Update app.py

app.py

@@ -5,7 +5,7 @@ import os
 
 app = FastAPI()
 
-# CORS इनेबल करना ताकि आपका फ्रंटएंड इसे एक्सेस कर सके
+# CORS इनेबल करना
 app.add_middleware(
     CORSMiddleware,
     allow_origins=["*"],
@@ -13,13 +13,21 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
+# कॉन्फ़िगरेशन - Secrets से लें
 NVIDIA_API_KEY = os.getenv("NVIDIA_API_KEY")
 MODEL_NAME = os.getenv("MODEL_NAME", "microsoft/phi-4-multimodal-instruct")
 API_URL = "https://integrate.api.nvidia.com/v1/chat/completions"
 
+# सिस्टम प्रॉम्ट जो हमेशा सर्वर पर रहेगा
+SYSTEM_PROMPT = "You are Aura Gen 2.0, a state-of-the-art AI assistant developed by Divy Bhai. You possess world-class expertise in software engineering, advanced algorithmic logic, and complex problem-solving. Always provide highly optimized, secure, and precise solutions in English by default. Use Markdown formatting extensively (code blocks, tables, lists) to structure your answers clearly and effectively."
+
 @app.post("/v1/chat/completions")
 async def chat_proxy(request: Request):
     data = await request.json()
+    user_messages = data.get("messages", [])
+    
+    # हमेशा सिस्टम प्रॉम्ट को सबसे ऊपर जोड़ें
+    full_messages = [{"role": "system", "content": SYSTEM_PROMPT}] + user_messages
     
     async with httpx.AsyncClient() as client:
         response = await client.post(
@@ -30,10 +38,13 @@ async def chat_proxy(request: Request):
             },
             json={
                 "model": MODEL_NAME,
-                "messages": data.get("messages"),
+                "messages": full_messages,
                 "max_tokens": data.get("max_tokens", 4096),
                 "temperature": data.get("temperature", 0.6),
-                "top_p": 0.70
+                "top_p": 0.70,
+                "frequency_penalty": 0.00,
+                "presence_penalty": 0.00,
+                "stream": False
             },
             timeout=60.0
         )
@@ -41,4 +52,5 @@ async def chat_proxy(request: Request):
 
 if __name__ == "__main__":
     import uvicorn
+    # Hugging Face Spaces 7860 पोर्ट का उपयोग करता है
     uvicorn.run(app, host="0.0.0.0", port=7860)