app.py

import gradio as gr
import requests
import json
import os

API_KEY = os.getenv('API_KEY') 
INVOKE_URL = "https://api.nvcf.nvidia.com/v2/nvcf/pexec/functions/df2bee43-fb69-42b9-9ee5-f4eabbeaf3a8"
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):
    print(f"User message: {message}")
    history = history or []
    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
    }

    session = requests.Session()
    response = session.post(INVOKE_URL, headers=headers, json=payload)
    # Novo método de polling para verificar o status da resposta
    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()
    
    # Processar a resposta da API aqui
    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, top_k, repetition_penalty):
    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("## Your Chatbot Interface")
    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, 512, label="Max Tokens", step=20, value=500)
    temperature = gr.Slider(0.0, 1.0, label="Temperature", step=0.1, value=0.7)
    top_p = gr.Slider(0.0, 1.0, label="Top P", step=0.05, value=0.95)
    chat_history_state = gr.State([])

    def update_chatbot(message, chat_history):
        print("Updating chatbot...")
        # Ensure the user's message is not added twice
        if not chat_history or (chat_history and chat_history[-1]["role"] != "user"):
            chat_history = user(message, chat_history)
        chat_history, _ = chat(chat_history, system_msg.value, max_tokens.value, temperature.value, top_p.value, 40, 1.1)
        # Format messages for display, removing roles from content
        formatted_chat_history = [(msg["role"], msg["content"]) for msg in chat_history]
        return formatted_chat_history, chat_history, ""


    submit.click(
        fn=update_chatbot,
        inputs=[message, chat_history_state],
        outputs=[chatbot, chat_history_state, message]
    )

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

demo.launch()