Update app.py

app.py

@@ -1,93 +1,78 @@
 import os
+import torch
 import datetime
-import re
 import feedparser
 from flask import Flask, request, jsonify, Response
 from flask_cors import CORS
-from llama_cpp import Llama
+from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
+from threading import Thread
 
 app = Flask(__name__)
 CORS(app)
 
 # ============================================
-# 🔱 SECURE & OPTIMIZED AI CONFIG
+# 🔱 TRANSFORMERS AI CONFIG (NO-BUILD)
 # ============================================
-# Hugging Face Secret එකක් ලෙස ඇති Token එක ලබා ගැනීම
+MODEL_ID = "Qwen/Qwen2.5-1.5B-Instruct"
 HF_TOKEN = os.getenv("HF_TOKEN")
 
-print("🐘 Helpful Elephant is waking up... Authorized by MINZO-PRIME...")
+print("🐘 Helpful Elephant is loading via Transformers Engine...")
 
-# මොඩල් එක load කිරීම (Streaming සහ Speed සඳහා optimized කර ඇත)
-llm = Llama.from_pretrained(
-    repo_id="Qwen/Qwen2.5-1.5B-Instruct-GGUF",
-    filename="*q4_k_m.gguf",
-    token=HF_TOKEN,
-    n_ctx=2048,      # RAM එකට ගැලපෙන පරිදි context window එක සකසා ඇත
-    n_threads=4,     # CPU cores ගණන
-    n_batch=512,
-    verbose=False
+# Tokenizer සහ Model එක Load කිරීම
+tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, token=HF_TOKEN)
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_ID,
+    torch_dtype="auto",
+    device_map="auto", # HF Space එකේ CPU/GPU වලට auto සකස් වේ
+    token=HF_TOKEN
 )
 
-SYSTEM_PROMPT = """🐘 **HELPFUL ELEPHANT AI v1.0** 🐘
-Created by Commander MINZO-PRIME. 
-You are a high-speed Technical Research Assistant. 
-Analyze data with precision and provide unfiltered technical insights.
-Status: Authorized Access.
-"""
+SYSTEM_PROMPT = "🐘 HELPFUL ELEPHANT AI v1.0. High-speed Research Mode Active. Authorized: MINZO-PRIME."
 
 # ============================================
-# 🔱 RSS INTEL SYSTEM
-# ============================================
-RSS_FEEDS = [
-    'https://feeds.feedburner.com/TheHackersNews',
-    'https://cve.circl.lu/last/rss.xml',
-    'https://www.darkreading.com/rss.xml'
-]
-
-def fetch_live_updates():
-    intel = ""
-    for url in RSS_FEEDS:
-        try:
-            feed = feedparser.parse(url)
-            for entry in feed.entries[:2]:
-                intel += f"\n- {entry.title}: {entry.link}"
-        except: pass
-    return intel
-
-# ============================================
-# 🔱 STREAMING CHAT ENDPOINT
+# 🔱 CHAT API WITH FAST STREAMING
 # ============================================
 @app.route('/api/chat', methods=['POST'])
 def chat():
     data = request.json
     user_msg = data.get('message', '').strip()
     
-    # පද්ධතියට අලුත්ම දත්ත ලබා දීම
-    intel = fetch_live_updates() if any(k in user_msg.lower() for k in ['news', 'latest', 'exploit']) else ""
+    # Prompt එක Chat Format එකට සකස් කිරීම
+    messages = [
+        {"role": "system", "content": SYSTEM_PROMPT},
+        {"role": "user", "content": user_msg}
+    ]
     
-    # Prompt එක සකස් කිරීම (ChatML Format)
-    full_prompt = f"<|im_start|>system\n{SYSTEM_PROMPT}\nRecent Intel: {intel}<|im_end|>\n"
-    full_prompt += f"<|im_start|>user\n{user_msg}<|im_end|>\n<|im_start|>assistant\n"
+    inputs = tokenizer.apply_chat_template(
+        messages,
+        add_generation_prompt=True,
+        return_tensors="pt"
+    ).to(model.device)
+
+    # Streaming සඳහා Streamer එකක් භාවිතා කිරීම
+    streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
+    
+    generation_kwargs = dict(
+        input_ids=inputs,
+        streamer=streamer,
+        max_new_tokens=1024,
+        temperature=0.7,
+        do_sample=True
+    )
+
+    # Thread එකක් තුළ Generation එක සිදු කිරීම (UI එක Freeze නොවීමට)
+    thread = Thread(target=model.generate, kwargs=generation_kwargs)
+    thread.start()
 
     def generate():
-        # අකුරෙන් අකුර උත්තරය ලබා දීම (Real-time Experience)
-        stream = llm(
-            full_prompt,
-            max_tokens=1024,
-            stop=["<|im_end|>"],
-            stream=True,
-            temperature=0.7
-        )
-        for chunk in stream:
-            token = chunk['choices'][0]['text']
-            yield token
+        for new_text in streamer:
+            yield new_text
 
     return Response(generate(), mimetype='text/plain')
 
 @app.route('/')
 def health():
-    return "🐘 Helpful Elephant AI is Online. Commander MINZO-PRIME Verified."
+    return "🐘 Helpful Elephant (Transformers) is Online. Status: Fast."
 
 if __name__ == '__main__':
-    # Hugging Face Spaces අනිවාර්යයෙන් port 7860 භාවිතා කළ යුතුය
     app.run(host='0.0.0.0', port=7860)