import gradio as gr import requests import os import json # Carrega a chave da API do ambiente ou define diretamente 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 call_nvidia_api(history, system_message, max_tokens, temperature, top_p): messages = [{"role": "system", "content": system_message}] if system_message else [] messages.extend([{"role": "user", "content": msg[0]}, {"role": "assistant", "content": msg[1]}] for msg in history if msg[1]) payload = { "messages": messages, "temperature": temperature, "top_p": top_p, "max_tokens": max_tokens, "stream": False } 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() if response_body.get("choices"): assistant_message = response_body["choices"][0]["message"]["content"] # Retorna tanto a mensagem formatada para o usuário quanto a estrutura completa para o histórico da API return assistant_message, response_body["choices"][0] else: return "Desculpe, ocorreu um erro ao gerar a resposta.", None def chatbot_submit(message, chat_history_ui, chat_history_api, system_message, max_tokens_val, temperature_val, top_p_val): print("Updating chatbot...") # Chama a API da NVIDIA para gerar uma resposta assistant_message, api_response = call_nvidia_api(chat_history_api, system_message, max_tokens_val, temperature_val, top_p_val) # Atualiza o histórico da interface do usuário chat_history_ui.append([message, assistant_message]) # Atualiza o histórico da API se a resposta incluir o formato esperado if api_response: chat_history_api.append(api_response) return assistant_message, chat_history_ui, chat_history_api 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=1024) temperature = gr.Slider(0.0, 1.0, label="Temperature", step=0.1, value=0.2) top_p = gr.Slider(0.0, 1.0, label="Top P", step=0.05, value=0.7) # Gradio interface setup with gr.Blocks() as demo: chat_history_state_ui = gr.State([]) chat_history_state_api = gr.State([]) # Outros componentes da interface... chatbot = gr.ChatInterface( fn=chatbot_submit, inputs=[gr.Textbox(label="Your Message"), chat_history_state_ui, chat_history_state_api, system_msg, max_tokens, temperature, top_p], outputs=[gr.Text(label="Assistant Response"), chat_history_state_ui, chat_history_state_api], title="Chatbot Interface" description="""
Llama 2 is a large language AI model capable of generating text and code in response to prompts.
How to Use:
Powered by NVIDIA's cutting-edge AI API, LLAMA 2 70B offers an unparalleled opportunity to interact with an AI model of exceptional conversational ability, accessible to everyone at no cost.
HF Created by: @artificialguybr (Twitter)
Discover more: artificialguy.com
""", submit_btn="Submit", clear_btn="🗑️ Clear", ) demo.launch()