Update app_low.py

app_low.py

@@ -20,6 +20,7 @@ pipe = pipeline(
     device=device,  # 0 for GPU, -1 for CPU
 )
 
+
 # ============================================================
 # 2️⃣ Define the generation function (chat-template style)
 # ============================================================
@@ -51,6 +52,18 @@ def enhance_prompt(user_prompt, temperature, max_tokens, chat_history):
 
     return chat_history
 
+def extract_later_part(user_prompt, generated_text):
+    """
+    Remove the user prompt from the start of the generated text
+    and return only the new added details.
+    """
+    # Ensure matching is case-insensitive and strip extra spaces
+    user_prompt_clean = user_prompt.strip().lower()
+    gen_clean = generated_text.strip()
+
+    if gen_clean.lower().startswith(user_prompt_clean):
+        return gen_clean[len(user_prompt_clean):].strip(", ").strip()
+    return gen_clean
 
 # ===================== Prompt Enhancer Function =====================
 def enhance_prompt1(user_prompt, temperature, max_tokens, chat_history):
@@ -65,11 +78,12 @@ def enhance_prompt1(user_prompt, temperature, max_tokens, chat_history):
     output = pipe(prompt, max_new_tokens=256)
     enhanced_text = output[0]['generated_text']
     print(enhanced_text)
-
-    # Wrap in a chat message format
-    chat_history = chat_history or []
+    enhanced_text = output[0]['generated_text']
+    later_part = extract_later_part(user_prompt, enhanced_text)
     chat_history.append({"role": "user", "content": user_prompt})
-    chat_history.append({"role": "assistant", "content": enhanced_text})
+    chat_history.append({"role": "assistant", "content": later_part})
+
+
 
     return chat_history