app.py

import gradio as gr
import requests
import os
import json

API_KEY = os.getenv('API_KEY')
INVOKE_URL = "https://api.nvcf.nvidia.com/v2/nvcf/pexec/functions/0e349b44-440a-44e1-93e9-abe8dcb27158"
FETCH_URL_FORMAT = "https://api.nvcf.nvidia.com/v2/nvcf/pexec/status/"

headers = {
    "Authorization": f"Bearer {API_KEY}",
    "Accept": "application/json",
    "Content-Type": "application/json",
}

BASE_SYSTEM_MESSAGE = "I carefully provide accurate, factual, thoughtful, nuanced answers and am brilliant at reasoning."

def clear_chat(chat_history_state, chat_message):
    print("Clearing chat...")
    chat_history_state = []
    chat_message = ''
    return chat_history_state, chat_message

def user(message, history, system_message=None):
    print(f"User message: {message}")
    history = history or []
    if system_message:  # Check if a system message is provided and should be added
        history.append({"role": "system", "content": system_message})
    history.append({"role": "user", "content": message})
    return history

def call_nvidia_api(history, max_tokens, temperature, top_p):
    payload = {
        "messages": history,
        "temperature": temperature,
        "top_p": top_p,
        "max_tokens": max_tokens,
        "stream": False
    }

    print(f"Payload enviado: {payload}")  # Imprime o payload enviado

    session = requests.Session()
    response = session.post(INVOKE_URL, headers=headers, json=payload)

    while response.status_code == 202:
        request_id = response.headers.get("NVCF-REQID")
        fetch_url = FETCH_URL_FORMAT + request_id
        response = session.get(fetch_url, headers=headers)
    
    response.raise_for_status()
    response_body = response.json()

    print(f"Payload recebido: {response_body}")  # Imprime o payload recebido

    if response_body["choices"]:
        assistant_message = response_body["choices"][0]["message"]["content"]
        history.append({"role": "assistant", "content": assistant_message})
    
    return history

def chat(history, system_message, max_tokens, temperature, top_p):
    print("Starting chat...")
    updated_history = call_nvidia_api(history, max_tokens, temperature, top_p)
    return updated_history, ""

# Gradio interface setup
with gr.Blocks() as demo:
    with gr.Row():
        with gr.Column():
            gr.Markdown("LLAMA 70B Free Demo")
            description="""
<div style="text-align: center; font-size: 1.5em; margin-bottom: 20px;">
    <strong>Explore the Capabilities of LLAMA 70B</strong>
</div>
<p>Code Llama is an LLM capable of generating code, and natural language about code, from both code and natural language prompts.
</p>
<p> <strong>How to Use:</strong></p>
<ol>
    <li>Enter your <strong>message</strong> in the textbox to start a conversation or ask a question.</li>
    <li>Adjust the <strong>Temperature</strong> and <strong>Top P</strong> sliders to control the creativity and diversity of the responses.</li>
    <li>Set the <strong>Max Tokens</strong> slider to determine the length of the response.</li>
    <li>Use the <strong>System Message</strong> textbox if you wish to provide a specific context or instruction for the AI.</li>
    <li>Click <strong>Send message</strong> to submit your query and receive a response from LLAMA70B.</li>
    <li>Press <strong>New topic</strong> to clear the chat history and start a new conversation thread.</li>
</ol>
<p> <strong>Powered by NVIDIA's cutting-edge AI API, LLAMA 70B offers an unparalleled opportunity to interact with an AI model of exceptional conversational ability, accessible to everyone at no cost.</strong></p>
<p> <strong>HF Created by:</strong> @artificialguybr (<a href="https://twitter.com/artificialguybr">Twitter</a>)</p>
<p> <strong>Discover more:</strong> <a href="https://artificialguy.com">artificialguy.com</a></p>
"""
    gr.Markdown(description)
    chatbot = gr.Chatbot()
    message = gr.Textbox(label="What do you want to chat about?", placeholder="Ask me anything.", lines=3)
    submit = gr.Button(value="Send message")
    clear = gr.Button(value="New topic")
    system_msg = gr.Textbox(BASE_SYSTEM_MESSAGE, label="System Message", placeholder="System prompt.", lines=5)
    max_tokens = gr.Slider(20, 1024, label="Max Tokens", step=20, value=500, interactive=True)
    temperature = gr.Slider(0.0, 1.0, label="Temperature", step=0.1, value=0.7, interactive=True)
    top_p = gr.Slider(0.0, 1.0, label="Top P", step=0.05, value=0.95, interactive=True)
    chat_history_state = gr.State([])

    # Ajuste na definição da função update_chatbot para aceitar o valor atualizado do system_msg
    def update_chatbot(message, chat_history, system_message, max_tokens, temperature, top_p):
        print("Updating chatbot...")
        if not chat_history or (chat_history and chat_history[-1]["role"] != "user"):
            chat_history = user(message, chat_history, system_message if not chat_history else None)
        else:
            chat_history = user(message, chat_history)
        chat_history, _ = chat(chat_history, system_message, max_tokens, temperature, top_p)
    
        formatted_chat_history = []
        for user_msg, assistant_msg in zip([msg["content"].strip() for msg in chat_history if msg["role"] == "user"],
                                        [msg["content"].strip() for msg in chat_history if msg["role"] == "assistant"]):
            if user_msg or assistant_msg:  # Verify if either message is not empty
                formatted_chat_history.append([user_msg, assistant_msg])
    
        return formatted_chat_history, chat_history, ""
    
    submit.click(
        fn=update_chatbot,
        inputs=[message, chat_history_state, system_msg, max_tokens, temperature, top_p],
        outputs=[chatbot, chat_history_state, message]
    )

    clear.click(
        fn=clear_chat,
        inputs=[chat_history_state, message],
        outputs=[chat_history_state, message]
    )

demo.launch()