Add max_model_len field to ModelCard and ModelInfo for vLLM compatibility

- Added max_model_len field to ModelInfo dataclass in service_discovery.py
- Added max_model_len field to ModelCard in protocols.py
- Updated main_router.py to pass max_model_len when creating ModelCard
- This allows the router to properly forward the max_model_len field from vLLM endpoints

This fixes the issue where /v1/models endpoint was not showing the context window size for models.

Signed-off-by: llm-net <[email protected]>

Author: llm-net

src/vllm_router/protocols.py

@@ -49,6 +49,7 @@ class ModelCard(OpenAIBaseModel):
     owned_by: str = "vllm"
     root: Optional[str] = None
     parent: Optional[str] = None
+    max_model_len: Optional[int] = None
 
 
 class ModelList(OpenAIBaseModel):

src/vllm_router/routers/main_router.py

@@ -152,6 +152,7 @@ async def show_models():
                 created=model_info.created,
                 owned_by=model_info.owned_by,
                 parent=model_info.parent,
+                max_model_len=model_info.max_model_len,
             )
             model_cards.append(model_card)
             existing_models.add(model_id)

src/vllm_router/service_discovery.py

@@ -50,6 +50,7 @@ class ModelInfo:
     root: Optional[str] = None
     parent: Optional[str] = None
     is_adapter: bool = False
+    max_model_len: Optional[int] = None
 
     @classmethod
     def from_dict(cls, data: Dict) -> "ModelInfo":
@@ -62,6 +63,7 @@ def from_dict(cls, data: Dict) -> "ModelInfo":
             root=data.get("root", None),
             parent=data.get("parent", None),
             is_adapter=data.get("parent") is not None,
+            max_model_len=data.get("max_model_len", None),
         )
 
     def to_dict(self) -> Dict:
@@ -74,6 +76,7 @@ def to_dict(self) -> Dict:
             "root": self.root,
             "parent": self.parent,
             "is_adapter": self.is_adapter,
+            "max_model_len": self.max_model_len,
         }