Update app.py

app.py

@@ -5,10 +5,8 @@ from transformers import AutoTokenizer, AutoModelForCausalLM
 from peft import PeftModel
 
 # ─────────────────────────────────────────────
-# Конфиг
-# ─────────────────────────────────────────────
-BASE_MODEL  = "Qwen/Qwen3-8B"
-LORA_MODEL  = "crambrodev/dragonvineAI-qwen3-hytale"
+BASE_MODEL = "Qwen/Qwen3-8B"
+LORA_MODEL = "crambrodev/dragonvineAI-qwen3-hytale"
 
 SYSTEM_PROMPT = """You are DragonvineAI — an expert Hytale modding assistant.
 You help developers create plugins and mods for Hytale servers.
@@ -24,8 +22,6 @@ Key facts about Hytale modding:
 
 Always provide working, well-commented code examples."""
 
-# ─────────────────────────────────────────────
-# Загрузка модели
 # ─────────────────────────────────────────────
 print("Loading tokenizer...")
 tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL, trust_remote_code=True)
@@ -43,22 +39,14 @@ model = PeftModel.from_pretrained(base_model, LORA_MODEL)
 model.eval()
 print("Model ready!")
 
-# ─────────────────────────────────────────────
-# Генерация ответа
 # ─────────────────────────────────────────────
 def respond(message, history, thinking_mode, max_tokens, temperature):
-    # Собираем историю
     messages = [{"role": "system", "content": SYSTEM_PROMPT}]
+    messages += history
 
-    for user_msg, assistant_msg in history:
-        messages.append({"role": "user",      "content": user_msg})
-        messages.append({"role": "assistant", "content": assistant_msg})
-
-    # Thinking mode Qwen3
     prefix = "/think " if thinking_mode else "/no_think "
     messages.append({"role": "user", "content": prefix + message})
 
-    # Токенизируем
     text = tokenizer.apply_chat_template(
         messages,
         tokenize=False,
@@ -66,7 +54,6 @@ def respond(message, history, thinking_mode, max_tokens, temperature):
     )
     inputs = tokenizer(text, return_tensors="pt").to(model.device)
 
-    # Генерируем
     with torch.no_grad():
         outputs = model.generate(
             **inputs,
@@ -81,23 +68,16 @@ def respond(message, history, thinking_mode, max_tokens, temperature):
         skip_special_tokens=True,
     )
 
-    # Убираем thinking блок из ответа если он есть
     if "<think>" in response and "</think>" in response:
         response = response.split("</think>")[-1].strip()
 
     return response
 
 # ─────────────────────────────────────────────
-# Gradio UI
-# ─────────────────────────────────────────────
-with gr.Blocks(title="DragonvineAI — Hytale Modding Assistant") as demo:
-    gr.Markdown("""
-    # 🐉 DragonvineAI — Hytale Modding Assistant
-    Ask anything about creating Hytale plugins and mods!
-    """)
+with gr.Blocks(title="DragonvineAI — Hytale Modding Assistant", theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# 🐉 DragonvineAI — Hytale Modding Assistant\nAsk anything about creating Hytale plugins and mods!")
 
-    # 2. Добавляем type="tuples", чтобы вернуть совместимость с форматом списков:
-    chatbot = gr.Chatbot(height=500, label="Chat", type="tuples")
+    chatbot = gr.Chatbot(height=500, label="Chat", type="messages")
 
     with gr.Row():
         msg = gr.Textbox(
@@ -107,53 +87,42 @@ with gr.Blocks(title="DragonvineAI — Hytale Modding Assistant") as demo:
         )
         submit = gr.Button("Send 🚀", scale=1, variant="primary")
 
-    with gr.Accordion("⚙️ Settings", open=False):
-        thinking = gr.Checkbox(
-            label="🧠 Thinking mode (slower but smarter)",
-            value=False,
-        )
-        max_tok = gr.Slider(128, 1024, value=512, step=64, label="Max tokens")
-        temp    = gr.Slider(0.1, 1.0, value=0.7, step=0.1, label="Temperature")
+    with gr.Accordion("Settings", open=False):
+        thinking = gr.Checkbox(label="Thinking mode (slower but smarter)", value=False)
+        max_tok  = gr.Slider(128, 1024, value=512, step=64, label="Max tokens")
+        temp     = gr.Slider(0.1, 1.0, value=0.7, step=0.1, label="Temperature")
 
     gr.Examples(
         examples=[
             "How do I create a simple Hytale plugin with a /hello command?",
             "Show me how to listen to player join events in Hytale",
-            "How do I create a custom NPC in Hytale?",
             "What does a basic manifest.json look like for a Hytale plugin?",
             "How do I register a command in Hytale?",
         ],
         inputs=msg,
     )
 
-    def user_submit(message, history, thinking, max_tok, temp):
-        history = history + [[message, None]]
+    def user_submit(message, history):
+        history = history + [{"role": "user", "content": message}]
         return "", history
 
     def bot_respond(history, thinking, max_tok, temp):
-        user_message = history[-1][0]
-        response = respond(user_message, history[:-1], thinking, max_tok, temp)
-        history[-1][1] = response
+        user_message = history[-1]["content"]
+        prev_history = history[:-1]
+        response = respond(user_message, prev_history, thinking, max_tok, temp)
+        history = history + [{"role": "assistant", "content": response}]
         return history
 
     submit.click(
-        user_submit,
-        inputs=[msg, chatbot, thinking, max_tok, temp],
-        outputs=[msg, chatbot],
+        user_submit, inputs=[msg, chatbot], outputs=[msg, chatbot]
     ).then(
-        bot_respond,
-        inputs=[chatbot, thinking, max_tok, temp],
-        outputs=chatbot,
+        bot_respond, inputs=[chatbot, thinking, max_tok, temp], outputs=chatbot
     )
 
     msg.submit(
-        user_submit,
-        inputs=[msg, chatbot, thinking, max_tok, temp],
-        outputs=[msg, chatbot],
+        user_submit, inputs=[msg, chatbot], outputs=[msg, chatbot]
     ).then(
-        bot_respond,
-        inputs=[chatbot, thinking, max_tok, temp],
-        outputs=chatbot,
+        bot_respond, inputs=[chatbot, thinking, max_tok, temp], outputs=chatbot
     )
 
 demo.launch()