x

app/config.py

@@ -17,6 +17,7 @@ class Settings(BaseModel):
     port: int = 8000
     openai_host: str = "0.0.0.0"
     openai_port: int = 8001
+    vllm_openai_internal_url: str = "http://127.0.0.1:8001/v1"
     public_model_name: str = "Qwen_local_model"
     reasoning_enabled: bool = False
     model_root: str = "/opt/model"
@@ -46,6 +47,7 @@ def get_settings() -> Settings:
         port=runtime["port"],
         openai_host=runtime["openai_host"],
         openai_port=runtime["openai_port"],
+        vllm_openai_internal_url=runtime["vllm_openai_internal_url"],
         public_model_name=runtime["public_model_name"],
         reasoning_enabled=runtime["reasoning_enabled"],
         model_root=runtime["model_root"],

app/engine.py

@@ -25,6 +25,8 @@ class InferenceEngine:
         }
         if req.stop:
             payload["stop"] = req.stop
+        if req.enable_thinking is not None:
+            payload["chat_template_kwargs"] = {"enable_thinking": req.enable_thinking}
         response = self.client.post(
             f"{self.settings.vllm_openai_internal_url}/chat/completions",
             headers=headers,

app/model_catalog.py

@@ -63,11 +63,16 @@ def resolve_runtime_settings(content: dict[str, Any]) -> dict[str, Any]:
     api_service = dict(services.get("api", {}))
     openai_service = dict(services.get("openai", {}))
     models = dict(content.get("models", {}))
+    openai_port = _to_int(openai_service.get("port"), 8001)
+    internal_url = _to_str(content.get("vllm_openai_internal_url"))
+    if not internal_url:
+        internal_url = f"http://127.0.0.1:{openai_port}/v1"
     return {
         "host": str(api_service.get("host", "0.0.0.0")),
         "port": _to_int(api_service.get("port"), 8000),
         "openai_host": str(openai_service.get("host", "0.0.0.0")),
-        "openai_port": _to_int(openai_service.get("port"), 8001),
+        "openai_port": openai_port,
+        "vllm_openai_internal_url": internal_url.rstrip("/"),
         "public_model_name": _to_str(content.get("public_model_name"), "Qwen_local_model"),
         "api_key": str(content.get("api_key", "")).strip() or None,
         "reasoning_enabled": _to_bool(content.get("reasoning_enabled"), False),

app/schemas.py

@@ -10,6 +10,7 @@ class GenerateRequest(BaseModel):
     top_p: float = Field(default=0.95, gt=0.0, le=1.0)
     repetition_penalty: float = Field(default=1.0, ge=0.5, le=2.0)
     stop: Optional[List[str]] = None
+    enable_thinking: Optional[bool] = None
 
 
 class GenerateResponse(BaseModel):