Merge 28096412ef into 527d882508

litellm/router_python_core.py

@@ -3569,6 +3569,8 @@ class PythonCore(DeprecatedRouterMembers):
             else:
                 model_client = potential_model_client
 
+            self.total_calls[model_name] += 1
+
             ### DEPLOYMENT-SPECIFIC PRE-CALL CHECKS ### (e.g. update rpm pre-call. Raise error, if deployment over limit)
             ## only run if model group given, not model id
             if model in self._core_model_names or not self._core_has_model_id(model):
@@ -3603,6 +3605,8 @@ class PythonCore(DeprecatedRouterMembers):
             ):
                 response.fetch_sync_stream()
 
+            self.success_calls[model_name] += 1
+
             # Wrap streaming responses so MidStreamFallbackError (raised
             # during iteration) triggers the Router's fallback chain.
             if isinstance(response, CustomStreamWrapper):
@@ -3616,6 +3620,8 @@ class PythonCore(DeprecatedRouterMembers):
             return response
         except Exception as e:
             verbose_router_logger.info("litellm.completion(model=%s)\x1b[31m Exception %s\x1b[0m", model_name, e)
+            if model_name is not None:
+                self.fail_calls[model_name] += 1
             # Set per-deployment num_retries on exception for retry logic
             if deployment is not None:
                 self._set_deployment_num_retries_on_exception(e, deployment)

tests/unit/test_router/test_router.py

@@ -27738,6 +27738,7 @@ class TestRouterIndexManagement:
                 f"{'=' * 70}\n"
             )
 
+
     def test_model_names_is_set(self):
         """Verify that model_names uses a set for O(1) lookups, not a list (O(n))"""
         router = Router(model_list=[])
@@ -32396,3 +32397,45 @@ def test_image_routing_inspects_native_request_content_without_translation(reque
     original: Final = copy.deepcopy(request_kwargs)
     assert _candidate_ids(router, request_kwargs=request_kwargs) == ["vision"]
     assert request_kwargs == original
+
+
+def test_sync_completion_tracks_router_metrics():
+    """
+    Regression: sync router.completion() must update router.total_calls /
+    router.success_calls / router.fail_calls, same as router.acompletion() does.
+
+    Before the fix, _completion() (the sync entry point that completion()
+    reaches through function_with_fallbacks -> async_function_with_retries)
+    never touched any of the three counters, so the router metrics stayed at
+    zero for every sync call.
+    """
+    router = Router(
+        model_list=[
+            {
+                "model_name": "gpt-4o",
+                "litellm_params": {"model": "openai/gpt-4o", "num_retries": 0},
+            }
+        ],
+        num_retries=0,
+    )
+
+    ## success: total/success track the deployment model
+    router.completion(
+        model="gpt-4o",
+        messages=[{"role": "user", "content": "hi"}],
+        mock_response="Hello",
+    )
+    assert router.total_calls["openai/gpt-4o"] == 1
+    assert router.success_calls["openai/gpt-4o"] == 1
+    assert router.fail_calls.get("openai/gpt-4o", 0) == 0
+
+    ## failure: fail tracks the deployment model
+    with pytest.raises(Exception):
+        router.completion(
+            model="gpt-4o",
+            messages=[{"role": "user", "content": "hi"}],
+            mock_response="litellm.InternalServerError",
+        )
+    assert router.fail_calls["openai/gpt-4o"] == 1
+    assert router.total_calls["openai/gpt-4o"] == 2
+    assert router.success_calls["openai/gpt-4o"] == 1