From c1f79fdf862a531761162be1e04a64893b4e5384 Mon Sep 17 00:00:00 2001
From: Amr Mohammed <amrm88289@gmail.com>
Date: Sun, 27 Sep 2026 16:30:31 +0300
Subject: [PATCH] refactor(caching): remove redundant comments, restore
 warning-set state in test

---
 litellm/caching/caching.py          |  4 ----
 tests/local_testing/test_caching.py | 33 ++++++++++++-----------------
 2 files changed, 13 insertions(+), 24 deletions(-)

diff --git a/litellm/caching/caching.py b/litellm/caching/caching.py
index 50b29830f31..088922046e1 100644
--- a/litellm/caching/caching.py
+++ b/litellm/caching/caching.py
@@ -394,10 +394,6 @@ class Cache:
                     param_value = kwargs[param]
                     cache_key += f"{param}: {param_value}"
                 elif kwargs[param] is not None and param not in _warned_dropped_cache_params:
-                    # Provider-specific param (e.g. ollama num_ctx) is being excluded
-                    # from the cache key. Two requests differing only in this param
-                    # will collide and the second gets the first's cached response.
-                    # Warn once per param; set enable_caching_on_provider_specific_optional_params=True to include it.
                     _warned_dropped_cache_params.add(param)
                     verbose_logger.warning(
                         "litellm.cache: provider-specific param '%s' is excluded from the cache key by default, "
diff --git a/tests/local_testing/test_caching.py b/tests/local_testing/test_caching.py
index c6eda589096..8f809df457a 100644
--- a/tests/local_testing/test_caching.py
+++ b/tests/local_testing/test_caching.py
@@ -78,32 +78,25 @@ async def test_dual_cache_async_batch_get_cache():
         
         
 def test_cache_key_warns_on_dropped_provider_specific_param(caplog):
-    """Provider-specific params dropped from the cache key (default flag) should
-    emit a one-time warning. Regression for #42400."""
     import litellm
     from litellm.caching.caching import Cache, _warned_dropped_cache_params
 
     litellm.enable_caching_on_provider_specific_optional_params = False
-    _warned_dropped_cache_params.discard("num_ctx")  # reset in case another test warned
+    _warned_dropped_cache_params.discard("num_ctx")
     cache = Cache()
+    try:
+        with caplog.at_level(logging.WARNING):
+            cache.get_cache_key(model="ollama/llama3.2",
+                messages=[{"role": "user", "content": "hello"}], num_ctx=2048)
+        assert any("num_ctx" in r.message for r in caplog.records)
 
-    with caplog.at_level(logging.WARNING):
-        cache.get_cache_key(
-            model="ollama/llama3.2",
-            messages=[{"role": "user", "content": "hello"}],
-            num_ctx=2048,
-        )
-    assert any("num_ctx" in r.message for r in caplog.records)
-
-    # second call must NOT warn again (one-time behavior)
-    caplog.clear()
-    with caplog.at_level(logging.WARNING):
-        cache.get_cache_key(
-            model="ollama/llama3.2",
-            messages=[{"role": "user", "content": "hello"}],
-            num_ctx=4096,
-        )
-    assert not any("num_ctx" in r.message for r in caplog.records)
+        caplog.clear()
+        with caplog.at_level(logging.WARNING):
+            cache.get_cache_key(model="ollama/llama3.2",
+                messages=[{"role": "user", "content": "hello"}], num_ctx=4096)
+        assert not any("num_ctx" in r.message for r in caplog.records)
+    finally:
+        _warned_dropped_cache_params.discard("num_ctx")
 
 
 def test_dual_cache_batch_get_cache():