app.py

import gradio as gr
from huggingface_hub import InferenceClient
from typing import List, Tuple

# Default settings
class ChatConfig:
    MODEL = "google/gemma-3-27b-it"
    DEFAULT_SYSTEM_MSG = "You are a super intelligent and useful Chatbot."
    DEFAULT_MAX_TOKENS = 512
    DEFAULT_TEMP = 0.3
    DEFAULT_TOP_P = 0.95

client = InferenceClient(ChatConfig.MODEL)

def generate_response(
    message: str,
    history: List[Tuple[str, str]],
    system_message: str = ChatConfig.DEFAULT_SYSTEM_MSG,
    max_tokens: int = ChatConfig.DEFAULT_MAX_TOKENS,
    temperature: float = ChatConfig.DEFAULT_TEMP,
    top_p: float = ChatConfig.DEFAULT_TOP_P
) -> str:
    messages = [{"role": "system", "content": system_message}]
    
    # Conversation history
    for user_msg, bot_msg in history:
        if user_msg:
            messages.append({"role": "user", "content": user_msg})
        if bot_msg:
            messages.append({"role": "assistant", "content": bot_msg})
    
    messages.append({"role": "user", "content": message})
    
    response = ""
    for chunk in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = chunk.choices[0].delta.content or ""
        response += token
        yield response


def create_interface() -> gr.ChatInterface:
    """Create and configure the chat interface."""
    # Custom CSS for a modern look
    custom_css = """
    .chatbot .message { 
        border-radius: 12px; 
        margin: 5px; 
        padding: 10px; 
    }
    .chatbot .user-message { 
        background-color: #e3f2fd; 
    }
    .chatbot .bot-message { 
        background-color: #f5f5f5; 
    }
    .gr-button { 
        border-radius: 8px; 
        padding: 8px 16px; 
    }
    """

    # Custom chatbot 
    chatbot = gr.Chatbot(
        label="Gemma Chat",
        avatar_images=("./user.png", "./botge.png"),
        height=450,
        show_copy_button=True
    )

    # Chat interface
    interface = gr.ChatInterface(
        fn=generate_response,
        chatbot=chatbot,
        title="Tomoniai's chat with Google-Gemma-3",
        theme=gr.themes.Soft(),
        css=custom_css,
        additional_inputs=[
            gr.Textbox(
                value=ChatConfig.DEFAULT_SYSTEM_MSG,
                label="System Prompt",
                lines=2,
                placeholder="Enter system message..."
            ),
            gr.Slider(
                minimum=1,
                maximum=8192,
                value=ChatConfig.DEFAULT_MAX_TOKENS,
                step=1,
                label="Max Tokens",
                info="Controls response length"
            ),
            gr.Slider(
                minimum=0.1,
                maximum=1.0,
                value=ChatConfig.DEFAULT_TEMP,
                step=0.1,
                label="Temperature",
                info="Controls randomness"
            ),
            gr.Slider(
                minimum=0.1,
                maximum=1.0,
                value=ChatConfig.DEFAULT_TOP_P,
                step=0.05,
                label="Top-P",
                info="Controls diversity"
            )
        ],
        additional_inputs_accordion=gr.Accordion(label="Advanced Settings", open=False)
    )
    
    return interface

def main():
    app = create_interface()
    app.launch(
        server_name="0.0.0.0",
        server_port=7860,
        share=False,
        show_api=False,
        show_error=True,
        debug=True
    )

if __name__ == "__main__":
    main()