app.py

import gradio as gr
from huggingface_hub import InferenceClient

# Initialize the Inference Client for the GPT-2 model (or "gpttrash")
client = InferenceClient("gpt2")

def respond(message, history, max_tokens, temperature, top_p):
    messages = []

    # Add the conversation history (user and assistant exchanges)
    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})

    # Add the current user message to continue the conversation
    messages.append({"role": "user", "content": message})

    response = ""

    # Get the model's response using chat completion
    for response_chunk in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = response_chunk.choices[0].delta.content
        response += token
        yield response

# Create Gradio Blocks layout for Hugging Face Spaces
with gr.Blocks() as demo:
    with gr.Row():
        user_input = gr.Textbox(label="User Input")
        history = gr.State()  # Keeps conversation history
    with gr.Row():
        max_tokens_slider = gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens")
        temperature_slider = gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature")
        top_p_slider = gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)")
    with gr.Row():
        output = gr.Textbox(label="Model Output")

    # Set up the chatbot functionality
    user_input.submit(respond, [user_input, history, max_tokens_slider, temperature_slider, top_p_slider], output)

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