import gradio as gr
from huggingface_hub import InferenceClient
import os

# Твой токен из переменной окружения
token = os.getenv("Test2")

# Клиент HF — используем text-generation
client = InferenceClient(
    model="mistralai/Mistral-7B-Instruct-v0.3",
    token=token,
    provider="hf-inference"
)

def build_prompt(system_message, history, user_message):
    """
    Превращаем историю чата в обычный текстовый prompt,
    потому что text_generation НЕ поддерживает формат messages.
    """
    prompt = f"System: {system_message}\n\n"

    for user, assistant in history:
        if user:
            prompt += f"User: {user}\n"
        if assistant:
            prompt += f"Assistant: {assistant}\n"

    prompt += f"User: {user_message}\nAssistant:"
    return prompt


def respond(message, history, system_message, max_tokens, temperature, top_p):
    # Формируем текстовый prompt
    prompt = build_prompt(system_message, history, message)

    response = ""

    # Стримим text_generation
    for token in client.text_generation(
        prompt,
        max_new_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        text = token.token.text
        response += text
        yield response


# System prompt
my_value2 = """ты вежливый чат бот, отвечай коротко и по делу на русском языке."""

demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value=my_value2, label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p"),
    ],
)

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