app.py

from huggingface_hub import InferenceClient
import gradio as gr
import os 

# Klient für die Inferenz
client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")

# Geheime Eingabeaufforderung aus Umgebungsvariablen
secret_prompt = os.getenv("SECRET_PROMPT")

def format_prompt(new_message, history):
    prompt = secret_prompt
    for user_msg, bot_msg in history:
        prompt += f"[INST] {user_msg} [/INST]"
        prompt += f" {bot_msg}</s> "
    prompt += f"[INST] {new_message} [/INST]"
    return prompt

def generate(prompt, history, temperature=0.25, max_new_tokens=1024, top_p=0.95, repetition_penalty=1.0):
    # Konfiguration der Parameter
    temperature = float(temperature)
    if temperature < 1e-2:
        temperature = 1e-2
    top_p = float(top_p)

    generate_kwargs = dict(
        temperature=temperature,
        max_new_tokens=max_new_tokens,
        top_p=top_p,
        repetition_penalty=repetition_penalty,
        do_sample=True,
        seed=727,
    )

    formatted_prompt = format_prompt(prompt, history)
    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
    output = ""

    for response in stream:
        output += response.token.text
        yield output
    return output

# Chatbot ohne Avatare und mit transparentem Design
hesse_chatbot = gr.Chatbot(bubble_full_width=True, show_label=False, show_copy_button=False, likeable=False)

# Minimalistisches Theme und Konfiguration der Gradio-Demo
theme = 'syddharth/gray-minimal'
demo = gr.ChatInterface(fn=generate, chatbot=hesse_chatbot, title="Prompt-Assistant", theme=theme)

demo.queue().launch(show_api=False)