Update app.py

app.py

@@ -1,8 +1,7 @@
 from fastapi import FastAPI, Request
-import requests
-import time
+from transformers import pipeline
+import torch
 from ddgs import DDGS
-from datetime import datetime
 from fastapi.middleware.cors import CORSMiddleware
 import uvicorn
 
@@ -15,14 +14,16 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
-# 🔱 Engine Upgrade: Llama-3-8B-Instruct (Faster & Stable)
-API_URL = "https://api-inference.huggingface.co/models/meta-llama/Meta-Llama-3-8B-Instruct"
+# 🔱 Loading local engine (SmolLM2 - Very fast and smart)
+print("🔱 Initializing Internal Neural Core...")
+pipe = pipeline("text-generation", model="HuggingFaceTB/SmolLM2-1.7B-Instruct", device_map="auto")
+print("🔱 Engine Ready!")
 
 def web_search(query):
     try:
         with DDGS() as ddgs:
-            results = [r for r in ddgs.text(query, max_results=3)]
-            return "\n".join([f"Verified Info: {r['body']}" for r in results])
+            results = [r for r in ddgs.text(query, max_results=2)]
+            return " ".join([r['body'] for r in results])
     except: return ""
 
 @app.post("/generate")
@@ -30,44 +31,18 @@ async def generate(request: Request):
     try:
         data = await request.json()
         user_prompt = data.get("prompt")
-        search_results = web_search(user_prompt)
-        current_date = datetime.now().strftime("%Y-%m-%d")
         
-        # 🔱 Optimized Llama-3 Prompt Format
-        full_prompt = (
-            f"<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\n"
-            f"Today is {current_date}. You are Elephant AI Pro, a powerful assistant created by MINZO-PRIME. "
-            f"Use this web context to provide accurate answers: {search_results}<|eot_id|>"
-            f"<|start_header_id|>user<|end_header_id|>\n\n{user_prompt}<|eot_id|>"
-            f"<|start_header_id|>assistant<|end_header_id|>\n\n"
-        )
-
-        payload = {
-            "inputs": full_prompt, 
-            "parameters": {
-                "max_new_tokens": 800, 
-                "temperature": 0.6, 
-                "top_p": 0.9,
-                "return_full_text": False
-            }
-        }
-
-        # 🔱 Fast-Response Logic
-        for attempt in range(3):
-            response = requests.post(API_URL, json=payload)
-            output = response.json()
-            
-            # පිළිතුර ලැබුණහොත් වහාම ලබා දෙයි
-            if isinstance(output, list) and 'generated_text' in output[0]:
-                return {"response": output[0]['generated_text'].strip()}
-            
-            # API එක Load වෙමින් පවතී නම් පමණක් සුළු වෙලාවක් රැඳී සිටියි
-            if "estimated_time" in str(output):
-                time.sleep(3)
-            else:
-                break
-
-        return {"response": "System Core is ready. Please re-send your command."}
+        context = web_search(user_prompt)
+        
+        messages = [
+            {"role": "system", "content": f"You are Elephant AI Pro by MINZO-PRIME. Web Data: {context}"},
+            {"role": "user", "content": user_prompt},
+        ]
+        
+        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)
+        
+        return {"response": outputs[0]["generated_text"].split("<|im_start|>assistant")[-1].strip()}
     
     except Exception as e:
         return {"error": str(e)}