Fix CI lint and test failures

- Remove unused imports (datetime, timezone, ModelResponse) flagged by ruff.
- Update test_cost_tracking_adds_two_callbacks_when_prisma_set to expect 2
  callbacks on litellm.callbacks (not 1): ShadowEvalLogger now registers
  alongside _ProxyDBLogger in cost_tracking(). Test name already said 'two',
  now it actually tests for the correct count.
- Format ShadowEvalSection.tsx/.test.tsx per prettier.

Tests: 94/94 passing (lifecycle, shadow-eval, auto-router endpoints).
Lint: ruff + prettier all clean.

Co-Authored-By: Claude <noreply@anthropic.com>

litellm/integrations/shadow_eval_logger.py

@@ -25,7 +25,6 @@ import json
 import random
 import re
 from collections.abc import Callable
-from datetime import datetime, timezone
 from typing import TYPE_CHECKING, Any, Final, Optional, cast
 
 import litellm
@@ -35,7 +34,7 @@ from litellm.integrations.custom_logger import CustomLogger
 if TYPE_CHECKING:
     from litellm.proxy.utils import PrismaClient
     from litellm.router import Router
-    from litellm.types.utils import ModelResponse, StandardLoggingPayload
+    from litellm.types.utils import StandardLoggingPayload
 
 _JSON_FENCE_RE: Final = re.compile(r"```(?:json)?\s*(.*?)\s*```", re.DOTALL | re.IGNORECASE)
 

tests/test_litellm/proxy/proxy_server/test_lifecycle.py

@@ -485,6 +485,8 @@ def test_load_from_azure_key_vault_missing_uri_failure_is_swallowed(monkeypatch)
 
 
 def test_cost_tracking_adds_two_callbacks_when_prisma_set(monkeypatch):
+    """_ProxyDBLogger and ShadowEvalLogger both register on litellm.callbacks;
+    only _ProxyDBLogger also registers on _async_success_callback."""
     import litellm
 
     fake_prisma = MagicMock()
@@ -503,7 +505,7 @@ def test_cost_tracking_adds_two_callbacks_when_prisma_set(monkeypatch):
         "prisma_was_set": True,
     }
     assert normalize(observed) == {
-        "added_to_callbacks": 1,
+        "added_to_callbacks": 2,
         "added_to_async_success": 1,
         "prisma_was_set": True,
     }

ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/ShadowEvalSection.test.tsx

@@ -55,20 +55,16 @@ const job = (overrides: Partial<ShadowEvalJob> = {}): ShadowEvalJob => ({
   ...overrides,
 });
 
-const mockHooks = ({
-  jobs = [],
-  detail = undefined,
-}: {
-  jobs?: ShadowEvalJob[];
-  detail?: ShadowEvalJob;
-}) => {
+const mockHooks = ({ jobs = [], detail = undefined }: { jobs?: ShadowEvalJob[]; detail?: ShadowEvalJob }) => {
   vi.mocked(useShadowEvalJobs).mockReturnValue({ data: jobs, error: null } as unknown as ReturnType<
     typeof useShadowEvalJobs
   >);
   vi.mocked(useShadowEvalJob).mockReturnValue({ data: detail } as unknown as ReturnType<typeof useShadowEvalJob>);
-  vi.mocked(useStartShadowEval).mockReturnValue({ mutate: vi.fn(), isPending: false, error: null } as unknown as ReturnType<
-    typeof useStartShadowEval
-  >);
+  vi.mocked(useStartShadowEval).mockReturnValue({
+    mutate: vi.fn(),
+    isPending: false,
+    error: null,
+  } as unknown as ReturnType<typeof useStartShadowEval>);
   vi.mocked(useStopShadowEval).mockReturnValue({ mutate: vi.fn(), isPending: false } as unknown as ReturnType<
     typeof useStopShadowEval
   >);

ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/ShadowEvalSection.tsx

@@ -140,15 +140,16 @@ const StartForm: React.FC<{ accessToken: string | null }> = ({ accessToken }) =>
   const start = useStartShadowEval();
 
   const parsedPct = Number.parseFloat(percentage);
-  const valid = Boolean(accessToken) && apiKeyId.trim() !== "" && routerName.trim() !== "" && parsedPct > 0 && parsedPct <= 100;
+  const valid =
+    Boolean(accessToken) && apiKeyId.trim() !== "" && routerName.trim() !== "" && parsedPct > 0 && parsedPct <= 100;
 
   return (
     <Card size="sm">
       <CardHeader>
         <CardTitle className="text-sm font-medium text-foreground">Start a shadow eval</CardTitle>
         <p className="text-xs text-muted-foreground">
-          Duplicates a sampled slice of the key&apos;s traffic through the auto-router and has an LLM judge compare
-          both answers blind. The router&apos;s answers are never served to users. Judge calls bill to the proxy — an
+          Duplicates a sampled slice of the key&apos;s traffic through the auto-router and has an LLM judge compare both
+          answers blind. The router&apos;s answers are never served to users. Judge calls bill to the proxy — an
           estimate is shown before anything runs.
         </p>
       </CardHeader>