From 372e6c53343642d1166dcf0550be7dc4712fd718 Mon Sep 17 00:00:00 2001
From: apex-mochen <2756823972@qq.com>
Date: Wed, 30 Sep 2026 22:48:36 +0800
Subject: [PATCH] fix(cache): apply default_redis_ttl on the pre/post-call
 batch write path

_set_on_batch fed default_in_memory_ttl to Redis whenever no explicit ttl
was passed, so a configured default_redis_ttl was silently ignored on the
request/post-call pipeline path (behind async_set_cache_pre_call /
async_set_cache_post_call). Chain default_redis_ttl -> default_in_memory_ttl
the same way set_cache / async_set_cache / async_set_cache_pipeline do, and
add a default_redis_ttl=3600 case beside the #43779 post-call test.
---
 litellm/caching/dual_cache.py                 |  5 ++-
 .../test_request_redis_batch_post_call.py     | 37 +++++++++++++++++++
 2 files changed, 41 insertions(+), 1 deletion(-)

diff --git a/litellm/caching/dual_cache.py b/litellm/caching/dual_cache.py
index de63f91c011..ee829998c47 100644
--- a/litellm/caching/dual_cache.py
+++ b/litellm/caching/dual_cache.py
@@ -553,7 +553,10 @@ class DualCache(BaseCache):
         effective_ttl: Final = self.default_in_memory_ttl if ttl is None else ttl
         if self.in_memory_cache is not None:
             await self.in_memory_cache.async_set_cache(key, value, ttl=effective_ttl)
-        return batch.set(key, value, effective_ttl)
+        redis_ttl = self.default_redis_ttl if ttl is None else ttl
+        if redis_ttl is None:
+            redis_ttl = effective_ttl
+        return batch.set(key, value, redis_ttl)
 
     # async_batch_set_cache
     async def async_set_cache_pipeline(
diff --git a/tests/unit/caching/test_request_redis_batch_post_call.py b/tests/unit/caching/test_request_redis_batch_post_call.py
index fdd328a9a57..bd799c13f0a 100644
--- a/tests/unit/caching/test_request_redis_batch_post_call.py
+++ b/tests/unit/caching/test_request_redis_batch_post_call.py
@@ -278,6 +278,43 @@ async def test_a_slot_released_before_the_response_reaches_redis_at_once_not_on_
     assert client.pipelines == []
 
 
+@pytest.mark.asyncio
+async def test_a_deferred_response_cache_set_without_a_ttl_expires_in_redis_like_the_direct_path():
+    client = FakeClient(_ok_replies)
+    redis_cache = PostCallFakeRedisCache(client)
+    dual_cache = DualCache(redis_cache=redis_cache, in_memory_cache=InMemoryCache(), default_in_memory_ttl=300)
+
+    await dual_cache.async_set_cache("direct", {"id": "resp"})
+    with request_redis_batch_scope():
+        await dual_cache.async_set_cache_post_call("deferred", {"id": "resp"}, None)
+        await flush_post_call_redis_batches()
+
+    (command,) = client.pipelines[0].commands
+    assert (command[0], command[1], command[3]) == ("SET", "deferred", redis_cache.alone[0][2]["ttl"])
+    assert command[3] == 300
+
+
+@pytest.mark.asyncio
+async def test_a_deferred_response_cache_set_without_a_ttl_uses_default_redis_ttl_when_configured():
+    client = FakeClient(_ok_replies)
+    redis_cache = PostCallFakeRedisCache(client)
+    dual_cache = DualCache(
+        redis_cache=redis_cache,
+        in_memory_cache=InMemoryCache(),
+        default_in_memory_ttl=300,
+        default_redis_ttl=3600,
+    )
+
+    await dual_cache.async_set_cache("direct", {"id": "resp"})
+    with request_redis_batch_scope():
+        await dual_cache.async_set_cache_post_call("deferred", {"id": "resp"}, None)
+        await flush_post_call_redis_batches()
+
+    (command,) = client.pipelines[0].commands
+    assert (command[0], command[1], command[3]) == ("SET", "deferred", redis_cache.alone[0][2]["ttl"])
+    assert command[3] == 3600
+
+
 @pytest.mark.asyncio
 async def test_a_released_slot_is_free_locally_at_once_and_the_older_redis_count_does_not_overwrite_the_gauge():
     def replies(command: tuple[object, ...]) -> object: