feat: Session-reference-aware Unified Radix Cache for agentic multi-turn workloads (#29173)

Co-authored-by: Ishan Dhanani <ishandhanani@gmail.com>
Co-authored-by: hzh0425 <hzh0425@apache.org>
Co-authored-by: ispobock <ispobaoke@gmail.com>
This commit is contained in:
Mingjun Zhang
2026-08-02 13:43:55 +08:00
committed by GitHub
co-authored by Ishan Dhanani hzh0425 ispobock
parent 131bd51b01
commit 056474cdb0
23 changed files with 1056 additions and 373 deletions
@@ -1274,17 +1274,6 @@ class TestPrefillOnlyDisableKvCache(unittest.TestCase):
self._validate_prefill_only_args(kv_cache_dtype=kv_cache_dtype)
class TestSessionRadixCacheServerArgs(unittest.TestCase):
def test_requires_priority_radix_eviction_policy(self):
server_args = ServerArgs(
model_path="dummy",
enable_session_radix_cache=True,
radix_eviction_policy="lru",
)
with self.assertRaisesRegex(ValueError, "--radix-eviction-policy priority"):
server_args._handle_cache_compatibility()
class TestCudaGraphConfigDataclassAccess(CustomTestCase):
@patch(
"sglang.srt.model_executor.runner_backend."