import gradio as gr
import random
import os

from huggingface_hub import InferenceClient

# os.environ["HF_TOKEN"] = "insert_token"

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

css_code = """
    .gradio-container {
        font-family: 'Arial', sans-serif;
    }
    .message.user {
        background-color: #e0f2f7; /* Light blue for user messages */
        color: #333;
    }
    .message.bot {
        background-color: #f0f0f0; /* Light gray for bot messages */
        color: #555;
    }
    """

#def respond_yes_no_randomly(message, history):
    # responses = ["Yes", "No"]
    # return random.choice(responses)

def respond(message, history):
    messages = [{"role": "system", "content": "You are a cutesy chatbot."}]

    if history:
        messages.extend(history)

    messages.append({"role": "user", "content": message})
    
    response = client.chat_completion(
        messages,
        max_tokens=100
    )
    
    return response['choices'][0]['messages']['content'].strip()

chatbot = gr.ChatInterface(respond, type="messages", title="Cutesy Chatbot", description="This is a very cute chatbot!", css=css_code)

chatbot.launch()