[Core] Refactor omni prefix cache into Manager/Controller (#6654)

Signed-off-by: ZhengWG <zwg0606@gmail.com>
Co-authored-by: amy-why-3459 <wuhaiyan17@huawei.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
This commit is contained in:
authored and GitHub committed 2026-09-19 21:23:59 +08:00
1 parent ea84055507
commit 470ec60848
25 files changed
+5710 -2149

No files matched your search

@@ -74,7 +74,9 @@ def test_ar_runner_without_model_hook_stays_on_normal_path():
assert "exceptException" not in compact_request_source
set_context_source = source[source.index("def _set_runner_assisted_full_attention_metadata_context") :]
set_context_source = set_context_source[: set_context_source.index("def _deferred_prefix_cache_mm_keys")]
set_context_source = set_context_source[
: set_context_source.index("def _prepare_prefix_cache_pooler_payload_sources")
]
assert "except Exception" not in set_context_source