Update app.py

app.py

@@ -33,26 +33,38 @@ if model_path:
     )
 
 # --- 3. Style / System Prompts ---
+# These are the "Buttons" logic to change how the AI behaves
 STYLE_SYSTEM_PROMPTS = {
-    "Default": "You are a helpful, polite assistant.",
-    "Short answer": "Answer as concisely as possible, usually in 1-3 sentences.",
-    "Detailed explanation": "Give clear, structured and detailed explanations.",
-    "Step-by-step reasoning": "Think step by step and explain your reasoning clearly.",
+    "Default": (
+        "You are a helpful, polite AI assistant. "
+    ),
+    "Short answer": (
+        "Answer as briefly as possible, usually in 1-3 sentences. "
+        "Give only the core information needed to answer the question. "
+        "Do not add extra explanations, lists, or examples unless the user asks for more detail."
+    ),
+    "Detailed explanation": (
+        "Give a clear, structured, and detailed explanation. "
+        "Break your answer into short paragraphs or bullet points when helpful. "
+        "Explain what, how, and why, but avoid unnecessary repetition or filler."
+    ),
+    "Step-by-step reasoning": (
+        "Solve the problem step by step. "
+        "First restate the task in your own words, then explain your reasoning in numbered steps, "
+        "and finally give a short final answer at the end. "
+        "Keep the reasoning easy to follow and avoid unrelated digressions."
+    ),
 }
 
 def _extract_text(content):
     if isinstance(content, list):
-        return "\n".join(
-            b.get("text", "")
-            for b in content
-            if isinstance(b, dict) and b.get("type") == "text"
-        )
+        return "\n".join(b.get("text", "") for b in content if isinstance(b, dict) and b.get("type") == "text")
     return str(content)
 
 def chat_fn(message, history, max_new_tokens, style):
-    if not llm:
-        return "Error: Model not loaded."
+    if not llm: return "Error: Model not loaded."
     
+    # Select the specific system prompt based on the button chosen
     system_prompt = STYLE_SYSTEM_PROMPTS.get(style, STYLE_SYSTEM_PROMPTS["Default"])
     
     prompt = f"System: {system_prompt}\nConversation:\n"
@@ -60,13 +72,12 @@ def chat_fn(message, history, max_new_tokens, style):
         role = msg.get("role")
         txt = _extract_text(msg.get("content", ""))
         if txt:
-            if role == "user":
-                prompt += f"User: {txt}\n"
-            elif role == "assistant":
-                prompt += f"Assistant: {txt}\n"
+            if role == "user": prompt += f"User: {txt}\n"
+            elif role == "assistant": prompt += f"Assistant: {txt}\n"
             
     prompt += f"User: {message}\nAssistant:"
 
+    # Default internal values for randomness
     output = llm(
         prompt,
         max_tokens=int(max_new_tokens),
@@ -76,48 +87,32 @@ def chat_fn(message, history, max_new_tokens, style):
     )
     return output["choices"][0]["text"].strip()
 
-# --- 4. Background CSS ---
-BACKGROUND_URL = "https://images.template.net/269645/Cute-Christmas-Background-edit-online-1.jpg"
+# --- 4. UI Controls ---
 
-custom_css = f"""
-.gradio-container {{
-  background-image: url('{BACKGROUND_URL}');
-  background-repeat: no-repeat;
-  background-size: cover;
-  background-position: center;
-}}
-/* Optional: make main blocks a bit translucent so text is readable */
-.gradio-container .block, .gradio-container .panel {{
-  background-color: rgba(255, 255, 255, 0.85);
-}}
-"""
+# Slider for length
+max_new_tokens_slider = gr.Slider(
+    minimum=16, 
+    maximum=256, 
+    value=64, 
+    step=8, 
+    label="Max Response Length"
+)
 
-# --- 5. Build UI inside Blocks (so we can use css) ---
-with gr.Blocks(title="Lab 2 - Fine-tuned GGUF model", css=custom_css) as demo:
-    gr.Markdown(
-        "# Lab 2 - Fine-tuned GGUF model\n"
-        "Chat with the fine-tuned Llama model. Use the controls below to change the response style."
-    )
-
-    with gr.Accordion("Controls", open=True):
-        max_new_tokens_slider = gr.Slider(
-            minimum=16,
-            maximum=256,
-            value=64,
-            step=8,
-            label="Max Response Length",
-        )
+# The "Buttons" at the bottom for Style
+style_radio = gr.Radio(
+    choices=["Default", "Short answer", "Detailed explanation", "Step-by-step reasoning"],
+    value="Detailed explanation",
+    label="Answer Style"
+)
 
-        style_radio = gr.Radio(
-            choices=["Default", "Short answer", "Detailed explanation", "Step-by-step reasoning"],
-            value="Detailed explanation",
-            label="Answer Style",
-        )
-
-    gr.ChatInterface(
-        fn=chat_fn,
-        additional_inputs=[max_new_tokens_slider, style_radio],
-    )
+# --- 5. Launch App (Clean / No Theme) ---
+demo = gr.ChatInterface(
+    fn=chat_fn,
+    title="Lab 2 – Fine-tuned GGUF model",
+    description="Chat with the fine-tuned Llama model. Use the controls below to change the response style.",
+    additional_inputs=[max_new_tokens_slider, style_radio],
+    additional_inputs_accordion="Controls",
+)
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()