app.py

import json
from huggingface_hub import InferenceClient
import gradio as gr
import os

# Laden der Prompts aus der JSON-Datei
def load_prompts_from_json(file_path):
    with open(file_path, 'r') as file:
        return json.load(file)

# Laden der Prompts aus 'prompts.json'
prompts = load_prompts_from_json('prompts.json')

# Inference client
client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")

# Secret prompt from environment variable
secret_prompt = os.getenv("SECRET_PROMPT")

def format_prompt(new_message, history, prompt_type='default'):
    prompt = prompts.get(prompt_type, 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=512, top_p=0.95, repetition_penalty=1.0, prompt_type='default'):
    # 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, prompt_type)
    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, history + [(prompt, output)]  # Speichere die Konversation

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

# Dropdown für Prompt-Typen
prompt_type_dropdown = gr.Dropdown(choices=list(prompts.keys()), label="Prompt Type", value='default')

# Minimalistisches Theme und Gradio Demo Konfiguration
theme = 'syddharth/gray-minimal'
demo = gr.Interface(
    fn=generate,
    inputs=[
        gr.Textbox(lines=2, label="Eingabe"),
        "state"  # State input for conversation history
    ],
    outputs=[samir_chatbot],
    title="Tutorial Master",
    theme=theme
)

# Launch the demo with the queue
demo.queue().launch(show_api=False)