Update app.py

app.py

@@ -2,17 +2,25 @@ import os
 from flask import Flask, render_template, request, jsonify
 from flask_cors import CORS
 from huggingface_hub import InferenceClient
+from duckduckgo_search import DDGS # Direct import for stability
 
 app = Flask(__name__)
 CORS(app)
 
-# 1. Initialize the Official Client
-# This uses your HF_TOKEN secret automatically
 client = InferenceClient(
     model="meta-llama/Llama-3.2-3B-Instruct",
     token=os.getenv("HF_TOKEN")
 )
 
+def get_live_data(query):
+    """Fetches real-time snippets from the web."""
+    try:
+        with DDGS() as ddgs:
+            results = [r['body'] for r in ddgs.text(query, max_results=3)]
+            return "\n".join(results)
+    except Exception:
+        return "No real-time data available at the moment."
+
 @app.route('/')
 def index():
     return render_template('index.html')
@@ -23,11 +31,26 @@ def ask():
         data = request.get_json()
         user_query = data.get("query")
 
-        # 2. Simple, Direct Call to Llama
+        # 1. Get real-time info
+        web_context = get_live_data(user_query)
+
+        # 2. Build the 'Augmented' Prompt
+        system_prompt = f"""You are a helpful research assistant. 
+        Use the following real-time web context to answer the user's request. 
+        If the context is empty, rely on your internal knowledge.
+        
+        WEB CONTEXT:
+        {web_context}
+        """
+
+        # 3. Call Llama
         response = ""
         for message in client.chat_completion(
-            messages=[{"role": "user", "content": user_query}],
-            max_tokens=500,
+            messages=[
+                {"role": "system", "content": system_prompt},
+                {"role": "user", "content": user_query}
+            ],
+            max_tokens=800,
             stream=True,
         ):
             token = message.choices[0].delta.content
@@ -37,9 +60,7 @@ def ask():
         return jsonify({"answer": response})
 
     except Exception as e:
-        print(f"Error: {str(e)}")
-        # This will tell us EXACTLY if the token is the problem
-        return jsonify({"answer": f"System Status: {str(e)}"}), 500
+        return jsonify({"answer": f"System Error: {str(e)}"}), 500
 
 if __name__ == "__main__":
     app.run(host="0.0.0.0", port=7860)