diff --git a/deepinfer/structure_output/compiler.py b/deepinfer/structure_output/compiler.py
index 4bbcddf..829d4e4 100644
--- a/deepinfer/structure_output/compiler.py
+++ b/deepinfer/structure_output/compiler.py
@@ -10,6 +10,10 @@ from deepinfer import envs
 
 logger = logging.getLogger(__name__)
 
+#: The ``schema_type`` values ``dispatch_and_compile`` knows how to compile. Shared with the
+#: matcher builder so a wrong type is rejected at the request boundary, not inside the pool.
+SUPPORTED_SCHEMA_TYPES: tuple[str, ...] = ("json", "regex", "ebnf")
+
 
 class XGrammarCompiler:
     def __init__(self, tokenizer_info: TokenizerInfo, num_threads: int) -> None:
@@ -52,15 +56,18 @@ class XGrammarCompiler:
             self.close()
 
     def dispatch_and_compile(self, schema_type: str, schema: str) -> CompiledGrammar:
+        # Checked outside the try: the `except ValueError` below is for xgrammar's own input
+        # errors and used to swallow this one too, re-raising it as a RuntimeError that
+        # try_result() then reported as a timeout.
+        if schema_type not in SUPPORTED_SCHEMA_TYPES:
+            raise ValueError(f"unsupported schema_type {schema_type!r}; expected one of {SUPPORTED_SCHEMA_TYPES}")
         try:
             if schema_type == "json":
                 ctx = self.inner_compiler.compile_json_schema(schema, strict_mode=True)
             elif schema_type == "regex":
                 ctx = self.inner_compiler.compile_regex(schema)
-            elif schema_type == "ebnf":
+            else:  # "ebnf" — the guard above makes this exhaustive
                 ctx = self.inner_compiler.compile_grammar(schema)
-            else:
-                raise ValueError("Unsupported Grammar Type")
         except RuntimeError as e:
             raise RuntimeError(f"xgrammar compilation RE: {e}") from e
         except ValueError as e:
diff --git a/deepinfer/structure_output/so_matcher_builder.py b/deepinfer/structure_output/so_matcher_builder.py
index ab3b867..2afc2d7 100644
--- a/deepinfer/structure_output/so_matcher_builder.py
+++ b/deepinfer/structure_output/so_matcher_builder.py
@@ -7,7 +7,7 @@ from typing import TYPE_CHECKING
 
 from xgrammar import TokenizerInfo
 
-from deepinfer.structure_output.compiler import XGrammarCompiler
+from deepinfer.structure_output.compiler import SUPPORTED_SCHEMA_TYPES, XGrammarCompiler
 from deepinfer.structure_output.matcher import XGrammarMatcher, XGrammarMatcherFuture, validate_trigger_bound
 
 if TYPE_CHECKING:
@@ -65,6 +65,8 @@ class SOMatcherBuilder:
         if missing:
             raise ValueError(f"so_args is missing required key(s) {missing}; provide schema_type and schema")
         schema_type = so_args["schema_type"]
+        if schema_type not in SUPPORTED_SCHEMA_TYPES:
+            raise ValueError(f"unsupported schema_type {schema_type!r}; expected one of {SUPPORTED_SCHEMA_TYPES}")
         schema = so_args["schema"]
 
         trigger_tokens = so_args.get("trigger_token_ids")