Update main.py

main.py

@@ -25,6 +25,17 @@ BASE_URL = os.getenv("BASE_URL", "https://generativelanguage.googleapis.com/v1be
 EXPECTED_API_KEY = os.getenv("API_HUGGINGFACE")
 API_KEY_NAME = "Authorization"
 
+class Message(BaseModel):
+    role: str
+    content: str
+
+class ChatCompletionRequest(BaseModel):
+    model: str = "gemini-2.0-flash" 
+    messages: List[Message]
+    max_tokens: Optional[int] = 8196
+    temperature: Optional[float] = 0.8
+    stream: Optional[bool] = False
+
 api_key_header = APIKeyHeader(name=API_KEY_NAME, auto_error=False)
 app = FastAPI(title="OpenAI-SDK-compatible API", version="1.0.0", description="Un wrapper FastAPI compatibile con le specifiche dell'API OpenAI.")
 app.add_middleware(
@@ -94,17 +105,7 @@ def read_general():
 async def health_check():
     return {"message": "success"}
 
-class Message(BaseModel):
-    role: str
-    content: str
-    
 # ---------------------------------- Generazione Testo ---------------------------------------
-class ChatCompletionRequest(BaseModel):
-    model: str = "gemini-2.0-flash" 
-    messages: List[Message]
-    max_tokens: Optional[int] = 8196
-    temperature: Optional[float] = 0.8
-    stream: Optional[bool] = False
 
 @app.post("/v1/chat/completions", dependencies=[Depends(verify_api_key)])
 async def chat_completions(req: ChatCompletionRequest):