[Bugfix] validate grammar and throw 400 error instead of crashing the engine when xgrammar validation fails (#17623)

Signed-off-by: Jason Cheng <jasoncky96@gmail.com>
Co-authored-by: Russell Bryant <rbryant@redhat.com>

vllm/v1/engine/processor.py

@@ -188,8 +188,10 @@ class Processor:
                 validate_xgrammar_grammar(params)
                 params.guided_decoding.backend = "xgrammar"
             except ValueError:
-                # The request includes some jsonschema feature(s) that
+                # The request either failed validation
+                # or includes some jsonschema feature(s) that
                 # are not supported in xgrammar. Fall back to guidance.
+                validate_guidance_grammar(params, tokenizer=None)
                 params.guided_decoding.backend = "guidance"
             # Remember that this backend was set automatically
             params.guided_decoding.backend_was_auto = True

vllm/v1/structured_output/backend_xgrammar.py

@@ -282,6 +282,12 @@ def validate_xgrammar_grammar(sampling_params: SamplingParams) -> None:
         else:
             schema = gd_params.json
 
+        try:
+            xgr.Grammar.from_json_schema(schema)
+        except Exception as err:
+            raise ValueError("Failed to transform json schema into a grammar: "
+                             f"{err}") from err
+
         if has_xgrammar_unsupported_json_features(schema):
             raise ValueError("The provided JSON schema contains features not "
                              "supported by xgrammar.")