fix(test_caching.py): add longer delay for async test

litellm/tests/test_caching.py

@@ -178,11 +178,17 @@ def test_caching_with_default_ttl():
         pytest.fail(f"Error occurred: {e}")
 
 
-def test_caching_with_cache_controls():
+@pytest.mark.parametrize(
+    "sync_flag",
+    [True, False],
+)
+@pytest.mark.asyncio
+async def test_caching_with_cache_controls(sync_flag):
     try:
         litellm.set_verbose = True
         litellm.cache = Cache()
         message = [{"role": "user", "content": f"Hey, how's it going? {uuid.uuid4()}"}]
+        if sync_flag:
             ## TTL = 0
             response1 = completion(
                 model="gpt-3.5-turbo", messages=messages, cache={"ttl": 0}
@@ -190,11 +196,23 @@ def test_caching_with_cache_controls():
             response2 = completion(
                 model="gpt-3.5-turbo", messages=messages, cache={"s-maxage": 10}
             )
-        print(f"response1: {response1}")
+
-        print(f"response2: {response2}")
             assert response2["id"] != response1["id"]
+        else:
+            ## TTL = 0
+            response1 = await litellm.acompletion(
+                model="gpt-3.5-turbo", messages=messages, cache={"ttl": 0}
+            )
+            await asyncio.sleep(10)
+            response2 = await litellm.acompletion(
+                model="gpt-3.5-turbo", messages=messages, cache={"s-maxage": 10}
+            )
+
+            assert response2["id"] != response1["id"]
+
         message = [{"role": "user", "content": f"Hey, how's it going? {uuid.uuid4()}"}]
         ## TTL = 5
+        if sync_flag:
             response1 = completion(
                 model="gpt-3.5-turbo", messages=messages, cache={"ttl": 5}
             )
@@ -204,6 +222,17 @@ def test_caching_with_cache_controls():
             print(f"response1: {response1}")
             print(f"response2: {response2}")
             assert response2["id"] == response1["id"]
+        else:
+            response1 = await litellm.acompletion(
+                model="gpt-3.5-turbo", messages=messages, cache={"ttl": 25}
+            )
+            await asyncio.sleep(10)
+            response2 = await litellm.acompletion(
+                model="gpt-3.5-turbo", messages=messages, cache={"s-maxage": 25}
+            )
+            print(f"response1: {response1}")
+            print(f"response2: {response2}")
+            assert response2["id"] == response1["id"]
     except Exception as e:
         print(f"error occurred: {traceback.format_exc()}")
         pytest.fail(f"Error occurred: {e}")

litellm/utils.py

@@ -2716,7 +2716,7 @@ def client(original_function):
 
             # [OPTIONAL] CHECK CACHE
             print_verbose(
-                f"kwargs[caching]: {kwargs.get('caching', False)}; litellm.cache: {litellm.cache}"
+                f"SYNC kwargs[caching]: {kwargs.get('caching', False)}; litellm.cache: {litellm.cache}; kwargs.get('cache')['no-cache']: {kwargs.get('cache', {}).get('no-cache', False)}"
             )
             # if caching is false or cache["no-cache"]==True, don't run this
             if (
@@ -2724,17 +2724,14 @@ def client(original_function):
                     (
                         (
                             kwargs.get("caching", None) is None
-                            and kwargs.get("cache", None) is None
                             and litellm.cache is not None
                         )
                         or kwargs.get("caching", False) == True
                     )
-                    and (
+                    and kwargs.get("cache", {}).get("no-cache", False) != True
-                        kwargs.get("cache", None) is None
-                        or kwargs["cache"].get("no-cache", False) != True
-                    )
                 )
                 and kwargs.get("aembedding", False) != True
+                and kwargs.get("atext_completion", False) != True
                 and kwargs.get("acompletion", False) != True
                 and kwargs.get("aimg_generation", False) != True
                 and kwargs.get("atranscription", False) != True
@@ -3014,21 +3011,16 @@ def client(original_function):
 
             # [OPTIONAL] CHECK CACHE
             print_verbose(
-                f"kwargs[caching]: {kwargs.get('caching', False)}; litellm.cache: {litellm.cache}; kwargs.get('cache'): {kwargs.get('cache', None)}"
+                f"ASYNC kwargs[caching]: {kwargs.get('caching', False)}; litellm.cache: {litellm.cache}; kwargs.get('cache'): {kwargs.get('cache', None)}"
             )
             # if caching is false, don't run this
             final_embedding_cached_response = None
 
             if (
-                (
+                (kwargs.get("caching", None) is None and litellm.cache is not None)
-                    kwargs.get("caching", None) is None
-                    and kwargs.get("cache", None) is None
-                    and litellm.cache is not None
-                )
                 or kwargs.get("caching", False) == True
             ) and (
-                kwargs.get("cache", None) is None
+                kwargs.get("cache", {}).get("no-cache", False) != True
-                or kwargs["cache"].get("no-cache", False) != True
             ):  # allow users to control returning cached responses from the completion function
                 # checking cache
                 print_verbose("INSIDE CHECKING CACHE")
@@ -3074,7 +3066,6 @@ def client(original_function):
                             preset_cache_key  # for streaming calls, we need to pass the preset_cache_key
                         )
                         cached_result = litellm.cache.get_cache(*args, **kwargs)
-
                     if cached_result is not None and not isinstance(
                         cached_result, list
                     ):