import gradio as gr import requests import json import os # API and environment variables 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 BASE_SYSTEM_MESSAGE = "I carefully provide accurate, factual, thoughtful, nuanced answers and am brilliant at reasoning." def clear_chat(): """Clears the chat history and message state.""" print("Clearing chat...") chat_history_state.value = [] chatbot.textbox.value = "" def user(message, history): """Updates the chat history with the user message.""" print(f"User message: {message}") history = history or [] history.append({"role": "user", "content": message}) return history def call_nvidia_api(history, system_message, max_tokens, temperature, top_p): """Calls the NVIDIA API to generate a response.""" # Transforma o histórico de listas de listas para o formato esperado pela API messages = [{"role": "system", "content": system_message}] for msg in history: messages.append({"role": "user", "content": msg[0]}) if msg[1]: # Garante que não adicionamos mensagens vazias do assistente messages.append({"role": "assistant", "content": 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"] history.append(["", assistant_message]) # Adiciona a nova mensagem do assistente ao histórico return history def chatbot_submit(message, chat_history, system_message, max_tokens_val, temperature_val, top_p_val): """Submits the user message to the chatbot and updates the chat history.""" # Adiciona a mensagem do usuário ao histórico updated_history = chat_history + [[message, ""]] # Chama a API da NVIDIA para gerar uma resposta updated_history = call_nvidia_api(updated_history, system_message, max_tokens_val, temperature_val, top_p_val) # A última entrada no histórico atualizado contém a última mensagem do usuário e a resposta do assistente if updated_history and updated_history[-1][1]: assistant_message = updated_history[-1][1] else: assistant_message = "Desculpe, ocorreu um erro ao gerar a resposta." return assistant_message, updated_history chat_history_state = gr.State([]) 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) with gr.Blocks() as demo: chat_history_state = gr.State([]) chatbot = gr.ChatInterface( fn=chatbot_submit, additional_inputs=[system_msg, max_tokens, temperature, top_p], title="LLAMA 70B Free Demo", 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", ) def clear_chat(): chat_history_state.value = [] chatbot.textbox.value = "" chatbot.clear() demo.launch()