app.py

from ctransformers import AutoModelForCausalLM
import gradio as gr
import random
import time



def call_chat(message):
    llm = AutoModelForCausalLM.from_pretrained("mistralai/Mixtral-8x7B-Instruct-v0.1", model_type="Mistral")
    for text in llm(message, stream=True):
        return text


gr.ChatInterface(call_chat).launch()