Code Change

app.py

@@ -29,7 +29,7 @@ def chat():
             "temperature": 0.5,   # 🔥 Controls randomness (lower = more deterministic)
             "top_p": 0.9,         # 🎯 Focus on high-probability words
             "max_new_tokens": 50,  # ⏳ Limits response length
-            "stop_sequences": ["\nUser:", "[INST]"]  # ⛔ Stops response at natural points
+            "stop_sequences": ["User:"]  # ⛔ Stops response at natural points
         }
     }
         response = requests.post(API_URL, headers=headers, json=payload)
@@ -50,9 +50,7 @@ def chat():
         raw_output = data[0]['generated_text'] if isinstance(data, list) else data.get('generated_text', "No response")
         reply = raw_output.split("Bot:")[-1].strip()
 
-        # Remove trailing "User" if present
-        if reply.endswith("User"):
-            reply = reply[:-4].strip()
+        
         return jsonify({"reply": reply})
 
     except Exception as e: