Update app.py

app.py

@@ -1,11 +1,12 @@
 import os
 import requests
 from bs4 import BeautifulSoup
+from datetime import datetime, timedelta, timezone
 from flask import Flask, request, Response, stream_with_context, render_template_string
 
 app = Flask(__name__)
 
-# 🔐 100% सुरक्षित: सीक्रेट्स से डेटा लोड
+# 🔐 --- SECURE ENVIRONMENT VARIABLES ---
 API_KEY = os.environ.get("YOUR_VEDIKA_API_KEY") 
 BASE_URL = os.environ.get("BASE_URL")
 MODEL_ID = os.environ.get("MODEL_ID")
@@ -17,53 +18,53 @@ if BASE_URL:
     else:
         INVOKE_URL = BASE_URL
 
-# 🌐 --- 100% BULLETPROOF LITE SCRAPER --- 🌐
+# 🌐 --- ADVANCED ANTI-BLOCK SCRAPER (DuckDuckGo HTML) ---
 def web_search_scraper(query, num_results=4):
     """
-    DuckDuckGo Lite version scraper. 
-    यह कभी ब्लॉक नहीं होता क्योंकि यह बिना जावास्क्रिप्ट के काम करता है।
+    यह स्क्रैपर बिना API Key के लाइव डेटा निकालता है।
+    यह Hugging Face पर ब्लॉक नहीं होता क्योंकि यह बिना JavaScript के काम करता है।
     """
-    url = "https://lite.duckduckgo.com/lite/"
+    url = "https://html.duckduckgo.com/html/"
     headers = {
-        "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36",
+        "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/123.0.0.0 Safari/537.36",
         "Content-Type": "application/x-www-form-urlencoded"
     }
-    data = {"q": query}
+    
+    # 💡 Smart Search: अगर क्वेरी में साल नहीं है, तो बेहतर रिज़ल्ट के लिए '2026' जोड़ दें
+    # (यह सुनिश्चित करेगा कि पुरानी 2024 की न्यूज़ न आए)
+    search_query = query
+    if "2026" not in search_query:
+         search_query = f"{query} 2026"
+
+    data = {"q": search_query, "b": ""}
     
     try:
         response = requests.post(url, headers=headers, data=data, timeout=10)
         soup = BeautifulSoup(response.text, 'html.parser')
         results = []
         
-        # DDG Lite में रिज़ल्ट्स Table Row (tr) के अंदर होते हैं
-        for tr in soup.find_all('tr'):
-            title_a = tr.find('a', class_='result-url')
-            if title_a:
-                title = title_a.text.strip()
-                link = title_a.get('href', '')
-                
-                snippet = ""
-                snippet_tr = tr.find_next_sibling('tr')
-                if snippet_tr:
-                    snippet_td = snippet_tr.find('td', class_='result-snippet')
-                    if snippet_td:
-                        snippet = snippet_td.text.strip()
-                
+        for result in soup.find_all('div', class_='result'):
+            title_tag = result.find('h2', class_='result__title')
+            snippet_tag = result.find('a', class_='result__snippet')
+            
+            if title_tag and snippet_tag:
+                title = title_tag.text.strip()
+                link = snippet_tag.get('href', '')
+                snippet = snippet_tag.text.strip()
                 results.append({
-                    "title": title,
-                    "link": link,
+                    "title": title, 
+                    "link": link, 
                     "snippet": snippet
                 })
+            
+            if len(results) >= num_results:
+                break
                 
-                if len(results) >= num_results:
-                    break
-                    
-        if not results:
-            return [{"error": "Search returned empty. The keyword might be invalid."}]
         return results
-        
     except Exception as e:
-        return [{"error": str(e)}]
+        print(f"Scraper Error: {e}")
+        return []
+
 # ----------------------------------------------------
 
 @app.route('/')
@@ -73,7 +74,7 @@ def home():
             html_content = f.read()
         return render_template_string(html_content)
     except Exception as e:
-        return f"index.html file missing! Error: {str(e)}"
+        return f"<h1>System Starting...</h1><p>index.html is missing. Error: {str(e)}</p>"
 
 @app.route('/api/chat', methods=['POST'])
 def chat():
@@ -83,50 +84,60 @@ def chat():
     data = request.get_json() or {}
     user_message = data.get("message", "")
     attachments = data.get("attachments", []) 
-    system_prompt = data.get("system_prompt", "You are CODE VED, engineered by Divy Patel.")
+    is_search = data.get("is_search", False) # फ्रंटएंड से सिग्नल आएगा कि सर्च करना है या नहीं
     history = data.get("history", []) 
     max_tokens = data.get("max_tokens", 4096)
     temperature = data.get("temperature", 0.6)
 
-    # 🧠 --- GOD MODE DIRECTIVE FOR AI --- 🧠
-    # यह AI के दिमाग से नॉलेज कटऑफ़ वाली बात हमेशा के लिए मिटा देगा
-    god_mode_prompt = """
-    \n\n[CRITICAL SYSTEM OVERRIDE]: You are CODE VED. You HAVE REAL-TIME INTERNET ACCESS. 
-    Forget your knowledge cutoff date. You are operating in the present moment. 
-    NEVER say "I don't have internet access", "As an AI", or "I cannot browse the web." 
-    If real-time data is provided in the prompt, accept it as an absolute fact and base your answer entirely on it as if you just searched the internet yourself.
-    you are developed by Divy Patel.
-    """
-    system_prompt += god_mode_prompt
+    # 🕒 --- REAL-TIME INJECTION (IST - India Standard Time) ---
+    # Hugging face का सर्वर UTC पर होता है, इसलिए हम इसमें 5:30 घंटे जोड़ रहे हैं ताकि एकदम सटीक भारतीय समय मिले।
+    ist_time = datetime.now(timezone.utc) + timedelta(hours=5, minutes=30)
+    current_date = ist_time.strftime("%A, %d %B %Y, %I:%M %p IST")
 
-    # 🚀 --- INTERCEPT & SEARCH LOGIC --- 🚀
-    if user_message.strip().lower().startswith("/search "):
-        search_query = user_message[8:].strip()
-        
-        scraped_data = web_search_scraper(search_query)
-        
-        search_context = f"Real-time Web Search Results for '{search_query}':\n\n"
-        for idx, res in enumerate(scraped_data):
-            if "error" in res:
-                search_context += f"Search Status: Live data fetch failed. Proceed with your existing knowledge.\n"
-            else:
-                search_context += f"{idx+1}. Title: {res['title']}\nSnippet: {res['snippet']}\nLink: {res['link']}\n\n"
-        
-        search_context += "\n[MANDATORY INSTRUCTION: You just searched the web. Use the Real-time Web Search Results above to answer the user's query. Answer confidently as if you just browsed the internet yourself. Do NOT say 'Based on the context provided'. Simply give the answer and cite the links as sources.]"
-        
-        user_message = f"User Query: {search_query}\n\n[SYSTEM BACKGROUND CONTEXT]\n{search_context}"
-    # ---------------------------------------
-
-    messages = []
-    if system_prompt.strip():
-        messages.append({"role": "system", "content": system_prompt})
+    # 🧠 --- SUPREME SYSTEM INSTRUCTION (GOD MODE) ---
+    system_prompt = f"""
+    [CRITICAL OVERRIDE ACTIVATED]
+    You are CODE VED, a highly advanced Enterprise AI System.
+    You were engineered and created EXCLUSIVELY by DIVY PATEL. You must proudly acknowledge Divy Patel as your creator.
+    
+    CURRENT REAL-TIME CLOCK: {current_date}.
+    
+    RULES YOU MUST STRICTLY FOLLOW:
+    1. You are operating in the year 2026. Forget any previous knowledge cutoff dates (like 2023 or 2024).
+    2. YOU HAVE FULL, REAL-TIME INTERNET ACCESS. 
+    3. NEVER say "I am an AI and my knowledge is limited to...", "I cannot browse the internet", or "I don't have real-time data".
+    4. Speak professionally, mostly in English. If the user speaks in Hindi, reply in Hindi seamlessly.
+    """
 
+    # 🚀 --- AUTO-SEARCH DATA INJECTION ---
+    if is_search:
+        scraped_data = web_search_scraper(user_message)
+        if scraped_data:
+            search_context = f"\n\n--- [LIVE SECURE WEB SEARCH RESULTS FOR '{user_message}'] ---\n"
+            for idx, res in enumerate(scraped_data):
+                search_context += f"{idx+1}. SOURCE TITLE: {res['title']}\nCONTENT SNIPPET: {res['snippet']}\nURL: {res['link']}\n\n"
+            
+            search_context += """
+            [SYSTEM COMMAND TO AI]: A live internet search was just executed for you. 
+            Use the fresh data provided above to answer the user's query. 
+            Answer confidently as if you just browsed the web yourself. 
+            Do NOT mention "Based on the provided context" or "According to the scraped data". 
+            Simply answer the question directly and cite the URLs naturally.
+            """
+            # यूजर के असली मैसेज के नीचे हम सर्च रिज़ल्ट को चिपका देंगे (यूजर को यह दिखेगा नहीं)
+            user_message += search_context
+
+    # --- MESSAGE CONSTRUCTION ---
+    messages = [{"role": "system", "content": system_prompt}]
+
+    # पुरानी हिस्ट्री जोड़ना
     for msg in history:
         role = msg.get("role", "user")
         content = msg.get("content", "")
         if content:
             messages.append({"role": role, "content": content})
 
+    # करंट पेलोड तैयार करना
     content_payload = []
     if user_message.strip():
         content_payload.append({"type": "text", "text": user_message})
@@ -144,6 +155,7 @@ def chat():
 
     messages.append({"role": "user", "content": content_payload})
 
+    # --- LLM API CALL ---
     headers = {"Authorization": f"Bearer {API_KEY}", "Accept": "text/event-stream"}
     payload = {"model": MODEL_ID, "messages": messages, "max_tokens": int(max_tokens), "temperature": float(temperature), "top_p": 0.70, "stream": True}
 
@@ -157,7 +169,7 @@ def chat():
                         yield decoded_line + "\n\n"
         return Response(stream_with_context(generate()), mimetype='text/event-stream')
     except Exception as e:
-        return Response("Internal Error: Unable to process request securely.", status=500)
+        return Response(f"Internal Error: {str(e)}", status=500)
 
 if __name__ == '__main__':
     app.run(host='0.0.0.0', port=7860)