app.py

import gradio as gr
from huggingface_hub import InferenceClient

client = InferenceClient("Grandediw/lora_model")

def respond(message, history, system_message, max_tokens, temperature, top_p):
    # Convert tuple-based history to messages if needed
    messages = [{"role": "system", "content": system_message}]
    for user_msg, assistant_msg in history:
        if user_msg:
            messages.append({"role": "user", "content": user_msg})
        if assistant_msg:
            messages.append({"role": "assistant", "content": assistant_msg})
    messages.append({"role": "user", "content": message})

    response = ""
    for partial in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = partial.choices[0].delta.content
        response += token
        yield response

with gr.Blocks(title="Enhanced LORA Chat Interface") as demo:
    gr.Markdown(
        """
        # LORA Chat Assistant
        Welcome! This is a demo of a LORA-based Chat Assistant.  
        Start by entering your prompt below.
        """
    )

    with gr.Row():
        # System message and other parameters
        with gr.Column():
            system_message = gr.Textbox(
                value="You are a friendly Chatbot.",
                label="Initial Behavior (System Message)",
                lines=3,
                placeholder="Describe how the assistant should behave..."
            )
            max_tokens = gr.Slider(
                minimum=1, maximum=2048, value=512, step=1,
                label="Max new tokens"
            )
            temperature = gr.Slider(
                minimum=0.1, maximum=4.0, value=0.7, step=0.1,
                label="Temperature"
            )
            top_p = gr.Slider(
                minimum=0.1, maximum=1.0, value=0.95, step=0.05,
                label="Top-p (nucleus sampling)"
            )

        # Create the chat interface using tuple format
        # Note: `type='tuple'` preserves the (user, assistant) tuple format.
        chat = gr.ChatInterface(
            fn=respond,
            additional_inputs=[system_message, max_tokens, temperature, top_p],
            type='tuples'
        )

if __name__ == "__main__":
    demo.launch()