Update app.py

app.py

@@ -1,6 +1,5 @@
-from fastapi import FastAPI, HTTPException, Response
+from fastapi import FastAPI, HTTPException
 from fastapi.middleware.cors import CORSMiddleware
-from fastapi.middleware.base import BaseHTTPMiddleware
 from fastapi.responses import JSONResponse
 import os
 from dotenv import load_dotenv
@@ -14,15 +13,6 @@ load_dotenv()
 
 app = FastAPI()
 
-class ContentTypeMiddleware(BaseHTTPMiddleware):
-    async def dispatch(self, request, call_next):
-        response = await call_next(request)
-        if isinstance(response, JSONResponse):
-            response.headers["Content-Type"] = "application/json; charset=utf-8"
-        return response
-
-app.add_middleware(ContentTypeMiddleware)
-
 app.add_middleware(
     CORSMiddleware,
     allow_origins=["*"],
@@ -54,21 +44,11 @@ def clean_assistant_response(text: str) -> str:
         text = text[:-3].strip()
     return text
 
-class CustomJSONResponse(JSONResponse):
-    media_type = "application/json; charset=utf-8"
-
-    def render(self, content: Any) -> bytes:
-        return json.dumps(
-            content,
-            ensure_ascii=False,
-            allow_nan=False,
-            indent=None,
-            separators=(',', ':')
-        ).encode('utf-8')
-
 @app.get("/")
 async def root():
-    return CustomJSONResponse({"status": "FastFlowWrapper is running"})
+    response = JSONResponse({"status": "FastFlowWrapper is running"})
+    response.headers["Content-Type"] = "application/json; charset=utf-8"
+    return response
 
 @app.get("/v1/models")
 async def get_models():
@@ -92,7 +72,9 @@ async def get_models():
                 "system_fingerprint": "phi4-r1"
             })
         
-        return CustomJSONResponse({"object": "list", "data": models})
+        response = JSONResponse({"object": "list", "data": models})
+        response.headers["Content-Type"] = "application/json; charset=utf-8"
+        return response
     except requests.RequestException as e:
         raise HTTPException(status_code=500, detail=str(e))
 
@@ -130,7 +112,7 @@ async def create_chat_completion(request: ChatCompletionRequest):
         # Подсчитываем токены ответа
         completion_tokens = count_tokens(assistant_response)
         
-        return CustomJSONResponse({
+        response = JSONResponse({
             "id": "chatcmpl-" + os.urandom(12).hex(),
             "object": "chat.completion",
             "created": int(start_time),
@@ -154,5 +136,7 @@ async def create_chat_completion(request: ChatCompletionRequest):
             "stats": {},
             "system_fingerprint": "phi4-r1"
         })
+        response.headers["Content-Type"] = "application/json; charset=utf-8"
+        return response
     except requests.RequestException as e:
         raise HTTPException(status_code=500, detail=str(e))