app.py

import gradio as gr
from huggingface_hub import InferenceClient

client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")


def respond(message, chat_history, system_message, max_tokens, temperature, top_p):
    messages = [{"role": "system", "content": system_message}]
    
    for user_msg, bot_msg in chat_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 message_obj in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message_obj.choices[0].delta.content or ""
        response += token
        yield chat_history + [(message, response)]


def initialize():
    # Initialer Verlauf mit Begrüßung
    return [("", "Hallo! Ich bin ein hilfreicher Assistent. Ich beantworte alle Fragen auf Deutsch.")]


with gr.Blocks() as demo:
    chatbot = gr.Chatbot(value=initialize(), height=400)

    with gr.Row():
        msg = gr.Textbox(label="Nachricht", placeholder="Stelle mir eine Frage...")
        send_btn = gr.Button("Senden")

    system_message = gr.Textbox(
        value="Du bist ein hilfreicher Assistent. Antworte immer auf Deutsch.",
        label="System message"
    )
    max_tokens = gr.Slider(1, 2048, value=512, step=1, label="Max new tokens")
    temperature = gr.Slider(0.1, 4.0, value=0.7, step=0.1, label="Temperature")
    top_p = gr.Slider(0.1, 1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)")

    send_btn.click(
        respond,
        inputs=[msg, chatbot, system_message, max_tokens, temperature, top_p],
        outputs=chatbot
    )

    msg.submit(
        respond,
        inputs=[msg, chatbot, system_message, max_tokens, temperature, top_p],
        outputs=chatbot
    )

demo.launch()