[Reasoning][Frontend] Add model config to adjust_request in reasoning parser (#37848)

Signed-off-by: rishitdholakia13 <rishit+github@cohere.com>
Signed-off-by: rishitdholakia13 <123388671+rishitdholakia13@users.noreply.github.com>
Signed-off-by: Aaron Pham <contact@aarnphm.xyz>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
Co-authored-by: Aaron Pham <contact@aarnphm.xyz>

vllm/entrypoints/serve/render/serving.py

@@ -548,7 +548,9 @@ class OpenAIServingRender:
 
         if reasoning_parser is not None:
             tokenizer = renderer.get_tokenizer()
-            request = reasoning_parser(tokenizer).adjust_request(request=request)
+            request = reasoning_parser(
+                tokenizer, model_config=self.model_config
+            ).adjust_request(request=request)
 
         # tool parsing is done only if a tool_parser has been set and if
         # tool_choice is not "none" (if tool_choice is "none" but a tool_parser

vllm/reasoning/abs_reasoning_parsers.py

@@ -14,6 +14,7 @@ from vllm.utils.collection_utils import is_list_of
 from vllm.utils.import_utils import import_from_path
 
 if TYPE_CHECKING:
+    from vllm.config import ModelConfig
     from vllm.entrypoints.openai.chat_completion.protocol import ChatCompletionRequest
     from vllm.entrypoints.openai.engine.protocol import DeltaMessage
     from vllm.entrypoints.openai.responses.protocol import ResponsesRequest
@@ -32,6 +33,9 @@ class ReasoningParser:
 
     def __init__(self, tokenizer: "TokenizerLike", *args, **kwargs):
         self.model_tokenizer = tokenizer
+        # Optional vLLM ModelConfig from the server. Use get (not pop) so composite
+        # parsers can forward **kwargs to nested parsers.
+        self._model_config: ModelConfig | None = kwargs.get("model_config")
 
     @cached_property
     def vocab(self) -> dict[str, int]: