42

app.py

@@ -1,69 +1,158 @@
+import os
+import torch
 import gradio as gr
-from huggingface_hub import InferenceClient
+from transformers import AutoTokenizer, AutoModelForCausalLM
+from peft import PeftModel
 
+# ─────────────────────────────────────────────
+# Конфиг
+# ─────────────────────────────────────────────
+BASE_MODEL  = "Qwen/Qwen3-8B"
+LORA_MODEL  = "crambrodev/dragonvineAI-qwen3-hytale"
 
-def respond(
-    message,
-    history: list[dict[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-    hf_token: gr.OAuthToken,
-):
-    """
-    For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-    """
-    client = InferenceClient(token=hf_token.token, model="openai/gpt-oss-20b")
+SYSTEM_PROMPT = """You are DragonvineAI — an expert Hytale modding assistant.
+You help developers create plugins and mods for Hytale servers.
 
-    messages = [{"role": "system", "content": system_message}]
+Key facts about Hytale modding:
+- Plugins are written in Java or Kotlin
+- Entry point: extend JavaPlugin, implement setup() method
+- Manifest file: manifest.json defines plugin metadata
+- Build system: Gradle with the hytale-mod plugin
+- Commands: extend CommandBase, override executeSync()
+- Events: use event listener system to hook into game events
+- API package: com.hypixel.hytale.server.core.*
 
-    messages.extend(history)
+Always provide working, well-commented code examples."""
 
-    messages.append({"role": "user", "content": message})
+# ─────────────────────────────────────────────
+# Загрузка модели
+# ─────────────────────────────────────────────
+print("Loading tokenizer...")
+tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL, trust_remote_code=True)
 
-    response = ""
+print("Loading base model...")
+base_model = AutoModelForCausalLM.from_pretrained(
+    BASE_MODEL,
+    torch_dtype=torch.float16,
+    device_map="auto",
+    trust_remote_code=True,
+)
+
+print("Loading LoRA adapter...")
+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}]
+
+    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})
 
-    for message in client.chat_completion(
+    # Токенизируем
+    text = tokenizer.apply_chat_template(
         messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        choices = message.choices
-        token = ""
-        if len(choices) and choices[0].delta.content:
-            token = choices[0].delta.content
-
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-chatbot = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
-)
+        tokenize=False,
+        add_generation_prompt=True,
+    )
+    inputs = tokenizer(text, return_tensors="pt").to(model.device)
+
+    # Генерируем
+    with torch.no_grad():
+        outputs = model.generate(
+            **inputs,
+            max_new_tokens=max_tokens,
+            temperature=temperature,
+            do_sample=temperature > 0,
+            pad_token_id=tokenizer.eos_token_id,
+        )
+
+    response = tokenizer.decode(
+        outputs[0][inputs["input_ids"].shape[1]:],
+        skip_special_tokens=True,
+    )
+
+    # Убираем thinking блок из ответ�� 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", theme=gr.themes.Soft()) as demo:
+    gr.Markdown("""
+    # 🐉 DragonvineAI — Hytale Modding Assistant
+    Ask anything about creating Hytale plugins and mods!
+    """)
+
+    chatbot = gr.Chatbot(height=500, label="Chat")
+
+    with gr.Row():
+        msg = gr.Textbox(
+            placeholder="Ask about Hytale modding... e.g. 'How do I create a custom command?'",
+            label="Your question",
+            scale=4,
+        )
+        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")
+
+    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]]
+        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
+        return history
 
-with gr.Blocks() as demo:
-    with gr.Sidebar():
-        gr.LoginButton()
-    chatbot.render()
+    submit.click(
+        user_submit,
+        inputs=[msg, chatbot, thinking, max_tok, temp],
+        outputs=[msg, chatbot],
+    ).then(
+        bot_respond,
+        inputs=[chatbot, thinking, max_tok, temp],
+        outputs=chatbot,
+    )
 
+    msg.submit(
+        user_submit,
+        inputs=[msg, chatbot, thinking, max_tok, temp],
+        outputs=[msg, chatbot],
+    ).then(
+        bot_respond,
+        inputs=[chatbot, thinking, max_tok, temp],
+        outputs=chatbot,
+    )
 
-if __name__ == "__main__":
-    demo.launch()
+demo.launch()