Update app.py

app.py

@@ -4,14 +4,14 @@ from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
 import torch
 
 # =====================================================
-# 🔹 Load model Gemma 2B (cocok buat CPU di HF Space)
+# 🔹 Load model Gemma 2B
 # =====================================================
 model_name = "google/gemma-2b-it"
 
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
-    torch_dtype=torch.float32,  # biar kompatibel di CPU Space
+    torch_dtype=torch.float32,  # cocok buat CPU Space
     device_map="auto",
 )
 
@@ -22,7 +22,7 @@ pipe = pipeline(
     max_new_tokens=512,
     temperature=0.7,
     top_p=0.9,
-    repetition_penalty=1.1
+    repetition_penalty=1.1,
 )
 
 # =====================================================
@@ -34,23 +34,30 @@ def chat(user_input, history):
 
     history = history or []
 
-    # prompt dengan identitas pembuatnya
+    # build prompt dengan konteks
     prompt = (
         "Kamu adalah YuuVx, asisten AI yang dibuat oleh Yusuf Ramadhani. "
-        "Kamu cerdas, tenang, dan selalu berbicara dalam Bahasa Indonesia. "
-        "Gunakan gaya santai tapi sopan, jangan terlalu panjang, dan responsif.\n\n"
+        "Kamu cerdas, tenang, dan selalu berbicara dalam Bahasa Indonesia dengan nada santai tapi sopan. "
+        "Gunakan kalimat yang alami dan tidak terlalu panjang.\n\n"
     )
 
-    for role, text in history:
-        prompt += f"{role}: {text}\n"
-    prompt += f"User: {user_input}\nAssistant:"
+    for msg in history:
+        role = msg["role"]
+        content = msg["content"]
+        if role == "user":
+            prompt += f"User: {content}\n"
+        else:
+            prompt += f"YuuVx: {content}\n"
+
+    prompt += f"User: {user_input}\nYuuVx:"
 
     # generate jawaban
     output = pipe(prompt, max_new_tokens=256, do_sample=True)
-    response = output[0]["generated_text"].split("Assistant:")[-1].strip()
+    response = output[0]["generated_text"].split("YuuVx:")[-1].strip()
+
+    history.append({"role": "user", "content": user_input})
+    history.append({"role": "assistant", "content": response})
 
-    history.append(("User", user_input))
-    history.append(("Assistant", response))
     return history, history
 
 # =====================================================
@@ -60,9 +67,9 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
     gr.Markdown("<h1 style='text-align:center;'>🤖 YuuVx Chat - Gemma 2B</h1>")
     gr.Markdown("<p style='text-align:center;'>Dibuat oleh <b>Yusuf Ramadhani</b></p>")
 
-    chatbot = gr.Chatbot(label="💬 Chat", height=520)
-    msg = gr.Textbox(label="Ketik pesanmu di sini...")
-    state = gr.State()
+    chatbot = gr.Chatbot(label="💬 Chat dengan YuuVx", height=520, type="messages")
+    msg = gr.Textbox(label="Ketik pesanmu di sini...", placeholder="Tulis sesuatu...")
+    state = gr.State([])
 
     def respond(message, chat_history):
         result, chat_history = chat(message, chat_history)
@@ -71,7 +78,7 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
     msg.submit(respond, [msg, state], [msg, chatbot, state])
 
 # =====================================================
-# 🔹 Run app
+# 🔹 Run App
 # =====================================================
 if __name__ == "__main__":
     demo.launch()