-1"}
+    )
+
+    response: Final = await litellm.adecisions(
+        model="azure_ai/decision-1",
+        state="review",
+        questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}},
+        api_base=api_base,
+        api_key="foundry-key",
+    )
+
+    assert route.called
+    request: Final = respx_mock.calls[0].request
+    assert request.headers["authorization"] == "Bearer foundry-key"
+    assert "api-key" not in request.headers
+    assert json.loads(request.content) == {
+        "model": "decision-1",
+        "state": "review",
+        "questions": {"is_defect": {"type": "noul", "instructions": "Is this a defect?"}},
+    }
+    assert response.answers == {"is_defect": NoulAnswer(type="noul", noul=0.9)}
+    assert response._hidden_params["model"] == "azure_ai/decision-1"
+
+
+@pytest.mark.asyncio
+async def test_azure_ai_decision_reads_foundry_env_and_keeps_gateway_path_prefix(
+    monkeypatch: pytest.MonkeyPatch,
+    respx_mock: respx.MockRouter,
+) -> None:
+    monkeypatch.setenv("AZURE_AI_API_BASE", "https://gateway.example.com/foundry/models")
+    monkeypatch.setenv("AZURE_AI_API_KEY", "env-foundry-key")
+    route: Final = respx_mock.post("https://gateway.example.com/foundry/providers/microsoft/v1/systemone").respond(
+        json=_RESPONSE
+    )
+
+    await litellm.adecisions(
+        model="azure_ai/decision-1",
+        state="review",
+        questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}},
+    )
+
+    assert route.called
+    assert respx_mock.calls[0].request.headers["authorization"] == "Bearer env-foundry-key"
+
+
+@pytest.mark.asyncio
+async def test_azure_ai_decision_requires_api_base_before_http(
+    monkeypatch: pytest.MonkeyPatch,
+    respx_mock: respx.MockRouter,
+) -> None:
+    monkeypatch.delenv("AZURE_AI_API_BASE", raising=False)
+    monkeypatch.setenv("AZURE_AI_API_KEY", "foundry-key")
+
+    with pytest.raises(litellm.BadRequestError, match="Missing AZURE_AI_API_BASE"):
+        await litellm.adecisions(
+            model="azure_ai/decision-1",
+            state="review",
+            questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}},
+        )
+
+    assert len(respx_mock.calls) == 0

ui/litellm-dashboard/src/app/(dashboard)/model-hub-table/page.tsx

@@ -6,9 +6,17 @@ import useAuthorized from "@/app/(dashboard)/hooks/useAuthorized";
 import { isAdminRole } from "@/utils/roles";
 
 export default function ModelHubTablePage() {
-  const { accessToken, userRole, premiumUser } = useAuthorized();
+  const { accessToken, userRole, premiumUser, isViewOnly } = useAuthorized();
   if (!isAdminRole(userRole)) {
     return <PublicModelHub accessToken={accessToken} isEmbedded={true} />;
   }
-  return <ModelHubTable accessToken={accessToken} publicPage={false} premiumUser={premiumUser} userRole={userRole} />;
+  return (
+    <ModelHubTable
+      accessToken={accessToken}
+      publicPage={false}
+      premiumUser={premiumUser}
+      userRole={userRole}
+      canOpenPlayground={!isViewOnly}
+    />
+  );
 }

ui/litellm-dashboard/src/app/(dashboard)/models-and-endpoints/page.integration.test.tsx

@@ -16,7 +16,7 @@ vi.mock("./panels/ModelRetrySettingsPanel", () => ({ default: () => <div data-te
 vi.mock("./panels/ModelGroupAliasPanel", () => ({ default: () => <div data-testid="panel-alias" /> }));
 vi.mock("./panels/PriceDataPanel", () => ({ default: () => <div data-testid="panel-price" /> }));
 vi.mock("./panels/AccessGroupBudgetsPanel", () => ({ default: () => <div data-testid="panel-budgets" /> }));
-vi.mock("@/components/molecules/cost_optimization_feedback_banner", () => ({ default: () => null }));
+vi.mock("@/components/molecules/DecisionModelsBanner", () => ({ default: () => null }));
 vi.mock("@/components/model_info_view", () => ({
   default: ({ modelId }: { modelId: string }) => <div data-testid="model-info">model:{modelId}</div>,
 }));

ui/litellm-dashboard/src/app/(dashboard)/models-and-endpoints/page.test.tsx

@@ -21,7 +21,17 @@ vi.mock("./detailNavigation", () => ({
   useModelDetailRouting: () => ({ ...detailState, close: vi.fn(), openModel: vi.fn(), openTeam: vi.fn() }),
 }));
 
-vi.mock("@/components/molecules/cost_optimization_feedback_banner", () => ({ default: () => null }));
+vi.mock("@/components/molecules/DecisionModelsBanner", () => ({
+  default: ({ onAddModel }: { onAddModel?: () => void }) => (
+    <section aria-label="Decision models">
+      {onAddModel && (
+        <button type="button" onClick={onAddModel}>
+          Add a decision model
+        </button>
+      )}
+    </section>
+  ),
+}));
 vi.mock("@/components/model_info_view", () => ({
   default: ({ modelId }: { modelId: string }) => <div data-testid="model-info">model:{modelId}</div>,
 }));
@@ -97,6 +107,27 @@ describe("ModelsAndEndpointsPage", () => {
     expect(screen.queryByTestId("panel-all-models")).not.toBeInTheDocument();
   });
 
+  it("opens the Add Model tab from the decision models banner", async () => {
+    const user = userEvent.setup();
+    renderPage();
+    await user.click(screen.getByRole("button", { name: "Add a decision model" }));
+    expect(screen.getByTestId("panel-add")).toBeInTheDocument();
+    expect(screen.queryByTestId("panel-all-models")).not.toBeInTheDocument();
+  });
+
+  it("does not offer the banner's Add a decision model to a session that cannot add models", () => {
+    mockUseAuthorized.mockReturnValue(NON_ADMIN);
+    renderPage();
+    expect(screen.getByRole("region", { name: "Decision models" })).toBeInTheDocument();
+    expect(screen.queryByRole("button", { name: "Add a decision model" })).not.toBeInTheDocument();
+  });
+
+  it("does not show a view-only session the decision models banner, whose Playground it cannot open", () => {
+    mockUseAuthorized.mockReturnValue(VIEW_ONLY_ADMIN);
+    renderPage();
+    expect(screen.queryByRole("region", { name: "Decision models" })).not.toBeInTheDocument();
+  });
+
   it("renders the model detail overlay from the ?model drill-in and hides the tabs", () => {
     detailState.modelId = "abc-123";
     renderPage();

ui/litellm-dashboard/src/app/(dashboard)/models-and-endpoints/page.tsx

@@ -9,7 +9,7 @@ import { useUISettings } from "@/app/(dashboard)/hooks/uiSettings/useUISettings"
 import { all_admin_roles, internalUserRoles } from "@/utils/roles";
 import { autoRouterCreationScope, canCreateModels } from "@/utils/modelPermissions";
 import BetaBadge from "@/components/BetaBadge";
-import CostOptimizationFeedbackBanner from "@/components/molecules/cost_optimization_feedback_banner";
+import DecisionModelsBanner from "@/components/molecules/DecisionModelsBanner";
 import ModelInfoView from "@/components/model_info_view";
 import TeamInfoView from "@/components/team/TeamInfo";
 import { useModelDetailRouting } from "@/app/(dashboard)/models-and-endpoints/detailNavigation";
@@ -177,7 +177,9 @@ export default function ModelsAndEndpointsPage() {
           </div>
         </div>
 
-        <CostOptimizationFeedbackBanner />
+        {!isViewOnly && (
+          <DecisionModelsBanner onAddModel={canCreate && !modelId ? () => setActiveKey("add") : undefined} />
+        )}
 
         {modelId ? (
           <ModelInfoView