[Spec] Enable grammar overlap scheduling for STANDALONE speculative decoding (#32110)

python/sglang/srt/speculative/spec_info.py

@@ -136,7 +136,8 @@ class SpeculativeAlgorithm(Enum):
     def supports_grammar_overlap(self) -> bool:
         # Whether the worker advances the grammar FSM inside verify() (via the
         # scheduler's grammar barrier), letting spec + grammar decode overlap.
-        return self.is_eagle()
+        # STANDALONE inherits the EAGLE V2 worker's verify path, barrier included.
+        return self.is_eagle() or self.is_standalone()
 
     def has_draft_kv(self) -> bool:
         """Whether the draft phase writes KV chains. NGRAM does not (its tree

test/registered/spec/test_spec_standalone.py

@@ -1,19 +1,24 @@
 import unittest
 
 from sglang.test.ci.ci_register import register_cuda_ci
+from sglang.test.kits.json_constrained_kit import JSONConstrainedMixin
+from sglang.test.kits.regex_constrained_kit import RegexConstrainedMixin
 from sglang.test.server_fixtures.standalone_fixture import StandaloneServerBase
 from sglang.test.test_utils import CustomTestCase
 
 # V2 standalone speculative decoding tests (FA3, Triton, FlashInfer backends).
 # Non-V2 backends moved to test_spec_standalone_extra.py.
-register_cuda_ci(est_time=406, stage="base-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=450, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestStandaloneV2SpeculativeDecodingBase(StandaloneServerBase, CustomTestCase):
     attention_backend = "fa3"
 
 
-class TestStandaloneV2SpeculativeDecodingTriton(StandaloneServerBase, CustomTestCase):
+class TestStandaloneV2SpeculativeDecodingTriton(
+    StandaloneServerBase, CustomTestCase, RegexConstrainedMixin, JSONConstrainedMixin
+):
+    # Constrained mixins reuse this server; overlap on -> grammar barrier path.
     attention_backend = "triton"