[OpenAI] Extend VLLMValidationError to additional validation parameters (#31870)

Signed-off-by: Rehan Khan <Rehan.Khan7@ibm.com>
This commit is contained in:
R3hankhan
2026-01-07 14:45:49 +00:00
committed by GitHub
parent b665bbc2d4
commit 1ab055efe6
9 changed files with 89 additions and 56 deletions
+1 -1
View File
@@ -911,7 +911,7 @@ def build_app(args: Namespace) -> FastAPI:
@app.exception_handler(RequestValidationError)
async def validation_exception_handler(_: Request, exc: RequestValidationError):
from vllm.entrypoints.openai.protocol import VLLMValidationError
from vllm.exceptions import VLLMValidationError
param = None
for error in exc.errors():
+1 -30
View File
@@ -72,6 +72,7 @@ from pydantic import (
)
from vllm.entrypoints.chat_utils import ChatCompletionMessageParam, make_tool_call_id
from vllm.exceptions import VLLMValidationError
from vllm.logger import init_logger
from vllm.logprobs import Logprob
from vllm.sampling_params import (
@@ -131,36 +132,6 @@ class ErrorResponse(OpenAIBaseModel):
error: ErrorInfo
class VLLMValidationError(ValueError):
"""vLLM-specific validation error for request validation failures.
Args:
message: The error message describing the validation failure.
parameter: Optional parameter name that failed validation.
value: Optional value that was rejected during validation.
"""
def __init__(
self,
message: str,
*,
parameter: str | None = None,
value: Any = None,
) -> None:
super().__init__(message)
self.parameter = parameter
self.value = value
def __str__(self):
base = super().__str__()
extras = []
if self.parameter is not None:
extras.append(f"parameter={self.parameter}")
if self.value is not None:
extras.append(f"value={self.value}")
return f"{base} ({', '.join(extras)})" if extras else base
class ModelPermission(OpenAIBaseModel):
id: str = Field(default_factory=lambda: f"modelperm-{random_uuid()}")
object: str = "model_permission"
@@ -140,16 +140,16 @@ class OpenAIServingCompletion(OpenAIServing):
)
except ValueError as e:
logger.exception("Error in preprocessing prompt inputs")
return self.create_error_response(str(e))
return self.create_error_response(e)
except TypeError as e:
logger.exception("Error in preprocessing prompt inputs")
return self.create_error_response(str(e))
return self.create_error_response(e)
except RuntimeError as e:
logger.exception("Error in preprocessing prompt inputs")
return self.create_error_response(str(e))
return self.create_error_response(e)
except jinja2.TemplateError as e:
logger.exception("Error in preprocessing prompt inputs")
return self.create_error_response(str(e))
return self.create_error_response(e)
# Extract data_parallel_rank from header (router can inject it)
data_parallel_rank = self._get_data_parallel_rank(raw_request)
+1 -1
View File
@@ -754,7 +754,7 @@ class OpenAIServing:
if isinstance(message, Exception):
exc = message
from vllm.entrypoints.openai.protocol import VLLMValidationError
from vllm.exceptions import VLLMValidationError
if isinstance(exc, VLLMValidationError):
err_type = "BadRequestError"
+1 -1
View File
@@ -373,7 +373,7 @@ class OpenAIServingResponses(OpenAIServing):
NotImplementedError,
) as e:
logger.exception("Error in preprocessing prompt inputs")
return self.create_error_response(f"{e} {e.__cause__}")
return self.create_error_response(e)
request_metadata = RequestResponseMetadata(request_id=request.request_id)
if raw_request:
+1 -1
View File
@@ -12,7 +12,7 @@ import torch
from pydantic import Field
from vllm.config import ModelConfig
from vllm.entrypoints.openai.protocol import VLLMValidationError
from vllm.exceptions import VLLMValidationError
from vllm.inputs.data import EmbedsPrompt, TextPrompt, TokensPrompt
from vllm.inputs.parse import get_prompt_components, parse_raw_prompts
from vllm.tokenizers import TokenizerLike