app.py

import gradio as gr
import random
from huggingface_hub import InferenceClient

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

def respond(message, history):

    
    messages = [{"role": "system", "content": "You are a chatbot who chooses kindness, and serves daily, just like a diva."}]
    if history:
        messages.extend(history)
        
    messages.append({"role" : "user", "content" : message})    

    response = ""

    for message in client.chat_completion(
        messages, 
        max_tokens = 1000,
       # temperature= .9, #code decimal value between 0-2
       # top_p= .8 #0-1
        stream=True
    
    ):
        token = message.choices[0].delta.content
        response += token
        yield response
        
            

    
def echo(message, history):    
    choices = ["yes", "no","No, you're not eating", "Yes, ate down diva", "No, you're not slick","Not today hunny", "Yes, level 100 gyatt","Naur, Bad girl", "Naur, Bad boy","Oui,good girl","Oui, good boy"]
    #use random to select one of those messages
    random_word = random.choice(choices)
    return random_word

chatbot = gr.ChatInterface(respond, type = "messages")

chatbot.launch(debug=True)