Update app.py

app.py

@@ -1,83 +1,56 @@
 import os
 import requests
-import urllib.parse
-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__)
 
 # 🔐 --- SECURE ENVIRONMENT VARIABLES ---
-API_KEY = os.environ.get("YOUR_VEDIKA_API_KEY") 
-MODEL_ID = os.environ.get("MODEL_ID") # Default fallback
+API_KEY = os.environ.get("NVIDIA_API_KEY") or os.environ.get("YOUR_VEDIKA_API_KEY") 
+MODEL_ID = os.environ.get("MODEL_ID", "google/diffusiongemma-26b-a4b-it") # Default fallback
 
 # NVIDIA's official invoke URL
 INVOKE_URL = "https://integrate.api.nvidia.com/v1/chat/completions"
 
-# 🌐 --- SUPER-HYBRID UNBLOCKABLE SCRAPER --- 🌐
-def web_search_scraper(query, num_results=4):
+# 🔑 --- SERPAPI KEY (Provided by Divy Patel) ---
+SERPAPI_KEY = "df7dc67448cc9fe63ee6bbc7c20c50662ac18bc1f4682cd480da97bde1970381"
+
+# 🌐 --- SERPAPI GOOGLE SEARCH ENGINE (100% BULLETPROOF) --- 🌐
+def web_search_scraper(query, num_results=5):
     """
-    Yahoo Search + Google News + Wikipedia (Never blocks on Hugging Face)
+    यह SerpApi का उपयोग करके सीधे Google Search से एकदम सटीक और ताज़ा (JSON) डेटा लाता है।
+    इसे Hugging Face या Google कभी ब्लॉक नहीं कर सकता।
     """
     results = []
-    headers = {
-        "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0.0.0 Safari/537.36",
-        "Accept-Language": "en-US,en;q=0.9"
-    }
-
-    # 1. YAHOO SEARCH (Highly reliable for Datacenter IPs like Hugging Face)
     try:
-        url = f"https://search.yahoo.com/search?p={urllib.parse.quote(query)}"
-        res = requests.get(url, headers=headers, timeout=8)
-        soup = BeautifulSoup(res.text, 'html.parser')
+        params = {
+            "engine": "google",
+            "q": query,
+            "api_key": SERPAPI_KEY,
+            "num": num_results,
+            "hl": "en",
+            "gl": "in" # India Region for local context
+        }
         
-        for div in soup.find_all('div', class_='algo'):
-            title_tag = div.find('h3', class_='title')
-            desc_tag = div.find('div', class_='compText')
-            
-            if title_tag and desc_tag:
-                title = title_tag.text.strip()
-                link = title_tag.find('a').get('href', '') if title_tag.find('a') else ""
-                snippet = desc_tag.text.strip()
+        response = requests.get("https://serpapi.com/search", params=params, timeout=10)
+        data = response.json()
+        
+        # 'organic_results' से असली Google सर्च का डेटा निकालना
+        if "organic_results" in data:
+            for item in data["organic_results"]:
+                title = item.get("title", "")
+                link = item.get("link", "")
+                snippet = item.get("snippet", "")
                 
-                results.append({"title": title, "link": link, "snippet": snippet})
-            
-            if len(results) >= num_results:
-                break
+                if title and snippet:
+                    results.append({
+                        "title": title, 
+                        "link": link, 
+                        "snippet": snippet
+                    })
+                    
     except Exception as e:
-        print(f"Yahoo Scraper Error: {e}")
-
-    # 2. GOOGLE NEWS RSS FALLBACK (If Yahoo fails or for latest news)
-    if not results:
-        try:
-            rss_url = f"https://news.google.com/rss/search?q={urllib.parse.quote(query)}&hl=en-IN&gl=IN&ceid=IN:en"
-            res = requests.get(rss_url, headers=headers, timeout=5)
-            soup = BeautifulSoup(res.content, 'xml')
-            items = soup.find_all('item')
-            
-            for item in items[:num_results]:
-                title = item.title.text if item.title else ""
-                link = item.link.text if item.link else ""
-                pub_date = item.pubDate.text if item.pubDate else ""
-                if title:
-                    results.append({"title": title, "link": link, "snippet": f"News Published on: {pub_date}"})
-        except Exception as e:
-            print(f"News Scraper Error: {e}")
-
-    # 3. WIKIPEDIA FALLBACK
-    if not results:
-        try:
-            wiki_url = f"https://en.wikipedia.org/w/api.php?action=query&list=search&srsearch={urllib.parse.quote(query)}&utf8=&format=json"
-            res = requests.get(wiki_url, headers=headers, timeout=5).json()
-            for item in res.get('query', {}).get('search', [])[:num_results]:
-                clean_snippet = BeautifulSoup(item['snippet'], "html.parser").text
-                results.append({
-                    "title": item['title'], 
-                    "link": f"https://en.wikipedia.org/wiki/{urllib.parse.quote(item['title'])}", 
-                    "snippet": clean_snippet
-                })
-        except Exception as e:
-            pass
+        print(f"SerpApi Error: {e}")
 
     return results
 
@@ -94,7 +67,7 @@ def home():
 @app.route('/api/chat', methods=['POST'])
 def chat():
     if not API_KEY or not INVOKE_URL or not MODEL_ID:
-        return Response("Server Error: Secrets missing.", status=500)
+        return Response("Server Error: Secrets missing. Check NVIDIA API Key.", status=500)
 
     data = request.get_json() or {}
     user_message = data.get("message", "")
@@ -119,18 +92,18 @@ def chat():
     3. Do NOT say "Based on the provided search results". Just answer naturally and confidently, citing the sources/links if needed.
     """
 
-    # 🚀 --- AUTO-SEARCH INJECTION ---
+    # 🚀 --- AUTO-SEARCH INJECTION (USING SERPAPI) ---
     if is_search:
         scraped_data = web_search_scraper(user_message)
         
-        search_context = "\n\n--- [LIVE VERIFIED WEB DATA] ---\n"
+        search_context = "\n\n--- [LIVE VERIFIED GOOGLE SEARCH DATA] ---\n"
         if scraped_data:
             for idx, res in enumerate(scraped_data):
                 search_context += f"{idx+1}. TITLE: {res['title']}\nSNIPPET: {res['snippet']}\nURL: {res['link']}\n\n"
             
-            search_context += "[SYSTEM COMMAND: Use the above live data to answer the user accurately.]"
+            search_context += "[SYSTEM COMMAND: Use the above live Google data to answer the user accurately. Synthesize the info naturally without claiming you read it from a prompt.]"
         else:
-            search_context += "[SYSTEM ALERT: Live search blocked. Rely on your existing knowledge, but DO NOT hallucinate recent news.]"
+            search_context += "[SYSTEM ALERT: Live search did not return results. Rely on your existing knowledge, but DO NOT hallucinate recent news.]"
             
         user_message = f"USER QUERY: {user_message}\n\n{search_context}"