Update app.py

app.py

@@ -14,9 +14,14 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
-# 🔱 Loading local engine (SmolLM2 - Very fast and smart)
+# 🔱 Loading local engine (SmolLM2 - Stable Version)
 print("🔱 Initializing Internal Neural Core...")
-pipe = pipeline("text-generation", model="HuggingFaceTB/SmolLM2-1.7B-Instruct", device_map="auto")
+pipe = pipeline(
+    "text-generation", 
+    model="HuggingFaceTB/SmolLM2-1.7B-Instruct", 
+    device_map="auto",
+    torch_dtype=torch.bfloat16 # RAM එක ඉතිරි කර ගැනීමට
+)
 print("🔱 Engine Ready!")
 
 def web_search(query):
@@ -31,7 +36,6 @@ async def generate(request: Request):
     try:
         data = await request.json()
         user_prompt = data.get("prompt")
-        
         context = web_search(user_prompt)
         
         messages = [
@@ -40,9 +44,13 @@ async def generate(request: Request):
         ]
         
         prompt = pipe.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
-        outputs = pipe(prompt, max_new_tokens=500, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
+        outputs = pipe(prompt, max_new_tokens=500, do_sample=True, temperature=0.7)
+        
+        # පිළිතුර පමණක් වෙන් කර ගැනීම
+        generated_text = outputs[0]["generated_text"]
+        response = generated_text.split("<|im_start|>assistant")[-1].strip()
         
-        return {"response": outputs[0]["generated_text"].split("<|im_start|>assistant")[-1].strip()}
+        return {"response": response}
     
     except Exception as e:
         return {"error": str(e)}